Our data is released under various Creative Commons licenses.詳細情報
当プロジェクトにご関心のある皆さまからの寄付金の受付けを行っております。詳しくはこちらをご覧ください。
掲示板（スレッド数：6,959）
ヒント
質問をする際は、必ずあらかじめよくある質問をお読みください。
なお、Tatoebaは文明的な討論を行うために健全な雰囲気を維持することを目指しています。悪質な行為に対するルールも併せてお読みください。
sharptoothed
1日前
Cangarejo
1日前
Cangarejo
5日前
Thanuir
5日前
ondo
5日前
ddnktr
6日前
ondo
6日前
AlanF_US
9日前
Nandixer
10日前
cblanken
12日前
I've uploaded 1000+ new German audios from BraveSentry
and 200 audios in English by Nero
thanks a lot to them :)
Tatoeba day #4 - Results!
http://blog.tatoeba.org/2011/03...4-results.html
List of funny sentences:
http://blog.tatoeba.org/2011/03...sentences.html
List of inspiring sentences:
http://blog.tatoeba.org/2011/03...sentences.html
List of other sentences:
http://blog.tatoeba.org/2011/03...sentences.html
Enjoy :)
Enjoying Tatoeba!
Great, glad to have you here! Welcome!
Tatoeba day #4
http://blog.tatoeba.org/2011/03...26th-2011.html
---------------------
Tatoeba day #4 is scheduled on Saturday, March 26th. As usual, it will start at 00:00 and will end at 23:59, France time.
+ The theme
The theme for this 4th Tatoeba day will be... exploration!
It's a bit experimental. I have no idea how well it's going to work, but it's worth trying out :) The idea is to have everyone explore the corpus in search of the "best" sentences in Tatoeba. There will be 3 categories.
- Funny sentences. Sentences that make you laugh.
- Inspiring sentences. Sentences that make you want to change the world.
- Other sentences. Sentences that don't fit in the 2 previous categories but are interesting enough to be worth mentioning.
+ The rules
Each participant can create 3 types of lists. Each list must be named the following way:
- [TD4] Funny sentences (lang)
- [TD4] Inspiring sentences (lang)
- [TD4] Other sentences (lang)
Replace "lang" by the ISO code of the language. For instance, if it's a list of Swedish funny sentences, the name of the list would be: [TD4] Funny sentences (swe).
You have the right to create lists in more than 1 language. So you can have a list of funny Swedish sentences, a list of funny Hungarian sentences, a list of funny Hindi sentences. Same thing for inspiring sentences and for other sentences. However, you don't have to create a list in all the categories. You can only create lists of funny sentences if you only want to work on funny sentences.
Each list you create must have precisely 10 sentences. No more, no less. Do NOT create two lists with the same language and category. If you have found more than 10 sentences for a specific list, you will have to pick your 10 favorite ones. If you have found less than 10 sentences, you will have to hunt for more, or come up with your own sentences, or seek inspiration in other people's lists (because anyone can view the content of every list, whether they are public or not).
N.B. Do not set your lists as "public. Public lists are lists that ANYONE can modify (i.e. add or delete sentences from the list). Non-public lists are only not editable by others.
+ Deadline and results
You can start creating your lists as soon as you have read this announcement. You have until the end of Tatoeba day to fill up your lists. Anything you add or remove after that will not be taken into account. To make my life easier, please make sure your lists are named properly, and make sure that other participants lists are named properly!
On Sunday, I will publish the sentences that people have listed, for each category and for each language, ordered by "popularity" (i.e. number of people who listed the sentence). I will also publish stats about who owns the highest number of "best sentences" in Tatoeba :)
[not needed anymore- removed by CK]
thumbs up! thx a lot!
Tatoeba Day will start on 01:00, and end in 00:59. (Turkey, Belarus Time)
For people who wonder how to create a list, it's here: http://tatoeba.org/sentences_lists/index
Or, from the menu:
"Browse" > "Browse by list"
plus de 35 000 phrases allemandes, alors qu'il n'y en a que 55 000, ne sont pas traduites en français. De même, 56 000 phrases en espéranto ne sont pas traduites en français parmi 79 000 phrases. Sur 16 000 phrases en néerlandais, 13 000 ne sont traduites ni en allemand ni en français et plus de 10 000 ne le sont pas non plus en anglais...
Pourtant, les traductions sont là, dans d'autres langues, mais personne ne les lient jamais alors que les traducteurs sont disponibles.
On se retrouve donc avec un système, qui, en théorie, permet les traductions vers de multiples langues, mais qui, en réalité, ne propose que des traductions 2 à 2...
Ceci implique que chaque fois que l'on cherche des phrases à traduire dans une langue, on ne voit pas qu'en fait elles sont déjà existantes mais attachées à d'autres langues que celle dont on veut démarrer. Le travail de traduction est donc constamment ralenti par la nécessité de remonter les liens pour voir si les traductions n'existent pas déjà. C'est pire que de créer des doublons. On réalise donc constamment un travail inutile, en repassant éternellement sur les mêmes phrases qu'on n'a aucun moyen de filtrer...
Je pense ke sysko a dit que ca sera mieux avec le nouveu systeme de tatoeba. Il a dit qu'on vera toutes les traductions meme ceux qui sont liées indirectementes.
ben je pense qu'il y a forcément une limite à une requête procédurale.
La solution, c'est que les liens soient réalisés !
je ne sais pas si c'était explicite dans mon message précedémment, mais évidemment cela ne remplacera jamais le travail de liage, car même si on peut afficher des traductions indirectes au 65ème degré, seul le fait de les lier permet ensuite au quidam de savoir que les phrases sont traductions l'une de l'autre. (les traductions indirectes n'étant là que pour dire "bon on a aussi ça, mais c'est à prendre avec des pincettes)
Merci pour tes réponses. C'est clair. Mais je faisais moins de reproches à la technique qu'aux traducteurs...
d'un autre coté la technique ici est loin d'aider le traducteur, on peut comprendre que tout le monde n'ait pas envie de cliquer 5, 6 fois, charger autant de pages, noter le numéro, pour ensuite seulement pouvoir lier.
Idem pour cliquer une dizaine de fois sur random ou autre pour tomber sur une phrase allemande non traduite en français.
Dans le cas d'une base de donnée SQL, oui il y a une limite du fait de la nature même du langage. C'est avec cela en tête que l'on a fait notre propre serveur de base de donnée pour avoir quelque chose convenant à nos besoin. Dès lors notre base ne se base plus sur une représentation en table, mais une representation en graphe, et nous pouvons dès lors effectué une recherche de type BFS http://fr.wikipedia.org/wiki/Al...urs_en_largeur qui nous retourne l'ensemble de l'îlot de phrase sous forme d'un arbre.
Ce qui permet d'aller très très loin (en théorie aucune limite), et de remarquer par exemple que de traduction en traduction, certains phrases "bateaux" comme "ça va?" peuvent devenir "t'as un problème ?" au bout de 20 traductions successives tout en voyant la dite chaine de traduction (ce que ne permet pas le système actuel, ou l'on ne sait pas lorsque l'on a une traduction indirecte d'affichée, par quelle traduction directe elle est reliée à la phrase-pivot)
Évidemment cela ne résouds pas le problème entièrement, mais cela facilite le travail de liage, vu qu'il n'y aura plus besoin de sauter de lien en lien.
(donc pour ce qui est dit au dessus c'est déjà "possible" et codé, il me manque d'autres parties pour pouvoir remplacer tatoeba par cette nouvelle version, mais ça avance plutôt bien, à présent j'ai également un moteur de recherche qui indexe en temps réel etc.)
Après (là c'est de l'hypothétique rien de codé pour l'instant) je peux aussi essayer de représenter en "dur" ces îlots de phrases, de façon par exemple à pouvoir rechercher tous les îlots contenants une phrase en français mais pas de phrases en Allemand, ou les îlots ayant une phrase allemande mais n'étant une traduction directe d'une phrase française du dit îlot. Cela dans le premier cas permettrait de trouver plus facilement des "phrases" (indépendant de la langue, mais je ne trouve pas de mot pour l'instant pour dire cela) nouvelles à traduire, et dans le deuxième des phrases à lier. Encore une fois grâce au système évoqué en premier, avoir ensuite tout sur une page, facilitant grandement la tâche.
En l'état actuel de mes compétences en informatique, je n'ai que ça à proposer que je puisse ou ai réalisé pour vous aider, après évidemment si on pouvait avoir un algo magique (je dis magique car je n'ai aucune idée de comment le faire autrement que dans les grandes lignes), permettant de trouver des phrases similaires.
Oui, ce n'est pas un système très... robuste !
pas très "efficace", quoi...
You can't copy info, even from Wikipedia. Because the CC-BY-SA is not compatible with CC-BY.
Lojban: A logical language. Lojban will provide communication with computers in future.
Klingon: The official language of Star Trek.
Lojban is hard. People will not learn it. I think computers will use normal languages.
I think normal languages will adapt to computers...
they will meet half way.
I missed Tatoeba! But I'm a fan of Kylie Minogue.
Je suis une fan du Kylie Minogue.
Mi estas ŝatanto da Kylie Minogue.
Ich bin eine Fan von Kylie Minogue.
well, i´m a fan of magister perotinus´.
i´m lucky: his lyrics are all in the public domain and thus are addable to tatoeba. kylie´s aren´t :P
even the lyrics of the songs of magister perotinus (i don't know him) are copyrighted.
Please stop adding lyrics and other copyrighted things to Tatoeba!!!
they are not...he´s been dead for over 600 years.
no lyrics, i don't wanted to add lyrics. don't mix it. i told only I'm a fan of Kylie Minogue.
Tatoeba has been updated :)
http://blog.tatoeba.org/2011/03...21st-2011.html
- http://tatoeba.org/contributions/index shows all the contributions, including the links.
- The profile has been revamped, finally!
I feel like the new http://tatoeba.org/contributions/index is too cluttered. I would move the linking information to a separate page. I doubt that many users are interested in this information anyway.
It's back to how it was before except that there's a link to browse all the contributions.
That was quick :)
+1
+1
It seems impossible to edit the information in the upper profile box. It always says "Error: The requested address '/spa/user/save_basic' was not found on this server."
The description box works fine though.
Should be fixed :)
how can i change the language of the interface? if i put "eng" or "fra" in the place where "deu" is now, either the page won´t load or it automatically switches back to "deu".
Next to the "log out" icon :P
which, i think, i´ve never used up to now :)
thanks!
hey...how about creating a rule that says numbers have to be written out?
to me its quite annoying to have sentences with numbers in ciphers in languages in which i don´t know how to pronounce those ciphers.
example: "There are 3 apples in the basket." - "Hay 3 manzanas en la canasta." => a spanish native will read "There are (tres) apples in the basket." - and an english native "Hay (three) manzanas en la canasta."
I've chosen to write things out the way it'd normally be written in that language. On sentences with numbers or some other type of reading ambiguity I leave a comment and tag the sentence with "cmt on reading".
this appears to be a good way of dealing with it.
is there any sensible way this could be suggested to all users?
Documentation is always tricky. Mainly because no-one reads it. I've long wanted to create a screencast for Tatoeba, but never gotten the damn software to work.
CK and I have set up pages with tags for people to look up available tags at. Trusted users are, I think, just expected to do a bit of work browsing around for good practices.
... or ask here on the wall. :-)
I think We should add a rule that there needs to be a comment about the numbers while leaving the style of the sentences untouched. E.g. One could write "3 = tres" below the spanish frase.
The problem with comments is that they are not exportable, so you ca’t use them outside Tatoeba.
Something better is needed. :)
Comments are perfectly exportable. Tatoeba works with most computer mice to enable an innovative "highlighting" feature in which you can hold the left mouse button to drag across the words you want to export. Then, if you have a right mouse button, you can right-click it to open up a hella tight new menu that Sysko just rolled out, which includes a bold new Copy™ feature you can click on to insert the entire comment into a specially allocated spot in your computer's memory. Tatoeba's patented Copy™-based export function is specifically designed for plug-and-play compatibility with the cutting-edge pasting capabilities of modern text editors and database programs. This Copy™-paste paradigm offers significant improvement in natural language processing efficiency over the older method of printing comments individually, holding the printouts up to the monitor, closing one's eyes, and making a wish.
I understand your point, but I am opposed to this idea. I think sentences on Tatoeba should model actual written usage. Nobody's going to write "three hundred forty three million six hundred fifty two thousand nine hundred thirty one point nine two four".
However, there are other ways to solve the problem. For example, for Chinese and Japanese sentences that use the Chinese numeral characters (一, 二, 三, 四, 五, 六, 七, 八, 九, 十, etc.), you will see the automatically generated readings below.
The Arabic numerals (0, 1, 2, 3, 4, 5, 6, 7, 8, 9) are also logographs just like the Chinese numbers. We only don't provide readings because there are few other logographs in most alphabetic languages (the only other ones that come to mind are currency symbols). It would be mostly wasted space.
I see no reason that, in the future, this capability could be extended to alphabetic languages, as long as it was hidden by default. Sentences could have a button that, when clicked, would should readings. This could also be extended to all languages, so I'm not tempted with pinyin when I should be reading the characters. ;-)
(Also, on a related note, I would like the ability to see zhuyin/bopomofo on Chinese sentences instead of pinyin. I can read it faster, and it's much more more logical... Both methods could use the HTML5 ruby element, so that the readings appear near the characters, so you don't have to go back and count characters.)
I quite agree. Forcing people to write our numbers would make corpus less representative.
However, generating the readings automatically may turn out to be quite problematic. For example, it Russian it would require to do the syntax analysis to determine the case of the numeral.
BTW, Zifre, do you have any downloadable or online reference materials on Chinese in zhùyīn?
I don't know about other languages, but in formal written English, numbers one through nine are written out, but 10 and over are written as Arabic numerals.
Pompous prescriptivist pedants need not quote me.
It's similar in German. Style guides suggest to write out 1 through 12 and to use numerals for 13 and over.
But in general I agree with BraveSentry. I often try to read texts in languages I only have a very limited knowledge of and numerals, dates (and also abbreviations) are usually the stumbling blocks where my reading flow collapses. There should be a way to annotate them.