Skip to main navigation Skip to search Skip to main content

KnowPrefix-Tuning: A Two-Stage Prefix-Tuning Framework for Knowledge-Grounded Dialogue Generation

  • Jiaqi Bai
  • , Zhao Yan
  • , Ze Yang
  • , Jian Yang
  • , Xinnian Liang
  • , Hongcheng Guo
  • , Zhoujun Li*
  • *Corresponding author for this work
  • Beihang University
  • Tencent

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Existing knowledge-grounded conversation systems generate responses typically in a retrieve-then-generate manner. They require a large knowledge base and a strong knowledge retrieval component, which is time- and resource-consuming. In this paper, we address the challenge by leveraging the inherent knowledge encoded in the pre-trained language models (PLMs). We propose Knowledgeable Prefix Tuning (KnowPrefix-Tuning), a two-stage tuning framework, bypassing the retrieval process in a knowledge-grounded conversation system by injecting prior knowledge into the lightweight knowledge prefix. The knowledge prefix is a sequence of continuous knowledge-specific vectors that can be learned during training. In addition, we propose a novel interactive re-parameterization mechanism that allows the prefix to interact fully with the PLM during the optimization of response generation. Experimental results demonstrate that KnowPrefix-Tuning outperforms fine-tuning and other lightweight tuning approaches, and performs comparably with strong retrieval-based baselines while being faster during inference (The code is available at https://github.com/fantast4ever/KnowPrefix-Tuning.)

Original languageEnglish
Title of host publicationMachine Learning and Knowledge Discovery in Databases
Subtitle of host publicationResearch Track - European Conference, ECML PKDD 2023, Proceedings
EditorsDanai Koutra, Claudia Plant, Manuel Gomez Rodriguez, Elena Baralis, Francesco Bonchi
PublisherSpringer Science and Business Media Deutschland GmbH
Pages525-542
Number of pages18
ISBN (Print)9783031434143
DOIs
StatePublished - 2023
Event23rd Joint European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2023 - Turin, Italy
Duration: 18 Sep 202322 Sep 2023

Publication series

NameLecture Notes in Computer Science
Volume14170 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference23rd Joint European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2023
Country/TerritoryItaly
CityTurin
Period18/09/2322/09/23

Keywords

  • Dialogue generation
  • Knowledge-grounded dialogue
  • Parameter-efficient fine-tuning
  • Pre-trained language models

Fingerprint

Dive into the research topics of 'KnowPrefix-Tuning: A Two-Stage Prefix-Tuning Framework for Knowledge-Grounded Dialogue Generation'. Together they form a unique fingerprint.

Cite this