Related Experiment Videos
Extracting gene pathway relations using a hybrid grammar: the Arizona Relation Parser
Daniel M McDonald1, Hsinchun Chen, Hua Su
1Artificial Intelligence Laboratory MIS Department, University of Arizona, 1130 E. Helen St, Tucson, AZ 85721, USA. dmm@eller.arizona.edu <dmm@eller.arizona.edu>
Bioinformatics (Oxford, England)
|July 17, 2004
Summary
The Arizona Relation Parser uses a hybrid grammar to extract biological relations from biomedical text, achieving 89% precision. This tool enhances hypothesis generation by improving access to research findings.
Area of Science:
- Biomedical informatics
- Natural Language Processing
Background:
- Biomedical research is rapidly expanding, necessitating efficient methods for accessing new findings.
- Text-mining aims to improve accessibility and accelerate hypothesis generation.
Purpose of the Study:
- To introduce the Arizona Relation Parser (AZ parser).
- To evaluate the performance of a novel syntax-semantic hybrid grammar for relation extraction.
Main Methods:
- Developed a syntax-semantic hybrid grammar for relation extraction.
- Trained the AZ parser on 40 PubMed abstracts.
- Tested the parser on 100 unseen PubMed abstracts for precision and recall.
Main Results:
- The AZ parser achieved 89% precision in extracting biologically relevant relations.
- Recall with semantic filtering reached 35% and 61% (before filtering).
- The parser demonstrated robust performance across diverse writing styles and semantic content.
Conclusions:
- The AZ parser offers competitive performance compared to existing semantic parsers.
- Its hybrid grammar approach provides a balance of broad coverage and accuracy.
- Extracted relations are available for visualization, aiding biomedical research.