improving a recommender system through integration of user profiles: a semantic approach

Download Improving a Recommender System Through Integration of User Profiles: a Semantic Approach

Post on 09-Jun-2015

476 views

Category:

Technology

2 download

Embed Size (px)

DESCRIPTION

The users are present in multiple social networks/virtual communities and each one can be considered as a source of information about this user. In face to this question it is important a mechanism to integrate the user profiles. Through the integration of user profiles it is possible identifier more accurately their interests analyzing other data sources that they are present, possible reducing the cold-start problem. In this context, we present a semantic approach to help integrate data from multiple sources, for the construction and maintenance of user profiles that will be used to improve the quality of a recommender system. To integrate data from multiple sources, we defined a heuristic that quantifies the importance of each data source for a given user. To validate our approach, we perform a case study, where the solution was coupled into a recommender system of papers focused in Software Engineering domain. The user profiles were built extracting their information from the Brazilian Curriculum Vitae database named CV-Lattes, an academic platform, and Linkedin, a business network. We compared the quality of the recommendation based on the profiles integrated and non-integrated. The results show the superior quality of the recommendation based on integrated profile.

TRANSCRIPT

  • 1. IntroductionOur PurposeEvaluation and ResultsConclusion Improving a Recommender System ThroughIntegration of User Proles: a Semantic Approach Jonathas Magalhes, Cleyton Souza, Priscylla Silva, aEvandro Costa and Joseana Fechine TIPS GroupFederal University of Campina Grande, BrazilFederal University of Alagoas, BrazilJ. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 1

2. IntroductionOur PurposeEvaluation and ResultsConclusionIntroductionJ. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 2 3. IntroductionOur PurposeEvaluation and ResultsConclusionIntroductionOur purposeJ. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 3 4. IntroductionOur PurposeEvaluation and ResultsConclusionIntroductionQuestions: Q1 : Will the proles integration improve the quality of recommendation of a recommender system? Q2 :Will the proles integration reduce the cold-start problem in a recommender system?J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 4 5. IntroductionOur PurposeEvaluation and ResultsConclusionThe User Proles Integrations s |S|sspu = integration(pu1 , pu2 , ..., pu ) =pu au , (1) sS swhere auj : Represents the importance of the data source sj to the integrated prole; Is computed by the activity of the user u in the source data sj :|Iu,s |1 s tnow t|Iu,s | +j=1 (tj+1 tj )au = .(2)|Iu,s | + 1J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 20125 6. IntroductionOur PurposeEvaluation and ResultsConclusionEvaluation The case study focused in the Software Engineering Domain, using an domain ontology; The terms of ontology were learn using 100 papers for each concept; In this case study participated ten researches, where ve were new users; Their proles were built using three dierent strategies: Using the Lattes CV ; Using the Linkedin; The integration of both.J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 6 7. IntroductionOur PurposeEvaluation and ResultsConclusionEvaluation We coupled the three dierent strategies in a Recommender System of papers; The papers (40,855) were loaded from the digital library CiteerSeeX; We recommended 15 papers, ve for each strategy; The users evaluated the recommended papers in a scale from 0 to 4; We compare the strategies using the metric Normalized Discounted Cumulative Gain (nDCG).J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 7 8. IntroductionOur PurposeEvaluation and ResultsConclusionResultsWith all users:Figure: Comparing the three strategies to built the user prole.J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 8 9. IntroductionOur PurposeEvaluation and ResultsConclusionEvaluationWith all users: t1 with Ha,1,1 : integrated is greater than lattes t2 with Ha,1,2 : integrated is greater than linkedinTable: Comparison between the three strategies using the Students t-test( = 0.05) with all users. T-testAlternativep-value Meaning Hypothesis t1 Ha,1,10.03152 Ha,1,1 accepted t2 Ha,1,20.02903 Ha,1,2 acceptedThe integrated prole improved the quality of the recommendersystem.J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine a SRS UMAP 20129 10. IntroductionOur PurposeEvaluation and ResultsConclusionEvaluationWith new users: t3 with Ha,2,1 : integrated is greater than lattes t4 with Ha,2,2 : integrated is greater than linkedinTable: Comparison between the three strategies using the Students t-test( = 0.05) with the new users. T-testAlternativep-value Meaning Hypothesis t3 Ha,2,10.03485 Ha,2,1 accepted t4 Ha,2,20.04208 Ha,2,2 acceptedThe integrated prole reduced the cold-start problem.J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine a SRS UMAP 201210 11. IntroductionOur PurposeEvaluation and ResultsConclusionConclusion and Future Work We conrm the superiority of the integrated prole; But the results can be better. Planning and execution of an experiment with: More volunteers; More data sources. The recommender system as support to learners in an Intelligent Tutoring System.J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 11 12. IntroductionOur PurposeEvaluation and ResultsConclusionThanks!!J. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 12 13. IntroductionOur PurposeEvaluation and ResultsConclusionFor more information visit the TIPS Group: http://tip.ic.ufal.br/site/You are welcome to Macei!oJ. Magalhes, C. Souza, P. Silva, E. Costa and J. Fechine aSRS UMAP 2012 13