BYOC: Personalized Few-Shot Classification with Co-Authored Class Descriptions
This addresses the challenge for end-users who need tailored text classification but lack resources for large-scale annotation or expertise in prompt design, offering a practical and accessible solution.
The paper tackles the problem of enabling end-users to build personalized few-shot text classifiers without large annotated datasets or extensive prompt engineering, achieving classifiers within 82% of the performance of models trained with much larger datasets and a 15% higher accuracy than state-of-the-art approaches in user studies.
Text classification is a well-studied and versatile building block for many NLP applications. Yet, existing approaches require either large annotated corpora to train a model with or, when using large language models as a base, require carefully crafting the prompt as well as using a long context that can fit many examples. As a result, it is not possible for end-users to build classifiers for themselves. To address this issue, we propose a novel approach to few-shot text classification using an LLM. Rather than few-shot examples, the LLM is prompted with descriptions of the salient features of each class. These descriptions are coauthored by the user and the LLM interactively: while the user annotates each few-shot example, the LLM asks relevant questions that the user answers. Examples, questions, and answers are summarized to form the classification prompt. Our experiments show that our approach yields high accuracy classifiers, within 82% of the performance of models trained with significantly larger datasets while using only 1% of their training sets. Additionally, in a study with 30 participants, we show that end-users are able to build classifiers to suit their specific needs. The personalized classifiers show an average accuracy of 90%, which is 15% higher than the state-of-the-art approach.