I realize that the annotation manager is a long way out. However, I got to thinking about my company’s use case. While prodigy enables much faster model development with few people, sometimes you just need a lot of data to get the results you need.
Thus enter my idea. The annotation manager had add-ons for services like mechanical turk. It could format tasks for the service and merging results on one source from multiple workers (for quality).
This gives even small shops the ability to accumulate large data sets with out big infrastructure cost. I wouldn’t necessarily expect this kind of feature to be free either.