Few-Shot Text Ranking with Meta Adapted Synthetic Weak Supervision
This work is significant for researchers and practitioners in information retrieval who need to deploy Neu-IR models in new domains with scarce labeled data, offering a method to improve their performance.
This paper addresses the challenge of applying Neural Information Retrieval (Neu-IR) models in domains with limited training data. The authors propose MetaAdaptRank, a method that synthesizes weak supervision signals from label-rich source domains and meta-learns to reweight these signals based on their contribution to target-domain ranking accuracy. Experiments on three TREC benchmarks demonstrate significant improvements in few-shot ranking accuracy for Neu-IR models.
The effectiveness of Neural Information Retrieval (Neu-IR) often depends on a large scale of in-domain relevance training signals, which are not always available in real-world ranking scenarios. To democratize the benefits of Neu-IR, this paper presents MetaAdaptRank, a domain adaptive learning method that generalizes Neu-IR models from label-rich source domains to few-shot target domains. Drawing on source-domain massive relevance supervision, MetaAdaptRank contrastively synthesizes a large number of weak supervision signals for target domains and meta-learns to reweight these synthetic "weak" data based on their benefits to the target-domain ranking accuracy of Neu-IR models. Experiments on three TREC benchmarks in the web, news, and biomedical domains show that MetaAdaptRank significantly improves the few-shot ranking accuracy of Neu-IR models. Further analyses indicate that MetaAdaptRank thrives from both its contrastive weak data synthesis and meta-reweighted data selection. The code and data of this paper can be obtained from https://github.com/thunlp/MetaAdaptRank.