Hi, I'd like to annotate text with POS tags, and then another more specific Features Tag on top of that. For example, "I eat apples", "eat" would get a [VERB] POS tag and the annotator could also add another layer of tags to the [VERB] tag in the same session, for instance [PRES][3][SING] (present tense, 3rd person, singular). Is double layer of annotation possible in the same session? Thanks!
Hi! At the moment, there's no feature for this out-of-the-box – although you could probably implement something yourself using custom JavaScript by listening to the prodigyspanselected event (fired whenever the user highlights a span) and then add form fields that write the data specific to that span to the underlying JSON object. I haven't tried this yet, but it should work.
That said, we'd typically recommend doing layered annotations like this in at least two steps and focusing on getting the top-level decision correct first, before doing more fine-grained annotation. It also lets you review the annotations from the first step, resolve conflicts and fix problems early (e.g. an annotator misunderstanding the objective). If you do it all in one and the top-level decision VERB is wrong, all other annotations related to it will be wrong, too, and the time spent on them will have been wasted. It also means that the annotator will have to keep switching focus between deciding on the POS tags and thinking about the specifics of all the different POS tags, which can slow down the process or introduce mistakes if the task is complex.
If you already have the coarse-grained tags and you're asking about the morphology separately, you could also change the order and focus on verbs first, then nouns etc. You could even group verbs with the same lemmas together (should be pretty easy to do programmatically, especially if you already have the POS tag) so that the annotator can work through variations of "to eat" and then move on to the next word. Sorting by frequency could be useful, as well.
1 Like
Thank you very much for the comprehensive reply! I'll try that and see 
1 Like