一种反映词语相关度语义库的构建方法

Constructing Semantic Library to Reflect Word Interrelationship

为了反映词语间的语义相关程度,提出了一种基于向量空间模型的构建语义库的新方法.在构建语义库时,对大量语料文本进行迭代式学习,在学习过程中引入淘汰算法,并综合考虑了诸如共同出现次数、平均出现距离、信息熵以及单字语义信息等多种对词语间语义关系产生影响的因素.实验证明,用该方法得到的相关语义库能够较好地反映现实世界中词语之间的相关程度.

In order to reflect word interrelationship, a new method based on a vector space model was presented to construct a semantic library. The semantic library is trained with large number of texts. This training is iterative, makes use of elimination algorithm, and takes into consideration of many relevant factors that may have effects on semantic relationship between words. The factors considered in the construction include co occurrence, average distance, information entropy, and meanings of Chinese characters. The experiments show that this library is able to reflect the interrelationships among words in the real world.