Resources

Biomedical text mining for knowledge extraction tutorial @ ISMB 2025

This tutorial provides a hands-on introduction to the core tasks in biomedical natural language processing (BioNLP) for extracting knowledge from text. These include named entity recognition to identify mentions of important concepts, entity linking to connect mentions with ontologies and relation extraction methods to understand what text is saying about the entities. It alsos explore the strengths and weaknesses of large language models for information extraction. Visit the tutorial website for the slides and Jupyter notebooks.

Mini Code Tutorials

We have a number of short code tutorials that we have made to help explain small concepts. We hope they're helpful.

NLP related
- HuggingFace tokenizer tricks: Showing some of the options for using and customizing the functionality of HuggingFace tokenizers
- Creating a custom token classification model: Illustrates creating a custom model in HuggingFace
- Further pretraining a language model and generating text with it: Demonstrates training a causal language model and generating text
- Creating a HuggingFace dataset object: One way to create a HuggingFace Dataset object that works with the Trainer
Knowledge graph related
- Two-Hop Link Prediction for a Knowledge Graph: Describes the problem of link prediction in a knowledge graph and implements a basic rule-based approach
Biomedical related
- Getting MeSH Tags with Entrez API: Illustrates using the Entrez API to access the MeSH tags on biomedical documents
- Loading BioC XML files plus dealing with an archive of them: Shows how to use the bioc package which is often used to store biomedical documents and associated annotations
- Loading a PubMed XML file: Shows how to load a PubMed XML file available through the bulk download system and extract title, abstract and other metadata
Entity linking and information retrieval related
- Candidate Generation for Entity Linking: Shows different methods for identify candidate entities given a mention
- Illustrating a Bi-Encoder architecture: Shows a method for calculating document similarity using PyTorch
Deep learning related
- Example of PyTorch Training Loop with Custom Hugging Face Model: Shows a standard PyTorch training loop and all the set up to get a batch of data ready