地球科学是数据密集型科学，信息检索是地学研究的必要步骤。文中针对目前主流的信息检索系统空间语义感知能力不足的问题，设计了顾及空间和语义的检索系统架构。在地学知识库的支持下，采用命名实体识别、实体消歧等自然语言处理技术，将非结构化文档空间化，采用WebGIS地理空间信息技术对查询过程与检索结果可视化。文中以石油勘探为例，在开源检索平台SOLR和空间数据库PostgreSQL的基础上实现了方法验证。经过系统测试和用户使用的验证，该系统提高了地球科学信息检索性能与用户体验。%Earth science is data-intensive science,and information retrieval is a necessary step in earth sciences. As the mainstream infor-mation retrieval system can identify space semantic weakly,the architecture taking into account spatial entities and thematic information is put forward. The spatialization of unstructured documents are implemented with natural language processing including named entity recog-nition and disambiguation with the help of geoscience knowledge base. The visualization method of the query and retrieved results is put forward by using of WebGIS. A demo system based on SOLR,an open source information retrieval platform and PostgreSQL,a spatial database,is implemented to verify the method. Practices show that the system improves the performance and user experience by data test and validation.
展开▼
