基于信息内容的词林词语相似度计算

IC-based approach for calculating word semantic similarity in CiLin

针对哈尔滨工业大学《同义词词林》扩展版的层次结构不能有效反映词语之间信息内容含量差异性的问题进行了研究,进行了《同义词词林》作为词语相似度计算本体的结构改造,增加了原编码信息节点的语义,提出了一种较为适合改造后本体的相似度计算策略。经实验证明,修改后的本体更能体现词语在本体中信息内容含量的差异性,提出的相似度计算策略应用在改进后的本体上时,得出的相似度计算结果准确程度达到了较高水平,具有较好的实用价值。

This paper resolved the problem that the hierarchical structure of CiLin provided by HIT （ Harbin Institute of Tech- nology） could not effectively reflect the difference of information content among the words, extracted a word for a node that only has the encoding, optimized the structure of CiLih, and proposed a semantic similarity computing strategy. The experimental resuits show that the optimized CiLin can better reflect the difference among the words. The proposed new strategy is applied to the optimized CiLin, the accuracy of the calculation results reach a high level, and has good practical value.