Estimation in Semantic Similarity of Texts

• Published in: Journal of Information Science and Engineering Vol. 37; no. 3; pp. 617 - 633
• Main Authors: Nguyen, Manh Hung, Tran, Dinh Que
• Format: Journal Article
• Language: English
• Published: Taipei 社團法人中華民國計算語言學學會 01.05.2021; Institute of Information Science, Academia Sinica
• Subjects: Artificial intelligence; Natural language processing; Semantic web; Semantics; Sentences; Similarity; Texts; semantic computing; sentence similarity; text mining; word similarity; text similarity
• ISSN: 1016-2364
• DOI: 10.6688/JISE.202105_37(3).0008

The semantic similarity of texts or documents has been widely studied in various areas including natural language processing, document comparison, artificial intelligence, semantic web, etc: Several similarity measures have been proposed but they are usually tied to special application domains or to data representation of various types. The purpose of this paper is to present a model for estimation in semantic similarity of texts based on similar sentences in structure of subjects, verbs and objects. And in turn, the semantic similarity of these components in the structure of sentences is estimated by means of the basic semantic similarity of words. The model is evaluated with two experiments: direct similarity and relative similarity among texts. The experimental results indicate that the proposed model is better than some baseline models in some circumstances.