基于语义框架填充过程的语义相似度计算

Semantic similarity calculation based on semantic frame of filling process

为了解决传统相似度计算结果准确度较低的问题,提出了一种基于语义框架填充过程的语义相似度计算方法。在知识表示层面利用语义框架为两个文档之间的相似度评价建立相似语境,该框架采用多层次语义槽表示文档的信息结构和内容。通过逻辑推理将语义框架中缺省的语义槽信息进行挖掘,进而计算出两者的相似度。实验结果表明,该方法相比传统的基于词频的相似度计算方法在相似度结果准确性方面有显著的提高。

To solve the problem of low similarity calculation result in traditional technology, a method using semantic similarity calculation based on filling procession of the semantic frame is proposed. First, similar context for similarity evaluation of the two documents using semantic framework at knowledge representation level is established, and multiple levels of semantic slots are used to represent the structure and content of the document. Then, the default semantic slot of the semantic framework is mined using logical reasoning to achieve the similarity. The experimental results show that the proposed method can achieve much better similarity results than traditional methods.