| 陈博立,鲜国建,赵瑞雪,黄永文,李娇,曹雨晴,孙坦.科技文献问答式智能检索总体设计与关键技术探析[J].中国图书馆学报,2023,49(3):92~106 |
| 科技文献问答式智能检索总体设计与关键技术探析 |
| Overall Design and Key Technology of Q&A Style Intelligent Retrieval for Scientific and Technical Literature |
| 投稿时间：2022-08-12 修订日期：2023-04-25 |
| DOI： |
| 中文关键词: 检索方式无法承载用户检索需求中丰富的语义信息,也无法满足高效、精准、智能检索海量科技文献的要求。本文分析科技文献问答式智能检索的场景需求,提出设计问答式智能检索应当遵循通用性、模块化、可移植三项基本原则,设计总体技术方案,包括掌握问答语言特征、构建知识图谱、研究问答式智能检索交互技术三大步骤。在此基础上,从科技文献知识图谱构建、检索意图理解与识别、检索意图形式化转换、人机自然语言对话管理、检索结果呈现与交互等方面论述实现科技文献问答式智能检索需要重点突破的关键技术,并提出可行的技术迭代方案。本文提出的科技文献问答式智能检索,结合了科技文献知识图谱和自然语言处理等人工智能技术,更加智能精准地理解用户多维复杂的文献检索需求,为用户提供具有高度相关性的检索结果,较传统科技文献检索在用户输入交互、智能检索效率等方面具有优势。图6。表2。参考文献37。关键词 科技文献检索 智能问答 知识图谱 语义检索 任务型对话 |
| 英文关键词: Scientific and technical literature retrieval Intelligent question and answer Knowledge graph Semantic search Task oriented dialogue |
| 基金项目: |
|
| 摘要点击次数: 319 |
| 全文下载次数: 391 |
| 中文摘要: |
| 科技文献是人类记录、学习、传承知识的重要手段。大数据时代,传统的基于字符匹配的 |
| 英文摘要: |
Scientific and technical literature is the most important means for human beings to record,learn and inherit knowledge. In the era of big data,the traditional keyword search method based on string matching cannot carry the rich semantic information in users complex search requirements,which leads to challenges in achieving efficient and accurate tracking and discovery from massive scientific and technical literature.
Therefore,based on the research progress of scientific and technical literature retrieval services and natural language Q&A,this paper puts forward the concept of Q&A style intelligent retrieval for scientific and technical literature. Firstly,we analyze the users demand of Q&A style intelligent retrieval scenes for scientific and technical literature,including literature,author,fund,subject,institution,journal,time,etc. And it is proposed that generality,modularity,and portability are the three basic principles to be followed in the design of Q&A intelligent retrieval. Then,we give a three step technical solution for mastering the language characteristics of question and answering,designing knowledge graph,and researching the interactive technology of Q&A intelligent retrieval. After this,we discuss the critical technologies that need breakthroughs,including the construction of knowledge graph for scientific and technical literature,understanding and recognition of retrieval intent,formal conversion of retrieval intent,human machine natural language dialogue management,presentation and interaction of retrieval results,and a feasible technical iteration scheme is proposed,as well as details that need to be paid attention to in practice. This paper shows a prototype verification system constructed by our team according to the above path,achieving successful parsing of Chinese questions such as “please help me find the literature on white backed plant hoppers,published in 2017 and included in the Chinese core journal of Peking University,supported by the general project of the Natural Science Foundation of China in which Academician WAN Jianmin participated”,etc. At last,we make an outlook for its further optimization and improvement.
The Q&A style intelligent retrieval for scientific and technical literature proposed in this paper aims to combine scientific and technical literature knowledge graph and natural language processing,and other artificial intelligence technologies to more intelligently and accurately understand users multi dimensional complex literature retrieval demands,and to provide users with highly relevant retrieval results. Compared with traditional scientific and technical literature retrieval,it has advantages in user input interaction and intelligent retrieval efficiency. It can provide a reference for constructing a next generation high end exchange platform for scientific and technical literature and information. 6 figs. 2 tabs. 37 refs.
|
|
查看全文
查看/发表评论 下载PDF阅读器 |
|
|
|
