Curriculum-Style Fine-Grained Adaption for Unsupervised Cross-Lingual Dependency Transfer

Guo, Peiming; Huang, Shen; Jiang, Peijie; Sun, Yueheng; Zhang, Meishan; Zhang, Min

doi:10.1109/taslp.2022.3224302

Cited by 4 publications

(1 citation statement)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, such direct transfer approaches are insufficient for target domains with large gaps compared to the source domain. Self-training (Yarowsky, 1995;McClosky et al, 2006;Yu et al, 2015;Ramponi and Plank, 2020;Guo et al, 2022) is a simple and early bootstrapping approach for domain adaptation, applied to tasks such as classification Dong and Schäfer (2011); Ye et al (2020), sequence labeling Wang et al (2021Wang et al ( , 2020, dependency parsing Yu et al (2015); Rotman and Reichart (2019); Guo et al (2022), sequence generation He et al (2019), and QA Sachan and Xing (2018) and so on. However, self-training has not been applied to cross-domain constituency parsing.…”

Section: Related Workmentioning

confidence: 99%

LLM-enhanced Self-training for Cross-domain Constituency Parsing

Li,

Zhang,

Guo

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

Self-training has proven to be an effective approach for cross-domain tasks, and in this study, we explore its application to cross-domain constituency parsing. Traditional self-training methods rely on limited and potentially lowquality raw corpora. To overcome this limitation, we propose enhancing self-training with the large language model (LLM) to generate domain-specific raw corpora iteratively. For the constituency parsing, we introduce grammar rules that guide the LLM in generating raw corpora and establish criteria for selecting pseudo instances. Our experimental results demonstrate that self-training for constituency parsing, equipped with an LLM, outperforms traditional methods regardless of the LLM's performance. Moreover, the combination of grammar rules and confidence criteria for pseudo-data selection yields the highest performance in the crossdomain constituency parsing 1 .

show abstract

Section: Related Workmentioning

confidence: 99%

LLM-enhanced Self-training for Cross-domain Constituency Parsing

Li,

Zhang,

Guo

et al. 2023

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

View full text Add to dashboard Cite

show abstract

Filtered data augmentation approach based on model competence evaluation

Yu,

Zhang

2024

Physical Communication

View full text Add to dashboard Cite

LLM-Assisted Data Augmentation for Chinese Dialogue-Level Dependency Parsing

Zhang,

Jiang,

Liu

et al. 2024

Computational Linguistics

View full text Add to dashboard Cite

Dialogue–level dependency parsing, despite its growing academic interest, often encounters underperformance issues due to resource shortages. A potential solution to this challenge is data augmentation. In recent years, large language models (LLMs) have demonstrated strong capabilities in generation which can facilitate data augmentation greatly. In this study, we focus on Chinese dialogue–level dependency parsing, presenting three simple and effective strategies with LLM to augment the original training instances, namely word–level, syntax–level and discourse–level augmentations, respectively. These strategies enable LLMs to either preserve or modify dependency structures, thereby assuring accuracy while increasing the diversity of instances at different levels. We conduct experiments on the benchmark dataset released by Jiang et al. (2023) to validate our approach. Results show that our method can greatly boost the parsing performance in various settings, particularly in dependencies among elementary discourse units (EDUs). Lastly, we provide in–depth analysis to show the key points of our data augmentation strategies.

show abstract

Curriculum-Style Fine-Grained Adaption for Unsupervised Cross-Lingual Dependency Transfer

Cited by 4 publications

References 44 publications

LLM-enhanced Self-training for Cross-domain Constituency Parsing

LLM-enhanced Self-training for Cross-domain Constituency Parsing

Filtered data augmentation approach based on model competence evaluation

LLM-Assisted Data Augmentation for Chinese Dialogue-Level Dependency Parsing

Contact Info

Product

Resources

About