A multi-source entity-level sentiment corpus for the financial domain: the FinLin corpus

Area of Science:

• Natural Language Processing
• Financial Informatics
• Behavioral Finance

Background:

• Existing financial sentiment corpora lack comprehensive coverage of the automobile industry.
• Sentiment analysis in finance requires nuanced datasets for accurate modeling.

Purpose of the Study:

• Introduce FinLin, a novel financial sentiment corpus.
• Facilitate research in financial sentiment analysis and its applications in behavioral science.
• Provide a publicly available resource for the automobile industry.

Main Methods:

• Compilation of diverse financial texts: investor reports, company reports, news articles, and StockTwits microblogs.
• Annotation of text spans with sentiment and relevance scores (-1.0 to 1.0 and 0.0 to 1.0).
• Inclusion of text spans for sentiment to enhance interpretability.

Main Results:

• Creation of FinLin, a unique corpus focused on the automobile industry.
• Detailed sentiment and relevance scores provided for annotated text.
• Availability of text spans offers insights into annotation reasoning.

Conclusions:

• FinLin serves as a valuable, publicly accessible resource for financial sentiment analysis.
• The corpus supports advancements in understanding market sentiment and investor behavior.
• Encourages further research at the intersection of finance, NLP, and behavioral science.