CAREER: Towards a Data-driven Understanding of Online Sentiment

This project promotes the progress and advancement of scientific knowledge by exploring new methods for understanding and modeling online sentiment. As the Web has become critical to society, impacting our personal lives as well as the world at large, a greater need to understand how people interact. In particular, research into how to measure people’s opinion on various

topics or types of content has failed to address the new ways and types of content used on the evolving social Web. This project addresses this lack of understanding by developing new tools to help quantify online sentiment using large-scale, empirical data. In addition to these new tools, output from the project, including datasets and code artifacts will help other researchers

advance our understanding of social media. In this project, the investigator seeks to achieve four research objectives. The first is the creation of a multi-platform social media dataset. In particular, the investigator will develop a series of tools leveraging prior experience in large-scale data collection to perform continuous

identification and collection of multimedia data from emerging social media platforms. Next, the investigator will develop data-driven techniques to understand coded language used in social media. These techniques will focus not just on textual content, e.g., slang words, but also coded visual language, e.g., memes. The third objective is developing a new, explainable system for

rating the sentiment of content. This method departs from existing approaches by treating the classification task as a game between two pieces of content and learning a total ordering of the content devised from the Elo rating system used in chess and video games. Finally, the investigator 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.