Extracting Biomedical Relations Using Conditional Random Fields
Author Information
Author(s): Bundschus Markus, Dejori Mathaeus, Stetter Martin, Tresp Volker, Kriegel Hans-Peter
Primary Institution: Ludwig-Maximilians-University Munich
Hypothesis
Can Conditional Random Fields effectively extract semantic relations between biomedical entities from text?
Conclusion
The study demonstrates that Conditional Random Fields can successfully extract and classify semantic relations in biomedical literature.
Supporting Evidence
- The cascaded CRF model achieved an F-measure of 72% for named entity recognition.
- The model identified 34,758 semantic associations between 4,939 genes and 1,745 diseases.
- The study demonstrated competitive performance compared to other state-of-the-art methods.
Takeaway
This study shows how computers can read medical papers and find connections between diseases and genes, helping scientists understand more about health.
Methodology
The study used Conditional Random Fields to extract and classify relations from biomedical texts, benchmarking on two datasets: disease-treatment relations from PubMed abstracts and gene-disease relations from GeneRIF phrases.
Limitations
The study's performance may be limited by the quality of the training data and the complexity of the biomedical language.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website