Why These Documents? Explainable Generative Retrieval with Hierarchical Category Paths

TL;DR

HyPE introduces a hierarchical category path-based generative retrieval method that enhances explainability and improves retrieval performance by generating semantic category paths before decoding document identifiers.

Contribution

The paper proposes HyPE, a novel hierarchical category path-enhanced generative retrieval approach that provides explanations and boosts retrieval accuracy.

Findings

HyPE achieves high explainability in document retrieval.

HyPE improves retrieval performance over baseline models.

HyPE effectively utilizes external semantic hierarchies for training.

Abstract

Generative retrieval directly decode a document identifier (i.e., docid) in response to a query, making it impossible to provide users with explanations as an answer for ``why is this document retrieved?''. To address this limitation, we propose Hierarchical Category Path-Enhanced Generative Retrieval (HyPE), which enhances explainability by first generating hierarchical category paths step-by-step then decoding docid. By leveraging hierarchical category paths which progress from broader to more specific semantic categories, HyPE can provide detailed explanation for its retrieval decision. For training, HyPE constructs category paths with external high-quality semantic hierarchy, leverages LLM to select appropriate candidate paths for each document, and optimizes the generative retrieval model with path-augmented dataset. During inference, HyPE utilizes path-aware ranking strategy to aggregate diverse topic information, allowing the most relevant documents to be prioritized in the final ranked list of docids. Our extensive experiments demonstrate that HyPE not only offers a high level of explainability but also improves the retrieval performance.