| 杨 旸1 蒋丽琼1 俞 旻1 王海洋1 彭 睿1 甘克勤2 张 勇2.基于两阶段排序的部分领域标准文献搜索优化方法研究[J].中国科技资源导刊,2024,(6):56~64 |
| 基于两阶段排序的部分领域标准文献搜索优化方法研究 |
| Research on Standard Literature Search Optimization Method Based on Two Stage Ranking |
| 投稿时间：2024-06-27 |
| DOI： |
| 中文关键词: 标准知识检索;深度学习;两阶段排序;深度语义 |
| 英文关键词: standard knowledge retrieval, deep learning, two-stage ranking, deep semantic information |
| 基金项目:中国标准化研究院基本科研业务项目“数字标准馆标准体系构建及关键标准研制与应用”(252023Y-10411) |
| 作者 | 单位 | | 杨 旸1 蒋丽琼1 俞 旻1 王海洋1 彭 睿1 甘克勤2 张 勇2 | (1. 国家管网集团西南管道有限责任公司，四川成都 610041
2. 中国标准化研究院，北京 100191) |
|
| 摘要点击次数: 20 |
| 全文下载次数: 26 |
| 中文摘要: |
| 当前的标准知识检索主要依赖较为浅显的语义信息处理方法，侧重于词语的表面共现特性，要求查询结果严格包含用户输入的关键词，而未能充分挖掘用户查询意图与标准文献深层次知识之间的语义关联，比如同义词、近义词及上下位词，从而影响了检索效果。基于此，创新性地提出基于传统词语信息和深度学习的基础召回、利用深度学习进行精准的检索结果排序两阶段排序相结合的策略。深度学习能够深入解析查询与文献间的深度语义信息，有效提升检索系统的智能化水平与检索效率。 |
| 英文摘要: |
| The current standard knowledge retrieval mainly relies on relatively simple semantic information processing methods, focusing on the surface co-occurrence characteristics of words, requiring the query results to strictly include the keywords input by the user, and failing to fully explore the semantic association between the user’s query intention and the deep level knowledge of the standard literature, such as synonyms, synonyms, and super-resolution words, thereby limiting the retrieval effect. The purpose of this article is to innovatively propose a strategy that combines traditional word information and deep learning based recall with precise retrieval result sorting through deep learning. Deep learning can deeply analyze the deep semantic information between queries and literature, effectively improving the intelligence level and retrieval efficiency of retrieval systems. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
