뉴스/소셜 미디어 텍스트와 투자자 기대

This paper analyzes whether the investor expectation implied by the text in news articles or social media forums affects stock returns in the Korean market. Our model, trained on 640,457 input articles and forum posts, classifies each post as positive or negative, employing word embedding based on the Word2Vec and bi-directional long short-term memory network to construct the investor expectation indices. We find that the expectation index constructed from news articles and the index from social media forums can explain stock return movements. Interestingly, the investor expectation extracted from social media forums outperforms the expectation from news articles.