This has been done in patch currently up for review https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikispeech/+/666395/2/includes/SpeechoidConnector.php.

I still get this when running locally:

$ composer doc
> DOXYGEN_FILTER='php ../../maintenance/mwdoc-filter.php' doxygen
/vagrant/mediawiki/extensions/Wikispeech/includes/Api/ApiWikispeechSegment.php:23: warning: Compound MediaWiki::Wikispeech::Api::ApiWikispeechSegment is not documented.
/vagrant/mediawiki/extensions/Wikispeech/includes/Api/ApiWikispeechListen.php:37: warning: Compound MediaWiki::Wikispeech::Api::ApiWikispeechListen is not documented.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/benchmark.php:19: warning: member with no name found.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/benchmark.php:476: warning: member with no name found.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/flushUtterances.php:18: warning: member with no name found.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/flushUtterances.php:184: warning: member with no name found.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/flushUtterancesByExpirationDateOnFile.php:16: warning: member with no name found.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/flushUtterancesByExpirationDateOnFile.php:82: warning: member with no name found.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/benchmark.php:20: warning: Member $IP (variable) of namespace MediaWiki::Wikispeech is not documented.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/benchmark.php:21: warning: Member if($IP===false) (variable) of namespace MediaWiki::Wikispeech is not documented.
/vagrant/mediawiki/extensions/Wikispeech/maintenance/benchmark.php:477: warning: Member $maintClass (variable) of namespace MediaWiki::Wikispeech is not documented.

@Aklapper, is there any way to have this only trigger when there is a change in the projects and not for other things. For example this changed the subscribers and that triggered the rule: T192433#6850156?

Images now updated. Only one remained that actually differed in appearance with the new UI. I realized this after making the other one, but now that one is a bit prettier at least.

It's up again now. Did @kalle (or anyone else) do anything to fix this?

Moodle is now sad:

I've added "ipa" as a parameter to the listen API. I did some dependency injection and added a couple of tests while I was at it. Still need to clean up and expand parameter validation. May make this its own patch.

I opted to go for a OOUI solution without an HTML Form, which was the initial design. There are a few things that will need to be supported that a form can't do, such adding an audio element (not there yet, but needed for T274486) and requests before submitting. While this may be possible to do with an HTML form with added Javascript, I feel like you lose the benefit of having it.

I added a group for editing with the edit right and added reading to all users by default. It's possible that reading is technically needed for editing somehow, but that can be changed if that's the case.

After a look through what rights exist in core and other extensions, it doesn't seem like reading from lexicon in and of itself would need a separate right. The only cases where I could find read or view rights was when dealing with sensitive information, either private or concerning security. I don't think that the information that will be presented when viewing lexicon entries will differ from normal pages in that regard.

Stumbled upon this little cute class that should be usable for language selection: HTMLSelectLanguageField. Maybe subclass it for filtering.

This was reported here: https://sv.wikipedia.org/wiki/Wikipediadiskussion:Gemenskapens_tekniska_%C3%B6nskelista/Automatiserad_artikelf%C3%B6rb%C3%A4ttringsbed%C3%B6mning#Syns_knappt. Don't know the version of Edge.

I think it is still reasonable and an improvement. I haven't seen any followup on my review.

Leaving this since we decided no (further) code review was needed.

Bahnhof server is now up and running at 176.10.250.192. Using minimum specs:

• 1 CPU
• 5 GB disk (half full with Ubuntu)
• 512 MB RAM (maybe 510, Bahnhof can't quite decide).

I've ran some initial tests and it seems to work well. This should be all we need at the moment. Thanks!

For now I am getting the available languages from the "WikispeechVoices" config variable. Not sure of this is what we want in the end, but it is reasonable to believe that you only want to edit lexicons for languages that you have voices for.

The old form had a field for lexicon, but I guess that will be taken care of in the backend so I am dropping it. In the future we may want to bring this back.

Speechoid requests look like this, not all parameters are needed:

/lexicon/addentry?lexicon_name=wikispeech_lexserver_testdb:sv&entry={
    "strn": "flesk",
    "language": "sv-se",
    "partOfSpeech": "NN",
    "morphology": "SIN-PLU|IND|NOM|NEU",
    "wordParts": "flesk",
    "lemma": {
	"strn": "flesk",
	"reading": "",
	"paradigm": "s7n-övriga ex träd"
    },
    "transcriptions": [
	{
	    "strn": "\" f l E s k",
	    "language": "sv-se"
	}
    ]
}

Using XMLHttpRequest instead of $.get solved it. It gets a Blod right away.

Looks like I still have not managed to get the correct data. It is not actually audio data, it seems. I tried with a website that managed to transform a sound file into data and it gets a different result.

We have updated to the new version now, but I don't know what the input should look like. Can you give an example? It would also be good to add one to the server documentation.