Social Query: A Query Routing System for Twitter
0 likes505 views
The document presents a social query routing system designed for Twitter, aiming to enhance the process of directing questions to suitable respondents within online social networks. It outlines the system's methodology, evaluation of its effectiveness through volunteer feedback, and plans for future improvements including a mobile app. The findings highlight a significant portion of unanswered questions on Twitter and propose a multi-criteria decision-making approach to improve response rates.
![Related Work & Background
• The differential of our research
– We are proposing a Query Routing to an OSN context
• Previous work usually focused on CQA context
• We are proposing a solution to a pre-existent and popular
context: Twitter
• Most part of questions asked on Twitter are not answered
(more than 80%) [Paul et al. 2012]
– We lead with the recommendation as multi-criteria
decision making problem
• Previous work usually apply probabilistic or Information
Retrieval-based models;
Cleyton Souza - ICIW 2013
5](https://image.slidesharecdn.com/socialqueryaqueryroutingsystemfortwitter-131113095319-phpapp02/85/Social-Query-A-Query-Routing-System-for-Twitter-5-320.jpg)
Social Query: A Query Routing System for Twitter
- 1. Social Query A Query Routing System for Twitter Cleyton Souza Jonathas Magalhães, Evandro Costa, and Joseana Fechine Laboratory of Artificial Intelligence – LIA Federal University of Campina Grande - UFCG
- 2. Introduction • Query Routing (QR) is the process of directing questions to appropriate responders – Community Question and Answering Services (CQA) – Online Social Networks (OSN) • We are proposing an Expertise Finding System to automatically routing questions on Social Networks Cleyton Souza - ICIW 2013 2
- 3. Introduction • Our goal is to present the Social Query System • How does it work? • How does the usual Q&A process is affected? • Talk about our preliminary results • Talk about our planning for the future Cleyton Souza - ICIW 2013 3
- 4. Agenda • • • • • • Introduction Related Work & Background Usual Q&A Social Query System: How it works Evaluation & Results Future Work Cleyton Souza - ICIW 2013 4
- 5. Related Work & Background • The differential of our research – We are proposing a Query Routing to an OSN context • Previous work usually focused on CQA context • We are proposing a solution to a pre-existent and popular context: Twitter • Most part of questions asked on Twitter are not answered (more than 80%) [Paul et al. 2012] – We lead with the recommendation as multi-criteria decision making problem • Previous work usually apply probabilistic or Information Retrieval-based models; Cleyton Souza - ICIW 2013 5
- 6. Usual Q&A on OSN • Sharing a public question Fig. 1: Sharing a Public Question Cleyton Souza - ICIW 2013 6
- 7. Q&A on OSN • Directing the question Fig. 2: Directing the Question Cleyton Souza - ICIW 2013 7
- 8. Q&A on OSN • Routing the question Fig. 3: Routing the Question Cleyton Souza - ICIW 2013 8
- 9. Social Query System • Works outside Twitter – Questioner’s Followers are Expert Candidates – Questions and Answers without size limitations Fig. 4: SocialCleyton Souza - ICIW 2013 Query System’s Homepage 9
- 10. “New Question” Page • Three text fields, two mandatory Fig. 5: “New Question” Page Cleyton Souza - ICIW 2013 10
- 11. “Recommendation List” Page • Questioner chooses who will “receive” the question Fig. 6: “Recommendation List” Page Cleyton Souza - ICIW 2013 11
- 12. Question’s Tweet • Questioner tweets the following message Fig. 7: Question’s Tweet Cleyton Souza - ICIW 2013 12
- 13. “New Answer” Page • Three options of answer Fig. 8: “New Answer” Page Cleyton Souza - ICIW 2013 13
- 14. “I don’t Know” & “I know Someone” Fig. 9: “I don’t know” Tweet Fig. 10: “I know someone” Tweet Cleyton Souza - ICIW 2013 14
- 15. I want answer • When the expert clicks on the “I want answer” button Fig. 11: “I want answer” Page Cleyton Souza - ICIW 2013 15
- 16. Tweeting about the Answer Fig. 12: “I just answered” Tweet Cleyton Souza - ICIW 2013 16
- 17. “New Evaluation” Page Fig. 12: “New Evaluation” `Page Cleyton Souza - ICIW 2013 17
- 18. How does it work? • (1) The questioner accesses our System and (2) informs his question; • (3) The System recommends potential responders and (4) the questioner chooses to whom direct the question; • (5) Those chosen access our System, (6) answers the question, (7) and informs the questioner about his answer; • (8) The questioner access our System, (9) see the answer, and (10) evaluates it. Cleyton Souza - ICIW 2013 18
- 19. Evaluation • Nine Volunteers evaluated ten recommendations for a couple of questions a) b) Looking for a new band to listen during weekend, does anyone have an indication? Going to the movie theater after years LOL. What is the best movie in theaters? • Each recommendation was labeled as good (relevance 1), neutral (relevance 0) and bad (relevance 0). • These labels reflect the opinion of the volunteers about the recommendation Cleyton Souza - ICIW 2013 19
- 20. Results Cases Best case for Question “a” Amount of Followers 192 % of good %of bad nDCG 50% 10% 0.63 Worst case for Question “a” 129 30% 60% 0.25 Best case for Question “b” 121 60% 0% 0.74 Worst case for Question “b” 68 30% 0% 0.18 Average for Question “a” 110 41% 28% 0.41 Average for Question “b” 110 50% 20% 0.51 Cleyton Souza - ICIW 2013 20
- 21. Future Work • Where are we? • Mobile App • Volunteer’s feedback – Follow Back Filter – Thesaurus • Real case study Cleyton Souza - ICIW 2013 21
- 22. Social Query System A System for Query Routing on Twitter Thank You! Lia TIPS Laboratory of Artificial Intelligence Group of Intelligent Social and Customizable Technologies Cleyton Souza Jonathas Magalhães, Evandro Costa, and Joseana Fechine Laboratory of Artificial Intelligence – LIA Federal University of Campina Grande - UFCG