Presentation on theme: "Jorge Gaete-Villegas, Dongman Lee, Meeyoung Cha, In-Young Ko Korean Advance Institute of Science and Technology Department of computer science TraMSNet."— Presentation transcript:
Jorge Gaete-Villegas, Dongman Lee, Meeyoung Cha, In-Young Ko Korean Advance Institute of Science and Technology Department of computer science TraMSNet -A Mobile Social Network Application for Tourism- 4th International Workshop on Location Based Social Networks Pittsburgh, Pennsilvania, USA 8 th September 2012
Presentation Outline 1) Introduction Motivation, hypothesis & research question 2) Related Work Tourist, Existing solutions & Matching algorithms 3) Approach Ranking Function 4) Evaluation Methodology, results & analysis 5) Results Findings 6) Final words Conclusions, future work 2:11
Introduction _motivation 3:11 User Place constraints certain activities: Some goals are more likely Similar users link by a mobile social network.Users share a location and purpose.Different profiles.Possible synergies Is it meaningful to always match co-located users based on their similarities? How does the location affect (or should affect) the matching process?
Introduction _Hypothesis 4:11 What is important for users? Online questionnaire, open answer: Top 5 concerns when looking for a travel partner" Sample Female... 9 Male....... 11 Gender Under 20.....3 20-25.........11 26-30...........4 31-30...........2 America...9 Europe.....5 Africa........4 Asia..........2 Where? Results ConcernsAnswersExamplesSimilarityComplementarity Language20Spoken, NativeYes Age/Gender11Age, GenderYesNo Experience in location10In location, similarNoYes Personality10Shy, funnyYesNo Nationality9Nationality, LocalYes Knowledge background10Major, hobbiesYes Hypothesis In a LBSN applications focused on tourism, matching users in terms of their similarities does not fulfill the needs of tourists
Related Work _tourism domain 6:11 Basically Cyber tour guides: Such as Cyberguide, GUIDE, LoL@, Deep Map, SmartKom, REAL, TellMaris, CRUMPET Leveraging Location information and Geo referred information To provide onsite information, recommendationss based on similar users, tours Commercial tourism apps Users Matching Similarities between users to generate a match Leveraging Profile information : User similarities Location history: User similar behaviors Social network analysis
Approach _Complementary skills 7:11 R ranks user j to user i user similarity Users Afinity given the place l Done by matching users characteristics Similarity between j and complementary user to i complementary user to i in the given location If i(n) and j(n) are numeric values If i(n) and j(n) are numeric range If i(n) and j(n) are lists of string
Evaluation _Design 8:11 What to evaluate? Focus on evaluating the recommendation function: It is the contribution and differentiation from existing solutions i) Analysis: Users perception on the correctness of a suggestion ii) Opinion when the user is exposed to both suggestions under the same circumstance Why survey? i) Analysis: Users perception on the correctness of a suggestion ii) Opinion when the user is exposed to both suggestions under the same circumstance Survey Mechanism Procedure 1. On-line questionnaire asking for his/her profile information 2. After completion, a fictional scenario is presented to the user 3. After reading, simultaneously two list of users are displayed. Lists showing user profile 4. Finally, the surveyed will be asked: "Which one of the lists would you preferred to received in the given situation.
Results 10:11 Homophily and Complementarism (α,β) Local vs Overseas touristsOverall responses OverseasLocalTotal% (cumulative) (0.0; 1.0) 30314.29 (0.25; 0.75) 11223.81 (0,5; 0,5) 60652.38 (0.75; 0.25) 41576.19 (1.0; 0.0) 235100 Total16521---- There is no difference between the 3 mayor preffered options Overseas prefer combined model. Locals prefer homophilly
To conclude 11:11 Conclusions Users want to be matched with others for multiple reasons Out of the exploratory survey, at least two Homophilly is not always enough Its appropiatness depends on the relationship between users and locations Our approach gives a general model for including similarities and differences between users Our ranking algorithm includes both drivers for matching, therefore is a general form of the homophilly based matching algorithms
Questions? 11:11 Thank You! ! Muchas gracias!
Jorge Gaete-Villegas, Dongman Lee, Meeyoung Cha, In-Young Ko Korean Advance Institute of Science and Technology Department of computer science TraMSNet -A Mobile Social Network application for tourism- 4th International Workshop on Location Based Social Networks Pittsburgh, Pennsilvania, USA 8 th September 2012