A practical competition of different filters used in automatic term extraction
Le An Ha
Research Group in Computational Linguistics, School of Humanities, Languages and Social Sciences, University of Wolverhampton
This paper discusses an experiment where different filters used in automatic term extraction (ATE) are practically compared. In the experiment, 8 filters, belong to three groups (lexical syntactic, statistical and semantic filters), are used to extract terms from two corpora from the domain of chemistry and of cancer research. The performance of each individual filter, and similarity among them are calculated. The experiment shows that: 1) simple filters maybe very efficient ones; 2) those filters are really different from each others; 3) the choice of which filters to be used is a domain, genre, and application-specific issue.
Automatic term extraction, evaluation