In this talk, the speaker will describe how conversational agency models defined during the last few decades have been gradually integrated within industrial "chatbot" solutions. Indeed, while first-generation chatbots (e.g. ELIZA in 1968) were designed with the sole purpose of maintaining a conversation and therefore simply reacted to patterns in the user’s utterances, today's virtual assistants are task-oriented and cooperate with users to achieve an information-seeking or transaction goal. Their natural language understanding is achieved through machine learning models that recognize entities in context or analyze utterance syntax. The talk will focus on customer care and related use cases in several Swiss market domains to explain how natural language understanding and dialogue management are approached in industrial solutions. The speaker will also point out some challenges that remain open for this technology.


Time: 10:45 - 11:15

Track 4
organised by SGAICO

Speaker: Silvia Quarteroni