Semantic Similarity Based On Interval Intuitionistic Fuzzy Sets Is Studied

Semantic similarity as a key technology in Chinese information-disposing, has been widelyapplied in the field of automatic classification, automatic clustering, machine translation,information retrieval, information filtering. To express the rich semantic information is one ofthe most difficult problems in traditional semantic similarity algorithm. So the calculated resultshave a certain gap with the person’s subjective understanding. Interval-valued intuitionistic fuzzysets have a powerful ability to describe the fuzzy information and adopt inter-valuedintuitionistic fuzzy numbers for semantic fuzzy information. So semantic similarity based onIVIFS describes object in detail and mines implied meaning, which can improve the accuracy ofsimilarity calculation and break through the bottleneck in semantic representation of traditionalsemantic similarity.This paper introduces domestic and international situation of semantic similarity at first andraises the main problem that the current semantic similarity not fully reflects the semanticinformation. The article introduces the IVIFSs descript semantic information and elaborated onhow to use IVIFSs shows the level, depth, density, semantic transmission etc. On this basis, webuild fuzzy matrices. The problem of semantic similarity is transformed into a fuzzy matrix. Andcombination with idea from HowNet word is integrated by sememe", the paper realizes thesemantic similarity based on IVIFSs. Finally,30pairs of words are tested to prove the validity ofthe algorithm.The main content is:（1）The paper dwells on the necessary of the research of semantic similarity, and analysisdomestic and international situation of semantic similarity.（2）Analyzes and classifies the main semantic similarity algorithm, and proposes theinadequacies of current semantic similarity algorithm.(3) Introduces IVIFSs, inter-valued intuitionistic fuzzy numbers, inter-valued intuitionisticfuzzy relations, the calculation of the semantic pass package as well as HowNet, which laid thefoundation for calculation based on IVIFSs.(4) Proposes an algorithms thinking about semantic similarity based on IVIFSs.Comprehensive consideration of the hierarchy, depth, density builds a fuzzy matrix to revealsemantic relations of hierarchy. And combination with units which is the smallest unit inHowNet, words are separated into original units. The semantic similarity of words is integratedby semantic similarity of units. (5) Verify and analysis semantic similarity algorithms. Compare results with Liu Qun’s testdata and the subjective experience in order to verify the rationality and effectiveness of theproposed algorithm in the article.The paper launches a research about semantic similarity based IVIFSs, proposescorresponding algorithm and validates its effectiveness. To improve aspects of semanticsimilarity and promote the algorithm to sentence similarity and document similarity is the goal ofour future.