Chat More: Deepening and Widening the Chatting Topic via A Deep Model

The past decade has witnessed the boom of the human-computer interaction, particularly via dialog systems. In this paper, we study the task of response generation in the open-domain multi-turn dialog systems. It is worth mentioning that many efforts have been dedicated to study the dialog systems, yet few of them shed light on deepening or widening the chatting topic in a conversational session, which is the key to retain users to talk more. To achieve this goal, in this paper, we present a novel deep scheme comprising of three channels, namely global, wide and deep ones. The global channel encodes the complete historical information within the given context, the wide one employs attention-based recurrent neural network to predict extensional keywords that may not appear in the historical context, and the deep one trains a Multilayer Perceptron model to select some keywords for an in-depth discussion. Thereafter, our scheme integrates the outputs of these three channels to generate the desired responses. To justify our model, we conducted extensive experiments to compare our model with several state-of-the-art baselines on two datasets: one is constructed by ourselves and the other is a well-known public benchmark dataset. Experimental results demonstrate that our model yields promising performance by widening or deepening the topic of interest.