CAREER: Towards a Data-driven Understanding of Online Sentiment

点击翻译按钮获取中文摘要

This project promotes the progress and advancement of scientific knowledge by exploring newmethods for understanding and modeling online sentiment. As the Web has become critical tosociety, impacting our personal lives as well as the world at large, a greater need to understandhow people interact. In particular, research into how to measure people’s opinion on varioustopics or types of content has failed to address the new ways and types of content used on theevolving social Web. This project addresses this lack of understanding by developing new toolsto help quantify online sentiment using large-scale, empirical data. In addition to these newtools, output from the project, including datasets and code artifacts will help other researchersadvance our understanding of social media.In this project, the investigator seeks to achieve four research objectives. The first is thecreation of a multi-platform social media dataset. In particular, the investigator will develop aseries of tools leveraging prior experience in large-scale data collection to perform continuousidentification and collection of multimedia data from emerging social media platforms. Next, theinvestigator will develop data-driven techniques to understand coded language used in socialmedia. These techniques will focus not just on textual content, e.g., slang words, but also codedvisual language, e.g., memes. The third objective is developing a new, explainable system forrating the sentiment of content. This method departs from existing approaches by treating theclassification task as a game between two pieces of content and learning a total ordering of thecontent devised from the Elo rating system used in chess and video games. Finally, theinvestigator will explore user and community level modeling of online sentiment.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.