基于潜在语义分析的本体空间表示模型研究

RESEARCH ON LSA-BASED ONTOLOGICAL SPACE TEXT REPRESENTATION MODEL

针对传统的向量空间模型和潜在语义分析方法应用于计算机辅助评估时存在的问题,提出一种将领域本体、一阶逻辑和潜在语义分析方法相结合的本体空间表示模型。该模型采用一阶逻辑表示从短文问题得到的二元关系并建立索引,使用潜在语义分析来计算关系集合中关系和包含段落的文档的相似度,从而得到段落在关系子集的平均相似度。实验结果表明,与向量空间模型相比,该模型的表示效果更好。

An ontological space text representation model combining domain ontology,first order logic and latent semantic analysis is presented. First order logic is applied to express binary relation derived from essay question, and latent semantic analysis is used to calculate the similarity between relations in subset and the documents containing paragraphs. Finally ,an average similarity value is obtained for paragraphs over the number of relations in the subset. The experimental result shows that the representation effect of ontological space model is better than vector space model.