大语言模型领域意图的精准性增强方法

TP391; 目前通用大语言模型(如GPT)在专业领域问答应用中存在不稳定性和不真实性.针对这一现象,提出了一种在通用大语言模型上耦合领域知识的意图识别精准性增强方法(EIRDK),其中引入了三个具体策略:a)通过领域知识库对GPT输出结果进行打分过滤;b)训练领域知识词向量模型优化提示语句规范性;c)利用GPT的反馈结果提升领域词向量模型和GPT模型的一致性.实验分析显示,相比于标准的GPT模型,新方法在私有数据集上可以提升25％的意图理解准确性,在CMID数据集上可以提升12％的意图理解准确性.实验结果证明了 EIRDK方法的有效性....

Full description

Saved in:
Bibliographic Details
Published in计算机应用研究 Vol. 41; no. 10; pp. 2893 - 2899
Main Authors 任元凯, 谢振平
Format Journal Article
LanguageChinese
Published 江南大学人工智能与计算机学院,江苏无锡 214122%江南大学人工智能与计算机学院,江苏无锡 214122 2024
江南大学人机融合软件与媒体技术江苏省高校重点实验室,江苏无锡 214122
Subjects
意图精准性增强
GPT反馈学习
knowledge Q&A with large language models
大语言模型知识问答
domain knowledge inte-gration
intent recognition accuracy enhancement
feedback learning from GPT
领域知识集成
Online AccessGet full text
ISSN1001-3695
DOI10.19734/j.issn.1001-3695.2024.02.0022

Cover

Abstract TP391; 目前通用大语言模型(如GPT)在专业领域问答应用中存在不稳定性和不真实性.针对这一现象,提出了一种在通用大语言模型上耦合领域知识的意图识别精准性增强方法(EIRDK),其中引入了三个具体策略:a)通过领域知识库对GPT输出结果进行打分过滤;b)训练领域知识词向量模型优化提示语句规范性;c)利用GPT的反馈结果提升领域词向量模型和GPT模型的一致性.实验分析显示,相比于标准的GPT模型,新方法在私有数据集上可以提升25％的意图理解准确性,在CMID数据集上可以提升12％的意图理解准确性.实验结果证明了 EIRDK方法的有效性.
AbstractList TP391; 目前通用大语言模型(如GPT)在专业领域问答应用中存在不稳定性和不真实性.针对这一现象,提出了一种在通用大语言模型上耦合领域知识的意图识别精准性增强方法(EIRDK),其中引入了三个具体策略:a)通过领域知识库对GPT输出结果进行打分过滤;b)训练领域知识词向量模型优化提示语句规范性;c)利用GPT的反馈结果提升领域词向量模型和GPT模型的一致性.实验分析显示,相比于标准的GPT模型,新方法在私有数据集上可以提升25％的意图理解准确性,在CMID数据集上可以提升12％的意图理解准确性.实验结果证明了 EIRDK方法的有效性.
Abstract_FL Large language models(such as GPT)exhibit instability and inauthenticity in professional domain Q&A applica-tions.To address this issue,this paper proposed a method to enhance intent recognition by domain knowledge(EIRDK)for large language models.The method involved three specific strategies:a)scoring and filtering the GPT output using a domain knowledge base,b)training the domain knowledge word vector mode to optimize prompt,c)utilizing feedback from GPT to im-prove the coherence between the domain word vector model and the GPT model.Experimental analysis demonstrates that,com-pared to the standard GPT model,the new method achieves a 25％improvement in intent understanding accuracy on the pri-vate dataset and a 12％increase on the CMID dataset.The results validate the effectiveness of the EIRDK method.
Author 任元凯
谢振平
AuthorAffiliation 江南大学人工智能与计算机学院,江苏无锡 214122%江南大学人工智能与计算机学院,江苏无锡 214122;江南大学人机融合软件与媒体技术江苏省高校重点实验室,江苏无锡 214122
AuthorAffiliation_xml – name: 江南大学人工智能与计算机学院,江苏无锡 214122%江南大学人工智能与计算机学院,江苏无锡 214122;江南大学人机融合软件与媒体技术江苏省高校重点实验室,江苏无锡 214122
Author_FL Ren Yuankai
Xie Zhenping
Author_FL_xml – sequence: 1
  fullname: Ren Yuankai
– sequence: 2
  fullname: Xie Zhenping
Author_xml – sequence: 1
  fullname: 任元凯
– sequence: 2
  fullname: 谢振平
BookMark eNrjYmDJy89LZWBQMzTQM7Q0NzbRz9LLLC7O0zM0MDDUNTazNNUzMjAy0TMw0jMwMDJiYeCES3AwcBUXZxkYmBgZWhpwMhg_XbL8xfq1L1Y0PFux8Om87peL2p7On_-spf_p7H3PZ7U837TvaXvbs4blTxfNe7pn17NpO59tnsrDwJqWmFOcyguluRk03FxDnD10yxPz0hLz0uOz8kuL8oAy8VnFWZWVlVkgtwAdYGBkTIJSAPqZU6Q
ClassificationCodes TP391
ContentType Journal Article
Copyright Copyright © Wanfang Data Co. Ltd. All Rights Reserved.
Copyright_xml – notice: Copyright © Wanfang Data Co. Ltd. All Rights Reserved.
DBID 2B.
4A8
92I
93N
PSX
TCJ
DOI 10.19734/j.issn.1001-3695.2024.02.0022
DatabaseName Wanfang Data Journals - Hong Kong
WANFANG Data Centre
Wanfang Data Journals
万方数据期刊 - 香港版
China Online Journals (COJ)
China Online Journals (COJ)
DatabaseTitleList
DeliveryMethod fulltext_linktorsrc
Discipline Computer Science
DocumentTitle_FL Intention recognition accuracy enhancing method for large language model
EndPage 2899
ExternalDocumentID jsjyyyj202410002
GrantInformation_xml – fundername: 国家自然科学基金
  funderid: (62272201)
GroupedDBID -0Y
2B.
4A8
5XA
5XJ
92H
92I
93N
ABJNI
ACGFS
ALMA_UNASSIGNED_HOLDINGS
CCEZO
CUBFJ
CW9
PSX
TCJ
TGT
U1G
U5S
ID FETCH-wanfang_journals_jsjyyyj2024100023
ISSN 1001-3695
IngestDate Thu May 29 04:34:00 EDT 2025
IsPeerReviewed false
IsScholarly true
Issue 10
Keywords 意图精准性增强
GPT反馈学习
knowledge Q&A with large language models
大语言模型知识问答
domain knowledge inte-gration
intent recognition accuracy enhancement
feedback learning from GPT
领域知识集成
Language Chinese
LinkModel OpenURL
MergedId FETCHMERGED-wanfang_journals_jsjyyyj2024100023
ParticipantIDs wanfang_journals_jsjyyyj202410002
PublicationCentury 2000
PublicationDate 2024
PublicationDateYYYYMMDD 2024-01-01
PublicationDate_xml – year: 2024
  text: 2024
PublicationDecade 2020
PublicationTitle 计算机应用研究
PublicationTitle_FL Application Research of Computers
PublicationYear 2024
Publisher 江南大学人工智能与计算机学院,江苏无锡 214122%江南大学人工智能与计算机学院,江苏无锡 214122
江南大学人机融合软件与媒体技术江苏省高校重点实验室,江苏无锡 214122
Publisher_xml – name: 江南大学人工智能与计算机学院,江苏无锡 214122%江南大学人工智能与计算机学院,江苏无锡 214122
– name: 江南大学人机融合软件与媒体技术江苏省高校重点实验室,江苏无锡 214122
SSID ssj0042190
ssib001102940
ssib002263599
ssib023646305
ssib051375744
ssib025702191
Score 4.6808124
Snippet TP391;...
SourceID wanfang
SourceType Aggregation Database
StartPage 2893
Title 大语言模型领域意图的精准性增强方法
URI https://d.wanfangdata.com.cn/periodical/jsjyyyj202410002
Volume 41
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://utb.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMw1V1LaxRBEG5CAuLFR1R8s4Itguy68-ie7mP37gxB0FOE3MLuzqyyhxXM5rA55RDiwYMnPSgkGITkIoKeFMm_mU38F1bV9O6OEkz0JgxNUVNT9XVXMlXTO1PF2O0wS0OZyU5VprpTDVUmq63My6rtbpR261kaRdRs4uEjufA4fLAklmZm50tvLa0O2rXO2pHflfyLV4EHfsWvZP_CsxOlwAAa_AsjeBjGE_mYx4KbkJuIx4qbhJsmEQpfX4glEsZDGR1zZXmsufG5ksRJ8AAZBblkQhzLbczjiGuDTCCsTxzBVURXSVSLtgTq0XTKNrg1eEpLbjUSNuBFP8txykuQYkISIaEjki8uFDhqMgcjAEaZuuMYze1k45DHIbcWD0QER-CgmWQqoritIzjE2sAVQQOgJShvbvjTbU0C7NFSgK4mYZssqcD1NJIMmzFasCcIvyaOQhy2STJAx8dM1inUeDnA9BslAAp9pMgpmhYBxQCJd8_3Qs8HafE_wS1FO3yfLpDur8KFw6IO2fjfvl4ObqpoZukSJXxUPzII6ygIKQqjkdrESA0dXBTIdSB-LXTeW-kNh8MeSuGPTpBXzflwq4BQNWds0ybTLB-S4nLVRx8LKk2fqrElgiyFMezTCHF5EsaEF0SCmj4UCVsIJ4uiJQ7oKXZnPI37f5wEfejX77b6T0o56eI5dsY9TFZMcWc4z2bWns6zs-NGLRUXty-wIP-we_jp4-He-mjvfb718sfOZr69Pdp4lb_bP3i7cfB5P3-xOVrfzXe28u_fRm--jr68vsjuJvFiY6HqbC-7O9LK8u8rGFxis_1n_ewyq0Qt0c5UmGYtPw073bRd78huC6sVwuQDIa-wW8equ3oCmWvsNNLFJuV1Njt4vprdgLR90L7p3PgTLjCoeg
linkProvider EBSCOhost
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=%E5%A4%A7%E8%AF%AD%E8%A8%80%E6%A8%A1%E5%9E%8B%E9%A2%86%E5%9F%9F%E6%84%8F%E5%9B%BE%E7%9A%84%E7%B2%BE%E5%87%86%E6%80%A7%E5%A2%9E%E5%BC%BA%E6%96%B9%E6%B3%95&rft.jtitle=%E8%AE%A1%E7%AE%97%E6%9C%BA%E5%BA%94%E7%94%A8%E7%A0%94%E7%A9%B6&rft.au=%E4%BB%BB%E5%85%83%E5%87%AF&rft.au=%E8%B0%A2%E6%8C%AF%E5%B9%B3&rft.date=2024&rft.pub=%E6%B1%9F%E5%8D%97%E5%A4%A7%E5%AD%A6%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD%E4%B8%8E%E8%AE%A1%E7%AE%97%E6%9C%BA%E5%AD%A6%E9%99%A2%2C%E6%B1%9F%E8%8B%8F%E6%97%A0%E9%94%A1+214122%25%E6%B1%9F%E5%8D%97%E5%A4%A7%E5%AD%A6%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD%E4%B8%8E%E8%AE%A1%E7%AE%97%E6%9C%BA%E5%AD%A6%E9%99%A2%2C%E6%B1%9F%E8%8B%8F%E6%97%A0%E9%94%A1+214122&rft.issn=1001-3695&rft.volume=41&rft.issue=10&rft.spage=2893&rft.epage=2899&rft_id=info:doi/10.19734%2Fj.issn.1001-3695.2024.02.0022&rft.externalDocID=jsjyyyj202410002
thumbnail_s http://utb.summon.serialssolutions.com/2.0.0/image/custom?url=http%3A%2F%2Fwww.wanfangdata.com.cn%2Fimages%2FPeriodicalImages%2Fjsjyyyj%2Fjsjyyyj.jpg