BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Quizz: Targeted Crowdsourcing with a Billion (Potential) Users
DTSTART:20150623T150000
DTEND:20150623T160000
DTSTAMP:20261005T021951Z
UID:441c3b7c4220913319df623fdc0282b135a1559d39545e24c05306c5
CATEGORIES:Conferences - Seminars
DESCRIPTION:Panos Ipeirotis is an Associate Professor and George A. Kellne
 r\nFaculty Fellow at the Department of Information\, Operations\, and\nMan
 agement Sciences at Leonard N. Stern School of Business of New York\nUnive
 rsity\, and he is also a visiting scientist at Google. His recent\nresearc
 h interests focus on crowdsourcing and on mining user-generated\ncontent o
 n the Internet. He received his Ph.D. degree in Computer\nScience from Col
 umbia University in 2004. He has received six “Best\nPaper” awards (IE
 EE ICDE 2005\, ACM SIGMOD 2006\, WWW 2011\, ICIS 2012\,\nHCOMP 2014\, Mana
 gement Science 2011-14)\, three “Best Paper Runner Up”\nawards (JCDL 2
 002\, ACM KDD 2008\, INFORMS Data Mining Contest 2014)\, and is also a re
 cipient of a CAREER award from the National Science\nFoundation and of sev
 eral other grants.\nWe describe Quizz\, a gamified crowdsourcing system th
 at simultaneously\nassesses the knowledge of users and acquires new knowle
 dge from them.\nQuizz operates by asking users to complete short quizzes o
 n specific\ntopics\; as a user answers the quiz questions\, Quizz estimate
 s the\nuser’s competence. To acquire new knowledge\, Quizz also incorpor
 ates\nquestions for which we do not have a known answer\; the answers give
 n\nby competent users provide useful signals for selecting the correct\nan
 swers for these questions. Quizz actively tries to identify\nknowledgeable
  users on the Internet by running advertising campaigns\,\neffectively lev
 eraging “for free” the targeting capabilities of\nexisting\, publicly 
 available\, ad placement services. Quizz quantifies\nthe contributions of 
 the users using information theory and sends\nfeedback to the advertising 
 system about each user. The feedback\nallows the ad targeting mechanism to
  further optimize ad placement.\nOur experiments\, which involve over ten 
 thousand users\, confirm that\nwe can crowdsource knowledge curation for n
 iche and specialized\ntopics\, as the advertising network can automaticall
 y identify users\nwith the desired expertise and interest in the given top
 ic. We present\ncontrolled experiments that examine the effect of various 
 incentive\nmechanisms\, highlighting the need for having short-term reward
 s as\ngoals\, which incentivize the users to contribute. Finally\, our\nco
 st-quality analysis indicates that the cost of our approach is below\nthat
  of hiring workers through paid-crowdsourcing platforms\, while\noffering 
 the additional advantage of giving access to billions of\npotential users 
 all over the planet\, and being able to reach users\nwith specialized expe
 rtise that is not typically available through\nexisting labor marketplaces
 .
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR