Predicting sentence-level polarity labels of financial news using abnormal stock returns
作者：
Highlights：
• We predict polarity labels of sentences in financial news based on stock returns.
• Multi-instance learning to transfer information from documents to sentences.
• Approach outperforms benchmark methods by 5.10% in terms of predictive accuracy.
• Method assists investors and helps companies communicating their messages as intended.
摘要
•We predict polarity labels of sentences in financial news based on stock returns.•Multi-instance learning to transfer information from documents to sentences.•Approach outperforms benchmark methods by 5.10% in terms of predictive accuracy.•Method assists investors and helps companies communicating their messages as intended.
论文关键词：Financial news,Expert systems,Natural language processing,Multi-instance learning,Decision-making
论文评审过程：Received 23 August 2019, Revised 17 November 2019, Accepted 18 January 2020, Available online 25 January 2020, Version of Record 5 February 2020.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.113223
