Skip to content

Course Description

This file is the seed document used by the learning-graph-generator skill to enumerate concepts, build the dependency graph, and assign concepts to a taxonomy. Keep it focused, concrete, and free of marketing language.

A good course description includes:

  • Title — same as site_name in mkdocs.yml
  • Audience — who this book is for and what they already know
  • Prerequisites — concepts the reader is assumed to have mastered
  • Topics — the major areas the book covers (typically 8–20 topics)
  • Bloom's Taxonomy outcomes — what the reader should be able to remember, understand, apply, analyze, evaluate, and create by the end

Run the course-description-analyzer skill to validate completeness, then run learning-graph-generator to enumerate ~200 concepts with dependencies.

Title

Learning Knowledge Graphs

Why This Book Matters

Knowledge graphs sit at the center of two converging trends: they are the backbone that grounds large language models in verifiable facts (via GraphRAG and similar retrieval patterns), and they are increasingly used inside professional-services organizations — including law firms — to connect matters, documents, people, and precedent. A reader who understands both the classical semantic-web foundations and the modern LLM-integration patterns is equipped to evaluate, build, or commission a knowledge graph project in either a technical or a legal knowledge-management setting.

Audience

This book is written for two overlapping groups of tech-forward readers:

  • Legal professionals — lawyers, legal knowledge managers, legal operations staff, and law librarians who are comfortable adopting new technology and want to understand knowledge graphs well enough to evaluate, commission, or contribute to a legal knowledge graph project (matter graphs, citation networks, expertise location, conflicts checking).
  • Scientists, engineers, and developers who want a structured, hands-on introduction to knowledge graphs — from modeling through querying, analytics, and integration with modern LLM-based tools.

Both groups are assumed to be curious about the underlying technology rather than looking for a purely conceptual overview — the book includes runnable code and real query examples throughout, not just diagrams.

Prerequisites

  • Comfort with basic set and relationship concepts (e.g., categories, hierarchies, many-to-many relationships) — no formal mathematics beyond that is assumed.
  • Familiarity with at least one programming language is helpful for the hands-on chapters (examples use Python) but is not required to follow the conceptual chapters.
  • No prior knowledge of RDF, OWL, SPARQL, Cypher, graph databases, or the semantic web is assumed — these are introduced from first principles.
  • Prior exposure to taxonomies, thesauri, or document classification systems (e.g., from legal knowledge management, library science, or information architecture) is helpful but not required.

Topics

  1. What is a knowledge graph — nodes, edges, and triples, contrasted with relational tables and document/folder-based systems
  2. From taxonomies to ontologies — controlled vocabularies, thesauri, and formal class hierarchies
  3. Graph data models — RDF triples vs. labeled property graphs, and when to use each
  4. Modeling concepts and relationships — entities, classes, properties, cardinality, and schema design
  5. Semantic web standards — RDF, RDFS, OWL, SKOS, and SPARQL
  6. Building a knowledge graph from data — sources, ETL, entity extraction, and entity resolution/deduplication
  7. Querying knowledge graphs — SPARQL and Cypher, with Python client examples
  8. Graph algorithms and analytics — centrality, community detection, and path-finding
  9. Knowledge graphs and machine learning — embeddings, graph neural networks, and link prediction
  10. Knowledge graphs meet large language models — GraphRAG, LLM-grounded retrieval, and LLM-assisted graph construction
  11. Case studies across industries — search, biomedical research, enterprise knowledge management, and recommendation systems
  12. Legal and enterprise knowledge graphs — matter/document graphs, citation networks, expertise location, and conflicts checking
  13. Governance, quality, and maintenance — versioning, provenance, data quality, and keeping a graph current over time
  14. Capstone — designing, building, querying, and visualizing an original knowledge graph project

Topics Not Covered

  • General relational database administration and performance tuning
  • Deep distributed-systems or cloud-infrastructure operations for hosting graph databases at scale
  • Legal doctrine, case law, or substantive legal analysis beyond illustrative examples
  • General statistics or machine learning theory beyond what is needed to understand embeddings and graph neural networks conceptually
  • Building or fine-tuning large language models from scratch

Learning Outcomes

By the end of this book, the reader will be able to:

Remember

  • List the core graph primitives: node, edge, triple, and property.
  • Name the major semantic web standards: RDF, RDFS, OWL, SKOS, and SPARQL.
  • Identify the two dominant graph data models: RDF triple stores and labeled property graphs.
  • Recall the stages of building a knowledge graph, from source data through governance.

Understand

  • Explain the difference between a taxonomy, an ontology, and a full knowledge graph.
  • Explain when a property graph model is preferable to an RDF triple store, and vice versa.
  • Describe how entity resolution and deduplication address the same real-world entity appearing under different labels or identifiers.
  • Explain how GraphRAG grounds large language model outputs in a knowledge graph.

Apply

  • Model a real-world domain (e.g., a legal matter or a scientific dataset) as a graph schema.
  • Write basic SPARQL queries to retrieve and traverse RDF data.
  • Write basic Cypher queries to retrieve and traverse property-graph data.
  • Use a Python graph client library to load and query a small dataset.

Analyze

  • Apply centrality measures to identify the most influential entities in a graph.
  • Apply community detection to find clusters of related entities.
  • Diagnose data-quality and entity-resolution issues in a constructed graph.
  • Trace a multi-hop path between two entities and explain what it reveals about their relationship.

Evaluate

  • Compare RDF vs. property graph models for a given use case and justify a recommendation.
  • Assess whether manual curation or LLM-assisted extraction better fits a project's accuracy and cost constraints.
  • Critique a knowledge graph schema for missing, ambiguous, or redundant relationships.
  • Judge whether a legal or enterprise knowledge graph's governance plan is sufficient to keep it current.

Create

  • Design a graph schema (classes, properties, cardinality) for an original domain.
  • Build and populate a knowledge graph from a real or realistic dataset.
  • Query and visualize the resulting graph to answer a specific question.
  • Draft a governance plan (versioning, provenance, update cadence) for keeping the graph current.