Embedding-based Zero-shot Retrieval through Query Generation

TL;DR

This paper introduces a novel synthetic data generation method for embedding-based zero-shot passage retrieval, significantly outperforming traditional BM25 in multiple datasets by leveraging query generation techniques.

Contribution

The work presents a new approach for generating synthetic training data to enhance neural retrieval models, enabling effective zero-shot retrieval without extensive labeled datasets.

Findings

Outperforms BM25 on 5 out of 6 datasets

Average Recall@1 improvement of 2.45 points

Synthetic data can sometimes surpass real data for training

Abstract

Passage retrieval addresses the problem of locating relevant passages, usually from a large corpus, given a query. In practice, lexical term-matching algorithms like BM25 are popular choices for retrieval owing to their efficiency. However, term-based matching algorithms often miss relevant passages that have no lexical overlap with the query and cannot be finetuned to downstream datasets. In this work, we consider the embedding-based two-tower architecture as our neural retrieval model. Since labeled data can be scarce and because neural retrieval models require vast amounts of data to train, we propose a novel method for generating synthetic training data for retrieval. Our system produces remarkable results, significantly outperforming BM25 on 5 out of 6 datasets tested, by an average of 2.45 points for Recall@1. In some cases, our model trained on synthetic data can even outperform the same model trained on real data