本文主要阐述了基于DSpace构建传统蒙古文学科机构知识库的难点以及解决的技术路线，包括蒙古文数字资料的采集、存储、检索以及显示等。针对蒙古文的构词和语法等方面的特点，对开源搜索引擎Lucene进行改进——采用B树管理Term、简化了特征词权值的计算、采用EC方法确定了蒙古文停用词表，实现了基于Lucene的蒙古文检索。

The paper focuses on the difficulties of building the Institutional Repository for Traditional Mongolian literature with DSpace and points out a technical route to solve the problem，including Mongolian digital data acquisition, storage, retrieval and display. Through analyzing the Mongolian characteristics in morphology and syntax, we improve on Lucene to implement the Mongolian search, including managing Term with  B tree，simplifying the calculation of Term weight，and determining the Mongolian stop list.