Building a financial sentiment bank on the investment behavior of foreign investors

Abstract

This study focuses on building a Financial Sentiment PhraseBank on Foreign Investors’ Trading Behavior in the Vietnamese Stock Market. 4,065 financial news articles containing information about eight typical trading behaviors, including “buying, selling, net buying, net selling, accumulating, offloading, and locking in profits,” were analyzed. The study successfully extracted a database comprising 7,130 key sentences from these articles and assigned sentiment labels to the content. The final product, the Financial Sentiment PhraseBank on Foreign Investors' Trading Behavior, compiles 5,126 sentences with corresponding sentiment labels. This PhraseBank holds significant value for advancing future research, particularly as a resource for developing models and tools for financial news sentiment analysis based on the BERT architecture, similar to the FinBERT tool. Furthermore, the study provides empirical evidence that applying Sentence Embedding techniques at the sentence level, instead of traditional Word Embedding methods for individual words, greatly enhances analytical efficiency and opens new directions for deeper exploration in the future. This research makes a substantial contribution to the body of knowledge on foreign investors' trading behavior and offers valuable tools and datasets to support both scientific research and practical investment activities.