Practice Questions - exam-ai-900 Flashcards

Question

Which AI service can you use to interpret the meaning of a user input such as `Call me back later?` A. Translator B. Text Analytics C. Speech D. Language Understanding (LUIS)

Answer 1

D. Language Understanding (LUIS) DISCUSSION: The question asks for the service that interprets the *meaning* of user input. Language Understanding (LUIS) is designed specifically for this purpose, understanding the intent behind the text. Although LUIS is being replaced by Conversational Language Understanding (CLU), it is still the correct answer based on the question's options. A. Translator is for converting text from one language to another, not understanding intent. B. Text Analytics can extract information like sentiment, but not necessarily the overall meaning or intent of a request. C. Speech deals with converting spoken audio to text, not understanding the meaning of text.

Answer 2

The correct matches are: * **Reliability and Safety** with **AI systems need to be reliable and safe in order to be trusted. It's important for a system to perform as it was originally designed and for it to respond safely to new situations. Its inherent resilience should resist intended or unintended manipulation. Rigorous testing and validation should be established for operating conditions to ensure that the system responds safely to edge cases, and A/B testing and champion/challenger methods should be integrated into the evaluation process.** * **Privacy and Security** with **A data holder is obligated to protect the data in an AI system, and privacy and security are an integral part of this system. Personal needs to be secured, and it should be accessed in a way that doesn't compromise an individual's privacy. Azure differential privacy protects and preserves privacy by randomizing data and adding noise to conceal personal information from data scientists.** These matches are based on the descriptions provided in the linked Microsoft documentation, which directly relate the definitions to the AI principles.

Answer 3

* **Identify objects in an image** matches with **Custom Vision** * **Automatically import data from an invoice to a database** matches with **Form Recognizer** * **Identify people in an image** matches with **Face** **Explanation:** * **Custom Vision** is designed for identifying objects in images by training custom models. * **Form Recognizer** is used to extract data from forms and documents, such as invoices. * **Face** is specifically for detecting, recognizing, and analyzing human faces in images.

Answer 4

A. sentiment analysis **Explanation:** Sentiment analysis is the process of determining the emotional tone behind a series of words, which perfectly describes the function of the app described in the question. * **B. speech recognition:** This involves converting spoken language into text. * **C. key phrase extraction:** This involves identifying the most important phrases in a text. * **D. entity recognition:** This involves identifying and categorizing key entities (e.g., people, organizations, locations) within a text.

Answer 5

A. Regression. Regression is used to predict a numeric value. The number of vehicles is a numeric value. Translation is changing something from one language to another. Classification is assigning something to a category. Clustering is grouping similar data points together.

Answer 6

* **Entity recognition** maps to **Identify the names of people and places mentioned in a news article.** * **Sentiment analysis** maps to **Determine whether customer feedback is positive or negative.** * **Translation** maps to **Convert a document from English to Spanish.**

Answer 7

B. Take a training sample that is representative of the population in the United Kingdom. To minimize bias, the training data should accurately reflect the population being modeled. This ensures the model's predictions are valid for all drivers in the UK. Option A, while potentially helpful in reducing direct discrimination, might not address indirect biases. Option C could introduce bias due to varying driving conditions and laws in different countries. Option D might still lead to bias if the original data pool isn't representative.

Answer 8

B. Register the model. DISCUSSION: Registering the model allows you to store and version your models in the Azure cloud, within your workspace. The model registry helps you organize and keep track of your trained models. A. Explaining the model provides documentation but does not track versions. C. Registering the training data is important for reproducibility but doesn't directly track model versions. D. Provisioning an inference cluster is a deployment step, not a version tracking mechanism.

Answer 9

No Yes No **Explanation** * **Statement 1: Language Service's question answering can directly query a SQL database.** This is **incorrect**. Language Service's Question Answering is designed to extract answers from pre-defined knowledge bases or documents, not to directly query databases. * **Statement 2: Language Service's question answering can provide the same answer to a question, even when asked in different ways.** This is **correct**. Question Answering uses the knowledge base and natural language understanding to provide consistent answers to semantically similar questions. * **Statement 3: Language Service's question answering can determine the intent of a user utterance.** This is **incorrect**. Intent recognition is typically handled by other NLP components or services like LUIS (Language Understanding Intelligent Service). Question Answering focuses on extracting answers from text and documents.

Answer 10

B. named entity recognition DISCUSSION: The correct answer is B. Named entity recognition (NER) is crucial for identifying key pieces of information within a user's text, such as product names (for orders), document titles (for retrieval), and order numbers (for status updates). Therefore, NER is essential for all three of the chatbot's functions. A. Sentiment analysis focuses on determining the emotional tone of text, which is not directly relevant to the tasks of accepting orders, retrieving documents, or providing status updates. C. Translation involves converting text from one language to another, which is not required for the chatbot's described functions. D. While language modeling is a fundamental part of NLP and is used under the hood in many NLP tasks, it's not the specific type of NLP that directly addresses the requirements of the chatbot. Language modeling is used to predict the next word in a sequence and create coherent responses but does not directly perform the actions requested.

Answer 11

C. A chatbot is the most suitable AI solution for providing customer support across email, phone, and live chat channels. Chatbots are designed to handle text-based or voice-based interactions, providing answers, resolving issues, and guiding users through various processes. A is incorrect because machine learning is a broader field that enables systems to learn from data, but it's not a specific solution for customer support channels. B is incorrect because computer vision deals with enabling computers to "see" and interpret images, which is not relevant to text or voice-based customer support. D is incorrect because natural language processing (NLP) is a component used within chatbots to understand and process human language, but it's not a complete customer support solution by itself.

Answer 12

Yes, Yes, No DISCUSSION: The question tests the understanding of supervised and unsupervised learning, specifically clustering and classification. * **Statement 1:** "You need to group a set of customer profiles based on purchase history and browsing behavior." This is a clustering scenario because the goal is to find natural groupings within the data without any prior knowledge of what those groups should be. Therefore, the answer is Yes. * **Statement 2:** "You need to organize patients into different groups based on symptoms and diagnostic results." Similar to the first statement, this is a clustering scenario because we are grouping patients based on their features (symptoms and diagnostic results) without predefined labels. Therefore, the answer is Yes. * **Statement 3:** "You need to predict whether a customer will buy a product based on demographic data." This is a classification problem because the goal is to predict a specific category (buy or not buy) based on input features (demographic data). Since classification is a supervised learning task, the answer is No.

Answer 13

C. regression **Explanation:** The question states that the machine learning model predicts the value of 'ColumnE' based on other numeric columns. Since 'ColumnE' contains numeric values, this is a regression problem. Regression models are used to predict continuous numerical values. * **A. analysis:** Analysis is a broad term and not a specific type of machine learning model. * **B. clustering:** Clustering is used to group similar data points together, not to predict a specific numerical value.

Answer 14

A. [Image](https://img.examtopics.com/ai-900/image230.png) DISCUSSION: The image depicts using facial analysis to infer the emotional state of kiosk users. This aligns with the concept of facial analysis as a method to understand emotions through facial expressions. Therefore, option A, which illustrates this scenario, is the correct answer.

Answer 15

A. Language Service B. Azure Bot Service DISCUSSION: The correct answers are A and B. * **A. Language Service:** Azure Cognitive Service for Language provides Natural Language Processing (NLP) features, which are essential for understanding and analyzing user queries in natural language. * **B. Azure Bot Service:** Azure Bot Service provides the platform and tools necessary to build, deploy, and manage intelligent bots. It integrates well with NLP services like Language Service. The incorrect answers are: * **C. Form Recognizer:** Form Recognizer is used for extracting data from forms and documents, which is not directly relevant to processing natural language queries against a knowledge base. * **D. Anomaly Detector:** Anomaly Detector is used for identifying unusual patterns in data, which is not directly relevant to building a bot that understands natural language.

Answer 16

Object detection

Answer 17

No, Yes, Yes **Explanation:** Based on the provided answers and discussion, the correct answer is No, Yes, Yes. The discussion confirms that Document Translation is a feature of the Azure Translator service and part of the Azure Cognitive Services.

Answer 18

Yes, No, Yes

Answer 19

D. key phrase extraction DISCUSSION: The question asks for identifying the main talking points in a collection of documents using the Language service. Key phrase extraction is designed to identify the main concepts and talking points within text. Therefore, option D is the correct choice. A. Language detection identifies the language of the text, but not the talking points. B. Sentiment analysis determines the emotional tone of the text, but not the talking points. C. Entity recognition identifies specific entities within the text (e.g., people, organizations), but not necessarily the main talking points.

Answer 20

* Box 1: Yes * Box 2: No * Box 3: Yes **Explanation:** * **Box 1:** The statement is correct. Azure Cognitive Service for Language includes language detection capabilities. * **Box 2:** The statement is incorrect. While OCR can handle some forms of handwriting, specifically trained models or other services (like Ink Recognizer) are typically used for robust handwritten detection. * **Box 3:** The statement is correct. Named Entity Recognition (NER) is a feature of Azure Cognitive Service for Language.

Answer 21

B. a compute resource Explanation: A compute resource is required to execute the pipeline steps, providing the necessary processing power for tasks like data transformation, training, and inference. Option A is incorrect because a registered model is needed after training, not before running the pipeline. Option C is incorrect because while Jupyter notebooks are helpful for experimentation, they are not required to run a pipeline in Azure Machine Learning designer.

Answer 22

Here's the mapping of AI workloads to the scenarios, based on the provided information and links: * **Extracting text from images:** Computer Vision (Specifically, Optical Character Recognition - OCR) * **Analyzing customer reviews to determine positive and negative feedback:** Natural Language Processing (Sentiment Analysis) * **Detecting unusual patterns in network traffic:** Anomaly Detection * **Grouping customers based on purchasing behavior:** Machine Learning (Clustering) **Explanation:** * **Computer Vision** is used for tasks related to understanding and analyzing images, including extracting text using OCR. * **Natural Language Processing (NLP)** is used for understanding and analyzing text, including sentiment analysis to determine the emotional tone of text. * **Anomaly Detection** is used for identifying unusual patterns or outliers in data, such as network traffic. * **Machine Learning (Clustering)** is used for grouping similar data points together, such as customers with similar purchasing behavior.

Answer 23

A. Clustering **Explanation:** The question asks about identifying groups of rows with similar numeric values. This is the core concept of clustering in machine learning. Clustering algorithms aim to group similar data points together into clusters based on their features. * **A. Clustering:** This is the correct answer because clustering is specifically designed for grouping similar data points together. * **B. Regression:** Regression is used for predicting a continuous numerical value based on input features. It's not suitable for identifying groups of similar rows. * **C. Classification:** Classification is used for assigning data points to predefined categories or classes. While it involves grouping, the groups are already known beforehand, unlike clustering where the groups are discovered.

Answer 24

[Image](https://img.examtopics.com/ai-900/image224.png) DISCUSSION: The image shows an example where the "sale price" is predicted based on other features. Since "sale price" is a numeric value, this is a regression problem. Regression is used to predict numeric values. The other options are not applicable here: classification is for categorical data, clustering is for grouping similar data points, and reinforcement learning is for training agents to make decisions in an environment.

Answer 25

The correct answer is **Reliability and Safety**. AI systems need to be reliable and safe in order to be trusted, perform as designed, and respond safely to new situations. The scenario given refers to self-driving cars making decisions based on data, where failures could cause accidents. The other options are less relevant.

Answer 26

D. A machine learning technique that analyzes data over time and identifies unusual changes. DISCUSSION: The correct answer is D because anomaly detection focuses on identifying unusual changes in data over time. Options A and C describe image classification, while option B describes natural language processing, none of which are directly related to anomaly detection.

Answer 27

D. fairness The discovery of gender bias in the AI-based loan approval system directly violates the principle of fairness. AI systems should treat all individuals equitably and avoid disparate impact on similarly situated groups. Options A, B, and C are incorrect because while important responsible AI principles, they don't directly address the issue of bias in outcomes. Accountability refers to taking ownership for the AI system's behavior. Reliability and safety focus on the system's consistent and secure performance. Transparency involves understanding how the AI system works and makes decisions.

Answer 28

[Image](https://img.examtopics.com/ai-900/image207.png)

Answer 29

* **Azure portal** with **Create an Azure Machine Learning workspace** * **Azure Machine Learning designer** with **Visually connect datasets and modules** * **Automated ML** with **Identify the best machine learning model for your data** **Explanation:** * The **Azure portal** is the primary interface for creating and managing Azure resources, including Azure Machine Learning workspaces. * **Azure Machine Learning designer** provides a drag-and-drop interface to build machine learning pipelines by visually connecting datasets and modules. * **Automated ML** automates the process of finding the best machine learning model for a given dataset by trying out different algorithms and hyperparameters.

Answer 30

1. Ingest and prepare dataset 2. Split data randomly into training and validation data 3. Train model 4. Evaluate model against validation dataset The steps listed are the standard process for creating and validating a machine learning model. First, the data must be brought in and cleaned. Second, to test the model's effectiveness, the data must be split into training and validation sets. Third, the model is trained using the training data. Fourth, the model is tested using the validation data to see how well it performs on data it has not seen before.

Answer 31

D. Optical character recognition (OCR) **Explanation:** Optical character recognition (OCR) is the correct answer because it's a technology that converts images of text (like scanned newspaper articles) into machine-readable text. * **A. Object detection:** Object detection identifies and locates specific objects within an image, but it doesn't convert text into a digital format. * **B. Facial recognition:** Facial recognition identifies and verifies faces in images or videos, which is irrelevant to digitizing newspaper articles. * **C. Image classification:** Image classification categorizes an entire image into a predefined class (e.g., "cat," "dog," "newspaper"), but it doesn't extract text from the image.

Answer 32

* **Identify specific content inside images** -> Custom Vision * **Extract text and data from documents** -> Form Recognizer * **Detect and analyze faces** -> Face

Answer 33

D. Optical character recognition (OCR) is the correct answer. Since the app is for scanning and storing expenses, OCR is needed to recognize the text on receipts and other expense documents. A. Face detection is used to identify faces in images, which is not relevant to scanning expense documents. B. Image classification categorizes an entire image into a predefined class (e.g., "receipt," "invoice"), but it doesn't extract the textual information needed for expense recording. C. Object detection identifies and locates specific objects within an image (e.g., identifying a "logo" or a "date field"), but it doesn't recognize the text itself, which is crucial for expense capture.

Answer 34

* **Speech** -> Transcribe audio into text * **Language Service** -> Extract intents from conversation * **Language Service** -> Translate text from one language to another **Explanation:** * **Speech:** The Azure Speech service is designed for speech-to-text and text-to-speech conversion. Therefore, it's the correct service for transcribing audio into text. * **Language Service:** The Language Service offers various natural language processing (NLP) features, including intent extraction from conversations (Conversational language understanding (CLU)) and text translation.

Answer 35

[Image](https://img.examtopics.com/ai-900/image228.png)

Answer 36

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0010200002.jpg) DISCUSSION: The question asks to complete the sentence about automatically extracting handwritten information. The correct answer is OCR (Optical Character Recognition), as this technology is specifically designed for converting images of text (handwritten or typed) into machine-readable text. The other options are not relevant to this process.

Answer 37

- **Speech** -> Convert spoken questions into text - **Language service** -> Interpret the meaning of questions - **Speech** -> Provide spoken responses **Explanation:** * **Speech (Convert spoken questions into text):** The Speech service offers speech-to-text functionality, which is essential for converting spoken questions into a textual format that the personal assistant can process. * **Language service (Interpret the meaning of questions):** The Language service, specifically Conversational Language Understanding (CLU), is designed to understand natural language. This enables the personal assistant to interpret the user's intent and extract relevant information from the questions. * **Speech (Provide spoken responses):** The Speech service also provides text-to-speech capabilities, which allow the personal assistant to generate spoken responses that the user can hear.

Answer 38

A. Splitting the data into training and testing sets and using the testing set for evaluation is the standard practice to assess how well the model generalizes to unseen data. Using the training set for evaluation (option D) can lead to overfitting and an overly optimistic assessment of the model's performance. Option B is incorrect because it also uses the training dataset for evaluation. Option C is incorrect because clustering does not involve labels.

Answer 39

C. Answer questions from multiple users simultaneously. The bot is built on a knowledge base of FAQs. Therefore, it can readily answer questions from multiple users simultaneously by retrieving and providing information from that knowledge base. Options A, B, and D would require the bot to perform actions beyond simply retrieving information, such as registering data or generating RMA numbers, necessitating additional skills or integrations.

Answer 40

- A customer's personal information must be visible only to staff who are involved in the decision-making process to ensure the privacy and security of sensitive data. **-> Privacy and security** - The decision-making process must be recorded so that staff can identify the reasoning behind a particular quote, promoting transparency and accountability. **-> Transparency** - The system must be accessible to customers who use screen readers or other assistive technology, ensuring inclusiveness and providing equal access to all users. **-> Inclusiveness** The question asks us to match the Microsoft responsible AI principles to specific requirements in the context of an automated insurance quote system. * **Privacy and security:** This principle directly relates to protecting customer data and ensuring it's only accessible to authorized personnel. * **Transparency:** This principle refers to the need for the system's decision-making process to be understandable and auditable. Recording the process allows staff to understand the reasoning behind quotes. * **Inclusiveness:** This principle emphasizes making the system accessible to all users, including those with disabilities who rely on assistive technologies.

Answer 41

A. regular expression DISCUSSION: The correct answer is A. Regular expressions (regex) are used to define patterns for identifying specific formats, like phone numbers. Options B, C, and D are incorrect because while machine learning, lists, and Pattern.Any entities have their uses in NLP, regular expressions are the best choice for the specific task of identifying a phone number pattern.

Answer 42

C. Language DISCUSSION: The question asks about extracting intent from a user input like "Call me back later." The Language service, specifically Conversational Language Understanding (CLU) or Language Understanding (LUIS), is designed for this purpose. It can understand the meaning and intent behind user text input. While Speech service can also perform intent recognition, the question does not specify that the input is audio, making Language a more general and appropriate answer. Azure Cognitive Search is for indexing and searching data, and Translator is for translating text from one language to another; thus, A and B are incorrect. Therefore, the correct answer is C.

Answer 43

B. Add utterances to the None intent DISCUSSION: The question asks how to handle utterances that are outside the intended scope of a Language Understanding model. Option B is correct because the None intent is specifically designed to categorize utterances that do not belong to any of the other custom intents. This allows the model to recognize when an utterance is outside of its intended scope. Option A is incorrect because exporting the model does not help in detecting out-of-scope utterances. Exporting is for backup or deployment purposes. Option C is incorrect because creating a prebuilt task entity focuses on extracting specific types of information from the text, not identifying out-of-scope utterances. Option D is incorrect because creating a new model would not address the issue of identifying utterances that don't fit within the existing model's scope. It would simply create a separate model.

Answer 44

confusion matrix

Answer 45

[Image](https://img.examtopics.com/ai-900/image260.png) DISCUSSION: The correct task to select is "Regression". The question states that we need to predict vehicle *sale values*, which are numerical values. Regression is the appropriate machine learning task for predicting numerical values. The other options are incorrect because they are used for different types of predictions (e.g., classification predicts categories).

Answer 46

Microsoft Copilot Studio, which includes the capabilities previously found in Power Virtual Agents, offers a no-code solution for building chatbots. It provides a guided, no-code graphical interface that allows subject matter experts to easily build bots without the need for coding or AI expertise. It simplifies the creation of conversational AI, enabling users to automate common inquiries and manage customer interactions more efficiently. Additionally, it supports rich, natural conversations with end-users, facilitating quick resolution of issues and improving customer satisfaction. [Image](https://img.examtopics.com/ai-900/image278.png)

Answer 47

[Image](https://img.examtopics.com/ai-900/image310.png)

Answer 48

C. Speech synthesis Speech synthesis (also known as text-to-speech) converts text into spoken audio. This is exactly what is required to generate narration from the provided scripts. Options A, B, and D are incorrect because they do not describe the process of converting text to speech. Language modeling predicts the next word in a sequence, speech recognition converts audio to text, and translation converts text from one language to another.

Answer 49

Yes No Yes **Explanation:** Based on the discussion, the correct answer is Yes, No, Yes. The question is identical to another question in a different order, and multiple users have confirmed this answer. Without the specific statements, it's impossible to provide a more detailed explanation of *why* each answer is correct, but the community consensus points to this solution.

Answer 50

B. helping users understand the decisions made by an AI system The principle of transparency in responsible AI focuses on making AI systems understandable and explainable to users. Option B directly reflects this by emphasizing the importance of helping users understand the decisions made by an AI system. Options A, C, and D relate to fairness, accountability, and security, respectively, but not transparency.

Answer 51

A. Face DISCUSSION: The question asks for a service to automatically tag *images of your friends*. The Azure Face service is specifically designed for face detection, recognition, and analysis. This makes it the most suitable choice for tagging images of friends. Option B, Form Recognizer, is used for extracting text and data from forms and documents. Option C, Language, is for natural language processing. Option D, Computer Vision, is a more general-purpose service for image analysis, but Face is more appropriate when the task is specifically related to identifying people.

Answer 52

* A smart device in the home that responds to questions such as “When is my next appointment?” is an example of conversational AI. (x) Yes ( ) No * An interactive webchat feature on a company website can be implemented by using Azure Bot Service. (x) Yes ( ) No * Automatically generating captions for pre-recorded videos is an example of conversation AI. ( ) Yes (x) No **Explanation:** * **Statement 1: Correct.** Conversational AI refers to technologies like chatbots and virtual assistants that can simulate human conversation. A smart device responding to questions falls under this category. * **Statement 2: Correct.** Azure Bot Service is designed to build, test, and deploy conversational bots for various platforms, including webchat features. * **Statement 3: Incorrect.** Automatically generating captions for videos is primarily an application of Natural Language Processing (NLP) and Speech-to-Text, not Conversational AI. While related, Conversational AI focuses on interactive dialogue, whereas caption generation focuses on transcribing audio into text.

Answer 53

D. inclusiveness The question explicitly mentions "unintentionally leave out specific user groups". This aligns directly with the principle of inclusiveness, which focuses on ensuring AI systems are accessible and usable by diverse populations, including those who might be marginalized or excluded due to factors like language, accent, or physical abilities. Options A, B, and C are incorrect because: * **Accountability** refers to establishing responsibility for the outcomes and decisions made by AI systems. * **Fairness** focuses on avoiding bias and ensuring equitable treatment across different groups. While related, inclusiveness goes further to consider usability and accessibility. * **Privacy and security** deals with protecting user data and ensuring the security of AI systems.

Answer 54

Object detection

Answer 55

[Image](https://img.examtopics.com/ai-900/image264.png)

Answer 56

The correct answer is the image that shows "Natural Language Processing". Natural Language Processing is usually associated with text processing. The question is asking to select the choice that correctly completes the sentence, and Natural Language Processing is the correct completion.

Answer 57

B. Form Recognizer **Explanation:** * **B. Form Recognizer:** This is the correct answer. Azure AI Form Recognizer (now known as Document Intelligence) is specifically designed to extract information from forms and documents like receipts. It can accurately identify and extract data fields such as dates, merchant names, totals, and taxes, which can then be used to populate a spreadsheet. * **A. Custom Vision:** Custom Vision is used for image recognition tasks, such as identifying objects or classifying images. It's not designed for extracting structured data from documents. * **C. Face:** Face service is used for facial recognition and analysis. It's irrelevant to processing receipts. * **D. Language:** Language service provides natural language processing capabilities, such as sentiment analysis and language detection. While it could potentially extract some information from receipt text, it's not optimized for this task like Form Recognizer is.

Answer 58

A. assigning the color pixels in an image to object names E. describing the contents of an image DISCUSSION: Options A and E are correct. Computer vision can be used to assign color pixels to object names (semantic segmentation or image labeling) and to describe the contents of an image (image captioning or image analysis). Option B is incorrect because detecting inconsistencies and anomalies in a stream of data is typically handled by anomaly detection algorithms, not computer vision. Option C is incorrect because creating visual representations of numerical data is typically handled by data visualization tools. Option D is incorrect because creating photorealistic images using three-dimensional models is more related to computer graphics and image generation than computer vision. Computer vision is about understanding existing images, not creating new ones.

Answer 59

B. Computer Vision Image Analysis capabilities

Answer 60

B. natural language processing DISCUSSION: The question describes a device that understands and responds to a question asked in natural language. This is the core function of Natural Language Processing (NLP). Option A is incorrect because knowledge mining focuses on extracting information from large datasets. Option C is incorrect because computer vision deals with processing visual data. Option D is incorrect because anomaly detection is about identifying unusual patterns in data.

Answer 61

B. Custom Vision Custom Vision allows users to train custom image recognition models using their own images. This aligns perfectly with the requirement of training the solution on images provided by the company. The other options are incorrect because: * Form Recognizer is used for extracting information from documents. * Face service is used for detecting and analyzing human faces. * Computer Vision provides pre-trained models, but does not allow training on custom datasets.

Answer 62

C. The system identifies authorized personnel, which directly implies facial detection and recognition. * **A is incorrect** because optical character recognition is used for reading text in images. * **B is incorrect** because semantic segmentation is used to classify each pixel in an image, it's not directly related to identifying individuals. * **D is incorrect** because image analysis is a broad term. Facial detection and recognition is a more specific and accurate description of what the system does.

Answer 63

B. Custom Vision object detection **Explanation:** The correct answer is B because object detection models identify the class of objects in an image, provide a probability score for the classification, and, most importantly, return the coordinates of a bounding box for each object. This is precisely what the app needs to identify the product's coordinates on the shelf. A is incorrect because Custom Vision classification only categorizes an entire image, not individual objects within it, and doesn't provide coordinate information. C is incorrect because Computer Vision Read extracts text from images, which is not the primary function of the app. D is incorrect because Computer Vision OCR (optical character recognition) also extracts text from images but doesn't identify and locate objects.

Answer 64

C. translation and key phrase extraction The solution processes manuscripts in different languages and categorizes them by topic. Since the manuscripts are in different languages, translation is necessary to understand the content. To categorize the manuscripts, key phrase extraction is used to identify the main topics or themes. Speech recognition is not needed because the input is text, not audio. Sentiment analysis is not directly related to categorizing by topic. Entity recognition could be used but key phrase extraction is more directly related to the categorization task. Language modeling is not necessary for the described task.

Answer 65

C. Form Recognizer DISCUSSION: Form Recognizer (now known as Document Intelligence) is designed to analyze text in documents and forms, including identifying sensitive information such as Personally Identifiable Information (PII). Option A is incorrect because Custom Vision is used for image recognition, not text analysis. Option B is incorrect because Conversational Language Understanding is focused on understanding the intent of user input in a conversational setting.

Answer 66

[Image](https://img.examtopics.com/ai-900/image284.png)

Answer 67

* Classification - This model predicts a category. * Clustering - This model identifies groupings in data. * Regression - This model predicts a numerical value. **Explanation:** * **Classification:** The goal of classification models is to predict which category a new observation belongs to, based on a training set of data containing observations whose category is already known. Spam detection is a typical example. * **Clustering:** Clustering models group similar data points together into clusters. This is an unsupervised learning technique, meaning the data is not labeled. Customer segmentation is a typical example. * **Regression:** Regression models are used to predict a continuous numerical value. House price prediction is a typical example.

Answer 68

B. regression

Answer 69

* Generate captions for images. -> Object detection * Extract movie title names from movie poster images. -> Optical character recognition (OCR) * Locate vehicles in images. -> Object detection **Explanation:** * **Generate captions for images -> Object detection:** Object detection identifies and locates specific objects within an image. By recognizing objects and their relationships, it can generate descriptive captions. While image classification can categorize an image, it doesn't provide the detail needed for a caption. Image analysis is the best answer, but it is not a choice, object detection is the next best thing. * **Extract movie title names from movie poster images -> Optical character recognition (OCR):** OCR is specifically designed to extract text from images. This is the most appropriate choice for recognizing and extracting movie titles from a poster. * **Locate vehicles in images -> Object detection:** Object detection excels at identifying and locating specific objects, like vehicles, within an image. Image classification would only categorize the entire image (e.g., "road scene"), not pinpoint individual vehicles.

Answer 70

B. Language **Explanation:** The Azure Language service (formerly LUIS) is designed to understand user intent from text inputs, which is crucial for chatbot development. * **A. Translator:** This service is for translating text between languages, not understanding intent. * **C. Azure Cognitive Search:** This service is for search capabilities within datasets, not interpreting user intent. * **D. Speech:** This service focuses on converting speech to text and vice versa, not directly understanding intent.

Answer 71

[Image](https://img.examtopics.com/ai-900/image292.png)

Answer 72

A. Azure Bot Service Bots created using Azure Bot Service can be integrated with voice-based solutions such as Cortana, Alexa, or Google Assistant. Azure Bot Service provides the necessary channels and interfaces to connect bots to these services, allowing them to process and respond to voice commands. This integration enables users to interact with the bots through natural language voice commands, extending the bot’s capabilities to various voice-enabled platforms. The other services do not provide the core bot functionality and channel integration needed.

Answer 73

- Identifies the location of the birds in the image: Object detection - Identifies the species of the birds in the image: Image classification

Answer 74

Regression.

Answer 75

D. Azure Bot Service **Explanation:** * **D. Azure Bot Service:** Azure Bot Service is designed for building, deploying, and managing intelligent bots that can interact with users across various channels like social media, phones, and digital assistants. This directly addresses the requirement of enabling customers to query order status via these channels. * **A. an Azure Machine Learning model:** Azure Machine Learning is used for building and deploying machine learning models, but it doesn't inherently provide the conversational interface needed for user interaction via multiple channels. * **B. the Translator service:** The Translator service is for translating text between languages and doesn't help with building a conversational interface for order status queries. * **C. a Form Recognizer model:** Form Recognizer is used to extract data from forms and documents. While it can be part of a larger solution, it doesn't directly provide the functionality for querying order statuses through various channels.

Answer 76

[Image](https://img.examtopics.com/ai-900/image274.png)

Answer 77

* **Predict the price of a house based on square footage and location:** Regression * **Segment customers into different groups to support a marketing initiative:** Clustering * **Predict whether or not a student will complete a university course:** Classification **Explanation:** * **Predict the price of a house based on square footage and location:** Regression is used for predicting a continuous numerical value (price). * **Segment customers into different groups to support a marketing initiative:** Clustering is used to group similar data points together (customers). * **Predict whether or not a student will complete a university course:** Classification is used for predicting a categorical outcome (complete/not complete, yes/no).

Answer 78

Classification DISCUSSION: The question asks about predicting whether a loan will be repaid. Since the outcome is binary (yes or no), this is a classification problem. Regression is used for predicting continuous values, and clustering is used for grouping data points. Therefore, classification is the correct answer.

Answer 79

C. **Explanation:** Classification is a supervised machine learning task where the goal is to assign data points to predefined categories or classes. * **A:** Predicting the number of cups of coffee is a regression problem because the output is a continuous numerical value. * **B:** Grouping images based on similar colors is a clustering problem, an unsupervised learning task that groups similar data points together. * **C:** Predicting whether someone uses a bicycle is a classification problem because the output is a binary outcome (yes/no or 0/1). * **D:** Predicting race time is a regression problem because the output is a continuous numerical value (time in minutes).

Answer 80

B. Regression DISCUSSION: Regression is used for predicting a continuous numerical value, such as the number of gift cards sold. Classification is used for predicting categories, and clustering is used for grouping data points based on similarity. Therefore, regression is the appropriate choice here.

Answer 81

The number of true positives (TP) is **11**. The number of false negatives (FN) is **1033**. **Explanation:** The confusion matrix shows: * True Positives (TP): 11 (Predicted 1, Actual 1) * False Positives (FP): 5 (Predicted 1, Actual 0) * False Negatives (FN): 1033 (Predicted 0, Actual 1) * True Negatives (TN): 13951 (Predicted 0, Actual 0) Therefore, the correct answers are: * TP = 11 * FN = 1033 Options B and D are incorrect because they represent the number of false positives and true negatives, respectively.

Answer 82

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0003300002.png)

Answer 83

B. Detect brands in an image. C. Detect the color scheme in an image DISCUSSION: Options B and C are the correct answers. Computer vision can be used to identify objects, logos, and other visual elements within an image, allowing it to detect brands. It can also analyze the color composition of an image to determine its color scheme. Option A is incorrect because predicting stock prices is a task typically associated with machine learning and time series analysis, not computer vision. Option D is incorrect because translating text between languages is a task related to natural language processing (NLP), not computer vision. Option E is incorrect because extracting key phrases from text is an NLP task, not a computer vision task.

Answer 84

* No * Yes * No **Explanation:** * **Predicting house prices:** This is a regression problem, as we are predicting a continuous numerical value (house price). Thus, it is not anomaly detection. * **Finding unusual spending patterns on credit cards:** This is a classic example of anomaly detection, where we are looking for outliers in spending behavior that deviate from the norm. * **Predicting whether a patient will develop diabetes based on the patient's medical history:** This is a classification problem, where we are predicting a binary outcome (whether the patient will develop diabetes or not). It is not anomaly detection, as we are not looking for unusual or unexpected data points, but rather classifying based on historical patterns.

Answer 85

B. Randomly split the data into rows for training and rows for evaluation. DISCUSSION: Option B is the correct answer. In machine learning, the dataset is typically split into training and evaluation (or testing) sets. The training set is used to train the model, and the evaluation set is used to assess the model's performance on unseen data. Splitting the data randomly by rows ensures that each subset is representative of the overall dataset and avoids bias that might occur if the data was split in a non-random way, or by columns. Option A is incorrect because training a model requires both features and labels. Option C is incorrect because it reverses the roles of features and labels; features are the input variables, and labels are the output variables we are trying to predict. Option D is incorrect because splitting by columns would separate the features from the labels, making it impossible to train or evaluate the model properly.

Answer 86

Here's the breakdown of the correct matches: * **Sentiment analysis**: Determine whether a customer is satisfied or unsatisfied based on ticket text. * *Explanation:* Sentiment analysis is designed to detect the emotional tone behind text, making it suitable for gauging customer satisfaction. * **Key phrase extraction**: Identify important concepts in text, including key phrases and named entities such as people, places, and organizations. * *Explanation:* Key phrase extraction helps to quickly identify the main topics and subjects discussed in the tickets. * **Entity Recognition**: Identify key dates in the text of support tickets. * *Explanation:* Named Entity Recognition (NER) can identify and classify specific entities like dates, which is helpful for tracking timelines or scheduling related to the tickets.

Answer 87

B. A webchat bot is designed to handle common customer queries, thus automating tasks previously done by customer service agents. This leads to a reduction in their workload, allowing them to focus on more complex or critical issues. Option A is incorrect because while a chatbot might indirectly contribute to sales by improving customer experience, it's not the primary expected benefit. Option C is incorrect because a webchat bot has no direct impact on the reliability of the company's products.

Answer 88

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0000900002.png) DISCUSSION: The correct answer is "Reliability and safety." The provided text explicitly states that AI systems need to be reliable and safe to be trusted, and that rigorous testing is essential to ensure they respond safely in unanticipated situations and edge cases. The other options, while potentially relevant in other contexts of AI development, are not the primary focus of the given text. The text emphasizes the importance of the system performing as designed and responding safely to new situations, which aligns directly with reliability and safety.

Answer 89

* **Box 1:** Reliability and safety * **Box 2:** Accountability * **Box 3:** Privacy and security **Explanation:** * **Box 1:** The description emphasizes consistent and safe operation, aligning perfectly with the principle of Reliability and Safety. * **Box 2:** The description focuses on human oversight and responsibility for AI systems, which is the core of Accountability. * **Box 3:** The description highlights the importance of data protection and compliance with privacy laws, directly relating to Privacy and Security.

Answer 90

B. Detect faces in an image. C. Recognize handwritten text. DISCUSSION: Options B and C are correct. The Computer Vision service can be used to detect faces in an image (B) and recognize handwritten text (C) using the OCR and Read APIs. Option A is incorrect because training a custom image classification model is a task for the Custom Vision service, not the Computer Vision service. Option D is incorrect because text translation is not a function of the Computer Vision service. This would typically be a function of a translation service in Azure Cognitive Services.

Answer 91

Classify Email messages The image provided depicts an email inbox and actions, which suggests that the task involves analyzing and categorizing emails. Since NLP is used for document categorization, this is the most appropriate answer. Other options do not directly relate to the use of NLP.

Answer 92

A. entity recognition DISCUSSION: The output in the exhibit clearly shows identified entities (e.g., organizations, people, location) along with their categories. This is a characteristic of entity recognition. Key phrase extraction would return a list of phrases, sentiment analysis would return a sentiment score, and translation would return the text in a different language.

Answer 93

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0004500001.png)

Answer 94

C. computer vision DISCUSSION: The question asks for the type of AI workload to identify bottles of the correct shape. Computer vision is the most appropriate choice because it deals with teaching machines to "see" and interpret images, which is exactly what's needed to identify the shape of a bottle. * **A. anomaly detection:** Anomaly detection is used for identifying unusual patterns or outliers in data, not for identifying shapes. * **B. conversational AI:** Conversational AI is used for building chatbots and other systems that can interact with humans using natural language, which is not relevant here. * **D. natural language processing:** Natural language processing is used for understanding and processing human language, not for identifying shapes.

Answer 95

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0009400002.png) DISCUSSION: The question describes a scenario where a developer needs to build, deploy, and improve their own image classifiers using custom labels based on the visual characteristics of images. This aligns perfectly with the description of Azure Custom Vision. Custom Vision allows you to train a model with your own images and labels, which is essential for building tailored image classifiers. In contrast, the Computer Vision service offers pre-trained models, which are not suitable when custom labels and specific training are required.

Answer 96

* Yes * No * No **Explanation:** * **Statement 1: Yes** - Labeled data, by definition, is data that has been annotated or tagged to indicate the target variable that the model should predict. This annotation process is often referred to as data labeling. * **Statement 2: No** - The statement is not presented. * **Statement 3: No** - While accuracy is a common and easily understandable metric, it is not *always* the primary metric for evaluating a classifier, especially when dealing with imbalanced datasets or when different types of errors have different costs. Other metrics like precision, recall, F1-score, AUC-ROC, etc., can be more appropriate in those scenarios.

Answer 97

* **Scenario 1: A chat bot that provides answers to customer questions:** Conversational AI * **Scenario 2: A system that automatically identifies products in images:** Computer Vision * **Scenario 3: A system that analyzes customer reviews to determine sentiment:** Natural Language Processing **Explanation:** * **Scenario 1:** Conversational AI focuses on creating systems that can engage in natural, human-like conversations. Chatbots are a prime example of this. * **Scenario 2:** Computer Vision is the field of AI that enables computers to "see" and interpret images. Identifying products in images falls squarely within this domain. * **Scenario 3:** Natural Language Processing (NLP) deals with enabling computers to understand and process human language. Sentiment analysis, which involves determining the emotional tone of text, is a key application of NLP.

Answer 98

* Do all the faces belong together? - A. Grouping * Who is this person in this group of people? - B. Identification * Do two images of a face belong to the same person? - C. Verification * Does this person look like the other people? - D. Similarity **Explanation:** * **Grouping:** This task focuses on determining if a set of faces logically belong together, implying a need to cluster or categorize faces based on shared characteristics or relationships. * **Identification:** This task deals with pinpointing the identity of an individual within a collection of faces, akin to facial recognition for security or tagging purposes. * **Verification:** This task is about confirming if two facial images represent the same person. It's a binary decision—yes, it's the same person, or no, it's not. * **Similarity:** This task involves gauging how alike one person's face is to others, often resulting in a similarity score or ranking.

Answer 99

B. the trip distance of individual taxi journeys DISCUSSION: The question asks about features to predict taxi fare. Option A is incorrect because the total number of taxi journeys in the dataset is a static property of the dataset, not a characteristic of an individual journey that would influence the fare. Option B is correct because the trip distance is a characteristic of the journey that directly influences the fare. This is a standard feature used in fare prediction models. Option C is incorrect because the fare is the target variable that we are trying to predict, not a feature used for prediction. It is the label. Option D is incorrect because the trip ID is a unique identifier for each journey and does not provide any predictive power for the fare.

Answer 100

D. Optical character recognition (OCR) DISCUSSION: The correct answer is D. OCR (Optical Character Recognition) is the appropriate computer vision technique to use because the primary function is to extract text from images, which would be essential for reading and storing expense data from receipts. A. Semantic segmentation is incorrect because it involves classifying each pixel in an image, which is not necessary for this task. B. Image classification is incorrect because it only categorizes the entire image, rather than extracting specific text data. C. Object detection is incorrect because, while it can identify objects within an image, it does not extract text, which is crucial for processing expense data.

Answer 101

A. a smart device in the home that responds to questions such as “What will the weather be like today?” B. a website that uses a knowledge base to interactively respond to users' questions DISCUSSION: Options A and B are correct because they both describe systems that engage in conversation with users. A smart device answering weather questions and a website using a knowledge base to answer user questions are both examples of conversational AI. Options C and D are incorrect because they describe automated processes that do not involve natural language interaction. Assembly line machinery and temperature monitoring systems perform specific tasks without engaging in conversation.

Answer 102

- Household Income: Feature - House Price Category: Label Explanation: In machine learning, features are the input variables used to predict the output, which is the label. In this case, household income is a feature used to predict the house price category, which is the label.

Answer 103

A. true positive rate DISCUSSION: The correct answer is A. True Positive Rate is a metric used to evaluate classification models. MAE, R2, and RMSE are metrics used to evaluate regression models. * **A. True Positive Rate:** This is a valid metric for evaluating classification models, representing the proportion of actual positives that are correctly identified. * **B. Mean Absolute Error (MAE):** MAE is a metric used for regression models, measuring the average magnitude of errors in a set of predictions, without considering their direction. * **C. Coefficient of Determination (R2):** R2 is also used for regression models, indicating the proportion of variance in the dependent variable that can be predicted from the independent variable(s). * **D. Root Mean Squared Error (RMSE):** RMSE is another regression metric, measuring the square root of the average of the squared differences between predictions and actual values.

Answer 104

A, D DISCUSSION: The Azure Machine Learning designer allows users to visually create machine learning models by connecting datasets and modules on an interactive canvas. A. **Correct:** Datasets are the foundation of any ML process, providing the data used for training and evaluation. They can be dragged onto the canvas to be used as inputs. B. **Incorrect:** While compute is essential for running pipelines, it is a configuration setting, not a component that is dragged onto the canvas. C. **Incorrect:** Pipelines are created by connecting modules and datasets; you don't drag a pre-existing pipeline onto the canvas to start. D. **Correct:** Modules are pre-built or custom components that perform specific tasks like data transformation, model training, or evaluation. They are dragged onto the canvas and connected to form the ML pipeline.

Answer 105

C. Split Data Explanation: The 'Split Data' module in Azure Machine Learning designer is specifically designed for dividing a dataset into two or more subsets, commonly used for creating training and validation sets. This allows for model training on one subset and performance evaluation on another. A is incorrect because 'Select Columns in Dataset' is used for choosing specific columns from the dataset, not for splitting the data into subsets. B is incorrect because 'Add Rows' is used for combining data by adding rows from one dataset to another, not for splitting. D is incorrect because 'Join Data' is used for merging datasets based on common columns, not for splitting.

Answer 106

B. Enabling "Explain best model" directly addresses the transparency principle by providing insights into how the model works and why it makes certain predictions. This helps users understand the model's behavior and limitations. A. Setting Validation type to Auto relates to model evaluation, not transparency. C. Setting the Primary metric to accuracy focuses on model performance, not transparency. D. Setting Max concurrent iterations to 0 would prevent the automated ML process from running effectively and has no direct impact on transparency.

Answer 107

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0004300001.png)

Answer 108

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0004600002.png) **Explanation:** The question describes a scenario where a linear relationship is established between independent variables and a numeric outcome. This is the fundamental concept of linear regression, which is used to predict a numeric target. * **Regression** is correct because it directly aligns with the description of predicting a numeric target using a linear relationship. * **Classification** is incorrect because it focuses on categorizing data into classes, not predicting numeric values. * **Clustering** is incorrect because it involves grouping data points into clusters based on similarity, which is different from predicting a numeric outcome based on independent variables.

Answer 109

* Yes * Yes * No **Explanation:** * **Statement 1:** The Azure Machine Learning designer provides a visual interface to connect datasets and modules for creating machine learning models. This is a core functionality of the designer, making the statement true. * **Statement 2:** In the Azure Machine Learning designer, connecting modules creates a pipeline draft. The designer automatically saves progress as a pipeline draft during editing. This statement accurately describes the pipeline creation and saving process. * **Statement 3:** Azure Machine Learning Designer enables custom Python and R functions, not Javascript. This statement is therefore false.

Answer 110

* A data science team is **examining** the performance of a classification model and needs to understand the trade-offs between precision and recall. **Model evaluation** * A machine learning engineer is **splitting** a dataset's 'Date' column into 'Month', 'Day', and 'Year' to improve model accuracy. **Feature engineering** * A data scientist is **picking** a subset of the most relevant features from a high-dimensional dataset to reduce model complexity and improve generalization. **Feature selection** **Explanation:** * **Model evaluation:** This task focuses on assessing the performance of a trained model. The scenario explicitly mentions "examining the performance" and understanding metrics like precision and recall, which are key components of model evaluation. * **Feature engineering:** This involves creating new features from existing ones to improve model performance. Splitting the 'Date' column into 'Month', 'Day', and 'Year' is a clear example of feature engineering because it transforms the original data into potentially more informative features. * **Feature selection:** This task aims to select the most relevant features from a dataset. The scenario directly states "picking a subset of the most relevant features," which aligns perfectly with the definition of feature selection.

Answer 111

B. regression DISCUSSION: The question asks for a prediction of sea level in *meters*, which is a continuous numerical value. Regression is the machine learning technique used for predicting continuous numerical values. * **Classification** is used for predicting categorical values (e.g., spam or not spam). * **Clustering** is used for grouping similar data points together, not for prediction.

Answer 112

B. Object detection is the correct choice because it identifies and locates specific objects (cars in this case) within an image, providing the necessary information to estimate distances. A. Optical Character Recognition (OCR) is used to identify text in images, not objects. C. Image classification categorizes the entire image into a single class (e.g., "car scene") but doesn't locate individual cars. D. Face detection identifies faces, not cars.

Answer 113

B. sentiment analysis Sentiment analysis is the process of determining the emotional tone behind a series of words, used to gain understanding of the attitudes, opinions and emotions expressed within an online mention. Language detection identifies the language of the text. Key phrase extraction identifies the main talking points of the text. Entity recognition identifies specific entities (people, places, things) within the text.

Answer 114

B. key phrase extraction DISCUSSION: The question asks for identifying the *main talking points* in a collection of documents. Key phrase extraction is designed to identify the main concepts or phrases within a text, making option B the correct answer. A. Entity recognition identifies specific entities (people, places, organizations, etc.) but not necessarily the main talking points. C. Sentiment analysis determines the overall sentiment (positive, negative, neutral) of the text. D. Language detection identifies the language of the text.

Answer 115

B. a chatbot that provides users with the ability to find answers on a website by themselves C. telephone voice menus to reduce the load on human resources DISCUSSION: Options B and C are correct because they both involve interactive communication, which is a key aspect of conversational AI. A chatbot (B) directly engages in dialogue with users to provide answers. Telephone voice menus (C), also known as Interactive Voice Response (IVR) systems, use voice or keypad input to understand user needs and provide automated assistance. Option A is incorrect because a pre-recorded message is a one-way communication and does not involve any conversation. Option D is incorrect because creating FAQ documents, even by crawling websites, doesn't involve real-time interaction or conversation.

Answer 116

* Yes * Yes * No **Explanation:** * **Statement 1: Azure bot service can be integrated with the powerful AI capabilities with Azure Cognitive Services.** This is correct. Azure Bot Service is designed to work with Cognitive Services to provide AI-powered features. * **Statement 2: Azure bot service engages with customers in a conversational manner.** This is correct. The primary function of a bot service is to conduct conversations with users. * **Statement 3: The QnA Maker service creates knowledge base, not question and answers sets.** This is incorrect. QnA Maker specifically creates a knowledge base composed of question and answer sets. The core function of QnA Maker is to establish these Q&A pairs that a bot can then use to respond to user queries.

Answer 117

ACE DISCUSSION: The correct answers are A, C, and E. * **A is correct:** QnA Maker can generate question and answer pairs by extracting content from existing web pages. * **C is correct:** QnA Maker allows users to manually input question and answer pairs. * **E is correct:** QnA Maker allows importing pre-defined "chit-chat" datasets to provide conversational responses. * **B is incorrect:** While machine learning is involved, QnA Maker focuses on question-answer pairs, not just training on questions alone. * **D is incorrect:** Connecting to Cortana is a channel integration, not a method for creating QnA content.

Answer 118

A. QnA Maker **Explanation:** QnA Maker is specifically designed to create conversational question answering systems from structured content like FAQ documents. It extracts question and answer pairs and allows users to interact with the knowledge base in a conversational manner. * **A. QnA Maker:** Correct. * **B. Text Analytics:** This service is used for sentiment analysis, key phrase extraction, and language detection, not for building conversational systems from FAQs. * **C. Computer Vision:** This service is for analyzing images, which is not relevant to processing a text-based FAQ document. * **D. Language Understanding (LUIS):** LUIS is used for building more complex conversational AI that understands user intents and entities, but it is not the most direct solution for a simple FAQ-based chatbot. QnA Maker is a better fit for this scenario.

Answer 119

B. QnA Maker C. Azure Bot Service **Explanation:** * **B. QnA Maker:** QnA Maker is a cloud-based AI service that allows you to create a knowledge base of questions and answers. This is essential for a chatbot that needs to provide predefined answers to user queries. * **C. Azure Bot Service:** Azure Bot Service provides the environment and tools to build, test, deploy, and manage intelligent bots. It is the platform on which the QnA Maker knowledge base can be integrated to create a functional chatbot. **Why other options are incorrect:** * **A. Text Analytics:** Text Analytics is used for extracting insights from text, such as sentiment analysis or key phrase extraction. While it can be useful in some chatbot scenarios, it's not directly related to providing predefined answers. * **D. Translator:** Translator is used for translating text from one language to another. While useful for multilingual chatbots, it's not core to the basic functionality of answering questions with predefined answers.

Answer 120

A. QnA Maker

Answer 121

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0014100002.png) The correct answer is "Conversational AI". The question describes the capabilities of Microsoft's tools for creating and managing intelligent bots. These bots are designed to engage in natural conversations with users across various platforms. Therefore, "Conversational AI" is the most appropriate term to complete the sentence, accurately reflecting the purpose of these tools.

Answer 122

C. inclusiveness D. fairness F. reliability and safety DISCUSSION: The correct answers are C, D, and F. Microsoft's six guiding principles for responsible AI are fairness, inclusiveness, transparency, privacy and security, reliability and safety, and accountability. Options C, D, and F are three of these six principles. Options A, B, and E are not among the Microsoft guiding principles for responsible AI.

Answer 123

B. the Detect operation in the Face service DISCUSSION: The Detect operation in the Face service is the correct choice. It can identify faces in an image and also detect attributes like the presence of glasses. Option A, Verify, is used to confirm if two faces belong to the same person. Options C and D, Describe Image and Analyze Image, are part of the Computer Vision service and are not specifically designed to detect faces and their attributes with the same precision as the Face service.

Answer 124

C. Splitting the data into separate subsets, such as training and testing sets, allows you to evaluate the model's performance on unseen data. This provides a more realistic assessment of how well the model generalizes to new, real-world scenarios. Option A is incorrect because training the model twice does not guarantee better accuracy and might lead to overfitting. Option B is incorrect because splitting data into subsets is not primarily for training multiple models simultaneously to improve performance, but for evaluating the trained model.

Answer 125

A and D are correct. Form Recognizer (now Document Intelligence) is designed for extracting information from documents like invoices and receipts. Option A directly aligns with this by extracting the invoice number. Option D also aligns with this functionality because Form Recognizer can identify information such as the retailer from a receipt. Option B is incorrect because Form Recognizer is not a translation service. Option C is incorrect because Form Recognizer does not find images of products.

Answer 126

B. providing closed captions for recorded or live videos D. creating a transcript of a telephone call or meeting DISCUSSION: The question is asking about "speech recognition," which in this context, refers to converting speech to text. Option B is correct because speech recognition is used to automatically generate closed captions for videos by transcribing the spoken audio into text. Option D is correct because creating a transcript of a telephone call or meeting involves converting spoken language into written text, which is a common application of speech recognition technology. Option A is incorrect because an in-car system that reads text messages aloud uses text-to-speech (speech synthesis), not speech-to-text (speech recognition). Option C is incorrect because an automated public address system typically uses pre-recorded or synthesized speech rather than real-time speech recognition.

Answer 127

C. The Speech service provides text-to-speech functionality, which is required to read the recipe instructions aloud. A. Text Analytics is for understanding the sentiment or key phrases in text, not for converting text to speech. B. Translator is for translating text from one language to another. D. Language Understanding (LUIS) is for building conversational AI and understanding the intent of user input.

Answer 128

D. A webchat bot interacts with users through a website interface to answer questions. Option D describes this scenario directly. Options A, B, and C describe sentiment analysis, translation, and email routing, respectively, and do not involve a webchat bot.

Answer 129

Yes, No, Yes, No **Explanation:** * **Statement 1: Yes.** The definition of Automated ML (AutoML) is accurate. It automates iterative tasks in ML model development. * **Statement 2: No.** AzCopy's description is correct but irrelevant to AutoML. The question asks about AutoML related concepts. * **Statement 3: Yes.** This accurately describes how Azure Machine Learning trains models using AutoML, by iterating through algorithms and parameters to find the best fit. * **Statement 4: No.** This statement is partially correct but contains an error. AutoML does train and tune models based on a specified metric. However, the "label" is the column to predict, not the "features".

Answer 130

* Yes * No * Yes **Explanation** * **Statement 1:** The description aligns perfectly with the capabilities of Azure Content Moderator (now Azure AI Content Safety), which is designed for content moderation using machine learning. Thus, "Yes" is correct. * **Statement 2:** The description accurately reflects the functionality of Azure Computer Vision, a service specializing in image analysis. Therefore, "No" is correct. * **Statement 3:** The description correctly defines Natural Language Processing (NLP) and its applications, including sentiment analysis. Thus, "Yes" is correct.

Answer 131

C. classification DISCUSSION: The problem describes a scenario where the goal is to assign brain scan images to predefined categories (haemorrhage types). This is the definition of a classification problem in machine learning. * **Clustering** is unsupervised learning where the algorithm attempts to group similar data points together without prior knowledge of categories. * **Regression** is used to predict a continuous numerical value, not a category.

Answer 132

B. regression The Predicted vs. True chart displays predicted values against actual values, which is a common method for evaluating the performance of regression models. Regression models predict continuous numerical values. Option A is incorrect because classification models predict categorical values, and their performance is typically evaluated using metrics like accuracy, precision, and recall, often visualized with confusion matrices or ROC curves. Option C is incorrect because clustering models group data points based on similarity, and evaluation typically involves metrics like silhouette score or visual inspection of the clusters.

Answer 133

AC DISCUSSION: The question asks about tasks that specifically require an enterprise workspace in Azure Machine Learning. Option A is correct because the graphical user interface (GUI) for running automated machine learning experiments was a feature primarily associated with the enterprise workspace. Option C is correct because using the Azure Machine Learning designer, which provides a GUI for defining and running experiments, was also a key feature of the enterprise workspace. Option B is incorrect because creating a compute instance is a general capability available in both basic and enterprise workspaces. Option D is incorrect because creating a dataset from a CSV file is a basic data ingestion task that doesn't require an enterprise workspace. Note that the enterprise workspace is now deprecated, but the question is presented in the context of when it was relevant.

Answer 134

* Yes * No * Yes **Explanation:** * **Statement 1:** The Text Analytics API does provide language detection capabilities, returning a language code and confidence score. So, "Yes" is correct. * **Statement 2:** The Text Analytics API primarily deals with text analysis, not image analysis or handwriting recognition. Detecting handwritten signatures would require OCR or specialized handwriting recognition services like Azure Form Recognizer or Computer Vision. Therefore, "No" is correct. * **Statement 3:** Named Entity Recognition is a core function of the Text Analytics API, accurately described in the statement. So, "Yes" is correct.

Answer 135

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0001400002.jpg) DISCUSSION: The question describes AI systems needing to operate reliably, safely and consistently, even under unexpected conditions, to build trust. The answer choice "Reliability and safety" directly corresponds to this principle. The other options are not listed as a core principle for building trust in AI systems.

Answer 136

* **Identify handwritten letters:** Computer Vision * **Predict the sentiment of social media posts:** Natural Language Processing * **Detect fraudulent credit card payments:** Anomaly Detection * **Predict next month's sales:** ML (Regression) **Explanation:** * **Computer Vision:** Deals with enabling computers to "see" and interpret images, which includes tasks like object detection and Optical Character Recognition (OCR) to identify letters. * **Natural Language Processing (NLP):** Focuses on enabling computers to understand and process human language, allowing for tasks like sentiment analysis. * **Anomaly Detection:** Identifies unusual patterns or data points that deviate from the norm, which is useful for detecting fraudulent activities. * **ML (Regression):** A type of machine learning used to predict continuous numerical values, such as sales figures.

Answer 137

Yes / Yes / Yes **Explanation:** Based on the provided references and user discussions, all three statements are true. * **Transcribe a call to text - Speech Service:** The Speech service includes speech-to-text functionality. * **Extract call transcription to find key entity - Text Analytics:** Text Analytics provides entity recognition capabilities. * **Translate a call to a different language - Speech Service:** The Speech service offers speech translation features.

Answer 138

Yes Yes No

Answer 139

Object detection [Image](https://www.examtopics.com/assets/media/exam-media/04234/0001800002.png)

Answer 140

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0005100002.png) DISCUSSION: The question requires completing a sentence related to Azure Machine Learning designer. The designer allows users to build machine learning models by visually adding, connecting, and configuring modules on a drag-and-drop canvas. Therefore, the correct answer is "adding and connecting modules on a visual canvas."

Answer 141

A and C. **Explanation:** The goal is to predict the "Income Range" using other fields as features. "Age" and "Education Level" are the most relevant factors that can influence a person's income. Therefore, they should be used as features. * **A. Education Level:** Higher education often leads to higher-paying jobs. * **C. Age:** Generally, with age comes more experience and potentially higher earnings. **Why other options are incorrect:** * **B. Last Name:** Last name is unlikely to have a direct correlation with income. * **D. Income Range:** Income Range is the target variable we are trying to predict, so it cannot be used as a feature. * **E. First Name:** Similar to last name, first name is unlikely to have a direct correlation with income.

Answer 142

* Face detection that perceives faces and attributes in an image; person identification that matches an individual in your private repository of up to 1 million people; perceived emotion recognition that detects a range of facial expressions like happiness, contempt, neutrality, and fear; and recognition and grouping of similar faces in images. -> Facial recognition (This description matches the functionality of facial recognition as described in the provided context and external references.) * -> OCR (The Detect API applies tags based on the objects or living things identified in the image implying that this description is relevant to OCR.) * Object detection is similar to tagging, but the API returns the bounding box coordinates (in pixels) for each object found. For example, if an image contains a dog, cat and person, the Detect operation will list those objects together with their coordinates in the image. You can use this functionality to process the relationships between the objects in an image. It also lets you determine whether there are multiple instances of the same tag in an image. -> Object detection (This description explicitly describes object detection and its bounding box functionality.) * The Detect API applies tags based on the objects or living things identified in the image. There is currently no formal relationship between the tagging taxonomy and the object detection taxonomy. At a conceptual level, the Detect API only finds objects and living things, while the Tag API can also include contextual terms like "indoor", which can't be localized with bounding boxes. -> OCR (The Detect API applies tags based on the objects or living things identified in the image implying that this description is relevant to OCR.)

Answer 143

Yes, No, No. Analyzing customer reviews for themes and sentiment is a common AI application. Chatbots are also a common application of AI. A simple webform for password reset does not require AI.

Answer 144

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0001900002.png)

Answer 145

A. QnA Maker B. Azure Bot Service **Explanation:** * **A. QnA Maker:** QnA Maker (now Azure Cognitive Service for Language - Custom question answering) is used to create a knowledge base of questions and answers. This knowledge base can then be queried using natural language. It is a crucial part of the solution to understand user queries. * **B. Azure Bot Service:** Azure Bot Service provides the platform and tools to build, test, deploy, and manage intelligent bots. It provides the channel through which users interact with the knowledge base. * **C. Form Recognizer:** Form Recognizer is used to extract information from forms and documents, which is not directly relevant to querying a knowledge base using natural language. * **D. Anomaly Detector:** Anomaly Detector is used to identify anomalies in data, which is not directly relevant to querying a knowledge base using natural language.

Answer 146

D. Language Understanding (LUIS) DISCUSSION: The question asks for the service that determines a user's intent. Language Understanding (LUIS), now known as Conversational Language Understanding (CLU) under the Azure Cognitive Services Language service, is specifically designed for this purpose. It uses machine learning to understand natural language and predict user intent based on their utterances. Option A, Translator, is for translating text between languages. Option B, QnA Maker, is for creating a knowledge base of questions and answers. Option C, Speech, is used for speech-to-text and text-to-speech conversions. Therefore, only option D directly addresses the task of determining user intent.

Answer 147

* No * Yes * No **Explanation:** **Statement 1**: QnA Maker determines the intent of a user's utterance. - **No** QnA Maker focuses on providing answers to questions, not understanding the intent behind them. **Statement 2**: Language Understanding (LUIS) determines the intent of a user's utterance. - **Yes** LUIS is specifically designed to understand the intent of user utterances. **Statement 3**: In QnA Maker, utterances are equivalent to questions. - **No** While questions are the primary input for QnA Maker, "utterances" as a term is more closely associated with LUIS, which uses utterances to train its intent recognition models. In QnA maker, the user input is considered as question.

Answer 148

Yes Yes No **Explanation:** Based on the discussion and the provided links, the correct answer is Yes/Yes/No. * The first two statements are true regarding the capabilities of Azure Cognitive Services Face API. * The third statement is false, as the Face API can detect that two faces belong to the same person.

Answer 149

No Yes No **Explanation:** The consensus from the discussion is that the correct answers are No, Yes, and No for the three statements, respectively. * **Statement 1: You can insert any custom code into automated machine learning.** This is **incorrect (No)**. Automated ML handles feature engineering, algorithm selection, and hyperparameter tuning automatically, without requiring custom code. * **Statement 2: You must have programming skills to use automated machine learning.** This is **correct (Yes)**. AutoML selects the parameters and algorithms but you do not need programming abilities. * **Statement 3: In automated machine learning, you build a canvas and drag-and-drop modules.** This is **incorrect (No)**. The drag-and-drop interface with modules is a feature of Azure Machine Learning Designer, which is distinct from Automated ML.

Answer 150

A. No B. Yes C. Yes The first statement is false. When creating a Custom Vision project with Object Detection as the project type, you do not need to select a classification type. Classification types (Multiclass or Multilabel) are only relevant when the project type is Classification. The second statement is true. You can add images to a Custom Vision project either by uploading them from your local computer or by providing a URL to the image. The third statement is true. After uploading and tagging images, you must train the Custom Vision project so the model can learn to identify the objects or classifications you've defined. Only then can you test and evaluate its performance.

Answer 151

* No * Yes * Yes **Explanation:** * **You can communicate with a bot by using Cortana through Azure: No.** Microsoft has deprecated the Cortana channel in Bot Framework. * **You can communicate with a bot by using Email through Azure: Yes.** Email is a supported channel for Azure Bot Service. * **You can communicate with a bot by using Microsoft Teams through Azure: Yes.** Microsoft Teams is a supported channel for Azure Bot Service.

Answer 152

D. active learning **Explanation:** Active learning is the correct approach because it directly involves the bot learning from user feedback to improve the relevance of its responses. The bot presents possible answers to the user, and the user's selection is used to retrain the model, improving future responses. * **A. key phrase extraction:** This identifies important phrases but doesn't directly use user feedback to improve response relevance. * **B. sentiment analysis:** This determines the emotional tone of the user's input but doesn't, on its own, improve the accuracy of the bot's answers. * **C. business logic:** While business logic can be used to filter low scores, it does not involve active learning from the user's selection of appropriate answers.

Answer 153

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0010400002.png) DISCUSSION: The question is asking about converting speech to text. "Speech recognition" is the process of converting audio streams into text. Speech synthesis does the opposite - it converts text to speech.

Answer 154

B. Azure Bot Service Azure Bot Service is the correct answer because it is specifically designed for building, deploying, and managing conversational AI solutions across various channels. Text Analytics is used for understanding and extracting insights from text, Translator is used for translating text between languages, and Form Recognizer is used for extracting data from forms. These services do not provide the core conversational capabilities offered by Azure Bot Service.

Answer 155

* Yes * Yes

Answer 156

* Yes * Yes * No The first two statements are accurate descriptions of Custom Vision's image classification and object detection capabilities and how it trains. The third statement is incorrect because while Custom Vision directly processes image files, it's also possible to analyze video by extracting frames and submitting them as images.

Answer 157

A. Form Recognizer DISCUSSION: The question asks for a service that extracts text, key/value pairs, and table data from scanned documents. Option A, Form Recognizer (now Azure AI Document Intelligence), is specifically designed for this purpose. Option B, Text Analytics, is for analyzing text sentiment and extracting key phrases, but not specifically for document structure extraction. Option C, Language Understanding (LUIS), is for building conversational AI and understanding user intents. Option D, Custom Vision, is for image recognition and classification.

Answer 158

Box 1: No Box 2: Yes Box 3: No **Explanation:** * **Statement 1: The validation dataset is different from the test dataset that is held back from the training of the model. - No** * **Correct:** The validation set and the test set serve different purposes and are distinct datasets. The validation set is used during training to tune hyperparameters and prevent overfitting, while the test set is used to evaluate the final model's performance on unseen data. * **Statement 2: A validation dataset is a sample of data that is used to give an estimate of model skill while tuning model's hyperparameters. - Yes** * **Correct:** The validation dataset is specifically used to evaluate the model's performance during training, allowing for adjustments to hyperparameters to improve generalization. * **Statement 3: The Validation Dataset is a sample of data used to provide an unbiased evaluation of a final model fit on the training dataset. - No** * **Correct:** The test dataset, not the validation dataset, is used for an unbiased evaluation of the final model. The validation dataset is used during training and hyperparameter tuning, and therefore cannot give an unbiased evaluation of the final model.

Answer 159

Box 1: Select Columns in Dataset Box 2: Split Data Box 3: Linear Regression **Explanation:** * **Box 1: Select Columns in Dataset:** Before training a model, it's crucial to select the relevant features (columns) from the dataset. This helps to remove irrelevant or noisy data that could negatively impact the model's performance. * **Box 2: Split Data:** To properly evaluate the model, the data needs to be split into two sets: a training set and a testing set. The training set is used to train the model, and the testing set is used to evaluate its performance on unseen data. * **Box 3: Linear Regression:** Since the goal is to predict automobile prices (a continuous numerical value), a regression algorithm is appropriate. Linear regression is a common and suitable choice for this type of prediction task.

Answer 160

[LLM error]

Answer 161

C. Clustering **Explanation:** Clustering is the appropriate machine learning technique for identifying groups of individuals with similar characteristics, in this case, purchasing habits. Clustering algorithms group data points based on their similarity, allowing you to discover distinct segments within a customer base. * **A. Classification:** Classification is used to predict the category a data point belongs to based on labeled training data. For example, classifying emails as spam or not spam. This is not suitable for identifying groups without pre-defined labels. * **B. Regression:** Regression is used to predict a continuous numerical value. For example, predicting house prices based on size and location. It's not designed for grouping data points.

Answer 162

A. transparency DISCUSSION: The question describes a scenario where an AI system used for loan approval needs to have explainable factors in its decision-making process. This directly relates to the principle of transparency in responsible AI, as transparency emphasizes making AI systems understandable. Options B, C, and D are incorrect because: - Inclusiveness refers to ensuring AI systems consider diverse perspectives and avoid excluding certain groups. - Fairness focuses on avoiding bias and ensuring equitable outcomes across different groups. - Privacy and security involve protecting sensitive data and ensuring the AI system is secure from unauthorized access. While all of these are important aspects of responsible AI, the scenario specifically highlights the need for explainability, which aligns with transparency.

Answer 163

Feature engineering

Answer 164

- **Fairness**: Ensure system decisions don't discriminate. - **Privacy & Security**: Secure personal data. - **Transparency**: Record training-related assets and metrics involved in the experiment to identify why. **Explanation:** * **Fairness:** This principle focuses on ensuring AI systems do not discriminate against individuals or groups based on characteristics like gender, race, or religion. * **Privacy & Security:** This principle mandates the protection of personal data within the AI system, ensuring it is accessed and used in a way that respects individual privacy. * **Transparency:** This principle emphasizes understanding how an AI model was created, including the data, algorithms, and transformation logic used. Recording training assets and metrics helps in understanding and reproducing the model's behavior, thereby promoting transparency.

Answer 165

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0002800001.png)

Answer 166

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0006600001.jpg)

Answer 167

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0007300001.jpg)

Answer 168

The correct matching is as follows: * **Image classification**: Classifying an image as containing a cat or a dog. * **Object detection**: Identifying all cars and pedestrians in a photo. * **Semantic Segmentation**: Identifying regions of an image that correspond to roads, buildings, and trees. **Explanation:** * **Image classification** is used when the goal is to categorize an entire image into a single class (e.g., cat or dog). * **Object detection** is used when the goal is to identify and locate multiple objects within an image (e.g., cars and pedestrians). This involves both classifying the objects and drawing bounding boxes around them. * **Semantic Segmentation** is used when the goal is to classify each pixel in an image, grouping pixels into meaningful regions (e.g., roads, buildings, and trees). This provides a fine-grained understanding of the image's content.

Answer 169

A. Face DISCUSSION: The correct answer is A. Face. The Face service is specifically designed for facial recognition and can be trained to identify individuals. This makes it suitable for tagging images of your friends automatically. Options B, C, and D are incorrect because: - Form Recognizer is used for extracting information from forms and documents. - Text Analytics is used for understanding and analyzing text. - Computer Vision provides general-purpose image analysis but isn't as specialized for facial recognition as the Face service.

Answer 170

C. Key phrase extraction identifies the main concepts in text, making it suitable for identifying documents about similar topics. Options A, B, and D involve sentiment analysis, speech-to-text, and machine translation, respectively, which are different NLP tasks.

Answer 171

B. a chatbot DISCUSSION: The diagram depicts a question and answer interaction, which is characteristic of a chatbot. Therefore, option B is the correct answer. * **A. a sentiment analysis solution:** Sentiment analysis focuses on determining the emotional tone behind text, not interactive question-answering. * **C. a machine learning model:** While a chatbot may be powered by a machine learning model, the diagram specifically shows the interaction, not the underlying model itself. * **D. a computer vision application:** Computer vision deals with interpreting images or videos, which is not represented in the given diagram.

Answer 172

[Image](https://img.examtopics.com/ai-900/image294.png)

Answer 173

A. Custom Vision **Explanation:** The question specifies the use of a "custom model." Custom Vision is the Azure Cognitive Service specifically designed for building custom image recognition models. * **A. Custom Vision:** This is the correct answer because it allows you to train a model with your own data to recognize specific objects or features in images, which is essential for identifying competitor products. * **B. Form Recognizer:** This service is used for extracting text and data from forms and documents, not for general image recognition. * **C. Face:** This service is used for facial recognition and analysis, which is not relevant to identifying products. * **D. Computer Vision:** While Computer Vision can perform general image analysis, it does not provide the ability to train custom models to recognize specific products. It is better suited for pre-trained object detection, or analyzing general image content.

Answer 174

1. Data Prep 2. Train Model 3. Evaluate Model The correct order of processes before deploying a machine learning model is: Data Preparation, Model Training, and Model Evaluation. Data must be prepared and cleaned. The model is then trained on the prepared data. Finally, the model's performance is evaluated to ensure it meets the required standards before deployment.

Answer 175

B. Implement a process of AI model validation as part of the software review process C. Establish a risk governance committee that includes members of the legal team, members of the risk management team, and a privacy officer DISCUSSION: Options B and C align with responsible AI principles. * **B:** Implementing AI model validation as part of the software review process ensures reliability and safety by rigorously testing and validating the system's performance. * **C:** Establishing a risk governance committee addresses accountability by providing oversight, insights, and guidance about developing and deploying AI systems, ensuring privacy and security. Option A, implementing an Agile software development methodology, is a general software development practice but not specifically tied to responsible AI principles. While Agile can be used in developing AI systems, it doesn't directly address the core principles of responsible AI. Option D, preventing the disclosure of the use of AI-based algorithms for automated decision-making, is the opposite of transparency, which is a key responsible AI principle. Users should be informed when AI is being used to make decisions that affect them.

Answer 176

A. coefficient of determination (R2) C. root mean squared error (RMSE) **Explanation:** * **A. coefficient of determination (R2):** R-squared is a statistical measure that represents the proportion of the variance for a dependent variable that's explained by an independent variable or variables in a regression model. It is a common metric to evaluate the goodness of fit of a regression model. * **C. root mean squared error (RMSE):** RMSE is a standard way to measure the error of a model in predicting quantitative data. It represents the square root of the average squared difference between the predicted values and the actual values. Lower RMSE values indicate a better fit. * **B. F1 score:** F1 score is used to evaluate classification models, not regression models. * **D. area under curve (AUC):** AUC is a metric used to evaluate the performance of binary classification models, specifically the ability of the model to discriminate between the positive and negative classes. * **E. balanced accuracy:** Balanced accuracy is a metric used to evaluate classification models, especially when dealing with imbalanced datasets. It calculates the average of recall obtained on each class.

Answer 177

Regression The question describes regression. Regression tasks in machine learning involve predicting a continuous numerical value (the label) based on input features. Other machine learning tasks include classification (predicting a category) and clustering (grouping similar data points). Since the task is to predict a numerical *value*, regression is the correct choice.

Answer 178

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0007100001.jpg)

Answer 179

A. **Explanation:** The question describes a scenario where the computer vision system needs to locate specific objects (weeds) within an image. This falls under the definition of object detection. * **A. object detection:** This is the correct answer because the system is identifying the presence and location of weeds. * **B. optical character recognition (OCR):** OCR is used to recognize text within images, which is not relevant to the scenario. * **C. scene segmentation:** Scene segmentation involves dividing an image into different regions or segments, but it doesn't necessarily identify specific objects like object detection does. While segmentation could be used as part of a larger weed detection pipeline, object detection more directly describes the core task.

Answer 180

A. Identify the retailer from a receipt C. Extract the invoice number from an invoice **Explanation:** Form Recognizer is designed to extract information from structured and semi-structured documents like receipts and invoices. * **A is correct:** Form Recognizer can be used to identify key-value pairs and entities in receipts, including the retailer's name. * **C is correct:** Form Recognizer can accurately extract specific data points like invoice numbers from invoices. * **B is incorrect:** Translation is a function of translation services, not Form Recognizer. * **D is incorrect:** Finding images in a catalog is more suited for computer vision or image recognition services, not Form Recognizer.

Answer 181

A. Translator **Explanation:** The question asks for a service to translate written press releases into multiple languages. The Translator service is specifically designed for this purpose. * **A. Translator:** This is the correct answer as it directly addresses the need for language translation. * **B. Text Analytics:** This service is used for extracting insights and information from text, not for translation. * **C. Speech:** This service deals with converting text to speech or speech to text. * **D. Language Understanding (LUIS):** LUIS is used for building conversational AI and understanding user intents from text, not for translating entire documents.

Answer 182

D. Natural language processing

Answer 183

Yes, No, Yes

Answer 184

* **Box 1: Articles about your customers:** Entity Recognition * **Box 2: Determine sentiment of the article:** Sentiment Analysis **Explanation:** * **Entity Recognition:** This task is used to identify specific entities, such as the names of companies (your customers in this case), people, or locations, mentioned in the text. It helps to filter news articles specifically related to your customers. * **Sentiment Analysis:** This task is used to determine the sentiment (positive, negative, or neutral) expressed in the text. This allows you to identify negative articles (to alert employees) and positive articles (to add to a press book).

Answer 185

D. QnA Maker DISCUSSION: The correct answer is D. QnA Maker is designed to extract question and answer pairs from documents and web pages, making it suitable for creating a chatbot that can answer questions based on the provided documents. Option A is incorrect because Azure Bot Service is a platform for building and deploying bots but doesn't inherently process documents for Q&A. It would utilize services like QnA Maker. Option B is incorrect because Language Understanding (LUIS) focuses on understanding user intent from natural language, not extracting Q&A pairs from documents. Option C is incorrect because Text Analytics provides sentiment analysis, key phrase extraction, etc., but it doesn't build a Q&A knowledge base.

Answer 186

D. Add chit-chat **Explanation** * **Correct:** Adding chit-chat functionality to the bot involves including pre-built conversational responses for general small talk or non-FAQ related interactions. Chit-chat responses can enhance the user experience by providing friendly greetings, acknowledgments, and appropriate responses to casual or non-specific queries, making the bot more user-friendly. * **Incorrect A:** Increasing the confidence threshold of responses affects the accuracy of returned answers but doesn't add greetings or other user-friendly responses. * **Incorrect B:** Enabling active learning helps improve the bot's accuracy over time by learning from user interactions, but it doesn't directly add greetings or other user-friendly responses. * **Incorrect C:** Creating multi-turn questions refers to handling conversations with multiple interrelated questions and responses. While it can contribute to a more interactive experience, it does not address the specific requirement of adding professional greetings and other user-friendly responses.

Answer 187

B. Add utterances to the None intent Explanation: The None intent is specifically designed to capture utterances that fall outside the defined scope or domain of the LUIS model. By adding example utterances that are irrelevant to the e-commerce business, the model can learn to identify and classify such out-of-scope requests as belonging to the None intent. This allows the application to handle these requests appropriately, such as by prompting the user for clarification or directing them to relevant resources. Option A is incorrect because while testing is important, it doesn't directly address how the model learns to identify out-of-scope utterances. Option C is incorrect because prebuilt entities are for extracting specific types of information, not for identifying the overall intent of an utterance. Option D is incorrect because creating a new model is unnecessary; the existing model can be trained to recognize the None intent.

Answer 188

A. natural language processing **Explanation:** The question asks for the appropriate AI workload to extract key terms from text reports. Natural Language Processing (NLP) is specifically designed for analyzing and understanding human language, making it the correct choice for this task. Key term extraction is a common application of NLP. * **A. natural language processing:** This is the correct answer because NLP focuses on enabling computers to understand and process human language. Key term extraction and text summarization are standard NLP tasks. * **B. conversational AI:** Conversational AI focuses on building systems that can engage in conversations with humans, which is not directly relevant to extracting key terms from existing text reports. * **C. anomaly detection:** Anomaly detection is used to identify unusual data points or patterns, which is not the primary goal of extracting key terms for summarization. * **D. computer vision:** Computer vision deals with enabling computers to "see" and interpret images, which is not relevant to text-based insurance claim reports.

Answer 189

A. and D. **Explanation:** * **A. an automated voice that reads back a credit card number entered into a telephone by using a numeric keypad:** This is a valid use case for speech synthesis (text-to-speech). The numbers entered are converted to text, and then synthesized into audible speech. * **D. an AI character in a computer game that speaks audibly to a player:** This is also a valid use case. The AI character's dialogue is likely stored as text, which is then converted to speech using speech synthesis. * **B. generating live captions for a news broadcast:** This would require *speech recognition* (speech-to-text), not speech synthesis. * **C. extracting key phrases from the audio recording of a meeting:** This also requires *speech recognition* to first transcribe the audio into text, followed by natural language processing to extract the key phrases.

Answer 190

C. the authentication key D. the REST endpoint DISCUSSION: The correct answers are C and D. * **C. the authentication key:** When you publish an inference pipeline as a web service, you need an authentication key to securely access it. This key verifies that the requests are authorized. * **D. the REST endpoint:** The inference pipeline is exposed as a REST API, and you need the REST endpoint URL to send requests to it. Incorrect Options: * **A. the model name:** While the model is part of the inference pipeline, you don't directly use the model name to access the web service. The endpoint handles the model invocation. * **B. the training endpoint:** The training endpoint is used during the training phase of the model, not when consuming the deployed web service. You only need the inference endpoint after deployment.

Answer 191

See explanation. The correct answer is: Yes, Mass is a feature. Yes, Temp is a feature. No, QualityTest is a feature. Mass and Temp are variables/measurements and act as inputs to the model to predict the quality. Therefore, they are features. QualityTest is what is being predicted, making it a label, not a feature.

Answer 192

B. Azure Bot Service C. Language Service **Explanation:** * **B. Azure Bot Service:** This is correct because Azure Bot Service provides the platform for developing, deploying, and managing chatbots. It allows you to create the actual chatbot that interacts with users. * **C. Language Service:** This is correct because Language Service, specifically its custom question answering feature, allows you to create a knowledge base of question and answer pairs. The chatbot can then use this knowledge base to respond to user queries. This is the modern replacement for QnA Maker. * **A. Azure Machine Learning:** This is incorrect because while Azure Machine Learning can be used for more advanced chatbot features (like intent recognition or sentiment analysis), it is not necessary for a simple question and answer chatbot. * **D. Translator:** This is incorrect because Translator is primarily used for language translation, which is not the primary goal of this scenario.

Answer 193

Box 1: No - Build conversational experiences with Power Virtual Agents and Azure Bot Service Box 2: Yes - Box 3: Yes - You can configure your bot to communicate with people via Microsoft Teams. DISCUSSION: The correct answer is No, Yes, Yes. Box 1: False. Azure Bot Service provides an environment, but Power Virtual Agents is the low-code platform. Box 2: True. The statement is correct. Box 3: True. Azure bots can be configured to communicate via Microsoft Teams.

Answer 194

* No * Yes * No **Explanation:** * **A restaurant can use a chatbot to answer queries through Cortana. → No:** Cortana support has been dropped for chatbots. * **A restaurant can use a chatbot to answer inquiries about business hours from a webpage. → Yes:** Chatbots are commonly used on webpages to answer customer inquiries, including business hours. * **A restaurant can use a chatbot to automate responses to customer reviews on an external website. → No:** While possible in theory, it's not a typical or recommended use case. Automating responses to customer reviews on external websites is complex and could lead to inappropriate or ineffective responses.

Answer 195

Image Analysis

Answer 196

* **You can use Azure Cognitive Search's knowledge mining results and populate your knowledge base of your chatbot. -> Natural Language Processing** (While Knowledge Mining can contribute data, the core interaction and understanding of user queries in a chatbot relies on NLP.) * **-> Computer Vision** (This one is not associated with any scenario.) * **Natural language processing (NLP) is used for tasks such as sentiment analysis. -> Natural Language Processing** (Sentiment analysis is a key application of NLP.)

Answer 197

Yes No Yes

Answer 198

C. Combine multiple datasets. E. Remove records that have missing values. **Explanation:** The data ingestion and data preparation stage focuses on acquiring, cleaning, and transforming data for use in machine learning models. * **C. Combine multiple datasets:** This is a common data preparation task, often needed to consolidate data from different sources into a single, unified dataset. * **E. Remove records that have missing values:** Handling missing data is a crucial part of data cleaning. Removing records with missing values is one approach to address this. **Why the other options are incorrect:** * **A. Calculate the accuracy of the model:** This occurs during model evaluation, after the model has been trained. * **B. Score test data by using the model:** This also occurs after model training, as part of the model evaluation process. * **D. Use the model for real-time predictions:** This happens after the model is trained, evaluated, and deployed. It's part of the model operationalization phase.

Answer 199

A. regression DISCUSSION: Regression is used for predicting numeric values, such as population size. Clustering is used for grouping data points, and classification is used for predicting categories. Therefore, regression is the correct choice.

Answer 200

A. Python B. R DISCUSSION: The correct answers are A and B. Azure Machine Learning designer supports custom code written in Python and R. C# and Scala are not supported languages for custom code within the Azure Machine Learning designer.

Answer 201

* Yes * Yes * No **Explanation:** * **Statement 1: Correct.** Regression problems require a numeric label column to represent the continuous response variable. * **Statement 2: Correct.** K-means clustering is an unsupervised learning algorithm, meaning it doesn't require a label column. * **Statement 3: Incorrect.** While categorical values in the label column can be represented numerically, they don't *have* to be, therefore the statement is untrue.

Answer 202

A. object detection DISCUSSION: The image shows bounding boxes around different objects (fruits) within the image. This is a characteristic of object detection, which identifies and localizes multiple objects within an image. Option B is incorrect because face detection specifically focuses on identifying faces. Option C is incorrect because OCR is used to extract text from images. Option D is incorrect because image classification would only assign a single label to the entire image (e.g., "Fruits") without identifying individual objects or drawing bounding boxes.

Answer 203

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0002900001.jpg) DISCUSSION: The question asks for the term that describes the ethical principle of avoiding discrimination and bias in AI systems. The provided text defines fairness as a core ethical principle that ensures AI systems do not discriminate based on gender, race, sexual orientation, or religion. Therefore, the correct answer is fairness.

Answer 204

[Image](https://www.examtopics.com/assets/media/exam-media/04234/0008200002.jpg)

Answer 205

Here's the mapping of NLP workloads to scenarios: * **Entity recognition**: Identify the names of people, places, and things in a body of text. * **Sentiment analysis**: Determine whether customer feedback is positive or negative. * **Translation**: Convert a document from English to French. **Explanation:** * **Entity Recognition:** This workload focuses on identifying and categorizing key elements within text, such as names of people, locations, organizations, and quantities. The scenario "Identify the names of people, places, and things in a body of text" directly aligns with this function. * **Sentiment Analysis:** This workload is designed to gauge the emotional tone or attitude expressed in text. The scenario "Determine whether customer feedback is positive or negative" perfectly matches this purpose. Sentiment analysis is used to understand customer opinions and feelings. * **Translation:** This workload involves converting text from one language to another. The scenario "Convert a document from English to French" directly corresponds to the core function of translation services.

Answer 206

CD DISCUSSION: The AI solution needs to first convert the verbal commands into text, which is done by Speech-to-Text (STT). Once the audio is converted to text, the AI needs to understand the intent of the command to control the smart device accordingly. This is achieved through Language Modeling, which helps the AI to interpret the command and determine the appropriate action. Option A, text-to-speech, is used for generating spoken responses from text, not for understanding commands. Option B, key phrase extraction, could be used to identify the main concepts in the text, but language modeling is generally needed to better understand the overall intent. Option E, translation, is not needed because the user is assumed to be speaking in a language the system understands.

Answer 207

C. The Speech service provides text-to-speech functionality, which is needed to read the recipe instructions aloud. Options A, B, and D do not provide text-to-speech capabilities. The Language service provides natural language processing, the Translator service translates text from one language to another, and the Personalizer service helps choose the best content to display to users.

Answer 208

B. The Split Data module is used to divide a dataset into two or more distinct sets, often used for creating training and testing/validation datasets for machine learning models. Option A is incorrect because scaling numeric variables is typically done using modules like "Normalize Data". Option C is incorrect because diverting records with missing data is usually handled by modules like "Clean Missing Data". Option D is incorrect because selecting columns is done by modules like "Select Columns in Dataset".

Answer 209

B. a smart device in the home that responds to questions such as, "What will the weather be like today?" C. a website that uses a knowledge base to interactively respond to users' questions DISCUSSION: Options B and C are correct because they involve understanding and responding to human language, which is the core function of natural language processing. Option B describes a smart device using NLP to interpret a question about the weather. Option C describes a website using NLP to understand and answer user questions from a knowledge base. Option A is incorrect because monitoring temperature and triggering a fan is a simple rule-based automation task, not NLP. Option D is incorrect because autonomously inserting headlamps is a robotic automation task, not NLP.

Answer 210

B. Option B, "AI systems must be transparent and inclusive," aligns with Microsoft's responsible AI principles, which emphasize transparency in AI systems and inclusiveness in their design and impact. Options A, C, and D do not reflect Microsoft's stated principles for responsible AI development and deployment.

Answer 211

* Which people in this group are the same person? - Similarity * Does this person have access to the building? - Identification * Is this the same person as in this photo? - Verification **Explanation:** * **Which people in this group are the same person? - Similarity:** The 'Find Similar' operation performs face matching between a target face and a set of candidate faces to find similar faces, which would answer the question of which people in the group are the same. * **Does this person have access to the building? - Identification:** Face identification is used for "one-to-many" matching of one face to a set of faces in a secure repository, which is used to grant building or airport access to a certain group of people. * **Is this the same person as in this photo? - Verification:** Face verification confirms if the person is who they claim to be, determining if the person in one photo is the same as in another photo.

Answer 212

B. Natural language processing (NLP) is the correct choice because it deals with understanding and processing human language. Extracting key terms from text to generate summaries is a typical NLP task. A. Anomaly detection identifies unusual data points, not relevant for text extraction. C. Computer vision deals with images and videos, not text. D. Knowledge mining involves discovering knowledge from large datasets but relies on underlying AI workloads like NLP to extract the initial information from text.

Answer 213

B. an utterance DISCUSSION: The question "Which act is playing on the main stage?" is an example of an utterance because it represents a specific user input or query to the LUIS application. An utterance is a sentence or phrase that a user types or speaks to interact with the application. Option A is incorrect because an intent represents the purpose or goal behind a user's utterance. Option C is incorrect because a domain refers to a specific area or topic of knowledge. Option D is incorrect because an entity represents a specific piece of information within an utterance.

Practice Questions - exam-ai-900 Flashcards

(246 cards)