knowledge graph, tutorial

Knowledge Graphs from Text

Welcome

Let’s go, let’s create some knowledge graphs from publications.

• Nodes: “entities” extracted with Named Entity Recognition (NER)

• Edges: “relations” extracted with Relation Classification (RC)

Things we need:

• Publications, we can search for them on PubMed.

• Visualisation tools.

Other tutorials I used:

• Building a Knowledge Base from Texts (https://www.nlplanet.org/course-practical-nlp/02-practical-nlp-first-tasks/16-knowledge-graph-from-text)

• Construct a biomedical knowledge graph with NLP (https://towardsdatascience.com/construct-a-biomedical-knowledge-graph-with-nlp-1f25eddc54a0)

Tools:

• Text preprocessing: Natural Language Toolkit (NLTK)

• Named entity recognition: BERN2, PubTator 3.0 with AIONER

• Relation extractor: zero-shot relation extractor (zero_shot_re)

        graph LR
  A[NLTK] --> B[NER] 
  B --> C[Relation extractor]
    

Text preprocessing

Tool: Natural Language Toolkit (NLTK)

Named entity recognition (NER)

Servers:

• BERN2

• PubTator 3.0 with AIONER

Relation extractor

I used the code for zero-shot relation extractor based on the paper Exploring the zero-shot limit of FewRel from fractalego/zero-shot-relation-extractor pip install didn’t work, so in the end I just copied the codes from extractor.py and model.py.

Model: fractalego/fewrel_zero_shot

Construct Knowledge Base (KB)

Plot Knowledge Base (KB)

References