Abstract
Neural models of dialog rely on generalized latent representations of language. This paper introduces a novel training procedure which explicitly learns multiple representations of language at several levels of granularity. The multi-granularity training algorithm modifies the mechanism by which negative candidate responses are sampled in order to control the granularity of learned latent representations. Strong performance gains are observed on the next utterance retrieval task using both the MultiWOZ dataset and the Ubuntu dialog corpus. Analysis significantly demonstrates that multiple granularities of representation are being learned, and that multi-granularity training facilitates better transfer to downstream tasks.- Anthology ID:
- D19-1184
- Volume:
- Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)
- Month:
- November
- Year:
- 2019
- Address:
- Hong Kong, China
- Editors:
- Kentaro Inui, Jing Jiang, Vincent Ng, Xiaojun Wan
- Venues:
- EMNLP | IJCNLP
- SIG:
- SIGDAT
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 1752–1761
- Language:
- URL:
- https://aclanthology.org/D19-1184
- DOI:
- 10.18653/v1/D19-1184
- Cite (ACL):
- Shikib Mehri and Maxine Eskenazi. 2019. Multi-Granularity Representations of Dialog. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 1752–1761, Hong Kong, China. Association for Computational Linguistics.
- Cite (Informal):
- Multi-Granularity Representations of Dialog (Mehri & Eskenazi, EMNLP-IJCNLP 2019)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-2/D19-1184.pdf
- Data
- MultiWOZ, UDC