Text Similarity Based on Semantic Analysis
Junli Wang, Qing Zhou, Guobao Sun · 2016
One of the most important challenges in measuring text similarity is language variability: texts with the same meaning can be realized in several ways.A way to address the language variability is the notion of semantic similarity.This paper extracts the relevance of texts and terms through Singular Value Decomposition (SVD).According to Bayesian Network, we construct term-topic sets and then use Mutual Information (MI) to calculate the semantic similarity between terms.Finally, we use graph structures instead of term vectors to calculate text similarity.