基于检索重排序模型的文本差异化研究

门业堃; 钱梦迪; 于钊; 滕景竹; 陈少坤; 颜旭

门业堃,钱梦迪,于钊,滕景竹,陈少坤,颜旭.基于检索重排序模型的文本差异化研究[J].电测与仪表,2023,60(1):57-63.
Men Yekun,Qian Mengdi,Yu Zhao,Teng Jingzhu,Chen Shaokun,Yan Xu.Research of text differentiation based on retrieval reordering model[J].Electrical Measurement & Instrumentation,2023,60(1):57-63.

基于检索重排序模型的文本差异化研究

Research of text differentiation based on retrieval reordering model

DOI：10.19753/j.issn1001-1390.2023.01.009

中文关键词: 差异性检索排序学习文本相似度检索重排序设备质量评估自动化

英文关键词:differential retrieval, ranking learning, text similarity, retrieval reordering, automation of equipment quality assessment

基金项目:

作者中文名	作者英文名	单位
门业堃	Men Yekun	国网北京市电力公司电力科学研究院，北京 100075
钱梦迪	Qian Mengdi	国网北京市电力公司电力科学研究院，北京 100075
于钊	Yu Zhao	国网北京市电力公司电力科学研究院，北京 100075
滕景竹	Teng Jingzhu	国网北京市电力公司电力科学研究院，北京 100075
陈少坤	Chen Shaokun	北京恒华龙信数据科技有限公司，北京 100088
颜旭	Yan Xu	北京恒华龙信数据科技有限公司，北京 100088

摘要点击次数: 1317

中文摘要:

电力行业在设备质量评估中需要结合行业标准规范文件中指定的具体标准来对设备质量进行准确评估。文中通过基于检索重排序模型的文本差异化模型，建立自动化、信息化、智能化的标准差异化梳理技术，有效解决了当前设备质量评估时的费时费力问题，并提升了文本差异检索的准确率。文中主要围绕自动化、信息化、智能化的标准差异化梳理技术，通过基于检索重排序模型的信息检索模型，建立针对同一领域不同标准的文本检索比对，检索出不同文件对同一技术有着不同要求的差异性内容并进行预警提示。文中的创新点是，利用检索重排序精确性高的优点，在保留传统差异性检索召回模型的召回候选能力的基础上进一步提高精确性。模型在真实的电力行业技术标准文档上进行了交叉验证，验证了所提出的模型效果，结果表明模型具有良好的实用性，可以广泛应用于电力设备质量评估，供应商评价标准检索等领域。

英文摘要:

In the equipment quality assessment of the power industry, it is necessary to combine the specific standards specified in the industry standard specification documents to accurately assess the quality of the equipment. Through the text differentiation model based on the retrieval reordering model, an automated, informative, and intelligent standard differentiation combing technology is established, which effectively solves the time-consuming and laborious problem of current equipment quality assessment, and improves the accuracy rate of text difference retrieval. This paper mainly focuses on the standard differentiation combing technology of automation, informationization and intelligence. Through the information retrieval model based on the retrieval reordering model, the text retrieval comparison for different standards in the same field is established, and the retrieval of differential content of different documents has different requirements for the same technology, and provides early warning. The innovation in the paper is to take advantage of the high accuracy of retrieval reordering, and to further improve the accuracy on the basis of retaining the recall candidate ability of the traditional difference retrieval recall model. The model is cross-validated on the real power industry technical standard documents to verify the effect of the proposed model. The results show that the model has good practicability and can be widely used in power equipment quality evaluation, supplier evaluation standard retrieval and other fields.

查看全文查看/发表评论下载PDF阅读器

关闭