Entity Communication Networks
Extract structured communication metadata from documents and build entity networks in Neo4j
In this 2-hour course, you will learn
Welcome to the Entity Communication Networks course.
In this hands-on course, you will extract text from raw documents, parse it into structured records, and build a communication network in Neo4j. You'll handle OCR noise, build rule-based and LLM-based parsers, train an NER model, and combine them into a hybrid pipeline that imports a metadata graph of entities and their communications.
By the end, you'll also receive a prompt pack — an Go to glossary for large language model (opens in a new tab)A model trained on text to predict the next token, and so to generate language. project that examines your own data, references the techniques from this course, and builds a custom pipeline for any document source. The course teaches you the concepts; the pack applies them to your data.
Before taking this course, you should have:
- Completed the Cypher Fundamentals course, or equivalent Full definition for Cypher (opens in a new tab)Neo4j's implementation of GQL, the ISO standard query language for graph databases. It is declarative: you describe the pattern to find, and the database decides how to find it. experience
- Basic understanding of Neo4j Full definition for property graph (opens in a new tab)The data model Neo4j implements, in which nodes and relationships both carry properties as well as labels and types. concepts
- Familiarity with Python
PDF Extraction
Extract raw text and structured metadata from PDF documents using Python libraries, handling multi-page files and preserving sender and recipient information.
Document Parsing
Parse unstructured document content into discrete entities and relationships by applying rule-based techniques to identify names, organisations, and communication patterns.
LLM Parsing
Use large language models to interpret ambiguous or complex document sections, extracting structured communication metadata that rule-based parsers cannot reliably resolve.
Hybrid Pipelines
Build extraction pipelines that combine deterministic rules with LLM inference, routing document sections to the appropriate parser based on content complexity and confidence thresholds.
Neo4j Import
Import extracted entities and relationships into Neo4j using Cypher MERGE statements, constructing a queryable communication network that links senders, recipients, and documents.
2 modules, 2 hours.
- What's in your documents?5 min
- Types of Graph5 min
- Parsing Approaches5 min
- Layout-aware extraction5 min
- Parsing libraries5 min
- Rule-based parsing5 min
- Building templates5 min
- ML-based parsing5 min
- Parsing with an LLM5 min
- The hybrid pipeline5 min
- Normalize your data5 min
- Import to Neo4j5 min
- Investigate your graph5 min
- Prompt pack5 min