A dialogue system implemented with latent parameters

Abstract

When given a conversation, traditional dialogue systems mainly focus on the context that can be observed on the surface of sentences; concretely, they process and determine the output based on the grammar, visible keywords and structure of the sentence. However, the content we convey to others is affected by a multitude of latent parameters, such as emotional state, personal knowledge and personality. Therefore, we have attempted to validate the integrity of a dialogue system which takes these latent parameters into measure, and have sucessfully developed a dialogue system which utilizes latent parameters as input.