One of the challenges in CS projects is to feed our projects with new tasks. ACTION makes a platform to execute workflows available to pilots. Workflows are sequences of tasks executed to complete a process. In the context of CS projects, these workflows can be used to feed crowdsourcing platforms with new data, or publish their results in repositories. These workflows can be executed automatically without human intervention, making your project more sustainable.
One example can be found in Street Spectra. This project has a set of workflows that i) download data from Epicollect5, ii) transform this data into tasks and iiu) upload these tasks to Zooniverse. One of the difficulties found during this period of execution has been the adaptation of the system in the face of demand peaks. It is usual in CS projects that the demand of tasks is increased after communication campaigns. Normally, the workflows are designed to periodically upload tasks, but in this case, this approach is not the correct because it does not adapt to the demand of tasks.Thus, the system has been tune-fined to make adaptive workflows. In this case, these workflows explore periodically the number of tasks available and create new tasks based on a threshold established by the project coordinator.
The number of classifications per day varies drastically at Zooniverse due to communication campaigns.
Group of students classifying images