Knowledge and Language Flashcards
(12 cards)
What are the two paradigms of AI knowledge representation and their key characteristics?
Symbolic AI (Knowledge Graphs): explicit, structured knowledge; logical reasoning; human-interpretable; formal semantics.
Neural AI (Language Models): implicit, distributed knowledge; statistical pattern recognition; black-box reasoning; probability-based.
Define a Knowledge Graph.
A graph-structured knowledge base that integrates real-world entities and their relationships using semantic edges and properties, enabling structured, machine-processable, and human-interpretable knowledge.
What are the four key components of a Knowledge Graph?
- Entities (Nodes): real-world objects or concepts.
- Relationships (Edges): semantic links between entities.
- Attributes (Properties): data values describing entities.
- Ontologies (Schemas): definitions of entity types and allowed relationships.
What are the main benefits of using Knowledge Graphs in AI systems?
– Structured, verifiable facts
– Precise querying and logical inference
– Explainability and provenance
– Integration across domains
How do Language Models represent knowledge compared to Knowledge Graphs?
LLMs encode knowledge implicitly via neural network weights and probability distributions, offering flexibility and generalization but lacking explicit structure, verifiability, and precise querying.
Why combine Knowledge Graphs with Language Models?
To leverage the structured accuracy and reasoning of Knowledge Graphs with the flexible language understanding of LLMs, yielding improved factual accuracy, explainability, and natural interfaces.
What is KG-Enhanced LLMs integration strategy?
Use a KG to supply verified facts to an LLM during response generation (e.g., Retrieval-Augmented Generation) for factual accuracy and context grounding.
What is LM-Enhanced KG integration strategy?
Use an LLM to extract entities and relations from unstructured text to build or enrich a KG, or to translate natural language questions into structured queries (SPARQL).
What is a Hybrid Reasoning System in the context of KG and LLM integration?
An AI system orchestrating both an LLM module (for language tasks) and a KG module (for logical reasoning and fact retrieval), choosing the appropriate component per query and combining outputs.
Give an example of a retrieval-augmented generation (RAG) workflow.
- User asks a question.
- LLM formulates a query.
- Query retrieves facts from the KG.
- LLM uses retrieved facts to generate a natural-language answer.
Name two real-world applications of Knowledge Graph–LLM integration.
Examples include: Google Search Knowledge Panels (KG-enhanced QA) and IBM’s Project Debater (KG-backed argument generation).
Summarize the role of ontologies (schemas) in Knowledge Graphs.
Ontologies define the vocabulary (classes, properties, domains, ranges) that enforce structure, consistency, and constraints within the KG schema.
