Improving Biomedical Summarization with Word Sense Disambiguation
Author Information
Author(s): Laura Plaza, Antonio J. Jimeno-Yepes, Alberto Díaz, Alan R. Aronson
Primary Institution: Universidad Complutense de Madrid
Hypothesis
Using a WSD algorithm to choose between the candidate UMLS concepts returned by MetaMap will improve the performance of the summarizer.
Conclusion
The use of WSD techniques positively impacts the results of the graph-based summarizer, showing a correlation between WSD and summarization tasks.
Supporting Evidence
- WSD improves the performance of a graph-based summarization system.
- The best WSD algorithm tends to be the best for summarization.
- Errors in disambiguation depend on the importance of concepts in the document.
Takeaway
This study shows that understanding the meaning of words helps create better summaries of medical texts.
Methodology
The study evaluates three WSD methods and a graph-based summarizer using the UMLS Metathesaurus.
Limitations
The improvement in summarization is less than expected due to errors in WSD and the summarizer's implicit WSD.
Statistical Information
P-Value
0.013
Statistical Significance
p<0.01
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website