Markov logic networks based emotion classification for Chinese microblogs. (2016)

• Record Type: Journal Article
• Authors: Ruan, Dongru; Ping, Xiaoli; Gao, Kai
• Is Part Of: International journal of intelligent information and database systems. Volume 9:Number 2(2016)
• Page Start: 197
• Page End: 211
• Publication Date: 2016
• Subjects: Markov logic networks -- MLNs -- microblogs -- emotion classification -- statistical relational learning -- SRL -- China -- Web 20 -- Weibo -- emotions; Database management -- Computer programs -- Periodicals; Information retrieval -- Computer programs -- Periodicals; Information storage and retrieval systems -- Computer programs -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Intelligent agents (Computer software) -- Periodicals; 006.33
• Languages: English
• ISSNs: 1751-5858

Abstract

With the rapid development of Web 2.0, more and more people have begun to use Weibo as a platform for giving opinions and expressing their emotions. As the microblog rapidly increasing, the researchers pay more attention to it. In addition, emotion has come to be established as a new direction of classification. This paper classifies Chinese microblogs through this prism of emotion. When analysing a microblog, we classify it under several different states of emotion. However, such microblogs contain complex correlativities. The traditional approaches have been to classify each message independently, ignoring correlations that may exist between them. In order to overcome these problems arised from traditional approaches, this paper adopts a statistical relational learning (SRL) method, Markov logic networks (MLNs), to establish the model for Chinese microblogs and to perform the task of emotion classification. The experimental results on imbalanced datasets reveal that MLNs is effective on emotion classification and slightly better than the performance of the baseline; the experimental results on balanced datasets indicate that MLNs is influenced by dataset and the balanced datasets can effectively improve the performance of MLNs on emotion classification.