Skip to main content

Quiz

GraphRAG Implementation​

Question 1: What two technologies does GraphRAG combine to create a comprehensive knowledge representation system?

  • A. SQL databases and BM25.
  • B. Structured graph databases and vector-based retrieval.
  • C. Flat indexing and HNSW graphs.
  • D. Cross-Encoders and Bi-Encoders.

Answer: B

Question 2: In the GraphRAG architecture, what is the role of 'Entity Extraction'?

  • A. Converting PDFs to images.
  • B. Identifying key entities and relationships from documents.
  • C. Traversing the graph to find answers.
  • D. Storing structured data in Neo4j.

Answer: B

Question 3: Which database technology is explicitly mentioned for storing the graph data?

  • A. MongoDB
  • B. MySQL
  • C. Neo4j
  • D. Redis

Answer: C

Question 4: What is defined as a 'clear promise, obligation, or prohibition' in the extraction rules?

  • A. A constraint.
  • B. A clause.
  • C. A commitment.
  • D. A regulation.

Answer: C

Question 5: What must the information extraction engine do if something does not exist in the text chunk?

  • A. Invent hypothetical data.
  • B. Return an empty list.
  • C. Throw a system error.
  • D. Perform a web search.

Answer: B

Question 6: Which specific Python class is used to convert natural language questions into Cypher queries?

  • A. DocumentConverter
  • B. Neo4jGraph
  • C. GraphCypherQAChain
  • D. RecursiveCharacterTextSplitter

Answer: C

Question 7: In the predefined Pydantic models, what are the four possible options for a 'ConstraintUnit'?

  • A. day, week, month, year
  • B. hours, dong, percent, other
  • C. high, medium, low, none
  • D. true, false, yes, no

Answer: B

Question 8: What type of node represents affected parties in the knowledge graph?

  • A. PolicyClause nodes
  • B. Constraint nodes
  • C. Stakeholder nodes
  • D. Regulation nodes.

Answer: C

Question 9: Which Cypher keyword is used heavily in the ingestion script to prevent the creation of duplicate nodes?

  • A. CREATE
  • B. INSERT
  • C. UPDATE
  • D. MERGE.

Answer: D

Question 10: What relationship connects a Commitment node to a Constraint node in the Cypher queries?

  • A. [:AFFECTS]
  • B. [:REFERENCES]
  • C. [:CONTAINS]
  • D. [:HAS_CONSTRAINT]

Answer: D

Question 11: Why are Pydantic classes utilized during the document extraction phase?

  • A. To split text into chunks.
  • B. To connect to the Neo4j database.
  • C. To serve as validation schemas for structured output from LLMs, ensuring consistency.
  • D. To increase the temperature of the LLM.

Answer: C

Question 12: What does a 'Regulation' node specifically track?

  • A. Measurable limits.
  • B. Legal references.
  • C. Policy topics.
  • D. Employee obligations.

Answer: B

Question 13: Why might you want to create a custom agent instead of relying solely on GraphCypherQAChain?

  • A. Because GraphCypherQAChain cannot connect to Neo4j.
  • B. To validate and refine generated Cypher queries, apply domain-specific optimization, and implement fallback logic.
  • C. Because GraphCypherQAChain requires a local LLM to run.
  • D. To automatically chunk PDF documents.

Answer: B

Question 14: What specific problem does GraphRAG solve that vector similarity search alone cannot handle?

  • A. Generating high-quality images from text.
  • B. Answering questions that require explicit relationships to define how entities connect.
  • C. Translating documents efficiently.
  • D. Searching for exact BM25 keywords.

Answer: B

Question 15: In the prompt rules provided to the LLM for extraction, how should measurable numeric limits within a commitment be handled?

  • A. They should be ignored.
  • B. They should be extracted as constraints.
  • C. They should be saved as separate Regulation nodes.
  • D. They should be converted to standard SI units.

Answer: B

Question 16: How are multiple constraints connected to a single commitment within the graph schema?

  • A. Through direct links to Stakeholders.
  • B. By creating separate independent sub-graphs.
  • C. Constraints are linked to commitments for complex constraints tracking via the HAS_CONSTRAINT relationship.
  • D. They are concatenated into a single string within the Commitment node.

Answer: C

Question 17: What is listed as a potential drawback or cautionary note regarding this specific GraphRAG implementation?

  • A. It relies heavily on specific types of structured data to build an effective knowledge base.
  • B. It is too fast to monitor properly.
  • C. Neo4j does not support Python integrations.
  • D. It cannot handle HR policy documents.

Answer: A

Question 18: What is the fourth step in the answer generation process when querying the graph with natural language?

  • A. Graph Traversal
  • B. Question Processing
  • C. Answer Generation: Converts query results into readable responses.
  • D. Cypher Generation.

Answer: C

Question 19: What relationship is established between a PolicyClause node and a Stakeholder node?

  • A. [:REFERENCES]
  • B. [:CONTAINS]
  • C. [:AFFECTS]
  • D. [:HAS_CONSTRAINT]

Answer: C

Question 20: According to the tips provided, how should you adapt the Pydantic schemas if you were processing medical documents?

  • A. Use the HR schema exactly as is
  • B. Extract 'Symptoms', 'Diagnoses', 'Treatments', and 'MedicationConstraints' instead
  • C. Disable structured extraction completely.
  • D. Only extract Stakeholders and Regulations.

Answer: B