Discourse Emotion Analysis

I worked on the Pedagogical Discourse Project at the USC Information Sciences Institute in an internship capacity for several years, first as an unpaid undergraduate intern, and after my graduation as a paid intern. The purpose of the project is to develop methods for scaffolding and analyzing student online discussion boards in order to facilitate learning through the use of NLP and other techniques.

I came into the project in my capacity as a Cognitive Science student, as most of the team was trained and focused on the project from Computer Science and Linguistic perspectives. One of the goals of the project was to attempt to analyze discussion board content for emotional material, both for scholarly analysis and to assist instructors in monitoring their discussion boards for potential issues that needed their attention. Using my experiences in research, I performed several literature reviews and test analyses and developed several tagging criteria and methodologies that could identify certain important types of emotional content in a context-independent fashion. I personally tagged thousands of posts for NLP learning, and performed several types of preliminary analyses on these learning sets to attempt to identify trends. Using these results and results from the rest of my team, I copywrote a short paper for the Workshop on Computational Models for Natural Argument, , Int’l Joint Conference on AI (IJCAI), 2009, which ultra slim electronic cigarette was accepted and I presented. Furthermore, our team submitted, was accepted, and presented a poster and short paper at the Annual Meeting of the Cognitive Science Society, 2010.

I learned an enormous amount from this project, and not all of it simply technical. It was the first time I had worked with an academic team, and I was given an enormous amount of leeway in developing my own approach and implementation. It was very frightening to go out on a limb for the project, but it was wildly rewarding to see my work take shape and be recognized. I am incredibly grateful to my principles in the project, who were consistently supportive and instructive, and I credit them with the success I was able to recognize in the project.

Links: