The Power of Selecting Key Blocks with Local Pre-ranking for Long Document Information Retrieval

TL;DR

This paper introduces a novel method for long document retrieval that involves selecting key blocks through local pre-ranking, enabling effective processing by models like BERT despite the challenges posed by long documents.

Contribution

The paper proposes a new approach of local pre-ranking to select key blocks, improving long document retrieval efficiency and effectiveness over existing truncation, segmentation, and sparse attention methods.

Findings

The method outperforms traditional truncation and segmentation techniques.

It achieves comparable or better results with less computational cost.

Experimental results validate the effectiveness of key block selection in IR tasks.

Abstract

On a wide range of natural language processing and information retrieval tasks, transformer-based models, particularly pre-trained language models like BERT, have demonstrated tremendous effectiveness. Due to the quadratic complexity of the self-attention mechanism, however, such models have difficulties processing long documents. Recent works dealing with this issue include truncating long documents, in which case one loses potential relevant information, segmenting them into several passages, which may lead to miss some information and high computational complexity when the number of passages is large, or modifying the self-attention mechanism to make it sparser as in sparse-attention models, at the risk again of missing some information. We follow here a slightly different approach in which one first selects key blocks of a long document by local query-block pre-ranking, and then few blocks are aggregated to form a short document that can be processed by a model such as BERT. Experiments conducted on standard Information Retrieval datasets demonstrate the effectiveness of the proposed approach.