Optimizing the Micro-Tasking Workflow and Exploring its Usage Potential Within Geospatial Data

Abstract

Micro-tasking is the process of dividing a large task into smaller parts, making the task easier to distribute and lowering its complexity. This process is used commercially to solve tasks such as labeling images, translating text, and digitizing old books, using platforms such as Mechanical Turk. This thesis proposes to extend the usage of the micro-tasking method to involve geospatial tasks. While the OpenStreetMap community has leveraged micro-tasking for building imports in the US, and for the Humanitarian OpenStreetMap Team, the use of micro-tasking for geospatial data has seen little interest from the research community. This thesis explores how to partition a geospatial import-task into smaller parts and also investigates whether if individuals with no geospatial experience manage to solve these tasks.

An online web experiment was developed and implemented to gather data about how individuals solve the geospatial import-tasks. Statistical analysis is conducted on the collected data to answer the research questions. Results show that background does not affect how well the participants solved the tasks and that fewer elements in each task will increase the quality but probably also increase the task completion time. The author concludes that it is possible to give geospatial micro-tasks to inexperienced individuals and if the quality of the task results is important, fewer elements will provide better quality. Results show a promising future of micro-tasking geospatial data, also outside the OSM community.