Background

Textonic originated at New York University‘s Interactive Telecommunications Program as one of the final projects in Clay Shirky‘s Spring 2009 class Design for UNICEF. The class was focused on iterative project-based group work and was conducted in close coordination with UNICEF Innovation. The project was conceived and developed by Steven Lehrburger, Thomas Robertson, Lina Maria Giraldo, Yaminie Patodia, and Amanda Syarfuan.

Mechanical Turk is a crowdsourcing marketplace for programmatically co-ordinating the use of human intelligence to perform tasks which computers are unable to do. An example of such a task would be determining if a particular image contained a flower – a person can do this easily, yet it would be very difficult to write a software program that would acheive the same results. Requesters can create Human Intelligence Tasks, or HITs, either manually or using one of Amazon’s software interfaces for MTurk. Those tasks can then be completed by any one of thousands of MTurk workers, and that worker will receive a small payment upon approval of his or her work. Amazon refers to MTurk as “artificial artificial intelligence” and the service is named after a late 18th century chess-playing machine (more on Wikipedia).

Textonic was originally imagined as a plug-in to UNICEF’s RapidSMS software package. RapidSMS is designed to leverage the increasing penetration of mobile technologies in the developing world to improve UNICEF”s data gathering programs. RapidSMS allows UNICEF field workers to send data as specially-formatted SMS messages from their standard mobile phones directly into a centralized database. The technology has been successfully deployed in Ethiopia to track distribution of Plumpy’nut high-protein food supplies and in Malawi to monitor child malnutrition over time.

The data sent via text message to the RapidSMS server needs to be in a format that the computer can understand and extract numerical information from. Since the people are doing the data entry are often working under intense conditions and constrained to typing on small devices, it is not uncommon for the RapidSMS server to receive messages that were intended as data entry yet that the parser was unable to handle. Textonic was to receive these messages, send them to MTurk along with the proper format for the data, receive a message in a corrected format from the HIT workers, and then send that message back into RapidSMS for processing as usual. As a result, more useful data would be received by RapidSMS, and UNICEF’s projects would benefit, becoming more efficient and scalable at minimal expense.

As the project progressed it became clear that MTurk could be more useful as a means to efficiently classify/categorize other types of text-based input, and more information about it’s current focus can be found on the Description page.