GubaLex: Guba-Oriented Sentiment Lexicon for Big Texts in Finance

Trading in stock market depends mostly on investor's emotions though technical analysis is a viable tool there. In China, Guba is a typical platform for individual investors to share news and opinions on their favorite stocks. The texts posted in Guba by investors involve in richful emotions which can reflect their willingness on the stock. Few works focus on Guba sentiment analysis though numerous have been done on investor sentiment analysis in finance market for the purpose of understanding the market. Text mining is the most popular method to analyze the sentiment implied in the web text, which depends heavily on the lexicon. Existed lexicons for general purpose work badly on sentiment analysis for Guba messages. In this work, we construct a specified lexicon for Chinese Guba, named GubaLex, in considerations of the characteristics of the Guba text: short, emotion enriched, colloquial (informal), and stock market oriented. It is constructed by using the merge of HowNet and NTUSD as the basic sentiment lexicon, then adding stock terms from the Guba corpus and information in the area of stock market. Based on GubaLex, we develop the bullish lexicon GLBull and the bearish lexicon GL-Bear especially including bullish and bearish sentiment terms for further sentiment analysis. We also proposed an auto update module and sentiment classification algorithm for Guba texts. The experiments show the proposed lexicon works better in sentiment analysis than the previous, like HowNet and NTUSD.