Abstract
We describe a new biomedical search engine that enables visual searching and utilises predicates instead of phrases. We report on the development and evaluation of the triple parser, the most essential component of the search engine, which extracts the necessary predicates from the biomedical text. Using texts from three biomedical related sites (N = 180), we compared the parser's output with a gold standard independently created by a medical expert. The parser achieved more than 91% precision and recall. Its individual components showed different strengths with Finite State Automata being excellent for achieving high recall, while Support Vector Machines improved the precision.
Original language | English (US) |
---|---|
Pages (from-to) | 351-367 |
Number of pages | 17 |
Journal | International Journal of Biomedical Engineering and Technology |
Volume | 10 |
Issue number | 4 |
DOIs | |
State | Published - Jan 1 2012 |
Keywords
- Finite State Automata
- Kernel methods
- Search engine
- Support Vector Machines
- Text mining
- Triple
ASJC Scopus subject areas
- Biomedical Engineering