Doctoral Consortium on Natural Language Processing for Computational Social Science

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Computational methods such as natural language processing and machine learning have the potential to revolutionize social science. At the same time, as computer technology is increasingly embedded in everyday life, the social science implications are increasingly pressing. Unfortunately, existing research communities and educational pathways provide few opportunities for interdisciplinary exchange of ideas. The doctoral consortium and associated workshop that this proposal supports will fill this gap, by bringing together doctoral students working in natural language processing and those working in relevant areas of social science. Moreover, participating doctoral students will be paired with faculty mentors in disciplines outside of their current areas of study, thus providing for each student a novel perspective that is likely to be both challenging and inspirational. Empirical Methods in Natural Language Processing (EMNLP) is one of the premier annual conferences at the intersection of natural language processing and machine learning; it attracts some of the best computational researchers in this discipline. It has also recently made efforts towards outreach to the social sciences, for example by inviting a prominent social science researcher to give a keynote address in 2015. The doctoral consortium and accompanying workshop on Computational Social Science and Natural Language Processing, which will be held at the 2016 EMNLP conference in Austin, Texas in November 2016, will continue this outreach effort in a new direction. Funding will be used primarily to support students in the social sciences to attend the meeting, and they will be given an opportunity to present their current dissertation progress. In addition, keynote speakers from social science disciplines will be invited to the workshop. The invited speakers can then be paired with computer science doctoral students, and the social science doctoral students paired with EMNLP "regulars" that have expertise in relevant areas of natural language processing. These mentorship pairings are expected to significantly impact the direction of ongoing doctoral research, bringing natural language processing and social science closer together. Of particular interest is sociolinguistics, an empirical branch of linguistics, where the potential for impact from computational methods looks to be particularly significant. The Linguistics Program in the Division of Behavioral and Cognitive Sciences is co-funding this workshop.