Order machine translation duplicate results by likeness

[Anders Melander]

When performing machine translation and the result contains multiple matches, the closest match should be the first in the list. This way the action “Use first available translation” will be more likely to select the desired translation.

[Anders Melander]

This is already partially implemented: If there's an exact match between the value being translated and the source value in the translation provider (e.g. TM) then the match will be placed fist in the result list. The match has to be exact though, including case.
Edit: There’s a bug in the current implementation, so the above doesn’t work.

The task here will be to rank the lookup results based on likeness, so every "equalization" rule that has to be applied to a result in order to make it more like the source, will penalize the result rank.

[Anders Melander]

Fixes #3: Translation lookup results are now ranked by the similarity of the translation source value to the value being looked up. This applies to both TM lookup hints, the translation suggestion dropdown list and the "Translate from TM" action. It does not apply to other translation providers (e.g. the Microsoft ones) as these services supposedly already rank the lookup results.

→ <<cset a84832940c74>>