Wikispeech/Log
- This is intended to serve as a log over developments and decisions to make it easier to follow what is going on in the project. If you have any questions don't hesitate to get in touch at andre.costawikimedia.se.
March 2023
edit- Released version 0.1.10 of Wikispeech.
January 2023
edit- Increased supported version of MediaWiki to 1.39, which is a new LTS.
August 2022
edit- Updated the producer wiki (which serves the Wikispeech gadget on sv.wikipedia) to add some quality of life improvements for the UX.
April 2022
edit- The Wikispeech project is presented to the Swedish Post and Telecom Authority, the funder of the original project [Swedish]
August 2021
edit- The toolkit for organising events for speech collection is published [Swedish]
July 2021
edit- The final report of the project is published [Swedish]
- Improved pronunciation of numbers and dates in Swedish.
- Prior to April 2021 the log lived at wmse:Projekt:Wikispeech – Talresursinsamlaren 2019/Project log
April 2021
edit- Submitted proposal for lightning talk at Arctic Knot Conference 2021.
- WMF started performance and security review of the Wikispeech extension.
- Updated to version 0.1.8 of Wikispeech on the demo server.
Mars 2021
edit- The report on the fourth phase of the project is published [Swedish]
- Added specialpage for editing lexicon. It can be used to add new lexicon entries.
February 2021
edit- Recording annotations are now stored in wiki pages (using MCR) rather than directly in the database. This adds edit interface, revision editing and rollback features with traceability to authors as in any other standard wiki article.
January 2021
edit- Wikispeech presented at Norwegian Wikitreff
- Wikispeech demoed for Norwegian Nasjonalbiblioteket
November 2020
edit- Speechoid docker images are automatically built and deployed to the WMF Docker Registry. One result of this is that the process for hosting Speechoid as a service on your own infrastructure is a simple (and documented) task.
- It is now possible to hide the player again after you have activated it. This also disables the selection player.
- The report on the third phase of the project is published [Swedish]
- A benchmarking script is now available to measure the processing and storage requirements for segmenting and synthesising a given page.
- The Wikispeech page on Meta has undergone a much needed update turning it into a portal for all things Wikispeech.
October 2020
edit- Wikispeech hidden in views where it cannot be used (e.g. History or Edit view).
- Released version 0.1.7 of Wikispeech. Primarily internal changes in how Wikispeech works. You can test it on https://wikispeech.wmflabs.org.
September 2020
edit- Released version 0.1.6 of Wikispeech. Primarily internal changes in how Speechoid and Wikispeech interact. You can test it on https://wikispeech.wmflabs.org.
- Listening to a page is no longer interrupted if another user edits the page.
- Released version 0.1.5 of Wikispeech. Highlights include better control of when Wikispeech is activated and an updated UI. You can test it on https://wikispeech.wmflabs.org.
- Speechoid is now versioned to make it easier to map compatibility with the extension.
August 2020
edit- A new version of https://wikispeech.wmflabs.org is now up including up-to-date changes in Wikispeech and Speechoid.
- Speechoid has now been integrated in the Continuous Integration (CI) workflow. Final images are not being published but Speechoid, as deployed on https://wikispeech-tts-dev.wmflabs.org/ is now being built in the same way as CI will do in the future.
July 2020
edit- Documentation on how to install Speechoid using the new Blubber build.
- Documentation on what is required to add new voices and languages to Speechoid.
- The report on the second phase of the project is published [Swedish]
June 2020
edit- Check-in with LinguaLibre
- New Speechoid services deployed on https://wikispeech-tts-dev.wmflabs.org, built on Blubber
May 2020
edit- Re-implementing Speechoid packaging to be done using Blubber
April 2020
edit- PTS published a video about the project
- Initiating work on the march 2020 decision to store generated speech in Wikispeech extension
- Initiating work on converting Pronlex databases from SQLite to MySQL
- Received feedback that the proposed re-architecture of Pronlex will need to be modified to avoid the service-dependancy-loop
March 2020
edit- Documented the re-architecturing decision needed to store the generated speech in the Wikispeech extension rather than the Speechoid services.
- Looking into how Pronlex can be re-architectured to avoid the service being stateful (See notes in Swedish)
- Check-in with Mozilla Voice
- The report on the first phase of the project is published [Swedish]
February 2020
edit- On-boarded new developer wmse:User:Karl Wettin (WMSE)
- Analysed how to implement the Speech Data Collector (in Swedish). In particular if it should be a MediaWiki extension running on Wikipedia or a stand-alone tool. The conclusion was to implement it as a MediaWiki extension but to not require it to run on Wikipedia. That way it can either be used as a stand-alone tool (a separate wiki) or implemented directly on the wiki.
January 2020
edit- Site visit took place at WMDE offices, in part focusing on requirements needed to bring Wikispeech to production. See notes here (in Swedish).
- Events prior to 2020 ar not included in the log