A glance through Conversational agents
Wednesday May 22 11:30
Lina Maria Rojas Barahona, Orange Labs, Lannion
Conversational agents are once and again gaining great interest from academics and industrials. The availability of big data as well as the advances in processing units have made deep learning approaches feasible and promising, reviving the dream of creating artificial agents that can easily converse to people. Several solutions have been proposed since the first psychoanalyst chatbot (1966). From regular expressions and symbolic approaches (formal grammars and formal logics) to statistical approaches (probabilistic models and data-driven techniques, e.g. machine learning or deep learning). We have already obtained promising results. For instance, we know that machines can learn optimal strategies for simple tasks in small domains (task-oriented dialogue systems). Moreover, we are treating open domain dialogues by asking questions to online encyclopedias such as Wikipedia (conversational reading comprehension) and we are able to predict the best answer in chitchats (end to end neural approaches). In this talk I will review briefly the state of the art and highlight the open research problems on conversational agents.
Contact at the lab: Noël Nguyen