Information extraction

As explained in Graph Algorithms for Data Science

An information extraction pipeline is a multistep computational process that identifies entities and relations in text and produces structured information.

An information extraction pipeline turns unstructured text into structured facts by identifying entities, resolving references, extracting relations, and producing triples. A triple contains a head, relation, and tail; together, triples can construct a graph. section 12.0

A practical pipeline commonly combines these stages:

• Coreference resolution connects pronouns and other references to the entities they describe. section 12.0
• Named entity recognition and linking identifies people, organizations, and other entities and connects mentions to entities. section 12.0
• Relation extraction identifies links between entities, such as a person founding a company or an organization acquiring another organization. section 12.0 section 12.4
• Graph construction converts extracted triples into nodes and relationships, while some facts may be stored as node properties. section 12.0

Extracting facts about Apple

From text about Apple Inc., the pipeline can produce triples such as Steve Jobs–FOUNDED–Apple, Steve Wozniak–FOUNDED–Apple, and Apple–FOUNDED_ON–April 1, 1976. The evaluation text also describes Apple’s first product, founders, financing, and acquisitions. section 12.0 section 12.4

Figure 12.2 Information extraction pipeline design

Use the pipeline design figure to follow how text becomes extracted information and then a graph. section 12.0

How Graph Algorithms for Data Science covers it

How deep it goes

Walks through existing-model implementation and evaluation, while deferring custom model training and WikiData enrichment.

Examples

Uses Apple’s founding, products, founders, financing, and acquisitions as evaluation examples for extracting entities and relationships.

What it assumes

Assumes basic Python familiarity while avoiding custom NLP model training through pretrained models and beginner-friendly interfaces.

What you can do after

Build an initial pipeline with SpaCy, coreferee, transformers, and pretrained models, then convert extracted triples into a graph.

How it connects

Connects natural language processing with coreference resolution, named entity recognition, relation extraction, and knowledge graph construction.

liveBookmanning.com

Read Graph Algorithms for Data Science on liveBook Get the book at manning.com

Related concepts

• tupleproduces · A tuple is a finite, ordered collection of elements treated as a single composite value or record.
• spaCyused by · spaCy is an open-source Python library for natural language processing, providing tools and pretrained pipelines for tasks such as tokenization and linguistic analysis.
• named entity recognitionhas part · Named entity recognition is a natural-language-processing task that identifies textual mentions of entities and classifies them into categories such as people, places, organizations, or dates.
• relation extractionhas part · Relation extraction is a natural-language-processing technique that identifies semantic relationships between entities expressed in text.
• coreference resolutionhas part · Coreference resolution is the task of identifying expressions in text that refer to the same entity and linking those expressions to it.

This explainer was generated by Manning's knowledge graph from the text of Graph Algorithms for Data Science. The section links point to the parts of the book it draws on.