Natural Language Dialogues with Sequence-to-Sequence Learning
Dirk von Grünigen, Jan Deriu, Mark Cieliebak (ZHAW)
Chatbots have recently become a focus of broad interest, since they offer entertainment and business value at the same time. Recent advances in deep learning allow to generate chatbots (or dialogue systems) end-to-end, i.e. to train a model that generates genuine replies in a dialogue.
We examine different technological approaches to dialogue systems, and report our experiences in training a conversational agent on movie dialogues with sequence-to-sequence learning.