Nicole Novielli, Ph.D.

I got my Ph.D. in Computer Science in May 2010, from the University of Bari (advisors dr. Carlo Strapparava and prof. Vito Leonardo Plantamura), defending a thesis on 'Lexical Semantics of Dialogue Acts'.

The thesis project started in January 2007 as a follow-up of the research experience in the Humaine EU project (http://emotion-research.net/) under the supervision of prof. Fiorella de Rosis and has been conducted in cooperation with the Human-Language Technologies division of the Fondazione Bruno Kessler (FBK-irst) in Trento.

The work developed in the Ph.D. thesis falls in the field of Human Computer Interaction, Natural Language Processing and Affective Computing; it aims at exploiting techniques for modeling both the interaction and the cognitive and affective components of the state of mind of participants to a dialogue. The major contribution of the work is the definition of domain independent linguistic profiles that characterize dialogue acts by exploiting the lexical semantics of utterances. These profiles are then employed on an empirical study regarding the task of automatic labeling dialogues with the proper speech acts. An application of Dialogue Act annotation to the task of detecting the user's attitudes is also presented. In particular, I propose an approach based on Hidden Markov Models to describe the differences in the dialogue pattern due to the different level of engagement experienced by the users.

Research         Background         Publications         CV         Contact

My research interests are in the field of recognition of affective and cognitive states in human-computer interaction. In particular, my attention is focused on  'natural' dialogs with Embodied Conversational Agents. The long-term goal is to create an ECA that plays the role of an artificial therapist in health counseling dialogs. The agent will have to inform, persuade and engage the user during the conversation: to achieve this goal, it should be able to recognize the mental state of the user (cognitive and affective aspects) and adapt its behavior, interaction style and arguments accordingly. My specialty is in: i) defining markup languages for corpora annotation, ii) combining language analysis methods with Bayesian classification and iii) integrating recognition methods with cognitive modeling, based on dynamic Bayesian networks.
In the scope of my cooperation with the Intelligent Interfaces research group in Bari, I also work on Ambient Intelligence and Ambient Assisted Living, Recommender Systems with particular focus on the tourism domain and E-Learning.

At present, I am working on the application of markovian models to dialog pattern description and on unsupervised recognition of Dialogue Acts by exploiting their lexical semantics using empirical methods (i.e. Latent Semantic Analysis).

Home