题目：Textual Sentiment Analysis in Financial Social Media
主讲人：Tao Wang, Professor, Department of Economics, Queens College of City University of New York
汪滔博士，纽约城市大学女王学院全职教授，主要从事投资、金融计量、国际金融等相关领域的研究，并发表了二十余篇高水平国际论文，发表期刊涵盖Journal of Economic Studies，European Financial Management，Journal of Real Estate Research，The British Accounting Review，International Review of Financial Analysis等,并多次荣获奖项。
讲座简介：Over the past decade, with the exponential increase in computing power, textual sentiment analysis has been applied in accounting, economics, and finance. In particular, textual analysis was used in information source such as news articles, earnings conference calls, Securities and Exchange Commission (SEC) filings, social media messages, financial analyst reports, and FOMC statements. The sentiment expressed in these texts conveys information and opinions of market participants and commentators on many aspects of firms, institutions, and markets. However, compared to traditionally used quantitative methods, textual analysis is substantially less precise. We examine the main methods in textual analysis, the dictionary approach and machine learning approach, and compare their relative performance in sentiment/tone classification in the context of financial social media articles. Our findings suggest that in terms of classification accuracy and precision, the supporting vector machine (SVM) method outperforms the commonly used Naive Bayes method and the Loughran and McDonald dictionary (2011) in both balanced and unbalanced samples. Moreover, feature selection in machine learning appears to be important as well. We offer some suggestions in using machine learning approach for textual analysis.