Enhancing Sentiment Analysis based Investment by Large Language Models in Japanese Stock Market

This paper explores the utilization of text data as a new α return resource for individual stock investment. In particular, we propose a new construction scheme of a polarity dictionary based on ChatGPT, a state-of-the-art language model, which is also directly used for sentiment analysis. Our research focuses on analyzing non-English text data, expanding the application of sentiment analysis by ChatGPT to a broader range of markets and languages. Through extensive analysis of statistical test, we identify the mean reversion feature following the release of negative news in Japanese large-cap company.As a result, a simple investment strategy based on the sentiment analysis is superior to the market benchmark index and well-known traditional strategies such as mean-variance and risk parity portfolios