Qualitative Stock Market Predicting with Common Knowledge Based Nature Language Processing: A Unified View and Procedure

Abstract

There are many artificial intelligent applications. Some of them focus on the financial market. They often use a nature language processing method, e.g., To predict stock prices. However, most of them are inaccurate. There are two reasons. For one thing, computer programs are more effective in the syntax analysis than semantic analysis. For another, accurately predicting stock prices is beyond our knowledge and ability today. However, there are many valuable experiences in existing studies. Therefore, we propose a unified view and procedure to facilitate using these experiences. This procedure is based on the common knowledge, which is primarily expressed as keywords in this paper. It first recognizes name entities and then learns rules with the common knowledge and last inferences crucial features. These features, with other quantitative features in the stock market, may make the prediction more accurate. As a result, this view and process can be a framework for many (but not all) nature language processing applications in stock predicting.