week 4 - history Flashcards

Question

pre ai ocr

Answer 1

pre prgrammed set of rules ie, see vertical like with 2 diagnols to describe a K

Answer 2

provide computer with lots of examples "training data" different fonts, styles etc so it recognizes "k"

Answer 3

Neural network OCR looks at pictures of letters, finds patterns with CNNs, and reads the letters in order using RNNs to turn the image into text.

Answer 4

A type of neural network that looks at images and finds important shapes and patterns, like edges or letters.

Answer 5

A neural network that reads data in order, like letters in a word, to understand the sequence and meaning.

Answer 6

ground truth

Answer 7

HTR uses neural networks to read handwritten text.

Answer 8

It gives easy access to old handwritten archives. It helps quickly explore and analyze large amounts of text. It supports research across different fields. It allows using text data for machine learning and data mining.

Answer 9

We focus more on easy-to-get digital texts and ignore harder-to-find ones. This can keep one-sided stories in charge. OCR/HTR isn’t always perfect and can cause mistakes. People rely too much on searching by keywords only. But AI can help fix these problems.

Answer 10

topic modelling Named entity rec word embeddings

Answer 11

looks through lots of text to find groups of related ideas or themes that might not be obvious — like finding hidden stories or topics. "It groups letters by subject, like “family stuff” or “travel,” so you don’t have to read everything.

Answer 12

is a tool that finds names of people, places, or things in the text — it can help spot people who are usually overlooked or ignored. It finds all the names in a text, like people or places, even if they are not famous.

Answer 13

are ways AI understands the meaning and style of words by comparing them — this helps see how language is used differently or similarly, like spotting different opinions or ways of thinking.

Answer 14

help us understand big collections of texts faster and better. They find hidden stories, important people, and how language changes over time — things we might miss if we just search for words. This makes research smarter and more complete.

Answer 15

moving beyond just being able to find words in old documents (searchable)… …to being able to really study and analyze them with AI tools

Answer 16

1. historical docs 1. ground truth = correct text we use to train ai to read handwriting 2. loghi htr = tool uses ai to read and transcribe handwrittens docs 3. page XML files = store both the image of the page and the text layout

Answer 17

Finding names of people, places, or things in text (entity recognition) And connecting them to real-world info or databases (linking)

Answer 18

Understanding the deeper meaning of words in their context — not just reading them, but knowing what they really refer to or imply.

Answer 19

web of linked data way of making the internet smarter — by helping computers understand the meaning of information, not just read the words.

Answer 20

Subject – Predicate – Object help computers understand relationships between things. They're like building blocks of meaning

week 4 - history Flashcards

(44 cards)