Massive Data Storage And Full-text Search

With the rapid development of China's naval vessels, the development of weapon systems and logistics support generated a lot of technical information and project management documents. Because of poor management of these documents, resulting in duplication of information about equipment and a huge waste of human and material resources. Establish a secure, high availability, integrated management system of technical publications, which is to establish a sound mechanism for logistics support. In order to meet the actual needs for technical staff of naval vessels, the thesis made more in-depth and systematic study on integrated management system of technical publications, so as to provide various, more accurate information.This thesis presents mass data storage environment, analyzes the current storage mode, and puts focus on object-oriented storage technology. Storage mode of the object has good scalability, high performance, cross-platform and secures data sharing capabilities, which makes it an ideal mass data storage choice.Second, according to software engineering development processes and user needs, the thesis described in detail feasibility analysis and demand analysis of integrated management system for technical publications. On the basis, the thesis design outline of the logic architecture and physical architecture of the system, meanwhile, detailed design the logical structure of function modules. Finally, each module of the system was realized.In this thesis, based on the realization of the basic functions of the system, also full-text search of the system was optimized. These include:⑴For the full-text search technology, the thesis found the lack of existing technology about Chinese segmentation and made better the maximum matching algorithm;⑵With the base of inverted index, a full-text index model which is concerning incremental B+-Lists is adopted.⑶In order to improve both precision and recall rate in information retrieva, This thesis proposes a new query optimization method based on analysis of local classification and genetic algorithm. The thesis uses analysis of local classification which is query expansion method to expand the query, and then uses genetic algorithms to reweight the query vector which is expanded and does experimental verification on the effectiveness of the method. Experimental results show that our method is better than no query expansion and local context analysis. Because we have improved in many aspects.