Author(s)

Cai Chongchao, Xu Huahu, Wan Jie, Wu Jiaqi

Corresponding Author

Cai Chongchao

ABSTRACT

Sentiment analysis mainly refers to the mining and extraction of useful knowledge from subjective texts. Most of existing analysis methods involve either text classification through machine learning or polarity classification through sentiment lexicon. The limited scale of Chinese sentiment lexicon impedes the performance of sentiment analysis for social network. In this context, a domain-specific method is proposed to extract large-scale sentiment lexicon. To achieve this, social network data are extracted and cleaned. Next, the deep representation model Word2Vec is used to learn the extracted data in order to obtain the Chinese lexicon vector and determine the seed words of the stock domain. Experimental results show that in terms of large social network sentiment analysis, the proposed method for acquisition of domain-specific sentimental lexicon is superior to other methods for popular sentiment lexicons and small-scale domain-specific lexicons.

KEYWORDS

Sentiment Lexicon, Deep Learning, Word2Vec, Sentiment analysis