Zero-shot Text Classification With Generative Language Models
It addresses the challenge of adapting models to new tasks without training data, which is incremental as it builds on existing generative language model approaches.
This paper tackles the problem of zero-shot text classification by using generative language models with natural language task descriptions, achieving up to a 45% absolute improvement in accuracy over baselines on six benchmark datasets.
This work investigates the use of natural language to enable zero-shot model adaptation to new tasks. We use text and metadata from social commenting platforms as a source for a simple pretraining task. We then provide the language model with natural language descriptions of classification tasks as input and train it to generate the correct answer in natural language via a language modeling objective. This allows the model to generalize to new classification tasks without the need for multiple multitask classification heads. We show the zero-shot performance of these generative language models, trained with weak supervision, on six benchmark text classification datasets from the torchtext library. Despite no access to training data, we achieve up to a 45% absolute improvement in classification accuracy over random or majority class baselines. These results show that natural language can serve as simple and powerful descriptors for task adaptation. We believe this points the way to new metalearning strategies for text problems.