Abstract

This talk describes two ongoing projects in our attempt to apply natural language processing to social science tasks like history and social psychology. We first describe a system for detecting human social intentions from spoken conversation: whether a speaker is awkward, friendly, or flirtatious. We create and use a new spoken corpus of 991 4-minute speed-dates. Using rich dialogue, lexical, disfluency, and prosodic features, we are able to detect flirtatious, awkward, and friendly styles in noisy natural conversational data with above 70% accuracy, significantly outperforming the human interlocutors. We suggest that humans are very poor perceivers of flirtatiousness or friendliness in others, instead often projecting their own intended behavior onto their interlocutors. (Joint work with Dan McFarland, Education, and Rajesh Ranganath, Computer Science). We then describe our results on detecting how ideas spread in the history of science, by studying what factors lead scholars to cite one another and by tracing the spread of scientific topics from subfield to subfield in research paper anthologies. (Joint work with Steven Bethard, David Hall, and Chris Manning).