The Impact Of Listed Company News On Stock Price

In financial markets,there is a large amount of unstructured data,mainly in the form of text data,which contains a wealth of information not covered by price and volume data.However,to accurately mine such information often requires the use of sophisticated artificial intelligence algorithms and large amounts of labelled data.This paper focuses on news text data in financial markets and,introduces the Transformer framework.Combined with self-supervised learning techniques,we build our own large-scale financial BERT pretrained language model on a Chinese corpus.Through multiple comparison experiments,we attempt to create a preliminary solution and workflow for solving financial problems using large natural language processing models.Ultimately,this study uses this solution and model to analyze 1.7 million news articles in the A-share market.By integrating traditional financial paradigms,we construct new sentiment factors and find that this combination of financial technology yields good empirical results.By dividing and regressing investment portfolios with stocks accompanied by news in the A-share market,we verified the empirical result that financial media sentiment does indeed affect market portfolio returns.We also found that the more negative the media sentiment in an investment portfolio,the greater its expected return.Smaller individual stock portfolios are more affected by financial media sentiment.In addition,we conducted empirical tests separately in the Shanghai and Shenzhen markets and five different industries,finding our results to be quite robust.The sentiment factors of financial news media imply pricing information about the market.