The Design, Realization And Research For A Campus-Objected Entity And Social Search Engine

Big data era demands a faster data mining for large scale of diversified data. The recall and precision rates of searching information about an organization in the whole network are still low, which leads to a tow learning efficiency and a hard searching possibility. Users expect an automatic tool which improves the efficiency to learn knowledge about this organization.This thesis mainly focuses on the design, implementation and key research of a campus objected search engine(short for COSE). The work includes:Firstly, the thesis designs the framework and functions of COSE based on our campus-BUPT. COSE will return query-related entity cards including teachers, curriculums, students and frequent asked questions (FAQ) and traditional unstructured web pages as welL Users can find tweets about BUPT on SNS, browse hot topics and find relationship paths between two persons. The thesis deeply analyzes the demand of internal searching for organization members, proposes a novel classification and feature for organization’s entities as well.Secondly, in the research of entity relevance, this thesis tests a few algorithms and creatively proposes algorithm based on both co-occurrence frequency and distance.Thirdly, the thesis also puts forward a novel model to detect organization’s entities based on Words Activation Force(short for WAF), which has a better performance than Stanford NER tool.Finally, the thesis introduces the design and implementation of the entire system and two key modules, shows the functions designed by the author.