tailieunhanh - Biểu diễn và so sánh động hồ sơ cá nhân trong các mạng khoa học

Bài báo này giới thiệu một phương pháp tổng hợp hồ sơ mối quan tâm của các nhà khoa học thông qua quá trình tương tác của họ trên cộng đồng, và phương pháp so trùng các hồ sơ dựa trên các phân tích về mặt ngữ nghĩa. Các phương pháp này không cần sử dụng ontology, nhưng vẫn có khả năng thực hiện các so sánh liên quan đến ngữ nghĩa, dựa vào các phương pháp thống kê. | Science & Technology Development, Vol 14, 2011 DYNAMIC PROFILE REPRESENTATION AND MATCHING IN DISTRIBUTED SCIENTIFIC NETWORKS Pham Tran Vu University of Technology, VNU-HCM (Manuscript Received on Decmber 07th, 2010, Manuscript Revised April 21st, 2011) ABSTRACT: Finding people having similar interests in online community is an interesting but challenging problem. Especially, in distributed multidisciplinary research network, locating scientists who share common interests to collaborate and solve large scientific problems is becoming more important. This paper introduces a method for extracting and modeling scientists’ interest profiles from their day-to-day interactions and a method for semantically matching interest profiles based on latent semantic analysis. These methods exclude the necessity of having ontology for semantic matching of profiles, while still maintain the ability to reason about the semantic meaning of words. Keywords: Scientific network, semantic matching, profile representation. registration. This approach is simple, but 1. INTRODUCTION Finding people having similar research interests to initiate scientific collaboration is becoming important in distributed scientific communities. It is especially more essential in communities, where interdisciplinary research collaboration play an integral part. This is also suffers from a couple of limitations. Firstly, the users may not realize exactly their interests. Secondly, a user’s interests may change overtime. As the result, the initial profile registration will no longer be valid, unless it is updated regularly. a well-known research problem (commonly In recently years, different ways of known as expert finding or profile matching) in building user profiles dynamically has been information retrieval and has been studied by introduced. Instead of explicitly specified, many previous researches. profiles are implicitly extracted from different The problem of finding people .