1853300 - Tokens containing non-word characters are translated incorrectly

Status: RESOLVED MOVED

(Firefox :: Translations, defect, P3)

Description

[martin-kokos]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:109.0) Gecko/20100101 Firefox/118.0

Steps to reproduce:

Click Firefox Translations add-on to bring up a a translating window.
Translate German to English
Type in something like:

[e-mail]
Email:[e-mail]
[e-mail]
[e-mail]

Actual results:

@ sign gets translated sometimes to dash "-"
sometimes to "(at)"
in my specific case inside an email to "Ã3"

"[e-mail]" translates to "ffà mozmail.com"

Expected results:

Tokens containing special characters like @,:; maybe others should not be broken.

The "Ã3" probably indicates encoding problems in training dataset.
However tokens which contain more non-word characters (emails, URL's, ) probably shouldn't be translated

Comment 1

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug should belong to the 'Firefox::Translation' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 3

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:gregtatum, could you have a look please?

For more information, please visit BugBot documentation.

Comment 4

[Greg Tatum [:gregtatum] (REO Fx132)]

I can confirm this reproduces in about:translations. I believe the OpusTrainer work will address this: https://github.com/mozilla/firefox-translations-training/issues/161

Comment 5

[Greg Tatum [:gregtatum] (REO Fx132)]

We need to retrain models with bad robustness, moved to: https://github.com/mozilla/firefox-translations-training/issues/757