资源系统建设部知识库

【目的/意义】相关性检索机制作为知识搜索引擎最核心的部分之一,对提升机构知识库精准知识服务能力和用户满意度具有重要意义。本研究旨在为机构知识库优化相关性检索机制提供技术路线支撑。【方法/过程】首先,系统化梳理相关性检索研究的背景及进展,对关键的相关性计算模型及工具进行对比分析与述评。其次,以中国科学院机构知识库CSpace为试验环境,以开源引擎Apache Solr为辅助工具及TF/IDF、BM25算法等为模型基础,对机构知识库相关性检索机制进行了研究设计与试验。【结果/结论】Solr可以与BM25等多种相似度计算模型有效结合,为用户提供多维度的可配置化的相关性计算服务。文章未能对基于机器学习排序算法的相关性检索机制进行深入研究。

【Purpose/significance】As one of the core parts of knowledge search engine, the relevance search mechanism is of great significance to improve the accurate knowledge service capability and user satisfaction of the institutional repository. This study aims to provide technical support for the optimization of the relevance search mechanism for peer institutional repository construction.【Method/process】Firstly, systematically combing the background and progress of research on relevance retrieval, and comparing and reviewing the key model algorithms and tools. Secondly, The Chinese Academy of Sciences institutional repository platform CSpace is used as the online application environment, and the open source analysis engine Apache Solr is used as the tool, TF/IDF algorithm and BM25 algorithm as the model basis.  The article studies and designs the relevance search mechanism of institutional repository. 【Result/conclusion】Solr can be effectively combined with a variety of similarity calculation model like BM25. It can provide users with multi-dimensional configurable correlation computing services.