Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
High level overview
AIDR is a stream processing application in Java implemented through a series of modules. The following diagram illustrates at a very high level the relationships between the different modules.
Operators of the AIDR platform operate AIDR through the Manager. The first thing operators do is to use the Collector to start collecting items. Those items are passed to the Tagger, which automatically annotates then. Then, the Output module keeps a small buffer of up to 1K items for viewing, and the Persister stores the items on disk so they can be downloaded.
The Tagger also samples a small number of items in the task buffer to get human-provided labels for them. Those labels are provided by the operator of the platform, through the Trainer, or can be provided by external annotators through the PyBossa Trainer.
An illustrated guide
An illustrated guide to data flows in AIDR is a non-technical presentation of data flows in AIDR through a visual metaphor including gnus, Maasai, and robots.