Generating Rational Commonsense Knowledge-Aware Dialogue Responses With Channel-Aware Knowledge Fusing Network

Abstract

Dialogues systems endow machines with the ability to converse with humans using natural language. Nonetheless, previous Seq2Seq-based generative dialogue systems often generate safe but meaningless responses, such as &#x2018;I don&#x0027;t know&#x0027; or &#x2018;I think so&#x0027;. To this end, researchers proposed to infuse external knowledge into dialogue generation, and such knowledge-enhanced methods have achieved remarkable improvements in the open-domain dialogue systems. External knowledge is an exogenous input, where the estrangement inevitably exists between knowledge and dialogue context. Although previous knowledge-enhanced works can already use commonsense knowledge to generate informative responses, they always use knowledge in a single-channel paradigm, which is hard to accurately handle different data-flows and then tends to generate irrational dialogue responses. Thus, they tend to be confused and generate strange responses when infusing the knowledge into dialogue generation, such as &#x2018;I just ate a basketball&#x2019;, dramatically degrading the user experience. To address this problem, this paper proposes a novel <i>Channel-Aware Knowledge Fusing</i> Network (CAKF). Rather than following the traditional single-channel paradigm, CAKF employs three unique channels to handle different data-flows more clearly and rationally: a <i>base</i> channel serves like a vanilla Seq2Seq decoder; a <i>context</i> channel to utilize the contextual information, and a <i>knowledge</i> channel to infuse commonsense knowledge into the dialogue generation. Above such three channels, a <i>Sequential Manager</i> is built to maintain the global sequential decision state, aggregate the local data-flows, and make the final prediction. Experiments on two open-released datasets (a Chinese Weibo and an English Reddit) demonstrated the superior performance of this work against various state-of-the-art approaches.