Introduction
I am interested in how individuals interact as both speakers and addressees when engaged in a cooperative task (task-oriented dialogues). Important features of language emerge only when speakers have joint intentions and share a common visual ground. Not only can we study how listeners recover linguistic representations, but also to what degree a speaker and a listener align such representations using dialogue. In this context of language as action, I analyze utterances beyond the sentence boundary in naturally occurring dialogues both from an experimental and computational perspective, to investigate how speakers communicate using language and contextual knowledge mediated by cognitive aspects. I address this question from both a comprehension and production perspective. On the comprehension side, I ask: How do speakers interpret non-sentential utterances into larger semantic units? What interfaces between syntax, semantics and pragmatics exist that allow interpretation to happen incrementally? How can we model this process computationally? On the production side, I ask: How do speakers plan their utterances incrementally? What factors are speakers sensitive to when going from meaning to speech plan?
My work stems from my multidisciplinary background in Computer Science, Linguistics, and Brain and Cognitive Science. Consequently, I use several methodologies to explore issues in task-oriented dialogues such as eye tracking on visual world experiments, corpus analysis, user-system usability surveys, and computational and mathematical models. These methodologies
can be broadly classified under computational psycholinguistics. This emergent area of research contributes to the philosophy of language by providing tools to deal with the complexity of online understanding, production, and acquisition of language. Furthermore, computational psycholinguistics provides both symbolic and probabilitistic models that help our understanding of linguistic representations and processing. In this context, my Ph.D. dissertation focuses on developing a semantically annotated multimodal dialogue corpus, developing computational models for incremental language interpretation in dialogue systems, and developing mathematical models for incremental language production.