Currently,research work is increasingly dependent on the Internet,Researchers do their academic exchanges,scientific achievements sharing,academic retrieval and other work on line more frequently.Meantime,all kinds of research documentation system are developed.The expert profile system,as a subsystem of scientific research documentation systems,builds a connection between experts and experts,experts and institutions,experts and scientific research fields,etc.And thus has a good practical value and profound realistic significance.In this paper,we do our technology research based on the building of expert profile system.Mainly focus on metadata extraction and data mining work,we studied several key issues like PDF-based metadata extraction,research area mining,expert relationship mining,research papers clustering and tags extraction.Firstly,we studied PDF-based metadata extraction.Different from other classification methods,we adopt paragraph-based classification model instead of line-based,and design a mixed metadata extraction model which combines support vector machine and rules.Secondly,we studied relationships among the entities in the expert profile system,designed an algorithm based on frequent items mining and Union-Find methods to extract research areas.By using these research areas,we improved the K-Means algorithm to cluster the research papers.We also studied the extraction of system tags.We combine global and local lexical relations,make an improvement to typical Text-Rank model,and propose a new key words&key phrase Text-Rank model based on fused weighted vertices and edges,then get expert tags by adding co-author relations.Thirdly,on the basis of tags,we also studied the expert aggregation model.We use frequent items mining algorithm to aggregate co-authors.At last,we designed and implemented an expert profile system using the research achievements by this paper.We also use visual chart component to show those mining data and guarantee a good visuals in this expert profile system.The results show,a full-depth technology study has important implications for the usability,scalability,performance and display effects of expert profile systems. |