An In-depth Analysis of Passage-Level Label Transfer for Contextual Document Ranking

TL;DR

This paper investigates passage-level label transfer for document ranking with pre-trained language models, revealing label noise issues and proposing a weak supervision scheme that improves retrieval performance and efficiency.

Contribution

It provides a detailed analysis of splitting and label transfer strategies, identifying issues with label noise and proposing a novel weak supervision approach for better passage-level relevance labeling.

Findings

Direct label transfer introduces noise affecting effectiveness.

Fine-grained splitting increases query processing time.

Proposed weak supervision scheme improves nDCG scores by 3-14%.

Abstract

Pre-trained contextual language models such as BERT, GPT, and XLnet work quite well for document retrieval tasks. Such models are fine-tuned based on the query-document/query-passage level relevance labels to capture the ranking signals. However, the documents are longer than the passages and such document ranking models suffer from the token limitation (512) of BERT. Researchers proposed ranking strategies that either truncate the documents beyond the token limit or chunk the documents into units that can fit into the BERT. In the later case, the relevance labels are either directly transferred from the original query-document pair or learned through some external model. In this paper, we conduct a detailed study of the design decisions about splitting and label transfer on retrieval effectiveness and efficiency. We find that direct transfer of relevance labels from documents to passages introduces label noise that strongly affects retrieval effectiveness for large training datasets. We also find that query processing times are adversely affected by fine-grained splitting schemes. As a remedy, we propose a careful passage level labelling scheme using weak supervision that delivers improved performance (3-14% in terms of nDCG score) over most of the recently proposed models for ad-hoc retrieval while maintaining manageable computational complexity on four diverse document retrieval datasets.