Recent advances in deep neural network architectures have enabled tremendous success on a number of difficult machine learning problems. While these results are impressive, producing a deployable neural network–based conversation model that can engage in open domain discussion still remains elusive. A dialogue system needs to be able to generate meaningful and diverse responses that are simultaneously coherent with the input utterance and the overall dialogue topic. Unfortunately, earlier conversation models trained with naturalistic dialogue data suffered greatly from limited contextual information, and lack diversity. These problems often lead to generic and safe utterance in response to varieties of input utterance.

In this talk, we will explore some novel neural adversarial learning frameworks that mimic more of human behavior allowing us to generate more human-like dialogue responses than the previous work.