Learning to Retrieve Passages without Supervision
This work addresses the data efficiency problem for researchers and practitioners in open-domain question answering by reducing reliance on large labeled datasets.
The paper tackles the problem of dense retrievers' dependence on labeled question-passage pairs for open-domain question answering by proposing an unsupervised pretraining scheme called 'recurring span retrieval', which creates pseudo examples from recurring spans across passages. The resulting model, Spider, significantly outperforms other pretrained baselines in zero-shot settings, is competitive with BM25, and improves performance when used in hybrid retrievers or as initialization for supervised training.
Dense retrievers for open-domain question answering (ODQA) have been shown to achieve impressive performance by training on large datasets of question-passage pairs. In this work we ask whether this dependence on labeled data can be reduced via unsupervised pretraining that is geared towards ODQA. We show this is in fact possible, via a novel pretraining scheme designed for retrieval. Our "recurring span retrieval" approach uses recurring spans across passages in a document to create pseudo examples for contrastive learning. Our pretraining scheme directly controls for term overlap across pseudo queries and relevant passages, thus allowing to model both lexical and semantic relations between them. The resulting model, named Spider, performs surprisingly well without any labeled training examples on a wide range of ODQA datasets. Specifically, it significantly outperforms all other pretrained baselines in a zero-shot setting, and is competitive with BM25, a strong sparse baseline. Moreover, a hybrid retriever over Spider and BM25 improves over both, and is often competitive with DPR models, which are trained on tens of thousands of examples. Last, notable gains are observed when using Spider as an initialization for supervised training.