My name is Pierre Lison and I am currently working as a Postdoctoral Research Fellow in the Language Technology Group of the Department of Informatics, University of Oslo. My research is funded by a 3-years research grant from the Norwegian Research Council.
My main research interest is dialogue modelling and its applications to various language technology tasks. For my PhD, I developed a hybrid approach to (spoken) dialogue management which combined expert domain knowledge and statistical models into a unified framework. One offspring of my work was the release of the OpenDial toolkit to build robust and adaptive dialogue systems.
I am now focusing on the use of dialogue modelling for another application domain, namely machine translation. In spite of great progress in the recent years, machine translation remains poor at adapting translations to the relevant context. To translate a given dialogue (for instance, film subtitles), current systems generally operate one utterance at a time and ignore the global coherence and structure of the conversation. The goal of this research is to enhance the contextual awareness of translation systems by incorporating features from the dialogue context and exploiting these features to dynamically modulate the translation outputs.
My article “A hybrid approach to dialogue management based on probabilistic rules” will soon appear in Computer Speech & Language (see the “Publications” section). The article is essentially a summary of my PhD work and describes the formalisation of probabilistic rules, the statistical estimation of unknown parameters, and the empirical evaluation of the framework in a human-robot interaction domain.