| 林 毅1, 2 张均胜1, 2 刘志辉1, 2 王唯滢3.技术评审专家遴选方法在颠覆性技术专家预判平台上的应用[J].中国科技资源导刊,2024,(2):54~62 |
| 技术评审专家遴选方法在颠覆性技术专家预判平台上的应用 |
| Application of Technical Review Expert Selection Method on Disruptive Technical Expert Prediction Platform |
| 投稿时间：2023-06-13 |
| DOI： |
| 中文关键词: 专家遴选；标签排序；特征融合；颠覆性技术 |
| 英文关键词: expert selection, label rank, feature fusion, disruptive technology |
| 基金项目:中国科学技术信息研究所创新基金青年项目“颠覆性技术预判专家画像构建及遴选方法研究”(QN2022-11)；中国科学技术信息研究所创新基金青年项目“战略性新兴产业集群中的技术创新主体竞合网络构建及测度方法研究”（QN2023-07）；中国科学技术信息研究所创新基金青年项目“基于技术特异性的产业链核心企业识别方法研究”（QN2023-06）。 |
| 作者 | 单位 | | 林 毅1, 2 张均胜1, 2 刘志辉1, 2 王唯滢3 | （1. 中国科学技术信息研究所，北京 100038；2. 富媒体数字出版内容组织与知识服务重点实验室，北京 100038；3. 中国科普研究所，北京 100081） |
|
| 摘要点击次数: 309 |
| 全文下载次数: 402 |
| 中文摘要: |
| 评审专家遴选是技术评审中的关键环节。鉴于颠覆性技术专家预判平台预判系统对时效性和智能型的要求，专家遴选对预判结果具有决定性影响。通过学术专长匹配和专业遴选来选择符合要求的专家，可以降低成本，提高推荐效率与准确度，完成颠覆性技术的预测任务。基于学术网络表示学习的方法既可以避免大量特征工程，又可以方便不同类型的特征进行融合。利用异质网络表示学习方法和标签排序的学术专长画像方法构建专家库，并使用融合专家综合评价指标特征的匹配方法对待预判的颠覆性技术和专家专长进行匹配，为专家遴选提供一份专业背景匹配的候选专家列表。这种方法在Academic Social Network 数据集上进行模拟实验。实验结果表明，这种方法能提升项目评审专家学术专长匹配，在加入综合指标特征后，专家的综合指标特征能有效地反馈到实验结果中，从而提高评审系统的时效性和智能性。 |
| 英文摘要: |
| The selection of review experts is a key step in technical reviews. Given the requirements for timeliness and intelligence of the predictive platform for disruptive technology experts, expert selection has a decisive impact on the prediction results. By matching academic expertise and conducting professional selection to select experts who meet the requirements, costs can be reduced while improving recommendation efficiency and accuracy, thus completing the prediction task of disruptive technologies. Based on academic network representation can not only avoid a lot of feature engineering, but also facilitate the fusion of different types of features. This paper proposes to build an expert database and use a matching method integrating the characteristics of experts’ comprehensive evaluation indicators to match the predicted subversive technology and expert expertise on the basis of heterogeneous network representation learning methods and precise profiling methods of academic expertise in label ordering, so as to provide a list of candidate experts related to professional background for expert selection. The method proposed in this paper has been tested on the Academic Social Network dataset, and the simulation results show that the method can improve the academic expertise matching of technology evaluation experts. After adding the comprehensive index feature, the expert’s comprehensive index feature can be effectively fed back into the experimental results, thus improving the timeliness and intelligence of the evaluation system. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
