基于多重继承与信息内容的知网词语相似度计算

针对目前中文词语语义相似度方法中，基于信息内容的算法研究不足的问题，对知网信息模型上使用基于信息内容的中文词语相似度算法进行了研究。根据知网采用语义表达式表示知识而缺乏完整概念结构的特点，通过抽取知网语义表达式中的抽象概念，结合原知网义原树构建具有多重继承特征的知网义项网作为基于信息内容的计算本体。根据该义项网，对基于信息内容的词语相似度算法进行了改进，提出了新的信息内容含量计算方法。经过Miller&Charles（MC30）基准平台的测试，验证了基于信息内容方法在计算中文语义相似度方面的可行性，也证明了本文的计算策略和改进算法的合理性。

Based On Multi-Inheritance And IC Approach For Calculating Word Semantic Similarity In Hownet

For the issue of insufficient research on IC-based Chinese word semantic similarity approach, this paper researched the IC-based Chinese word semantic similarity measurement on Hownet information model. According to the characteristics of the Hownet, which takes the concept semantic expression to represent knowledge and lacks of complete conceptual structure, this paper constructed a multi-inheritance-based Hownet concept net as the ontology of IC-based measurement by extracting the abstract concepts from the concept semantic expression, and extending the original Hownet sense structure. Based on it, this paper studied the given IC-based Chinese word semantic similarity, and then improved the IC approach. The test results on MC30 show that the IC-based Chinese word semantic similarity measurement is feasible. The test results also verifie the validity of this method on Chinese semantic similarity measurement.