基于ERNIE的科技成果供需匹配模型研究

doi:10.16046/j.cnki.issn2096-5680.2026.02.011

河北水利电力学院学报 ›› 2026, Vol. 36 ›› Issue (2): 64-72.DOI: 10.16046/j.cnki.issn2096-5680.2026.02.011

• 计算机应用与自动控制 • 上一篇下一篇

基于ERNIE的科技成果供需匹配模型研究

李明亮¹, 刘佳², 曹颖², 李信冬²

1.河北地质大学信息中心,河北石家庄市槐安东路136号 050031;
2.河北地质大学信息工程学院,河北省石家庄市河北大道601号 052161

收稿日期:2025-01-03 修回日期:2025-03-14 出版日期:2026-06-30 发布日期:2026-06-22
作者简介:李明亮(1976-),男,河北石家庄人,教授,硕士生导师,主要研究方向:物联网应用技术。E-mail:59532499@qq.com
基金资助:
石家庄市科技计划项目(245790035A)

Research on the Supply and Demand Matching Model of Scientific and Technological Achievements Based on ERNIE

LI Mingliang¹, LIU Jia², CAO Ying², LI Xindong²

1. Information Center, Hebei GEO University, 050031, Shijiazhuang, Hebei, China;
2. School of Information Engineering, Hebei GEO University, 052161, Shijiazhuang, Hebei, China

Received:2025-01-03 Revised:2025-03-14 Online:2026-06-30 Published:2026-06-22

摘要/Abstract

摘要： 针对科技成果供需匹配准确率偏低的问题,为有效提升科技成果转化效率,构建了一种基于ERNIE的科技成果供需匹配模型。首先,对科技成果和企业需求的文本使用ERNIE模型进行知识深度挖掘,将词向量表示输入Transformer结构捕捉文本的深层语义信息,加入了多头注意力机制增强模型对复杂特征的处理,显著提高语义理解的准确性;其次,通过添加全局平均池化和多层前馈神经网络提升模型的语义表示能力;最后,引入了Sentence-BERT模型,孪生网络架构和对比损失函数对句子嵌入表示进行优化,并结合余弦相似度计算匹配相似度。经评估指标验证,模型的训练准确率高于其他模型2个百分点,实际准确率、损失率以及F1值提高近1个百分点,达到94.5%。

关键词: 科技成果, 供需匹配模型, Transformer, 注意力机制, Sentence-BERT, 文本相似度

Abstract: At present, the supply and demand matching of scientific and technological achievements faces the problem of low accuracy. In order to effectively improve the transformation efficiency of scientific and technological achievements, a model of supply and demand matching of scientific and technological achievements based on ERNIE is constructed. First, the ERNIE model is used for in-depth knowledge mining of scientific and technological achievements and enterprise requirements. Word vector representation is input into Transformer structure to capture deep semantic information of text. Multi-head attention mechanism is added to enhance the model’s processing of complex features and significantly improve the accuracy of semantic understanding. Secondly, the semantic representation ability of the model is improved by adding global average pooling and multi-layer feedforward neural network. Finally, the Sentence-BERT model, twin network architecture and contrast loss function are introduced to optimize the sentence embedding representation, and the matching similarity is calculated with cosine similarity. According to the evaluation index, the training accuracy of the model is 2 percentage points higher than that of other models, and the actual accuracy, loss rate and F1 value are increased by nearly 1 percentage point, reaching 94.5%.

Key words: scientific and technological achievements, Supply-demand matching model, Transformer, attention mechanism, Sentence-BERT, text similarity

中图分类号:

李明亮, 刘佳, 曹颖, 李信冬. 基于ERNIE的科技成果供需匹配模型研究[J]. 河北水利电力学院学报, 2026, 36(2): 64-72.

LI Mingliang, LIU Jia, CAO Ying, LI Xindong. Research on the Supply and Demand Matching Model of Scientific and Technological Achievements Based on ERNIE[J]. Hebei University of Water Resources and Electric E, 2026, 36(2): 64-72.

参考文献

[1] BENGIO Y, SENECAL J S. Adaptive importance sampling to accelerate training of a neural probabilistic language model[J]. IEEE transactions on neuralnetworks,2008,19(4):713-22.
[2] 毕云杉,钱亚冠,张超华,等.基于ERNIE模型的中文文本分类研究[J].浙江科技学院学报,2021,33(6):461-468.
[3] 邵继美.科研文献核心文本的自然语言处理[D].济南:济南大学,2014.
[4] 关志广,程乔.基于NLP的文本挖掘技术在提升电信客户满意度中的应用[J].无线互联科技,2023,20(5):117-119.
[5] 祖月芳,凌海风,吕永顺.基于NLP技术的装备故障文本匹配算法研究[J].兵器装备工程学报,2021,42(11):204-208.
[6] JIANG Y, ZHANG M, LI C, et al. Semantic text matching for long-form documents[C]//The World Wide Web Conference,2019:795-806.
[7] YANG L, ZHANG M, LI C, et al. Beyond 512 tokens: Siamese multi-depth transformer-based hierarchical encoder for long-form document matching[C]//Proceedings of the 29th ACM International Conference on Information & Knowledge Management, 2020:1725-1734.
[8] 余辉,夏文蕾,黄炜,等.基于文本匹配-组态优化的两阶段技术供需匹配方法研究[J].情报学报,2024,43(3):261-273.
[9] 钟琪,王中卿,王红玲.基于孪生网络文本语义匹配的多文档摘要[J].中文信息学报,2024,38(5):107-116.
[10] 王佳睿,彭程,范敏.面向长文本的两阶段文本匹配模型TP-TM[J].计算机应用,2023,43(S1):33-38.
[11] 蔺志,李原,王庆林.基于BERT改进的文化活动事件论元抽取研究[J].中文信息学报,2022,36(12):115-122.
[12] 代翔,孙海春,牛硕,等.融合互注意力机制与BERT的中文问答匹配技术研究[J].信息网络安全,2021,21(12):102-108.
[13] 夏飞,陈帅琦,华珉,等.基于改进BERT的电力领域中文分词方法[J].计算机应用,2023,43(12):3711-3718.
[14] CHURSIN A A, DUBINA N I, CARAYANNIS G E, et al.Technological platforms as a tool for creating radical innovations[J].Journal of the Knowledge Economy,2021,13(1):1-12.
[15] 林德萍,汪红娟.基于BERT和RNN的新闻文本分类对比[J].北京印刷学院报,2021,29(11):156-162.
[16] 张靖宜,贺光辉,代洲,等.融入BERT的企业年报命名实体识别方法[J].上海交通大学学报,2021,55(2):117-123.
[17] 罗峦,夏骄雄.融合ERNIE与改进Transformer的中文NER模型[J].计算机技术与发展,2022,32(10):120-125.
[18] 朱海东,郑虹,侯秀萍.基于MC_E_ImprovDPC_B_SV模型的中文情感分类方法[J].长春工业大学学报,2022,43(1):72-80.
[19] CUI W, LI L, CHEN G. Market-value oriented or technology-value oriented?:Location impacts of industry-university-research (IUR) cooperation bases on innovation performance[J]. Technology in Society, 2022,70:102025.
[20] TAEGYEOM L, KYUNGSEOP S. Performance Comparison of Natural Language Processing Model Based on Deep Neural Networks[J].The Journal of Korean Institute of Communications and Information Sciences,2019,44(7):1344-1350.
[21] 周自强,王文浩,李政翰,等.文本相似度算法在核电厂备件管理上的应用[J].河北水利电力学院学报,2022,32(1):67-71.
[22] 严尹彤,于璐,王泰彦,等.基于Jump-SBERT的二进制代码相似性检测技术研究[J].计算机科学,2024,51(5):355-362
[23] LU X L, NI B. BERT-CNN: a Hierarchical Patent Classifier Based on a Pre-Trained Language Model[EB/OL]. (2019-11-03)[2025-01-01]. https://doi.org//10.48550/arxiv.1911.06241.
[24] 吴浩,潘善亮.基于BERT-RCNN的中文违规评论识别研究[J].中文信息学报,2022,36(1):92-103.
[25] HAN H W, KE Z W, NIE X Y, et al. Multimodal fusion with dual-attention based on textual double-embedding networks for rumor detection[J]. Applied Sciences, 2023,13(8):4886.

基于ERNIE的科技成果供需匹配模型研究

Research on the Supply and Demand Matching Model of Scientific and Technological Achievements Based on ERNIE

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 2

编辑推荐

Metrics

[1]	吉庆昌, 刘强, 栗梦媛, 孙东来, 王继超. 便携收费终端条件下改进型车牌识别算法研究[J]. 河北水利电力学院学报, 2025, 35(2): 35-41.
[2]	周自强, 王文浩, 李政翰, 姜久超. 文本相似度算法在核电厂备件管理上的应用[J]. 河北水利电力学院学报, 2022, 32(1): 67-71.