A Novel Parsing-Based Automatic Domain Terminology Extraction Method

Abstract

As domain terminology plays a crucial role in the study of every domain, automatic domain terminology extraction method is in real demand. In this paper, we propose a novel parsing-based method which generates the domain compound terms by utilizing the dependent relations between the words. Dependency parsing is used to identify the dependent relations. In addition, a multi-factor evaluator is proposed to evaluate the significance of each candidate term which not only considers frequency but also includes the influence of other factors affecting domain terminology. Experimental results demonstrate that the proposed domain terminology extraction method outperforms the traditional POS-base method in both precision and recall.