A construction and self-learning method for intelligent domain sentiment lexicon. (22nd April 2020)

• Record Type: Journal Article
• Authors: Wu, Shaochun; Xiao, Qifeng; Gao, Ming; Zou, Guobing
• Is Part Of: International journal of information technology and management. Volume 19:Number 4(2020)
• Page Start: 318
• Page End: 333
• Publication Date: 2020-04-22
• Subjects: sentiment lexicon -- SO-PMI algorithm -- seed words -- LDA topic model -- word clustering -- incremental text processing; Management information systems -- Periodicals; Information technology -- Periodicals; Management -- Data processing -- Periodicals; 658.403805
• Languages: English
• ISSNs: 1461-4111

Abstract

A new method of building intelligent sentiment lexicon based on LDA and word clustering is put forward in this paper. In order to make seed words more representative and universal, this method uses LDA topic model to build the term vectors and select seed words. The improved SO-PMI algorithm has been used to calculate the emotional tendency of each sentiment word. In addition, the domain sentiment lexicon's automatic extension and update method is designed to deal with dynamic corpus data. Experiments show that the proposed method can build the sentiment lexicon with higher accuracy, and can reflect the change of words' emotional tendency in real time. It is proved in this paper that this method is more suitable for processing a large number of dynamic Chinese texts.