Accepted Papers

 

  • Recognition of New Sentiment Words Based on Relative Contextual Entropy
    Dongyu Zhang, Hongfei Lin, Peng Fei and Liang Yang, Dalian University of Technology, China
    ABSTRACT
    New sentiment words are the sentiment words that are not included in existing sentiment lexicons and emerge on Weibo every day. As the uncertainty of polarity of many new words will have a detrimental effect on the sentiment analysis of Weibo, how to extract new sentiment words automatically is important for the sentiment analysis of Weibo and the expansion of sentiment lexicons. Considering that the importance of determining a word’s boundary, we propose a method, based on relative contextual entropy, that focuses on the context and combines with adjacent information to recognize new sentiment words. Then, we design three experiments on the COAE2014 Weibo corpus, not only to compare different statistical methods with the proposed method but also to discuss the different methods of determining the polarity of new sentiment words and the effect of new sentiment words on the sentiment classification of Weibo. Experimental results show that the proposed method has high accuracy and effectiveness. Meanwhile, the extraction of new sentiment words is verified to achieve the promoted effect on the performance of Weibo sentiment classification.

 

                                                                                                                                                                Copyright ® ICAIT 2017                                                                                                      Designed By NnN