Hi all,
I wonder if I can use prodigy to annotate two tasks at once, as annotating NE and text classification or Relatedness score of a pair of sentences and also its inference label.
Thanks!
Hi! Part of Prodigy's core philosophy is to focus on single tasks at a time, instead of building workflows where each decision consists of multiple steps the annotator has to perform. This gives you more consistent data, reduces the potential for human error, makes it easier to iterate on the label scheme and use semi-automated workflows to label more efficiently, with different objectives that match the task type.
If you need to label text categories and named entities, there's not necessarily an advantage in labelling everything at the same time. You'll be training two separate components, and typically want to run separate experiments. You may even use slightly different datasets, or try out multiple different label schemes until you find the one that works best. That said, you can of course always build custom interfaces that collect several parts of information at once, if you want to.