Repairing Knowledge Graphs

Question 1

What is Knowledge Graph Repair/Refinement?

Answer 1

The process of detecting and correcting errors in a knowledge graph to ensure its reliability for downstream tasks.

Question 2

Name two broad approaches to KG repair.

Answer 2

Manual correction (human) and automatic correction (machine).

Question 3

Advantage and disadvantage of manual KG repair.

Answer 3

Advantage: Highly accurate; Disadvantage: Slow and labor-intensive.

Question 4

Advantage and disadvantage of automatic KG repair.

Answer 4

Advantage: Scalable and fast; Disadvantage: May introduce new inconsistencies.

Question 5

What is General Link Prediction?

Answer 5

A method learning entity and relation embeddings from text to predict missing links.

Question 6

What is Type-link Prediction?

Answer 6

Infers an entity’s semantic type by learning from known (entity, type) pairs.

Question 7

What is Identity-link Prediction?

Answer 7

Identifies when different labels refer to the same real-world entity, merging duplicates.

Question 8

What is Fact Validation?

Answer 8

Automatically checks KG facts against models or external data to detect anomalies.

Question 9

Why use embeddings in KG repair?

Answer 9

They enable similarity queries and anomaly detection for link/type prediction.

Question 10

Name two challenges in automatic KG repair methods.

Answer 10

Handling KG incompleteness and balancing precision vs. recall.