Out-of-Domain Semantics to the Rescue! Zero-Shot Hybrid Retrieval Models

TL;DR

This paper investigates the generalization challenges of deep retrieval models in out-of-domain scenarios and proposes a hybrid approach combining lexical and deep models, resulting in significant performance improvements.

Contribution

It introduces a simple hybrid framework that effectively combines lexical and deep retrieval models to enhance out-of-domain passage retrieval performance.

Findings

Deep models' performance drops significantly out-of-domain.

Lexical models are more robust across different domains.

Hybrid models outperform individual models in out-of-domain settings.

Abstract

The pre-trained language model (eg, BERT) based deep retrieval models achieved superior performance over lexical retrieval models (eg, BM25) in many passage retrieval tasks. However, limited work has been done to generalize a deep retrieval model to other tasks and domains. In this work, we carefully select five datasets, including two in-domain datasets and three out-of-domain datasets with different levels of domain shift, and study the generalization of a deep model in a zero-shot setting. Our findings show that the performance of a deep retrieval model is significantly deteriorated when the target domain is very different from the source domain that the model was trained on. On the contrary, lexical models are more robust across domains. We thus propose a simple yet effective framework to integrate lexical and deep retrieval models. Our experiments demonstrate that these two models are complementary, even when the deep model is weaker in the out-of-domain setting. The hybrid model obtains an average of 20.4% relative gain over the deep retrieval model, and an average of 9.54% over the lexical model in three out-of-domain datasets.