一种基于WordNet和Corpus Statistics的语义相似性计算方法

提出一种新的基于WordNet和文本集语义参数IC的计算方法,通过综合考虑概念在WordNet中语义信息以及数据集中的概率信息,即概念的自信息,同时利用新的参数考虑概念对在WordNet中的共享信息,设计了一种通用的概念语义相似性计算方法,该方法简化了传统语义相似性算法,并解决了语义相似性计算领域的相关问题,可以应用在信息抽取、信息检索、文档分类及本体学习中.领域通用的数据集RB数据实验结果表明,该方法在计算语义相似度问题上有效。

A Semantic Similarity Computing Approach Based on WordNet and Corpus Statistics

We first proposed a new method calculating semantic similarity parameter information content. The new algorithm is based on the conceptsemantic information in the knowledge base called WordNet and the probability in the corpus called self information. Then, considering the existing algorithmsare all domain related and the calculating processes are complicated, we proposed a universal method based on corpus statistics and WordNet calculating semantic similarity which can be used in information extraction, information retrieval, document clustering and ontology learning. The proposed method makes a substantial improvement experimenting on the benchmark data set R&B concept pairs.