Dependency-based Mixture Language Models

Zhixian Yang, Xiaojun Wan


Abstract
Various models have been proposed to incorporate knowledge of syntactic structures into neural language models. However, previous works have relied heavily on elaborate components for a specific language model, usually recurrent neural network (RNN), which makes themselves unwieldy in practice to fit into other neural language models, such as Transformer and GPT-2. In this paper, we introduce the Dependency-based Mixture Language Models. In detail, we first train neural language models with a novel dependency modeling objective to learn the probability distribution of future dependent tokens given context. We then formulate the next-token probability by mixing the previous dependency modeling probability distributions with self-attention. Extensive experiments and human evaluations show that our method can be easily and effectively applied to different neural language models while improving neural text generation on various tasks.
Anthology ID:
2022.acl-long.535
Volume:
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
May
Year:
2022
Address:
Dublin, Ireland
Editors:
Smaranda Muresan, Preslav Nakov, Aline Villavicencio
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7758–7773
Language:
URL:
https://aclanthology.org/2022.acl-long.535
DOI:
10.18653/v1/2022.acl-long.535
Bibkey:
Cite (ACL):
Zhixian Yang and Xiaojun Wan. 2022. Dependency-based Mixture Language Models. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7758–7773, Dublin, Ireland. Association for Computational Linguistics.
Cite (Informal):
Dependency-based Mixture Language Models (Yang & Wan, ACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.acl-long.535.pdf
Software:
 2022.acl-long.535.software.zip
Video:
 https://aclanthology.org/2022.acl-long.535.mp4
Code
 fadedcosine/dependency-guided-neural-text-generation
Data
Penn TreebankROCStories