menu
Tatoeba
language
Registriĝi Ensaluti
language Esperanto
menu
Tatoeba

chevron_right Registriĝi

chevron_right Ensaluti

Foliumi

chevron_right Montri hazardan frazon

chevron_right Foliumi laŭ lingvo

chevron_right Foliumi laŭ listo

chevron_right Foliumi laŭ etikedo

chevron_right Foliumi sonregistraĵojn

Komunumo

chevron_right Muro

chevron_right Listo de ĉiuj membroj

chevron_right Lingvoj de la membroj

chevron_right Denaskaj parolantoj

search
clear
swap_horiz
search
Lepotdeterre Lepotdeterre 2015-julio-01, modifita 2015-julio-01 2015-julio-01 05:59:42 UTC, modifita 2015-julio-01 06:00:16 UTC link Konstanta ligilo

Why doesn't the auto-detect recognise Macedonian? Whenever I try to use it for a Macedonian sentence, it classifies it as Bulgarian. Now, how, pray tell, could a sentence containing the љ, њ, ј, ќ, ѓ and/or џ be considered Bulgarian, when those letters are fully absent in Bulgarian? Isn't the basic step of language recognition looking at its alphabet, before going into the words themselves?

This doesn't really matter to me, as I don't use auto-detect - I just select Macedonian manually after logging in, and subsequently, it stays as my default language until I log out, which I generally don't do for entire weeks. However, the fact that the auto-detect feature is endowed with such an absurd imperfection makes me uneasy.

{{vm.hiddenReplies[23304] ? 'expand_more' : 'expand_less'}} kaŝi la respondojn montri la respondojn
User55521 User55521 2015-julio-01 2015-julio-01 06:26:50 UTC link Konstanta ligilo

I believe auto-detect is learning from the existing sentences, but it’s not doing that real-time: it needs to re-learn to take the new sentences into account. Probabely when the auto-detect data was generated last time, we had much more Bulgarian sentences than Macedonian.

> Isn't the basic step of language
> recognition looking at its alphabet,
> before going into the words themselves?

Not neccessarily. There are many algorithms available.

gillux gillux 2015-julio-01, modifita 2015-julio-01 2015-julio-01 06:36:12 UTC, modifita 2015-julio-01 06:36:39 UTC link Konstanta ligilo

> Why doesn't the auto-detect recognise Macedonian?

Because the autodetection algorithm is based on the Tatoeba corpus itself, and we still need to update it manually once in a while so that it takes new sentences into account. Three month ago, there were less than 200 sentences in Macedonian, which was not enough for the algorithm to work. Now you added about 50 000 Macedonian sentences, it will certainly work better once we update it. I’ll let you know when it’s done.

gillux gillux 2015-julio-01 2015-julio-01 07:11:38 UTC link Konstanta ligilo

It should work better now.

{{vm.hiddenReplies[23307] ? 'expand_more' : 'expand_less'}} kaŝi la respondojn montri la respondojn
Guybrush88 Guybrush88 2015-julio-01 2015-julio-01 07:47:41 UTC link Konstanta ligilo

actually it seems that it uses the language that is used the most by users. I just added #4322143 and #4322166 to test it and it firstly recognized those sentences as Italian

{{vm.hiddenReplies[23308] ? 'expand_more' : 'expand_less'}} kaŝi la respondojn montri la respondojn
Lepotdeterre Lepotdeterre 2015-julio-01 2015-julio-01 09:43:51 UTC link Konstanta ligilo

But I've never posted anything in Bulgarian, so that's clearly not the only factor.

Lepotdeterre Lepotdeterre 2015-julio-01 2015-julio-01 09:44:01 UTC link Konstanta ligilo

Thank you.