CrowdDb is a hybrid computer-human database that processes queries using crowd-sourcing if the data in the database is incomplete or if the query is asking about something that machines are notoriously bad at doing (ex. entity resolution, matching photos).
We've published two papers so far:
- VLDB '11 CrowdDb: Query Processing with the VLDB Crowd - VLDB Best Demo Award '11
- SIGMOD '11 CrowdDb: Answering Queries with Crowdsourcing
If you're interested in finding out more information, we'll be releasing a website soon, with screenshots and screencasts. You can also follow @crowddb on Twitter for updates.