ISBN: 978-981-11-3671-9 DOI: 10.18178/wcse.2017.06.070
A Mixed Approach of Expanding Sentiment Lexicon Based on word2vec and Phrase Dependency Tree
Abstract— As one of the basic and crucial tasks in sentiment analysis, sentiment lexicon generation has
great significance. Due to diversities of sentiment words, automatic expansion of sentiment lexicon has
become a challenging research topic in recent years. This paper proposes a new approach to expand
sentiment lexicon. At first, the approach selects some sentiment seed words, then uses word2vec to train
word embeddings and find out the words which have potential similarities with seed words to generate a
candidate sentiment lexicon, and then finds out the words which have conjunctive relations with seed words
based on the phrase dependency tree to generate another candidate sentiment lexicon; Finally, we take the
words appearing in the two candidate sentiment lexicons as final expanded sentiment lexicon. Experimental
results demonstrate the effectiveness of our approach and the advantages against state of-the-art baselines.
Index Terms— sentiment analysis, sentiment lexicon, word2vec, phrase dependency tree
Junxia Wang, Pu Zhang, Yinghao Wang
Department of Computer Science and Technology, Chongqing University of Posts and Telecommunications, CHINA
Cite: Junxia Wang, Pu Zhang, Yinghao Wang, "A Mixed Approach of Expanding Sentiment Lexicon Based on word2vec and Phrase Dependency Tree," Proceedings of 2017 the 7th International Workshop on Computer Science and Engineering, pp. 410-416, Beijing, 25-27 June, 2017.