DBCrowd 2013 - First VLDB Workshop on Databases and Crowdsourcing

Crowdsourcing has received more and more research attention lately for its potential in solving problems that are currently outside the capabilities of computers. Among the tasks that are known to be hard for computers to solve ? mainly because of their underlying ambiguity ? are: curation of content, entity resolution, clustering, question answering, or content rating. Unsurprisingly, efficiently and effectively employing the help of humans in the computation of the above tasks has the potential to improve the quality of the results and thus play an important role in the design of better products and applications. The interest of using crowds for computational tasks is compounded by the emergence of several crowd computing platforms and the explosion in the number of users on the Internet in the last few years, some of which are glad to lend a hand to solve such problems. To give some examples, Amazon Mechanical Turk helps task creators connect to and incentivize people in order to solve small tasks. The reCAPTCHA system is useful both as protection against spam for website owners yet is also a resource for building good training datasets for image recognition. Games like the ESP game or, more recently, Google’s mobile game Ingress generate useful data for their creators through the gameplay actions of the players. Finally, it can be argued that collaborative Web applications are a form of crowdsourcing, in the realm of online encyclopedias (e.g., Wikipedia), social question-answering (e.g, Aardvark, Yahoo! Answers, StackExchange), or even online microcredit sites (e.g., Kiva).
In parallel, the explosion of data available for analysis and processing (a consequence of the emergence of so-called “Web 2.0” applications and the numbers of people using them) constitutes an immense challenge for current database systems and has garnered considerable research interest for industry and academia equally. New distributed computing paradigms, e.g., MapReduce, help with processing such data. Yet, even large scale processing cannot solve problems that are hard to solve by computers. Indeed, the combination of modern, large-scale, data algorithms and paradigms with the help of crowds might arguably be one of the more promising directions for solving today’s data management challenges.
However, challenging research questions still lie at this intersection of databases and crowdsourcing. In addition to the classical computational problems of processing large-scale data, the research questions around incentivizing users to perform tasks, evaluating the truthfulness and reputation of these users, and declarative languages that use crowd task in their design, there is also a significant potential to exploit for systems able to combine the best of both worlds. Current crowdsourcing-enabled data management systems (e.g, Qurk, CrowdDB, CrowdForge, CrowdFlower, and Deco) are still in their infancy, but represent an important first step towards fully integrating database management systems and the power of the crowd.
This workshop aims to provide a forum for researchers and practitioners to disseminate and discuss new results and visionary ideas with the potential to advance the state-of-the-art in the research at the intersection of databases and crowdsourcing.
Topics
The topics of interest of the workshop include, but are not limited to:
Analyzing crowdsourcing worker behavior
Cleaning data obtained from crowdsourcing
Crowd database systems
Data analytics with crowds
Declarative languages for crowdsourcing
Effective tagging in crowdsourcing
Entity-resolution with crowds
Incentive allocation for crowdsourcing
Labeling objects with crowds
Probabilistic modeling of the crowd
Scalable crowdsourcing systems
Sorts and joins with crowds
Utilizing social network information to support crowdsourcing
Voting strategies for crowdsourcing data