InPars: Data Augmentation for Information Retrieval using Large Language Models
This work addresses the need for domain-specific training data in information retrieval, offering a method to enhance model performance without extensive labeled datasets, though it is incremental as it builds on existing few-shot capabilities of language models.
The authors tackled the problem of domain-specific training data scarcity in information retrieval by using large language models to generate synthetic data, showing that models fine-tuned on this data outperform BM25 and self-supervised dense retrieval methods, with improvements in zero-shot transfer when combined with supervised data.
The information retrieval community has recently witnessed a revolution due to large pretrained transformer models. Another key ingredient for this revolution was the MS MARCO dataset, whose scale and diversity has enabled zero-shot transfer learning to various tasks. However, not all IR tasks and domains can benefit from one single dataset equally. Extensive research in various NLP tasks has shown that using domain-specific training data, as opposed to a general-purpose one, improves the performance of neural models. In this work, we harness the few-shot capabilities of large pretrained language models as synthetic data generators for IR tasks. We show that models finetuned solely on our unsupervised dataset outperform strong baselines such as BM25 as well as recently proposed self-supervised dense retrieval methods. Furthermore, retrievers finetuned on both supervised and our synthetic data achieve better zero-shot transfer than models finetuned only on supervised data. Code, models, and data are available at https://github.com/zetaalphavector/inpars .