Annotated sentences improvements

[Jetske]

The following edits will be made:

• Ability to edit which glosses are representative
• Add that you can download an eaf file and upload an improved one on the edit annotated sentence page
• Underneath the video show the annotation gloss id rather than the lemma (e.g. PT -> PT-1 hand)
• Add the source of the sentence (that is, which corpus)

[uklomp]

The text fields for the translations seem to allow a limited number of characters. Can we expand it? I believe it might be 200 characters now, I'd like it to be 300.

[uklomp]

It should be possible for us to upload multiple sentences under one gloss.

[Jetske]

@uklomp Uploading multiple sentences under one gloss is already possible, or what do you mean?

[uklomp]

Ah, sorry! You're right! Ignore this.

Another point: what did we decide again on the left hand glosses? When there's overlap, only the right hand is selected, but otherwise the left hand is also included?

[Jetske]

Yes, all glosses from the right hand tier are added and only those from the left hand tier if they don't have overlap with a gloss on the right hand tier

[susanodd]

This is deployed on signbank and on signbank-dev

[uklomp]

Hi, I encountered a new issue: after my Windows update (??? don't know if it has anything to do with it but OK), the glosses are not immediately visible anymore.
// edit: sometimes it works, sometimes it doesn't... I don't know why haha.

[Jetske]

@uklomp it's always tricky when sometimes it works and sometimes not, haha. Until I find out why this happens, it should help to just refresh the page when it does not work. I don't think it is related to your update.

[uklomp]

Refreshing doesn't work. Uploading it again also doesn't work. I can send you an email with two .eafs that I could upload today and two that I couldn't, maybe you can spot the difference?

[Jetske]

Yes please do!

[Jetske]

Thanks for the files, that really helped to debug this. It looks like the error was caused by a "_" that should have been "__" in my code... anyway I hope this solves it!
Also made small improvements to the feedback in case it finds annotations that are not in the dataset.

[susanodd]

@Jetske are you able to add the gettext to the feedback so that it will be added to the PO files for translation? (Eventually.)

[uklomp]

Ok, I will try again soon.

I have a new request (please let me know if you want me to make separate issues for these things, now or in the future);
namely (actually like was done for the old example sentences connected to senses), if we could provide information if it's in question or a sentence containing negation. So, in the upload process, I'd like two boxes we can tick, one with 'question' ( 'vraagzin') and one with 'contains negation' ('bevat ontkenning'). If these boxes are ticked, I'd like this information to appear on signbank as well.

And a question: I see that there's a field now, where I think it is possible to say something about the corpus / source, was it the idea that we put the link there? And should we do this manually for now, or were you maybe still looking into a way to do this automatically?

[uklomp]

.eaf 43 and 72 still don't really upload so I'm afraid the bug isn't solved yet!

[uklomp]

Zou het mogelijk zijn om de video, als die is geupload, nog links of rechts in te korten met een soort crop-functie? Ik kan me voorstellen dat dit ingewikkeld is maar ik vraag het toch maar even.

[Jetske]

@uklomp the changes are not online yet :) when it's gone from the pull requests you can try again.

For new requests, please make a separate issue with some priority tag. Problems encountered afterwards can be added to the existing issue (like the bug you found here).

Yes the corpus field was meant to add the corpus name or a link to the corpus. I wouldn't know how to add it automatically but if you have ideas about that let me know!

For the question about cropping: I'll look into it a little bit but I think it is difficult to implement.

[Jetske]

For annotated sentences a difficulty with cropping is also that the annotations each have a start and end time. This would need to be adjusted when cropping from the beginning of a video. That's possible, just not very efficient maybe.

[susanodd]

This is live on master.

[susanodd]

I'm getting this warning after the new updates:

unable to import get_creation_time from CNGT_scripts

Maybe this function is needed in the Signbank Python files instead?

[Jetske]

@susanodd Any idea which function it occurred in?

[susanodd]

@susanodd Any idea which function it occurred in?

No idea. It was just a general run.
I was working on the feedback, which has nothing to do with eaf files. But I hadn't run a branch that included the sentences yet.

It looks like print statement is in the frequency.py file. (Which is really old.)
PyCharm does extra checking depending on the settings. So perhaps I have something different set on the Mac.
I think it does that when it wants to generate new pyc files,
So I guess not related.

[Jetske]

Yeah I think that it's unrelated to the new code

[uklomp]

I'm able to upload the files that I couldn't upload before, so that's great.

I made a new issue for the source things.

As for the cropping: you mean that we crop the video live on signbank, but then the .eaf-file doesn't fit anymore, right? So we would need to adjust the .eaf again. Maybe we make it ourselves a bit more easy and only crop in between sign boundaries. This would already be helpful because one of the common reasons to crop is that the video starts halfway a certain sign, that we don't need to include anyway. So then we just delete this gloss and crop the video online?

[Jetske]

@uklomp the cropping functionality crops both the video and eaffile, cutting out irrelevant annotations.

Then I believe this issue can be closed. Other changes are still to be done in other issues.