Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2022
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2205.13346 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866910868143341568 |
|---|---|
| author | Li, Mingzhe Lin, XieXiong Chen, Xiuying Chang, Jinxiong Zhang, Qishen Wang, Feng Wang, Taifeng Liu, Zhongyi Chu, Wei Zhao, Dongyan Yan, Rui |
| author_facet | Li, Mingzhe Lin, XieXiong Chen, Xiuying Chang, Jinxiong Zhang, Qishen Wang, Feng Wang, Taifeng Liu, Zhongyi Chu, Wei Zhao, Dongyan Yan, Rui |
| contents | Contrastive learning has achieved impressive success in generation tasks to militate the "exposure bias" problem and discriminatively exploit the different quality of references. Existing works mostly focus on contrastive learning on the instance-level without discriminating the contribution of each word, while keywords are the gist of the text and dominant the constrained mapping relationships. Hence, in this work, we propose a hierarchical contrastive learning mechanism, which can unify hybrid granularities semantic meaning in the input text. Concretely, we first propose a keyword graph via contrastive correlations of positive-negative pairs to iteratively polish the keyword representations. Then, we construct intra-contrasts within instance-level and keyword-level, where we assume words are sampled nodes from a sentence distribution. Finally, to bridge the gap between independent contrast levels and tackle the common contrast vanishing problem, we propose an inter-contrast mechanism that measures the discrepancy between contrastive keyword nodes respectively to the instance distribution. Experiments demonstrate that our model outperforms competitive baselines on paraphrasing, dialogue generation, and storytelling tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2205_13346 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Keywords and Instances: A Hierarchical Contrastive Learning Framework Unifying Hybrid Granularities for Text Generation Li, Mingzhe Lin, XieXiong Chen, Xiuying Chang, Jinxiong Zhang, Qishen Wang, Feng Wang, Taifeng Liu, Zhongyi Chu, Wei Zhao, Dongyan Yan, Rui Computation and Language Contrastive learning has achieved impressive success in generation tasks to militate the "exposure bias" problem and discriminatively exploit the different quality of references. Existing works mostly focus on contrastive learning on the instance-level without discriminating the contribution of each word, while keywords are the gist of the text and dominant the constrained mapping relationships. Hence, in this work, we propose a hierarchical contrastive learning mechanism, which can unify hybrid granularities semantic meaning in the input text. Concretely, we first propose a keyword graph via contrastive correlations of positive-negative pairs to iteratively polish the keyword representations. Then, we construct intra-contrasts within instance-level and keyword-level, where we assume words are sampled nodes from a sentence distribution. Finally, to bridge the gap between independent contrast levels and tackle the common contrast vanishing problem, we propose an inter-contrast mechanism that measures the discrepancy between contrastive keyword nodes respectively to the instance distribution. Experiments demonstrate that our model outperforms competitive baselines on paraphrasing, dialogue generation, and storytelling tasks. |
| title | Keywords and Instances: A Hierarchical Contrastive Learning Framework Unifying Hybrid Granularities for Text Generation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2205.13346 |