Consistent Response Generation with Controlled Specificity

We propose a method to control the specificity of responses while maintaining the consistency with the utterances for open-domain conversation systems. We first design a metric based on pointwise mutual information, which measures the co-occurrence degree between an utterance and a response. To control the specificity of the generated responses, we add the distant supervision based on the cooccurrence degree and a PMI-based word prediction mechanism to a sequence-to-sequence model. Using these mechanisms, our model outputs the words with desired specificity for a given specificity level. In experiments with open-domain dialogue corpora, automatic and human evaluation results confirm that our model controls the specificity of the responses more sensitively than the conventional model and can generate highly consistent responses.