Re-framing Incremental Deep Language Models for Dialogue Processing with Multi-task Learning

We present a multi-task learning framework to enable the training of one\nuniversal incremental dialogue processing model with four tasks of disfluency\ndetection, language modelling, part-of-speech tagging, and utterance\nsegmentation in a simple deep recurrent setting. We show that these tasks\nprovide positive inductive biases to each other with the optimal contribution\nof each one relying on the severity of the noise from the task. Our live\nmulti-task model outperforms similar individual tasks, delivers competitive\nperformance, and is beneficial for future use in conversational agents in\npsychiatric treatment.\n