Naše data jsou k dispozici pod různými typy licencí Creative Commons.Více informací
Pokud se vám tyto stránky líbí, můžete nás podpořit.
Zeď (7 139 témat)
Tipy
Před položením dotazu se podívejte na často kladené otázky – FAQ.
We aim to maintain a healthy atmosphere for civilized discussions. Please read our rules against bad behavior.
AmarMecheri
Před 3 hodinami
deniko
Před 5 dny
deniko
Před 5 dny
frpzzd
Před 5 dny
araneo
Před 5 dny
deniko
Před 5 dny
deniko
Před 5 dny
deniko
Před 5 dny
deniko
Před 5 dny
deniko
Před 5 dny
Hello everybody!
I'm a new member and I don't know how to listen a sentence in this web! Please help me!
Thanks you very much!
Bonvenon !
ĵus alklaki la botonon je la dekstra de la flago.
Sed malmultaj frazoj enhavas aŭdaĵoj.
Ĉi tie ĉiuj ili http://tatoeba.org/epo/sentences/with_audio
Thanks Sacred!
I heard it! ^^
Es inútil discutir con alguien que no quiere aceptar sus propios errores..!
Usted tiene razón. A propósito, puede agregar esa oración como frase de ejemplo aquí:
http://tatoeba.org/spa/sentences/add
The newest spam comes from linajohn023@yahoo.co.uk
+1
Hello everbody
Azul fell-ak (Hello [in Berber, North Africa]) ^v^
Please try my “Tatoeba Markdown Editor” userscript
--------------------------------------------------
It now features a (not yet very beautiful) language dropdown. You can choose the language you want to write in and thereby activate language-specific replacement patterns: e.g. when typing `some "quoted" text` it will propose `“quoted”` when english is selected, whereas it should propose `« quoted »` in french mode.
Please try it an give me feedback or tell me what replacement patterns you want to have for your language!
Thank you for your efforts. I don't know what I would do without some of your scripts. Still I failed to understand how this script works. I installed it but nothing seems to change. Where should I expect the editor to appear?
Firefox 3.6.28
Tatoeba English interface
(Suggest tags and linker scripts working fine)
It should create a green box above each textarea (e.g. the one you write comments or wallposts in), and a yellow and a white box below the textareas. The green box contains buttons like `bold`, `italic`, `Hyperlink`, `Numbered list`, `Heading`. The yellow box contains replacement suggestions like `"bonjour" => « bonjour »`. The white box contains a preview of the HTML version, e.g. `*italic*` gets `<em>italic</em>` (actually it’s emphasis, not italic) and then displayed as italic.
You can either update your firefox (which i strongly recomend if you don't have a really old computer (then im not sure)); or you can tell me what your javascript error console tells you. The first one would be a lot less effort for me ;)
PS: I didn't get the editor to work on wall replys yet, sorry for that, i'm working on it.
After a fresh install of Firefox 13 in virtual machine (XP) this script worked too. But I can't get it to work in FF 3.6.28 in real machine (W7). There are some compatibility issues of my addons with new generation FF, so I can't upgrade. I just don't understand how other scripts are working fine and this one doesn't.
Maybe I should add this one later when I begin to use latest release of firefox.
Thanks
I’ve updated the code with a little change that might be the reason it didn’t work in FF3. Could you check if it now works? Also, could you check if you’ve got the newest greasemonkey available for FF3?
Do you have ideas what replacement patterns you would like to have for turkish? Maybe some letters that are difficult to type on other keyboards or other special signs that should be used, but many don’t, because they type lazy? (like `dont't => don’t`)
I tried the new script in both machines (installed a fresh ff3 in VM too) with no luck. And yes GM is up-to-date, 0.9.20v. Maybe its time for an FF upgrade for me.
There are some commonly misspelled words. I can PM a list to you. And there are Turkish specific characters, too. But its impossible to devise a pattern for them because it takes a proofing tool to spot such spelling errors. (They are in almost every word). And your Tatoeba Symbol Insert Helper script already contains most of Turkish characters (I should provide another list for that too). One can use it for special characters if s/he doesn't have them in keyboard.
It now also works on wall replies!
Why does it take like 3 minutes to switch pages? I can't work well. I'm getting bored.
between which pages?
For example:
http://tatoeba.org/eng/sentence...rent/page:6750
and
http://tatoeba.org/eng/sentence...rent/page:6751
or
http://tatoeba.org/eng/sentence...rent/page:6749
I use to translate directly from these pages, but I always have to wait like 3 minutes to open them... And as long as other pages are loading in different tabs, the translations are not addded either.
Maybe the filters are too complex, though...
Yep unfortunately this set of pages, especially when getting in high number, start to be very slow because then the database can't use any optimization method as there's already a lot of filter. and for this I can't really do anything on the short term:(
The only things that maybe will indirectly speed it up is that today i discovered that for two weeks Google started back to crawl Tatoeba with their own determined speed (which is very high, like 2 request/second), which create an artificial heavy load on the server. I've made some hours ago a request for a lower crawl rate, it takes around a day to apply, so in one day you will certainly notice a speed improvement for this kind of pages.
the mid-term solution I'm thinking about is using the API , to have a desktop tatoeba client with the database locally (it's not that heavy, something like 300mo admitting you're interested by all languages and links etc.) that keep sync with something like pushsub, so that this kind of heavy filter will be done on your computer which will be far far faster as you're computer is being used only by you.
Thank you, sysko! That would be great!
I'll try to work on "lower pages".
About detecting languaje automaticaly.
What metod will use for detecting languaje?
I write in wall about trigrams or N-grams. Its seem a very good a simple metod.
I don't know if sysko know it.
In the message there the word http://detectlanguage.com/
A message was readed by sysko but other not. I don't find message, but maybe sysko have a searching tool for wall.
Maybe is't also interesting registre sentences which automatic clasification fail. For future training of sistem.
I hope enhance my english a little. Sorry for my wrinting.
Sysko planned this long time ago, but since he’s so busy nowadays...
He had an idea that the languages someone uses should affect probability. E.g. if you input some text, it is more likely to be Spanish than French, because you haven't added any French sentences before (or maybe languages would be selected in the profile).
So that’s one reason not to use off-site processing: if this is done on Tatoeba, it's going to be better.
But it seems no one has time to implement this. :) What is more, no one knows when new Tatoeba version will be ready. New Tatoeba will use its own database engine, so the code will be incompatible... So maybe we should wait for sysko to finish the new version first.
actually I have already implemented it and it works with a sufficient precision (more than 95%, i don't remember exactly, 98% maybe), i now need to turn it into a web service. The goal is to make it independent of Tatoeba itself, as it can be used by other people (and I think we're not the only ones with that problematic). So I think I will finish it first before continuing the new version (because it will be needed by the new version and it will not be hard to interface with the current code as we were already using an external service before)
Es genial Sysko!
Magnifique sysko!
What metod use for recognition.
In the first time I think about caracter estatistics. But this maybe inefectibe with short text like sentences.
n-grams work very well, I think.
In Spanish are theren't sh but in english yes. In think is a good metod for short sentences.
yep that is basically the methods I use (with some weights), when I will release it, I will take the times to explain it in details.
Thanks for explanation.
I wait for future interesting explanation.
Wow, cool! Sysko, you’re great! :)
the other reason to use our own service is that we do have a lot of languages that other services are very unlikely to be able to detect because Tatoeba is the largest easily available dataset for these languages (like Shanghainese or Berbere)
Suggest for login metod.
http://tatoeba.org/spa/wall/sho...#message_12719
un petit dédoublonnage ne serait pas du luxe...
hier dans le train m'emmenant au fin fond d'un trou perdu en Chine (18 heures de train avec heuresement une prise electrique) j'ai commence a pondre un script qui me converti le dump de la base en une version utilisable par la nouvelle base de donnee (le but est de resynchroniser la version de la base de donnee avec la nouvelle version pour mes tests, vu que le nombre de phrases a doublés) et il me faut pour cela un fichier sans doublon du coup ca m'a amenez a revoir cela, et la ce que j'ai ecrit est autrement plus elegant que ce que j'utilisais avant, il faut encore que je le finisse et apres je ferais un dedoublonnage.
Le principal a retenir est que je me remets à pondre du code, et vu que le plus dur c'est toujours de s'y remettre, c'est plutot bon signe...
Génial !
bon je l'ai fini et il est en train de tourner, je viens de remarquer que la plupart du temps passer a mettre a jour la base se passe dans les requetes qui fusionne les commentaires (et souvent vu que des commentaires il y en a pas tant que ca, ca ne fusionne rien), donc il doit y avoir moyen de diminuer le temps d'execution sur le serveur en faisant un check si la phrase a des commentaires ou pas.
Merci sysko !
bon là je tombe de fatigue, mais là normalement l'algo en lui même est visiblement fonctionnel, il reste plus qu'à écrire la partie qui génère les requetes à faire sur la base, mais ce ne sera pas le plus difficile vu que c'est déjà écrit dans mes précédents scripts.
Grandan dankon al dominiko pro lia valora subteno rilate al la trovado de eraroj!