Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AutoPCR: Automated Phenotype Concept Recognition by Prompting

Yicheng Tao, Yuanhao Huang, Yiqun Wang, Xin Luo, Jie Liu

TL;DR

AutoPCR is presented, a prompt-based phenotype CR method that does not require ontology-specific training and achieves the best average and most robust performance across both mention-level and document-level evaluations, surpassing prior state-of-the-art methods.

Abstract

Phenotype concept recognition (CR) is a fundamental task in biomedical text mining, enabling applications such as clinical diagnostics and knowledge graph construction. However, existing methods often require ontology-specific training and struggle to generalize across diverse text types and evolving biomedical terminology. We present AutoPCR, a prompt-based phenotype CR method that does not require ontology-specific training. AutoPCR performs CR in three stages: entity extraction using a hybrid of rule-based and neural tagging strategies, candidate retrieval via SapBERT, and entity linking through prompting a large language model. Experiments on four benchmark datasets show that AutoPCR achieves the best average and most robust performance across both mention-level and document-level evaluations, surpassing prior state-of-the-art methods. Further ablation and transfer studies demonstrate its inductive capability and generalizability to new ontologies.

AutoPCR: Automated Phenotype Concept Recognition by Prompting

TL;DR

AutoPCR is presented, a prompt-based phenotype CR method that does not require ontology-specific training and achieves the best average and most robust performance across both mention-level and document-level evaluations, surpassing prior state-of-the-art methods.

Abstract

Phenotype concept recognition (CR) is a fundamental task in biomedical text mining, enabling applications such as clinical diagnostics and knowledge graph construction. However, existing methods often require ontology-specific training and struggle to generalize across diverse text types and evolving biomedical terminology. We present AutoPCR, a prompt-based phenotype CR method that does not require ontology-specific training. AutoPCR performs CR in three stages: entity extraction using a hybrid of rule-based and neural tagging strategies, candidate retrieval via SapBERT, and entity linking through prompting a large language model. Experiments on four benchmark datasets show that AutoPCR achieves the best average and most robust performance across both mention-level and document-level evaluations, surpassing prior state-of-the-art methods. Further ablation and transfer studies demonstrate its inductive capability and generalizability to new ontologies.

Paper Structure

This paper contains 17 sections, 4 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Phenotype concept recognition
  4. Distantly-supervised biomedical entity linking
  5. Method
  6. Experiments
  7. Benchmarks and baselines
  8. Implementation details
  9. Main results
  10. Ablation study
  11. Generalizability
  12. Hyperparameter sensitivity
  13. Conclusion
  14. Conflicts of interest
  15. Funding
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Architecture of AutoPCR. It performs concept recognition in three stages: entity extraction using BioNER and syntactic entity extraction (e.g., extracted entities $e_1$, $e_2$, and $e_3$), candidate concept retrieval via SapBERT initialized from PubMedBERT (e.g., retrieved concepts $c_1$ and $c_2$ for $e_3$), and entity linking through prompting an LLM (e.g., linked concept $c_1$).
  • Figure 2: Sensitivity analysis of AutoPCR w.r.t. $\tau_1$ shows robustness.
  • Figure 3: Sensitivity analysis of AutoPCR w.r.t. $\tau_2$ and $k$ shows robustness.

Theorems & Definitions (4)

  • Definition 1: Concept recognition
  • Definition 2: Entity extraction
  • Definition 3: Candidate concept retrieval
  • Definition 4: Entity linking