ChatGPT as Data Augmentation for Compositional Generalization: A Case Study in Open Intent Detection

Introduction

Open intent detection, a crucial aspect of natural language understanding, involves the identification of previously unseen intents in usergenerated text. Despite the progress made in this field, challenges persist in handling new combinations of language components, which is essential for compositional generalization. In this paper, we present a case study exploring the use of ChatGPT as a data augmentation technique to enhance compositional generalization in open intent detection tasks. We begin by discussing the limitations of existing benchmarks in evaluating this problem, highlighting the need for constructing datasets for addressing compositional generalization in open intent detection tasks. By incorporating synthetic data generated by ChatGPT into the training process, we demonstrate that our approach can effectively improve model performance. Rigorous evaluation of multiple benchmarks reveals that our method outperforms existing techniques and significantly enhances open intent detection capabilities. Our findings underscore the potential of large language models like ChatGPT for data augmentation in natural language understanding tasks.

Experiments

To effectively integrate paraphrases into the training process of BERT with ADB (DA-ADB), three different strategies are evaluated. The first strategy involves synthesizing 10 paraphrases for each instance in the dataset (GPTAUG-F10), while the second strategy generates 4 paraphrases for each instance (GPTAUG-F4). The third strategy, on the other hand, focuses on instances that the model predicts incorrectly at the current iteration and synthesizes 10 paraphrases for each of these instances (GPTAUG-WP10).

./open_intent_detection/examples/run_ADB.sh
./open_intent_detection/examples/run_DA-ADB.sh
./open_intent_detection/examples/run_ADB-BOOST-F-10.sh
./open_intent_detection/examples/run_ADB-BOOST-F-4.sh
./open_intent_detection/examples/run_ADB-BOOST-WP-10.sh
./open_intent_detection/examples/run_DA-ADB-BOOST-F-10.sh
./open_intent_detection/examples/run_DA-ADB-BOOST-F-4.sh
./open_intent_detection/examples/run_DA-ADB-BOOST-WP-10.sh

Citation

If you use our source code in a research paper, please cite our work as follows:

@inproceedings{fang-etal-2023-chatgpt,
    title = "{C}hat{GPT} as Data Augmentation for Compositional Generalization: A Case Study in Open Intent Detection",
    author = "Fang, Yihao  and
      Li, Xianzhi  and
      Thomas, Stephen  and
      Zhu, Xiaodan",
    booktitle = "Proceedings of the Fifth Workshop on Financial Technology and Natural Language Processing and the Second Multimodal AI For Financial Forecasting",
    year = "2023",
    address = "Macao",
    url = "https://aclanthology.org/2023.finnlp-1.2",
    pages = "13--33",
}