Constructing Set-Compositional and Negated Representations for First-Stage Ranking

TL;DR

This paper develops methods for constructing set-compositional and negated query representations in zero-shot first-stage ranking, improving retrieval of complex and niche information needs using vector operations and addressing limitations of existing retrieval models.

Contribution

It introduces Disentangled Negation and Combined Pseudo-Term approaches for zero-shot compositional query encoding, enhancing retrieval performance without fine-tuning.

Findings

Zero-shot methods outperform some fine-tuned retrievers

Disentangled Negation improves negation handling in retrieval

Enhanced LSR models better attribute negative term scores

Abstract

Set compositional and negated queries are crucial for expressing complex information needs and enable the discovery of niche items like Books about non-European monarchs. Despite the recent advances in LLMs, first-stage ranking remains challenging due to the requirement of encoding documents and queries independently from each other. This limitation calls for constructing compositional query representations that encapsulate logical operations or negations, and can be used to match relevant documents effectively. In the first part of this work, we explore constructing such representations in a zero-shot setting using vector operations between lexically grounded Learned Sparse Retrieval (LSR) representations. Specifically, we introduce Disentangled Negation that penalizes only the negated parts of a query, and a Combined Pseudo-Term approach that enhances LSRs ability to handle intersections. We find that our zero-shot approach is competitive and often outperforms retrievers fine-tuned on compositional data, highlighting certain limitations of LSR and Dense Retrievers. Finally, we address some of these limitations and improve LSRs representation power for negation, by allowing them to attribute negative term scores and effectively penalize documents containing the negated terms.