Information Extraction from Online Text --- from Opinions to Arguments to Persuasion
A long line of research in Natural Language Processing (NLP),including our own, has addressed the task of identifying andextracting information about opinions with the goal of determiningwhat people (and other entities) are thinking or feeling. In thistalk, I'll present new research on argument mining, a relatively newarea of study in NLP that focuses less on extracting from text WHATpeople think or feel, but rather analyzing argumentative text tounderstand WHY they do so.Specifically, I will first present some of our new research on theautomatic analysis of informal, user-generated arguments in which weaim to expose the intended underlying structure of the argument.Next, we will examine the arguments on a public debate forum todetermine what makes one argument more convincing than another. Inparticular, recent studies in NLP have provided empirical evidencethat the language of the debaters and their patterns of interactionplay a key role in changing the mind of a reader/listener. On theother hand, research in psychology has shown that prior beliefs canaffect our interpretation of an argument. Using debate forum data, Iwill present the results of our work to determine the relative rolesof language use vs. prior beliefs in the creation of persuasiveargumentative text.