Generating High-Quality and Informative Conversation Responses with Sequence-to-Sequence Models

Yuanlong Shao, Stephan Gouws, Denny Britz, Anna Goldie, Brian Strope, Ray Kurzweil


Abstract
Sequence-to-sequence models have been applied to the conversation response generation problem where the source sequence is the conversation history and the target sequence is the response. Unlike translation, conversation responding is inherently creative. The generation of long, informative, coherent, and diverse responses remains a hard task. In this work, we focus on the single turn setting. We add self-attention to the decoder to maintain coherence in longer responses, and we propose a practical approach, called the glimpse-model, for scaling to large datasets. We introduce a stochastic beam-search algorithm with segment-by-segment reranking which lets us inject diversity earlier in the generation process. We trained on a combined data set of over 2.3B conversation messages mined from the web. In human evaluation studies, our method produces longer responses overall, with a higher proportion rated as acceptable and excellent as length increases, compared to baseline sequence-to-sequence models with explicit length-promotion. A back-off strategy produces better responses overall, in the full spectrum of lengths.
Anthology ID:
D17-1235
Volume:
Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing
Month:
September
Year:
2017
Address:
Copenhagen, Denmark
Venue:
EMNLP
SIG:
SIGDAT
Publisher:
Association for Computational Linguistics
Note:
Pages:
2210–2219
Language:
URL:
https://www.aclweb.org/anthology/D17-1235
DOI:
10.18653/v1/D17-1235
Bib Export formats:
BibTeX MODS XML EndNote
PDF:
http://aclanthology.lst.uni-saarland.de/D17-1235.pdf
Attachment:
 D17-1235.Attachment.pdf