Display Intervention Tagging in Extraction#441

I recently finished the Tagging module for a study and have now moved on to the Extraction module, but something that strikes me as a little inefficient is that I can’t see the tagging info in the Extraction module. Basically, I feel like I did the hard work of reading the articles and hunting down all the necessary information in the Tagging module, but now it’s not really paying off in the Extraction module because I just have to do the whole thing over again and hunt down all the information.

I’ll probably work around this by going into Study Inspector where I can easily flip back and forth between the Tagging and Extraction tabs, but I don’t think that’s necessarily the intended design.

I’m wondering if the module for Extraction could really foreground the tagging info (and de-emphasize the full PDF text) so that extraction is easier? Thanks!

Interesting idea, thanks for your perspective!

Are you familiar with tag icon we show in extraction? Each data element is shown with the little blue tag, if it was tagged.

• Hovering the tag shows the tagged excerpt
• Clicking the tag will jump you to the PDF annotation, if it was annotated.
• Data Elements that were also tagged are sorted to the top of the list.

If so, are you saying you want the full context of what was tagged, not just the Data Element you plan to extract?

Wow, had no idea that the tags would show the information if you hovered! I never would’ve assumed that, I just thought they were indicating that the data element had been tagged. This solves my problem!

Awesome :) I believe this was initially Kevin and/or @Izzet Akosman‘s idea, credit where it’s due!

I guess one place this does still come into play is when selecting the intervention and arm size––you have to input that information in order for the data elements to come up, but that information is most easily found in the tags. Is there a way to remedy this?

Good question– it may make sense to put the same flagging of tags in the Study Design panel? Basically replicate what we do for DEs for Interventions?

We could:

1. prefill the Arms table using tags (could even get cute & try to auto-extract # of patients from the excerpt, but probably a bad idea)
2. Show pills under the Arms table that the user could click
3. Sort/mark tagged interventions in the Intervention selectionn

With 2/3, we could trivially auto-navigate to the tag in the PDF, and with 2, we could display the tag excerpt when hovering.

I have another question about showing tagged text during Extraction: I don’t know if this is best practice, but there were a couple times I tagged multiple pieces of text with the same tag (e.g., in the Methods I tagged that mortality was planned for collection, and in the Results I tagged the actual mortality rate), but hovering over the tag icon in Extraction it’s not clear which tag I’m currently seeing.

I know that applying one tag to multiple parts of the PDF may not be best practice, but I got in the habit of doing this on this project because I noticed it was fairly common for things to be reported in the Methods but not in the Results, or vice versa, so I just wanted to document what I saw before moving on. Not totally sure how to address this

1. We do just arbtirarily choose one of the applied tags to show, if you applied it multiple times. We didn’t expect this to be common enough to build a solution.
2. Off the top of my head, I’d recommend creating + applying two different tags for those two different contexts. Maybe that’s not practical, and you’d end up with a mountain of concepts in your hierarchy.

Would a better solution be to not have to tag outcomes (in the Tagging module) in the first place? Perhaps you could tag & annotate outcomes while extracting.

Tagging outcomes during Extraction is an interesting idea! On one hand it seems efficient to go ahead and extract the data once you’ve located it in the text. On the other hand, it could feel cumbersome to perform multiple different actions in one step. Hard to say just based on an idea

I will say though that I can easily imagine a text box that contains multiple text tags that would show up when you hover over the blue tag icon. In my use case, it’d be pretty rare to apply the same tag to >2 pieces of text, so including both pieces of tagged text shouldn’t take up too much room

Note– part of the pain point here (the title & early discussion about showing which interventions are tagged) is being solved in https://nested-knowledge.nolt.io/316 as we speak! I think the only outstanding ‘ask’ seems to be the ability to tag and extract the same information at once, which can be accomplished from Inspector (toggling between modes) and which is somewhat mitigated by the jump-to Tag icons. Can I archive and ask that, if we truly want a “tag and extract at once” sequential mode, we address it in a separate issue? Thx!