UAlbertaALTLab / recording-validation-interface

Maskwacîs recordings validation interface
https://speech-db.altlab.app/
Other
1 stars 1 forks source link

Pre-fill recording update on issues #444

Open fbanados opened 3 months ago

fbanados commented 3 months ago

Recording update should have several improvements:

See also: #443

fbanados commented 1 month ago

I've updated the interface a bit to deal with the first part of this issue (pre-fill for speaker code and word, plus auto-complete for wrong word) Screenshot 2024-08-12 at 6 27 59 PM

This has been deployed. I made sure to restart the image after adding static files so it should not have the issues of last time (missing css)

fbanados commented 1 month ago

@aarppe Do we want to add fields for lexical category and stem or just make available the analysis field that is already in the database to be edited on issues?

fbanados commented 1 month ago

From our discussion today, something like this for an intermediate quality state? Screenshot 2024-08-26 at 5 53 13 PM

(will revisit the colour tomorrow to make it darker)

fbanados commented 1 month ago

Example: https://speech-db.altlab.app/maskwacis/search/?query=namôya+ê-ayamihât

fbanados commented 1 month ago

All currently ticked changes have been deployed.

fbanados commented 1 month ago

Last two items should be a separate issue.

aarppe commented 3 weeks ago

When selecting an issue directly from the speaker view, autocomplete doesn't seem to work, cf. below one would expect to see pahkonikêw in the replacement window, not pahkonêw.

image
aarppe commented 3 weeks ago

Same here - the comment gives the correct speaker code, i.e. BL2, but that isn't auto-copied to the field that shows BRI instead, cf.

image
fbanados commented 3 weeks ago

A refactoring performed in 2021 changed names of fields, so the suggestion fields for this cases (likely for all issues before Sep 2021) are empty. There should have been a data migration added, but it looks like this was not the case.

I will have to make a migration pass that tries to extract the information from the comment, which seems to be the only place where that information remains in the database.

fbanados commented 3 weeks ago

I was this close to saying "new issues should not have the problem", but it looks like they do too. I will look into this in further detail.

aarppe commented 3 weeks ago

Ok, in principle the information should be in the appropriate field, for the wrong speaker and wrong word cases.

As a reminder for ourselves, we do also have a general Comment field in the Linguistic view, which could contain all sorts of information (including stems, inflectional categories, and derivational morphological make-up), but not guaranteedly in an always systematic extractable format.

fbanados commented 3 weeks ago

Currently, new "wrong word" issues work, but not new "wrong speaker" issues, because wrong speaker issues only store the information in the comment. As we have to fix this anyways for old "wrong word", I'll do both on the same pass.

fbanados commented 3 weeks ago

(Transferred from incorrect issue)

A fix is ready to deploy, but I'll wait until Daniel tells me he's done (or tomorrow morning, whichever happens first) to avoid issues.

Just as a side note, Speaker Code suggestion is only available from June 2021 onwards. Word suggestions exist from almost the beginning.

fbanados commented 3 weeks ago

Updated entries on speech-db.

fbanados commented 3 weeks ago

@aarppe would you like the interface for issues to only show the field that should be changed in the issue? (say, only speaker or only phrase?), or keep it as it is?

aarppe commented 2 weeks ago

Storing this for reference. Sometimes the comment has the correct speaker code (here: ARL), but the pre-filled code is still the old one (here: BL2): https://speech-db.altlab.app/maskwacis/issues/5

image image
fbanados commented 2 weeks ago

This behaviour is expected. You are in a different issue than the one on the card. The current issue has no speaker: Screenshot 2024-09-17 at 12 14 39 PM

There is a different issue that does have the right speaker annotation:

Screenshot 2024-09-17 at 12 15 31 PM

If you go to the details of that issue, (Clicking on "More info"), the speaker is autocompleted.

fbanados commented 2 weeks ago

Request: resolve all issues of the same kind when clicking on resolve.

fbanados commented 2 weeks ago

Change Merge phrases to entries.

fbanados commented 2 weeks ago

Merge for "câh-cakataham mîtosa namôya nikî-nipân" is not working with a 403 forbidden

fbanados commented 1 week ago

Must generate the "linguist-admin" role that can merge items.