基于多策略的领域本体术语抽取研究

Terminology Extraction Research Based on Integrated Strategy Method

术语的抽取是领域本体构建的基础工作，决定了本体构建的质量。获取的术语除了要求有准确的短语识别率，还要求有较高的术语领域度。本文试图研究一种不依赖于背景语料的术语领域度筛选方法。本文的主要工作集中在两个方面：一是通过统计和规则相结合的方法从领域语料中抽取候选术语（短语），二是提出了通过候选术语的分布度、活跃度以及主题度进行计算的多策略术语抽取方法，并通过实验进行了验证和分析。实验结果表明，在小规模航空航天领域语料库上进行验证性实验后发现，在不大量增加计算时间复杂度的情况下，能够有效提高领域术语抽取的质量，获得令人较满意的结果。

Terminology extraction is one of the most important basic prepare work for ontology construction, which assured the qualification of ontologies for building. The acknowledged terminolo- gy should not only have high recognized precision, but also have high termhood in the domain. This paper tried to find a method for terminology extraction not relied on background corpus supported. Our work focused on two aspects, one is discussing a phrase recognized approach based on statistical and Chinese grammar rule, and the other is that we proposed an approach for termhood calculation of candidate terminology which synthesized three factors of distribution degree, activity degree and sub- ject degree. Experiment on testing corpus shows that our method can have good result in terms of precision and recall.