A Robust Dialogue System with Spontaneous Speech and Touch Screen

摘要

In this paper, we describe a robust dialogue system which is intended to understand spontaneous speech and also accepts the input by touch screen. It is difficult to recognize spontaneous speech, on which misrecognition occurs frequently, because spontaneous speech has many ambiguous phenomena such as interjections, ellipses, inversions, repairs, unknown words and so on. Also, a recognizer may output the sentences as recognition results that human being never speaks. Therefore, the interpretation part that receives recognized sentences must cope not only with spontaneous sentences but also with illegal sentences having recognition errors. We developed the robust interpretation method that uses some heuristics for understanding spontaneous or misrecognized sentences. The input by touch screen helps the disadvantage of the input by spontaneous speech. We report the evaluation results of our dialogue system.