SparTerm: Learning Term-based Sparse Representation for Fast Text Retrieval

TL;DR

SparTerm introduces a novel framework that transfers deep knowledge from pre-trained language models to term-based sparse text representations, enhancing semantic matching while maintaining efficiency and interpretability.

Contribution

The paper presents SparTerm, a new method that learns sparse text representations directly in the full vocabulary space, combining term importance prediction and gating control for improved retrieval.

Findings

Outperforms traditional sparse methods on MSMARCO dataset

Achieves state-of-the-art ranking among PLM-based sparse models

Effectively balances sparsity, interpretability, and semantic understanding

Abstract

Term-based sparse representations dominate the first-stage text retrieval in industrial applications, due to its advantage in efficiency, interpretability, and exact term matching. In this paper, we study the problem of transferring the deep knowledge of the pre-trained language model (PLM) to Term-based Sparse representations, aiming to improve the representation capacity of bag-of-words(BoW) method for semantic-level matching, while still keeping its advantages. Specifically, we propose a novel framework SparTerm to directly learn sparse text representations in the full vocabulary space. The proposed SparTerm comprises an importance predictor to predict the importance for each term in the vocabulary, and a gating controller to control the term activation. These two modules cooperatively ensure the sparsity and flexibility of the final text representation, which unifies the term-weighting and expansion in the same framework. Evaluated on MSMARCO dataset, SparTerm significantly outperforms traditional sparse methods and achieves state of the art ranking performance among all the PLM-based sparse models.