While the volume of information available to users is growing, the interaction capabilities of systems remain the same. Our research aims at designing personalized interaction.
ECA are proven to be intuitive and interactive interfaces, with promising results in several applications such as assistance, e-learning, remediation and so on. Unfortunately, dialogical capabilities of ECA are limited so far. We therefore propose computational models to recognize emotion and to manage dialogue for ECA.
Context is a key factor to personalized interaction. Interaction should be adapted to humans (by opposition to machines), to categories of humans, to individuals and to activity. To recognize such contexts, usual situations should be extracted from traces of activity or dialogue as recurrent patterns of behaviour.
Internet represents a perfect example of large amount of data, where information is difficult to find and analyse. We propose methods and tools to assist users in assessing and retrieve information or sources of information on Internet or Social Media such as Twitter. A link with ECA can be made as they are efficient assistants to help users while searching information.