Dynamic Injection of Entity Knowledge into Dense Retrievers

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yamada, Ikuya, Ri, Ryokan, Kojima, Takeshi, Iwasawa, Yusuke, Matsuo, Yutaka
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908523682594816
author Yamada, Ikuya
Ri, Ryokan
Kojima, Takeshi
Iwasawa, Yusuke
Matsuo, Yutaka
author_facet Yamada, Ikuya
Ri, Ryokan
Kojima, Takeshi
Iwasawa, Yusuke
Matsuo, Yutaka
contents Dense retrievers often struggle with queries involving less-frequent entities due to their limited entity knowledge. We propose the Knowledgeable Passage Retriever (KPR), a BERT-based retriever enhanced with a context-entity attention layer and dynamically updatable entity embeddings. This design enables KPR to incorporate external entity knowledge without retraining. Experiments on three datasets demonstrate that KPR consistently improves retrieval accuracy, with particularly large gains on the EntityQuestions dataset. When built on the off-the-shelf bge-base retriever, KPR achieves state-of-the-art performance among similarly sized models on two datasets. Models and code are released at https://github.com/knowledgeable-embedding/knowledgeable-embedding.
format Preprint
id arxiv_https___arxiv_org_abs_2507_03922
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Dynamic Injection of Entity Knowledge into Dense Retrievers
Yamada, Ikuya
Ri, Ryokan
Kojima, Takeshi
Iwasawa, Yusuke
Matsuo, Yutaka
Computation and Language
Dense retrievers often struggle with queries involving less-frequent entities due to their limited entity knowledge. We propose the Knowledgeable Passage Retriever (KPR), a BERT-based retriever enhanced with a context-entity attention layer and dynamically updatable entity embeddings. This design enables KPR to incorporate external entity knowledge without retraining. Experiments on three datasets demonstrate that KPR consistently improves retrieval accuracy, with particularly large gains on the EntityQuestions dataset. When built on the off-the-shelf bge-base retriever, KPR achieves state-of-the-art performance among similarly sized models on two datasets. Models and code are released at https://github.com/knowledgeable-embedding/knowledgeable-embedding.
title Dynamic Injection of Entity Knowledge into Dense Retrievers
topic Computation and Language
url https://arxiv.org/abs/2507.03922