[1]侯继辉,吴小忠,郑浩,等.基于模态注意力与知识库比对的招标文件编制方法[J].计算机技术与发展,2025,(05):36-44.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0407]
HOU Ji-hui,WU Xiao-zhong,ZHENG Hao,et al.Method for Bidding Document Compilation Based on Multimodal Attention and Knowledge Base Comparison[J].,2025,(05):36-44.[doi:10.20165/j.cnki.ISSN1673-629X.2024.0407]
点击复制
基于模态注意力与知识库比对的招标文件编制方法
《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]
- 卷:
-
- 期数:
-
2025年05期
- 页码:
-
36-44
- 栏目:
-
软件技术与工程
- 出版日期:
-
2025-05-10
文章信息/Info
- Title:
-
Method for Bidding Document Compilation Based on Multimodal Attention and Knowledge Base Comparison
- 文章编号:
-
1673-629X(2025)05-0036-09
- 作者:
-
侯继辉1; 吴小忠1; 郑浩1; 夏卓群2; 唐海东2*; 王欣2
-
1. 湖南湘能创业项目管理有限公司,湖南 长沙 410221;
2. 长沙理工大学 计算机与通信工程学院,湖南 长沙 410114
- Author(s):
-
HOU Ji-hui1; WU Xiao-zhong1; ZHENG Hao1; XIA Zhuo-qun2; TANG Hai-dong2*; WANG Xin2
-
1. Hunan Xiangneng Chuangve Project Management Co. ,Ltd. ,Changsha 410221,China;
2. School of Computer and Communication Engineering,Changsha University of Science and Technology,Changsha 410114,China
-
- 关键词:
-
招标文件; 智能编制; 知识库; 命名实体识别; 结构化
- Keywords:
-
bidding documents; intelligent compiling; knowledge base; named entity recognition; structured
- 分类号:
-
TP399
- DOI:
-
10.20165/j.cnki.ISSN1673-629X.2024.0407
- 摘要:
-
多模态自适应权重注意力命名实体识别与知识库比对的招标编制方法随着人工智能的发展,招投标业务中引入了大量的智能化手段。 招标文件编制作为招投标业务中的重要流程,编制结果的准确性和编制速率的高效性需要更好的保障。 传统招标文件编制中存在的文件编制过程复杂、耗时较长和人为编制的偶然错误问题,降低了文件编制的效率甚至影响了招标业务的流畅性。 为此提出了基于多模态自适应权重注意力机制的命名实体识别(MMAWA-NER)提取与知识库比对的招标文件编制方法。 首先以人工分析的形式将历史招标文件按照文章条目分为各个模块,每个模块使用包含了固定的条款信息并存入结构化标准库中;在文件编制阶段,根据标准库当中的模块内容生成标准化文件,编制人员随后填写关键信息;在编制完成后,将需校验的招标文件通过 MMAWA-NER 抽取关键信息与领域知识库进行对比,并返回评估结果给编制人员;最后使用了历史招标文件处理后的知识库数据生成新的招标文件数据集进行评估,并对比了其他模型的方法,实验证明该模型实现了 80% 的编制效率和 20% 的准确率提升。
- Abstract:
-
With the development of artificial intelligence, a large number of intelligent means have been introduced into the bidding business for multi - modal adaptive weighted attention named entity recognition and knowledge base comparison bidding preparation method. As an important process in the bidding business,the accuracy of the preparation results and the efficiency of the preparation rate need to be better guaranteed. The complicated and time-consuming process of document preparation and the occasional errors of manual preparation in traditional bidding document preparation reduce the efficiency of document preparation and even affect the fluency of bidding business. Therefore, a bidding document compilation method based on MMAWA - NER ( Multimodal Adaptive Weighted Attention Mechanism-Named Entity Recognition) extraction and knowledge base comparison is proposed. Firstly,the historical bidding documents are divided into modules according to article items in the form of manual analysis. Each module contains fixed terms information and is stored in the structured standard library. In the document preparation stage,standardized documents are generated according to the module content in the standard library,and the preparation personnel fill in the key information afterwards. After the preparation is completed,the key information of the bidding documents to be verified is extracted by MMAWA-NER for comparison with the domain knowledge base,and the evaluation result is returned to the preparation personnel. Finally,we use the knowledge base data after the historical bidding documents processing to generate a new bidding documents data set for evaluation,and compare the methods of other models. The experiment proves that the proposed model achieves 80% compilation efficiency and 20% accuracy improvement.
更新日期/Last Update:
2025-05-10