Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916546967764992 |
|---|---|
| author | Zhang, Ding Li, Yangning Bai, Lichen Zhang, Hao Li, Yinghui Lin, Haiye Zheng, Hai-Tao Su, Xin Shan, Zifei |
| author_facet | Zhang, Ding Li, Yangning Bai, Lichen Zhang, Hao Li, Yinghui Lin, Haiye Zheng, Hai-Tao Su, Xin Shan, Zifei |
| contents | Chinese grammatical error correction (CGEC) aims to detect and correct errors in the input Chinese sentences. Recently, Pre-trained Language Models (PLMS) have been employed to improve the performance. However, current approaches ignore that correction difficulty varies across different instances and treat these samples equally, enhancing the challenge of model learning. To address this problem, we propose a multi-granularity Curriculum Learning (CL) framework. Specifically, we first calculate the correction difficulty of these samples and feed them into the model from easy to hard batch by batch. Then Instance-Level CL is employed to help the model optimize in the appropriate direction automatically by regulating the loss function. Extensive experimental results and comprehensive analyses of various datasets prove the effectiveness of our method. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_00334 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction Zhang, Ding Li, Yangning Bai, Lichen Zhang, Hao Li, Yinghui Lin, Haiye Zheng, Hai-Tao Su, Xin Shan, Zifei Computation and Language Artificial Intelligence Chinese grammatical error correction (CGEC) aims to detect and correct errors in the input Chinese sentences. Recently, Pre-trained Language Models (PLMS) have been employed to improve the performance. However, current approaches ignore that correction difficulty varies across different instances and treat these samples equally, enhancing the challenge of model learning. To address this problem, we propose a multi-granularity Curriculum Learning (CL) framework. Specifically, we first calculate the correction difficulty of these samples and feed them into the model from easy to hard batch by batch. Then Instance-Level CL is employed to help the model optimize in the appropriate direction automatically by regulating the loss function. Extensive experimental results and comprehensive analyses of various datasets prove the effectiveness of our method. |
| title | Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2501.00334 |