Knowledge graphs and entity resolution

Learning outcome: Keep relationships precise when names are ambiguous.

A knowledge graph represents entities and relationships. It can support queries that depend on connections, such as which document belongs to which product. The usefulness of the graph depends on correct identities and relationships; a visually impressive network can still encode false links.

Entity resolution decides whether different references describe the same entity. Names alone are often insufficient. Use stable identifiers and record evidence for merges. Preserve uncertainty when two records may refer to different people or organisations. Graph-based retrieval can complement text retrieval, but the graph does not make unsupported claims true.

Worked example

A fictional catalogue contains “Cedar Ltd” as a supplier and “Cedar Training” as a customer. An automated merge treats them as one organisation and attaches the customer's contact to a supplier order. Keeping distinct organisation IDs prevents the name similarity from creating a false relationship.

Try it yourself

Draw a graph with two similarly named organisations, two documents and one product. Label each relationship and the evidence needed before merging any nodes.

You can do this exercise on paper. If using an assistant, keep to the fictional material and an approved tool.

Check your understanding

When should two similarly named records be merged?

1. Whenever their names share a word
2. Whenever merging makes the graph simpler
3. When sufficient identity evidence supports the merge

Workbook

Download the course workbook to record your answer and compare it after the knowledge check.

Further reading

• scikit-learn: common pitfalls
• scikit-learn: model evaluation
• NIST: AI Risk Management Framework
• scikit-learn: time series splits
• scikit-learn: novelty and outlier detection
• modAL: uncertainty sampling
• Neo4j: entity resolution