Authors: Arefeh Kazemi; Mohammad Ali Nematbakhsh; Mohammad Mehdi Keikha
Addresses: Computer Engineering Department, University of Isfahan, Hezarjerib St., Isfahan, Iran ' Computer Engineering Department, University of Isfahan, Hezarjerib St., Isfahan, Iran ' Computer Engineering Department, University of Isfahan, Hezarjerib St., Isfahan, Iran
Abstract: One of the most significant features of social networks is the connections made between new compatible individuals. Quite a few research projects have been done on making such connections more convenient and possible. Almost all of these systems interpret compatibility as similarity and consider only exact-matching of users' interests. This research proposes a different approach for finding 'new' compatible friends through social networks. We have proposed three new relations among users' interests. These are, firstly semantic similarity, secondly conceptual complement, and thirdly associative complement. We have used the first two relations in the current system and the third relation is left for future works. We chose 50 members of the LiveJournal and calculated the degree of compatibility between each pair. The results showing an average error of 0.21 which is acceptable in comparison with the previous exact-matching systems. In the latter, the average rate of error was 0.54.
Keywords: social networks; compatible persons; semantic relations; complementary relations; compatible users; compatible friends; social networking; semantic similarity.
International Journal of Social Network Mining, 2015 Vol.2 No.1, pp.1 - 18
Received: 10 Aug 2013
Accepted: 24 Mar 2014
Published online: 11 Jun 2015 *