User:Runab WMF/CX-Roadmap

Warning:

The contents of this page have moved to the CX-Roadmap page. Please do not make any changes to this page.

Content Translation Minimum Viable Product (MVP)

See below for the detailed development plan for each of these features

Feature Set: Translation View

1. Entry points
   1. Entry point: Red interlanguage link
2. Editor
   1. Basic editor to support side-by-side and section-by-section editing
   2. Progress bar
   3. Tools column
   4. Limited machine translation with support to detect lack of post-editing (correction of machine translation output)
   5. Reference adaption - unchanged copying and manual editing
   6. Rich-text manipulation with LinearDoc (partially flattened tree structure)
   7. Publish the content as a user sub-page after HTML to wikitext conversion with Parsoid
3. Link Adaptation
   1. Support for link adaptation (applied to source text only) (wikidata support)
   2. Support for link adaptation (applied to MT target text)
4. Machine translation support (MT)
   1. Beta deployment using Apertium MT: Spanish-Catalan
   2. MT translation warning and progress
   3. Store MT unchanged percentage on save (can use a category)
5. Dictionary support
   1. At least one bilingual dictionary based on dictd files from Freedict or similar #3639.
6. Templates support
   1. Copy other templates unchanged
   2. Do not allow editing templates, references or other alienated content. Editability must be whitelisted not blacklisted.
   3. Support for template “adaptation” in at least one language pair
7. Architecture (technical feature)
   1. Use aggressively cacheable architecture
   2. Server testing infrastructure

Production Deployment - Resources & Provisioning

WIP

---

Milestones

Completion Date/Milestones	Features	Sprints
March 20 - March 31 2014	Section alignment for the loaded article and edit support Dummy machine translation Server testing infrastructure Data modeling (Redis)	64
April 30 2014	Entry point: Red interlanguage link Progress bar Publish the content as a user sub-page after parsing with Parsoid Generic interfaces to language service providers - TM, MT, dictionaries, Wikidata	65-66
May 31 2014	Word information from dictionary and glossaries Use aggressively cacheable architecture Link adaption using LinearDoc Support for link adaptation (applied to source text only)	67-68
June 30 2014		69-70
July 2014		71-72
August 2014		73-74
September 2014		75-76

---

Development Plan

Mingle Story Board

Feature	Details
Entry Points	3631 Red interlanguage link
Editor	Layout and Design 3623 Side-by-side basic editor 3651 Section-by-section editor 3625 Small screen (responsive) support Progress bar 3624 Show translation progress 4120 Move the progress bar to the top of the tools column Tools column 3654 Add initial documentation to the sidebar Editing 3637 Support for non-translatable elements. Do not allow editing templates, references or other alienated content. Editability must be whitelisted not blacklisted. In the long run, VE solves a lot of problems that would otherwise need addressing. Our strategy includes using VE is the long run. We do not want to reinvent solutions for the issues VE is addressing already. 4154 - Use LinearDoc for all rich-text manipulation. Requires sub-sentence alignment using LinearDoc and reordering info from MT engine Do the link target lookup part of the link annotation before (in parallel with) the MT. Reference Adaptation 4156 Reference adaptation - copy unchanged 3636 Reference adaptation - edit references. Allow users to indicate how a source citation is translated. 4159 References, inline markup: apply to MT candidate text Automatic Translation 3957 Keeping text format 3634 Warn auto-translation abusers 3959 Control and revert from tools panel Publishing 4135 Publish translated article in wiki syntax to user space 3631 Confirm the article has been published
Link Adaptation	3635 - Support for link adaptation (applied to source text only). Allow users to indicate how a source link is translated 4158 - Support for link adaptation (applied to MT target text)
Machine Translation Support (mt)	4157 NEW CARD - Beta deployment using Apertium MT: Spanish-Catalan Spanish-Catalan because reordering is minimal. We could potentially add other language pairs, if Apertium is enhanced to output reordering information. MT translation warning and progress NEW CARD - Store MT unchanged percentage on save (can use a category).
Dictionary Support	3639 - At least one bilingual dictionary based on dictd files from Freedict or similar. Care must be taken about licence and provenance Node-based dictd wrapper http service (suitable for Varnish caching). Full dictionary definitions are *not* part of the MVP 4163 - Show most frequently used term in dictionary information
Templates Support	4160 - Support for {{cite...}}template “adaptation” in at least one language pair (but it can be a pair where no adaptation is required, and the template is copied unchanged) NEW CARD - Copy other templates unchanged (so they may not work)
Architecture (technical feature)	4161 - Use aggressively cacheable architecture diagrammed
Research and preliminary development	Segmentation 4025 Research on sentence segmentation cx-server 3998 Proof of concept implementation for CX client server communication model 4023 Improve the server as per the POC to handle requests using socket.io, have all the boiler plate and architecture code 4024 Data model preparation - first version 4098 ContentTranslation server testing infrastructure 4112 Tests for ContentTranslation segmentation