Based on T338606, identify languages where can be proposed to be default instead of other services.
Description
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | Pginer-WMF | T353140 Propose MinT as the default over proprietary services where community feedback on quality does not indicate otherwise | |||
Resolved | KCVelaga_WMF | T353145 Potential languages for MinT to be proposed as default, based on Nov 2023 MT usage report |
Event Timeline
Purely based on service usage at target language, there aren't a lot of languages are close to Google's usage on the respective Wikipedias, apart from the ones that are already identified. The following are the closest, where
the deletion rates are not higher than that of Google: as (5%), ta (11%), yi (10%) [percentage is the percent of MinT usage]
I was also looking at the percentage of MT content modified data, and at the following languages
- Percentage of "MT modified content is less than 50%" is more than that of Google: te, bn, ta, uz
- Percentage of "MT modified content is less than 10%" is more than that of Google: el, ha
My reasoning is that, translators would have to spend less time for a translation if there is no need to make a lot of changes to the machine translated text. The deletions rates of articles created with MinT in all these languages are less and similar to that of Google.
I believe that these would be good starting points to have a conversation with. If translators are more happy with MinT than Google, it will make a good case for MinT to be proposed default. I can help with data about which translators have been using Google prior to MinT, and now have been using MinT, if that'll be helpful to have to those conversations.
Thanks a lot for the analysis, @KCVelaga
I created a ticket to adjust the defaults for the languages identified: T353259: Set MinT as the default service in Content Translation for a set of languages based on their machine translation usage