[Freecats-Dev] RE: KBabel's secrets

[Kemper DOC (Nerim)]

Stanislav,

> > We would also be glad to receive feedback about KBabel's under
> > the hood secrets, not to mention POEdit, which development
> > team we did not contact - yet.
> >
>
> Aargh, that must be my fault. So better to write down at least some
> KBabel details right away:
>
> KBabel uses a plugin-based approach for handling matching. In fact, all
> the logic is stored in a plugin and the API is roughly:
> get exact translation, get translations according to the plugin settings
> and get fuzzy translations. Each returned translation has an associated
> score. The score is plugin dependent, but KBabel contains a library to
> calculate score based on 3-grams (chunks for 3 characters).
>
> Current plugins typically match according to these criteria:
> - case sensitivity
> - substring/superstring/prefix/suffix
> - 3-gram matching
>
> Also, the integrated TM supports "good keys". AFAIK it's used to
> speedup the search. TM keeps a list of "good keys", texts which are
> similar to the searched text, for example contains 50% of the queried
> text. This can be computed quite fast and only on these texts the
> plugin uses the rest of the matching.
>
> Another "unimplemented" features was to allow replacement/missing
> of particular number of words, so for example "Joe is here" would
> match "Mary is here", but not "Mary is there".
>
> That's basically it.

Thanks for your feedback. Its basic features look rather close from what I
was envisioning in my own draft document.

Oeuf corse, You may provide Tim and Michael with more details about the
source code if they need it.


Cheers,

Henri