%0 Conference Proceedings
%T Towards Less Generic Responses in Neural Conversation Models: A Statistical Re-weighting Method
%A Liu, Yahui
%A Bi, Wei
%A Gao, Jun
%A Liu, Xiaojiang
%A Yao, Jian
%A Shi, Shuming
%Y Riloff, Ellen
%Y Chiang, David
%Y Hockenmaier, Julia
%Y Tsujii, Jun’ichi
%S Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing
%D 2018
%8 oct nov
%I Association for Computational Linguistics
%C Brussels, Belgium
%F liu-etal-2018-towards-less
%X Sequence-to-sequence neural generation models have achieved promising performance on short text conversation tasks. However, they tend to generate generic/dull responses, leading to unsatisfying dialogue experience. We observe that in the conversation tasks, each query could have multiple responses, which forms a 1-to-n or m-to-n relationship in the view of the total corpus. The objective function used in standard sequence-to-sequence models will be dominated by loss terms with generic patterns. Inspired by this observation, we introduce a statistical re-weighting method that assigns different weights for the multiple responses of the same query, and trains the common neural generation model with the weights. Experimental results on a large Chinese dialogue corpus show that our method improves the acceptance rate of generated responses compared with several baseline models and significantly reduces the number of generated generic responses.
%R 10.18653/v1/D18-1297
%U https://aclanthology.org/D18-1297
%U https://doi.org/10.18653/v1/D18-1297
%P 2769-2774