Crowdsourcing Toolkit for Low-resource Communities

This repository provides code for a crowdsourcing platform that is specifically designed to be inclusive of users from low-resource communities. There are two barriers to inclusion: language and connectivity. Most users in rural communities in India do not speak/understand English, and many of them do not have constant and high-speed data connectivity. This platform address these challenges by 1) providing a fast mechanism to add support for different languages and 2) using a two-tier server architecture to enable people without any data connectivity to still participate on the platform.

Overview of the Platform Architecture

A typical crowdsourcing platform contains two components: a server running in the cloud and a web/mobile client running in the edge. Work requesters register with the server and submit new tasks. Workers register directly with the server to receive new tasks. The server assigns tasks to workers based on the type of task and the skills of the worker. Unfortunately, this setup requires workers to have (constant) access to internet.

To break this assumption, this platform uses a two-tier server architecture. The main server runs in the cloud and interfaces with work requesters to receive new tasks. The second component, the "box" server can be deployed on a device in the field and acts a local crowdsourcing server for the specific region. Workers with a smartphone can directly interact with the box server to receive tasks and submit their responses. The platform assumes intermittent connectivity between the box server and the main server during which time they exchange information. This platform architecture enables two benefits.

1. The box server can still be run as a cloud instance. As a result, if users in a region have good connectivity, one can setup a virtual instance of the box server for the users and they can participate in the platform seamlessly.

2. If users in a region do not have any connectivity, then the box server can be run on a physical device. We have currently experimented running the box server on a Raspberry Pi with a 4G dongle. (The codebase needs to be tweaked a little to support this setting. We will work on seamless support for this mode in subsequent releases.)

Code Organization

The code is split between two folders: server, and client.

The server folder implements the following three components: backend is the main server built on top of nodejs, frontend is a react webapp that allows admins and work requestors to interact with the main server, and box is the box server built on top of nodejs. The remaining folders core, common, and utils contain various modules that are shared between these three components of the server.

The client folder contains the Android app that allows workers to interact with the platform.

Setup Instructions

Please see INSTALL.md for installation and setup instructions.

License

We have released this code under the MIT license. Please see LICENSE.txt for details.

Contributors

Bulk of the initial code base was developed at Microsoft with contributions from the following people.

• Vivek Seshadri
• Danish Goel
• Pallav Karya
• Mrinal Das
• Anurag Shukla

Currently, the code is also being developed by Navana Tech with contributions from the following people.

• Aditya Wasan (https://github.com/skrilltrax)
• Amogh Lele (https://github.com/sphericalkat)

Microsoft Open Source Code of Conduct

This project has adopted the Microsoft Open Source Code of Conduct. Please visit Code of Conduct for more information or email opencode@microsoft.com with your queries.