基于知网语义相关度计算的词义消歧方法

歧义字段处理一直是中文信息处理领域中最关键也是最困难的问题之一,至今该问题仍没有得到完全而有效的解决,使得以此为基础的多个应用领域都难以取得突破性进展。传统的消歧方法--规则消歧和统计消歧都有不可避免的缺点：规则消歧存在规则的完备性与合理性问题,统计消歧则只取大概率事件而忽视小概率事件。在研究了知网表达汉语知识的基础上,改进了基于知网语义相关度的计算模型,并应用于汉语的歧义字段处理中。经大量例句作实验,以句子为单位的切分正确率可达到97.1%,验证了该消歧方法的有效性。

Word Sense Disambiguating Method Based on HowNet Semantic Relevancy Computation

As one of the most important and also the most difficult problems of Chinese information processing field, disambiguation has not been entirely solved until now. And because of this problem, some application fields based on it have not achieved breakthrough progress. As the traditional disambiguating methods, rule-based disambiguation and statistic-based disambiguation both have the.it inevitable vice： Rule-based disambiguation has the problem of maturity and rationality of rules, while statistic-based disambiguation only takes affairs with large probability and ignores them with little probability. On the basis of our research on HowNet representing Chinese knowledge, this article has ameliorated the model based on HowNet semantic relevancy computation, and utilizes it to disambiguate. After experiment on lots of examples, the segmenting precision can achieve 97.1% if taking sentence as unit; it shows that the disambiguating method is very good.