SQL Turk
I worked with Bogdan Alexe, a PhD student at UCSC on a project in a cloud computing course called “SQLTurk.” Bogdan and I looked into using Amazon’s Mechanical Turk as a platform for converting natural language queries into semantically correct SQL. To date, research has been rather stagnant in the attempt to do this conversion process, so we proposed using human processing units instead. In the end we came up with a set of workflows which allowed for query-writers on Mechanical Turk to either write a natural language query or vote on a group of semantically valid submissions from other workers. We were able to recover a rather good chunk of data which validated our approach of inserting humans as “mini-computers” in a completely automated form. Read the project for more info!