An Incremental Acquisition of Terms Using Automatic Selected Seeds and Extended Variation Rules

Abstract

An automatic pattern extraciton from very large corpora can be esaily applied to corpus-based NLP (Natural Language Processing) research using the domain characteristics. This paper focuses on a fully automatic acquisition of terms and its incremental processing. For the increment, we repeat the process that applies variations to original terms and uses the acquired terms as the original terms of the next iteration. We extend the variation, and combined variation that simultaneously uses both variations. This extension helps to produce more terms incrementally. Our experiment for Web documents shows that the number of newly discovered terms increases by 61.4%, when compared with original variation.