Using OntoGPT to boost ontology curation

OntoGPT is a Python package for extracting structured information from text with large language models (LLMs), instruction prompts, and ontology-based grounding.

Two different strategies for knowledge extraction are currently implemented in OntoGPT:

• SPIRES: A Zero-shot learning (ZSL) approach to extracting nested semantic structures from text
• TALISMAN (previously known as SPINDOCTOR): Summarizes gene set descriptions (pseudo gene-set enrichment)

More info about OntoGPT:

• Documentation: https://monarch-initiative.github.io/ontogpt/
• Source: https://github.com/monarch-initiative/ontogpt
• Preprints:
  • Caufield JH, Hegde H, Emonet V, Harris NL, Joachimiak MP, Matentzoglu N, et al. Structured prompt interrogation and recursive extraction of semantics (SPIRES): A method for populating knowledge bases using zero-shot learning. http://arxiv.org/abs/2304.02711
  • Joachimiak MP, Caufield JH, Harris NL, Kim H, Mungall CJ. Gene Set Summarization using Large Language Models. http://arxiv.org/abs/2305.13338

We are also working on a curation-specific tool, CurateGPT, a prototype web application and framework for performing general-purpose AI-guided curation and curation-related operations over collections of objects.

Please see this walkthrough to learn how to leverage OntoGPT for ontology curation.