Relevance feedback strategies for recall-oriented neural information retrieval

TL;DR

This paper introduces a recall-focused relevance feedback method for neural information retrieval, using BERT-based re-ranking and cumulative embedding summation to significantly reduce review effort while maintaining high recall.

Contribution

It proposes a novel relevance feedback strategy that enhances recall in neural IR systems by iteratively re-ranking with BERT embeddings and cumulative feedback, addressing false negatives.

Findings

Reduces review effort by up to 59.04% at fixed recall

Uses BERT-based dense-vector search for relevance ranking

Demonstrates effectiveness in recall-oriented IR applications

Abstract

In a number of information retrieval applications (e.g., patent search, literature review, due diligence, etc.), preventing false negatives is more important than preventing false positives. However, approaches designed to reduce review effort (like "technology assisted review") can create false negatives, since they are often based on active learning systems that exclude documents automatically based on user feedback. Therefore, this research proposes a more recall-oriented approach to reducing review effort. More specifically, through iteratively re-ranking the relevance rankings based on user feedback, which is also referred to as relevance feedback. In our proposed method, the relevance rankings are produced by a BERT-based dense-vector search and the relevance feedback is based on cumulatively summing the queried and selected embeddings. Our results show that this method can reduce review effort between 17.85% and 59.04%, compared to a baseline approach (of no feedback), given a fixed recall target