AugGPT: Leveraging ChatGPT for Text Data Augmentation

Abstract

Text data augmentation is an effective strategy for overcoming the challenge of limited sample sizes in many natural language processing (NLP) tasks. This challenge is especially prominent in the few-shot learning scenario, where the data in the target domain is generally much scarcer and of lowered quality. A natural and widely-used strategy to mitigate such challenges is to perform data augmentation on the training data to better capture the data invariance and increase the sample size. However, current text data augmentation methods either can not ensure the correct labeling of the generated data (lacking faithfulness) or can not ensure sufficient diversity in the generated data (lacking completeness), or both. Inspired by the recent success of large language models, especially the development of ChatGPT, which demonstrated improved language comprehension abilities, in this work, we propose a text data augmentation approach based on ChatGPT (named ChatAug). ChatGPT is trained on data with unparalleled linguistic richness and employs a reinforcement training process with large-scale human feedback, which endows the model with affinity to the naturalness of human language. Our text data augmentation approach ChatAug rephrases each sentence in the training samples into multiple conceptually similar but semantically different samples. The augmented samples can then be used in downstream model training. Experiment results on few-shot learning text classification tasks show the superior performance of the proposed ChatAug approach over state-of-the-art text data augmentation methods in terms of testing accuracy and distribution of the augmented samples.

Detecting Micromanagement During Pair Programming

“…• CBERT [15]: First, we utilize BERT's segment embeddings to condition the BERT model on the class labels during finetuning. 2 We then finetuned the model with the masked language model (MLM) objective which randomly masks some words in the sequences and aims to predict the original word using the context. Finally, we use the resulting model to predict and replace masked words in the training set.…”

“…• ChatGPTfew-shot [2]: We used few-shot prompting of ChatGPT for data augmentation to produce several paraphrases of each sentence in the training set.…”

“…Dai et al [2] utilized few-shot prompting of ChatGPT [8] for data augmentation to produce several variations of each sentence in the training sample. The generated sentences are similar in meaning but have different syntax.…”

ChatGPT: A Brief Narrative Review

“…ChatGPT can assist organizations in automating their customer care assistance procedures, lowering the demand for human agents and enhancing response times. ChatGPT can aid users with personal assistance chores like making appointments or looking for information online [35]. Finally, ChatGPT can be used to create content, such as text for social media postings or marketing initiatives.…”

The Distribution of Etiology and Clinical Presentations of Birth Defects: A Preliminary Assessment of China Birth Cohort Study

“…Recently, the emergence of ChatGPT has significantly advanced NLP tasks by enhancing the capabilities of conversational models, making it a valuable tool for businesses and organizations. Chataug et al [12] leverages ChatGPT to rephrase sentences for text data augmentation. Jiao et al [23] finds the translation ability of Chat-GPT performs competitively with commercial translation products on high-resource and low-resource languages.…”