Zengzhi Wang, Qiming Xie, Rui Xia
Nanjing University of Science and Technology, China
📄 [Paper] ☕ [Homepage on PaperWithCode] 🍟 [Paper Summary in Kudos] 📃 [Poster] 🎥 [5 Mins Presentation Video]
Updates
- 2023/4/24: We upload the domain-adaptive pre-training models (🤗restaurant-t5-base, 🤗laptop-t5-base, 🤗dutch-restaurant-mt5-small and 🤗french-restaurant-mt5-small) to 🤗 huggingface.
In this work, we introduce a simple yet effective framework called FS-ABSA, which involves domain-adaptive pre-training and textinfilling fine-tuning. Specifically,
- we approach the End-to-End ABSA task as a text-infilling problem.
- we perform domain-adaptive pre-training with the text-infilling objective, narrowing the two gaps, i.e., domain gap and objective gap, and consequently facilitating the knowledge transfer.
To run the code, please install all the dependency packages by using the following command:
pip install -r requirements.txt
NOTE: All experiments are conducted on NVIDIA RTX 3090 (and Linux OS). Different versions of packages and GPU may lead to different results.
NOTE: All experiment scripts are with multiple runs (three seeds).
## English Dataset: 14lap
$ bash script/run_aspe_fewshot_14lap.sh
## English Dataset: 14res
$ bash script/run_aspe_fewshot_14res.sh
## Dutch Dataset: 16res
$ bash script/run_aspe_fewshot_dutch.sh
## French Dataset: 16res
$ bash script/run_aspe_fewshot_french.sh
## English Dataset: 14lap
$ bash script/run_aspe_14lap.sh
## English Dataset: 14res
$ bash script/run_aspe_14res.sh
## Dutch Dataset: 16res
$ bash script/run_aspe_dutch.sh
## French Dataset: 16res
$ bash script/run_aspe_french.sh
Results on 14-Lap and 14-Res under different training data size scenarios
Comparison with SOTA under the full data setting
Results in two low-resource languages under different training data sizes
If you find this work helpful, please cite our paper as follows:
@inproceedings{10.1145/3539618.3591940,
author = {Wang, Zengzhi and Xie, Qiming and Xia, Rui},
title = {A Simple yet Effective Framework for Few-Shot Aspect-Based Sentiment Analysis},
year = {2023},
isbn = {9781450394086},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3539618.3591940},
doi = {10.1145/3539618.3591940},
abstract = {The pre-training and fine-tuning paradigm has become the main-stream framework in the field of Aspect-Based Sentiment Analysis (ABSA). Although it has achieved sound performance in the domains containing enough fine-grained aspect-sentiment annotations, it is still challenging to conduct few-shot ABSA in domains where manual annotations are scarce. In this work, we argue that two kinds of gaps, i.e., domain gap and objective gap, hinder the transfer of knowledge from pre-training language models (PLMs) to ABSA tasks. To address this issue, we introduce a simple yet effective framework called FS-ABSA, which involves domain-adaptive pre-training and text-infilling fine-tuning. We approach the End-to-End ABSA task as a text-infilling problem and perform domain-adaptive pre-training with the text-infilling objective, narrowing the two gaps and consequently facilitating the knowledge transfer. Experiments show that the resulting model achieves more compelling performance than baselines under the few-shot setting while driving the state-of-the-art performance to a new level across datasets under the fully-supervised setting. Moreover, we apply our framework to two non-English low-resource languages to demonstrate its generality and effectiveness.},
booktitle = {Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval},
pages = {1765–1770},
numpages = {6},
keywords = {few-shot learning, opinion mining, sentiment analysis},
location = {Taipei, Taiwan},
series = {SIGIR '23}
}
If you have any questions related to this work, you can open an issue with details or feel free to email Zengzhi(zzwang@njust.edu.cn), Qiming(qmxie@njust.edu.cn).
Our code is based on ABSA-QUAD. Thanks for their work.