Thanks for the comments. The page is indeed lacking in details related to the implementation. Some of it may be found in the pilot study but most isn't.
I've made a first answered at https://www.mediawiki.org/wiki /Extension_talk:Wikispeech. I also copied your (Bawolff) questions there, hope you don't mind. I would appreciate it if the follow-up was kept there to make it accessible to people off-list. Thanks, André On 18 December 2016 at 20:49, Brian Wolff <[email protected]> wrote: > On Friday, December 16, 2016, André Costa <[email protected]> wrote: > > *Cross posting on purpose, no excuses made.* > > > > Hi, > > > > At Wikimedia Sverige we have been working on an extension called > > Wikispeech. It will be a text-to-speech solution which aim to make the > > information on Wikipedia more accessible for people that have limited > > abilities to read. > > > > This is Wikimedia Sverige's first MediaWiki development project from > > scratch and it has been suggested to us that we should ask for > endorsements > > - as this will make the need clear if/when the extension needs support. > So, > > if you think that this sound like something important, please let > everybody > > know it! https://www.mediawiki.org/wiki/Wikispeech#Endorsement > > > > Please spread the word. Best, > > André Costa > > André Costa | Senior Developer, Wikimedia Sverige | > [email protected] > > | +46 (0)733-964574 > > > > Stöd fri kunskap, bli medlem i Wikimedia Sverige. > > Läs mer på blimedlem.wikimedia.se > > _______________________________________________ > > Wikitech-l mailing list > > [email protected] > > https://lists.wikimedia.org/mailman/listinfo/wikitech-l > > I agree with nemo, a more detailed implementation plan would be very > welcome. > > From what I gather from your existing implementation, your current plan is: > * Using a ParserAfterParse hook, do some complex regexes/DomDocument > manipulation to create "utterance" annotations of clean html for the tts > server. > * insert this utterance html at end of page html > * javascript posts this to a (currently) python api, that returns a json > response that contains a url for the current utterance (not sure how long > an utterance is, but im assuming its about a paragraph) > * javascript plays the file. > > Is this your general plan, or is the existing code more a proof of concept? > Im not sure im a fan of adding extra markup in this fashion if its only > going to be used by a fraction of our users. > > -- > bawolff > _______________________________________________ > Wikitech-l mailing list > [email protected] > https://lists.wikimedia.org/mailman/listinfo/wikitech-l > _______________________________________________ Wikitech-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/wikitech-l
