图书情报工作 ›› 2021, Vol. 65 ›› Issue (23): 70-78.DOI: 10.13266/j.issn.0252-3116.2021.23.008

• 情报研究 • 上一篇    下一篇

基于深度学习的网络科技信息情报价值计算方法研究

张敏1,2,3,4, 刘欢2,3, 丁良萍2,3, 范青5   

  1. 1. 中国科学院武汉文献情报中心 武汉 430071;
    2. 中国科学院文献情报中心 北京 100190;
    3. 中国科学院大学经济与管理学院图书情报与档案管理系 北京 100190;
    4. 科技大数据湖北省重点实验室 武汉 430071;
    5. 华中师范大学国家文化产业研究中心 武汉 430079
  • 收稿日期:2021-06-16 修回日期:2021-09-09 出版日期:2021-12-05 发布日期:2021-12-18
  • 作者简介:张敏,馆员,博士研究生,E-mail:zhangmin2012@mail.whlib.ac.cn;刘欢,博士研究生;丁良萍,博士研究生;范青,讲师,博士研究生。
  • 基金资助:
    本文系国家自然科学基金项目"基于CityGML的三维古建筑语义建模研究"(项目编号:41801295)和中国科学院文献情报能力建设专项项目"网络科技监测平台智能分析核心能力升级"(项目编号:Y9290906)研究成果之一。

Research on the Web Technology Information Value Calculation Method Based on Deep Learning

Zhang Min1,2,3,4, Liu Huan2,3, Ding Liangping2,3, Fan Qing5   

  1. 1. Wuhan Library, Chinese Academy of Sciences, Wuhan 430071;
    2. National Science Library, Chinese Academy of Sciences, Beijing 100190;
    3. Department of Library, Information and Archives Management, School of Economics and Management, University of Chinese Academy of Sciences, Beijing 100190;
    4. Hubei Key Laboratory of Big Data in Science and Technology, Wuhan 430071;
    5. National Cultural Industry Research Center, Central China Normal University, Wuhan 430079
  • Received:2021-06-16 Revised:2021-09-09 Online:2021-12-05 Published:2021-12-18

摘要: [目的/意义]针对当前科研人员无法从海量的网络科技信息中及时甄别有情报价值的情报内容的问题,建立一套综合性情报价值计算方法,从而对网络科技信息的情报价值进行计算判断,最终帮助科研人员快速而准确地发现有情报价值的网络科技信息。[方法/过程]综合考虑情报外部特征与文本语义内容特征,利用深度学习(预训练语言模型) BERT方法构建基于文本语义内容特征的情报价值计算模型,利用深度学习模型的预测输出完成打分,并结合基于情报外部特征的原始计算方法得到最终的综合评价得分。[结果/结论]实验结果显示,基于文本语义内容特征的情报价值计算模型可以对情报按照情报价值得分进行有效的星级区分,弥补了基于情报外部特征的原始计算模型中星级区分度差的问题,最终的综合评价结果表明本文提出的情报价值计算模型在实际应用中也能够很好地满足科研人员的需求。

关键词: 网络科技信息, 情报价值计算, 文本语义内容, BERT

Abstract: [Purpose/significance] In view of the problem that it's difficult for researchers to find valuable information from large amounts of scientific and technological information in the Web, this paper constructs a comprehensive calculation method for information value. It can calculate the information value of Web technology information and help researchers find Web technology information of information value quickly and accurately.[Method/process] Taking overall consideration of the external feature and textual semantic feature of the information, this paper used deep learning (pretrained language model) BERT to construct information value calculation model based on the textual semantic feature, used the predictive output of the deep learning model to complete the scoring, and combined the original calculation method of the external feature of the information to get the final information value score.[Result/conclusion] The experimental results show that the information value calculation model based on the textual semantic feature can rank the information to different levels according to their information value score, which makes up for the problem of poor star differentiation in the original calculation method only based on the external feature of the information. And the final comprehensive evaluation results show that the information value calculation model proposed in this paper can also meet the needs of researchers in the practical application.

Key words: Web technology information, information value calculation, textual semantics, BERT

中图分类号: