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：6,960）
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。
marafon
2日前
CK
3日前
sharptoothed
8日前
Cangarejo
8日前
Cangarejo
11日前
Thanuir
12日前
ondo
12日前
ddnktr
12日前
ondo
12日前
AlanF_US
16日前
I suggest that one should be able to (optionally) see all the translations that once where linked to a sentence. Sometimes i dont unlink a sentence that still needs adaption, because i fear that it will never be linked again and thereby get lost.
For example the sentence "The house is painted blue." could have a translation "The house is painted red."(in another language), which should be unlinked or corrected. So lets say it gets unlinked first and only corrected afterwards: Then we would have "The house is painted blue." and "The house is painted blue."(in another language) not linked to each other. Only in the changes history one could check all the sentences that got unlinked and link them again if they fit again, but this is very inconvenient.
There should be a button or an option to append the unlinked to the list (just like the indirect translations) or the unlinked sentences should be readable in full length when hovering thier number in the history.
> because i fear that it will never be linked again
It will, sooner or later...
> or later...
Thats the point ;)
If sentence B gets unlinked from sentence A, which has a translation C; then it might easily happen that B gets translated to C by adding a double and not by just linking them (which is extra useless effort).
I know that Sysko wants to develop alogrithms that identify similar sentences to avoid doubles and add more links, but until then I think it would reduce effort.
But currently, duplicates are merged every now and then by a batch procedure. I think it works very well. It doesn't matter much that there are a few duplicates at any given time between 2 deduplication procedures. It actually takes more time to check for duplicates than to create them, so contributors would not really benefit from a preemptive duplicate detection system, because to check for a duplicate, you must type it first, and that's what is time consuming...
(i wrote a long reply before, but it fot lost in the depth of my crappy mobile browser)
The problem to me is false links: if we have "A" and "B" then by linking someone verifies that they are translations of eachother ("A"="B" true), but if "A" changes to "a" then "a"="B" is false. So we should unlink them and not wait until someone fixes it. By suppling "relink" buttons for "ghost"-translations we can unlink "temporarily" because its easy to relink. So the work of varification has to be redone (else its not verified)(of course not formminor changes like punctuation).
But what I mean is, even if you don't "relink", if "a" is a valid translation of "B", someone, somewhere, will link them or retranslate them separately into each other, and through the mechanisms of deduplication/merging, they will be relinked the same way as before to the other sentences they were linked to. So, really, it doesn't matter that the same person unlinks and relinks. Linking will take place if it makes sense, at some point...So I don't think a "ghost link" is necessary.
Reversely, classic translation mistakes (like people translating the French "actuellement" into the English "actually", although the latter actually means "réellement" in French) : You may correct them hundreds of times and unlink them hundreds of times, they will reappear at some point...That is the downside of the "wisdom of crowds" that everybody can see on Wikipedia, with always the same false informations popping up time and again...
We live the myth of Sisyphus...
For that i think we should be able to comment our actions like "Unlinked the english translation because "actuellement" is not "actually"" or "tagged the sentence 'war' because the word 'army' suggests so" (beside not commenting actions at all).
So i would like to see the history of actions merged with the comments. Then also one wouldnt wonder why there is a comment saying "i would use 'crapword' here" when 'crapword' isnt even in the sentence but it got changed afterwards. You wouldnt need to seperately look at the history anymore.
The stupidity of the crowd would find its way also by retranslating instead of relinking i think, even more so if one would be made aware of the fact that relinking is a more delicate task than newly linking, because maybe someone DECIDED to unlink the sentences and not just unlined them because either was changed.
Problème difficile dans tous les cas.
Je constate qu'il est de plus en plus difficile au fur et à mesure qu'on avance :
Plus l'historique d'une phrase est long plus il devient complexe de comprendre pourquoi telle phrase a été détachée de telle autre, etc.
pour le coup des fautes de traductions, il y aurait surement moyen de faire comme sur la wikipedia, c'est à dire mettre pour les taches répétitive un système de bot.
J'espère pouvoir (et pour l'instant, mis à part encore un peu de temps, je ne vois pas de limitation technique qui l'en empêcherai), mettre en place la chose suivante:
L'api enverra via xmpp (ou autre), quand des evenements arrivent sur tatoeba (ajout de phrase/de traduction/ajout d'un tag (peut etre se limiter uniquement ))
le but principal serait de pouvoir avoir un tatoeba "hors ligne", installable sur son ordinateur (la nouvelle version peut tres bien fonctionner sans serveur apache ni mysql, donc on peut imaginer faire un .exe) qui se maintiendrai à jour via ce système de notification.
Mais on peut tres bien imaginer mettre a la place un petit programme qui ecouterait aussi ses evenements, et par exemple avoir un programme "verifications des ponctuations manquantes"
(il aurait un ensemble de regle en fonction, de la langue etc.). Et avoir la possiblité via l'api de poster un commentaire (le robot aurait comme sur wikipedia, un compte expliquant que c'est un robot, ses fonctions etc.)
Il serait donc possible de la meme maniere de faire un robot 'faute commune de traduction" , un autre "faute d'orthographe" (sûrement plus facile)
Pour la question pourquoi externaliser cela plutôt que de l'inclure dans tatoeba directement, c'est que cela permettra de garder le code du site "simple" et propre (le passage obligé par l'api évitant les tours de magie de noirs), et aussi de permetre à d'autre de developper ses bots en paralléle, indépendamment de ma disponibilité pour expliquer comment tatoeba fonctionne en interne, intégrer leur modifications etc. De plus cela permettra d'éviter le fameux "on ne le fait pas car le serveur ne peut pas le supporter en terme de puissance"
Évidemment, les robots devront obtenir un "permis" pour pouvoir accéder a des fonctionnalités comme "ajouter un commentaire"
Enfin je m'égare.
My bet is that if we allow for sufficient time, every sentence will end up being linked to any other at a given degree..."black" to "white" and "yes" to "no"...It's already the case in a few funny instances...
The corrolary of this is that at the nth degree, the number of linked sentences can be just huge, and recursively representing the complete trees would be really stressful for the server...
Germany sentences have sound option, can other users put sound on their sentence?
You can find some more information here: http://tatoeba.org/deu/faq#submit-audio
Acabo de adicionar algumas frases diversas (73) e algumas utilizando-me de lugares-comuns ou clichês (101). Os clichês são colocações sem originalidade, porém utilizados por muitas pessoas.
http://tatoeba.org/eng/sentence...nd/indifferent
My next prediction: 100.000 "Tom" sentences before the end of this year...
1 000 000 frazoj en Tatoeba!
Well done folks!
Proksimume je 0:23 h lau mezeuropa tempo cetere, 14-a de augusto 2011.
Environ 0:23 h (heure de Paris), le 14 août 2011.
At present pace, the number of sentences in Tatoeba should double in under a year...The million sentences could be reached before next summer.
BUMP for those who like history ;)
So, we continued on average "at present pace" and not exponantially growing.
And, above all: the weather in Europe wasn't better last year!
Yeah, so as I correctly predicted, the million took just under a year to be reached...It will be reached today, actually!
And indeed, the weather is even shittier...
>as I correctly predicted
you should've used "should" or "will" instead of "could"...
>the number of sentences in Tatoeba should double in under a year...
haha, yes, but in this sentence you precised "at present pace". So it was just a calculation.
ah oui, en effet, je ne lis pas dans les boules de verre...excuse-moi de ne me servir que de mon cerveau...
je ne dis rien contre.
étant mathématicienne, tu sais la différence entre « prédiction » et « divination »
tu n'as pas prédit que la million prendrait moins d'un ans. t'as prédit que SI CA CONTINUAIT À LA VITESSE ACTUELLE, la million prendrait moins d'un an.
ca fait une grande différence, c'est tout.
oui oui...
Je ne pense pas, en dépit de toutes les bonnes volontés du monde, que le chiffre d'un million sera atteint en 4 mois...
Il faudrait doubler le rythme actuel à 3500 nouvelles phrases/traductions par jour, 120 x 3500 = 420 000 et ce serait tout juste suffisant...car on est même pas à 500.000
Alors à moins d'un miracle...
You did the calculations?
Look at the timeline. We're currently adding a bit less than 2000 sentences/translations per day, and growing...
Where can you see those statistics?
Thanks
on the home page.
The growth should be exponential though. I bet you we double in 4-5 months. One million in 6 months.
Not exponential, because it comes in burst. There was a sudden burst in July (and I was part of it) because emails were sent in different communities. But bursts tend to tame down. History shows that some very active users at some point stopped suddenly. We will see if current pace will be sustained, come september...
It should be exponential. Bursts are perturbations, which get evened out in the long run, like you said. I fit it, and the exponential model fits quite well.
According to this, Tatoeba may very well reach 1,000,000 sentences on the 1023rd day of its existence. C'est-à-dire, in 273 days, or roughly 9 months.
But that's an even-weighted model. If we give the recent few months more weight (though I don't know any justification for doing so), it might fall down to 5 months or even less...
I don't think the current months are any representative. Many contributors here are students and they will be busy soon enough and also the weather in Europe this august is particularly shitty...
but I agree with your analyse, I don't think it will become exponential, for some reason:
1 - as sacredceltic a lot of contributors have come in a burst, and we're still in the "high motivation" part, most of the time it last one month or two, and after their contribution rate drop, so it suppose not only new comers will come to "replace" them, but new ones too.
2 - I don't think for the present time tatoeba is able to support 10000 contributions a day ^^ (but I can be wrong, and I hope I'm wrong)
Anyway, maybe we can start a betting pool...
I say May 5, 2011.
No, I still hold that it should be. The word "exponential" shouldn't be taken to mean that it follows some sort of super rapid growth. It just means that when taking all the data of Tatoeba's growth, from the day of its inception, plotting that data, and fitting the curve, the exponential fit works better than anything else. That's why there were so few contributions for such a long time, and then the steady growth in the past half-year or so.
Again, bursts play no role here. The fit ignores the bursts, since it looks at the entire two years.
10000/day is exactly what you might get on day 1023 (you did the calc, didn't you?)... Well, you have almost a year to build up the server.
spoiler: this year I will have to take care of hundred of chinese students in China, and one of my attribution will be to teach them French, and to help the French students there learn chinese, so guess what I will use ...
That could be a booster...at least for these languages.
I think also promoters of Tatoeba should take contact with defenders of minority languages such as Amerindians who sometimes are desperate to save their near defunct languages and could see Tatoeba as an opportunity with their young.
There are contacts to be taken with universities across America but also Australia, India, etc...
In fact, as I will have (I hope), some free times, I will take time to see native speakers of minority languages in China (mainly Shanghainese and other Wu dialect as I will be in this region)
But I agree, rather than a "absolute" adding rate, my "hope" is not to reach a million, 10 million, but to boost contributions in our current minority languages, and add new ones, especially endangered ones.
Then we'll probably be there in nine months.
OK, you inspired me. I'm going to do it in Excel right now, and tell you when the million will come.
Une fonctionnalité vraiment sensationnelle, ça serait de pouvoir chercher une phrase dans la langue A contenant le mot M et traduite en langue B par une phrase ne contenant pas le mot N.
ou l'inverse : une phrase dans la langue A ne contenant pas le M et traduite en B par une phrase contenant N.
Ça permettrait de chercher des traductions rares, parce qu'on est trop souvent submergés de traductions classiques trop fréquentes...
Careful!
maureenbaby is a spammer. I've just received his/her message.
Thanks, I've deleted the messages.
Mine got through anyways. (Finally! Somebody likes me!)
Perhaps there should be a message sending throttle.
got sth suspicious from @maureenbaby. same story?
I think maureenbaby is anything BUT suspicious...
*** Statistikoj pri kontribuantoj ***
Mi analizis la unajn cent kontribuantojn.
Analizo per aĝo: ne konata: 39 , meznombro el 61 konataj: 38,6 jaroj
Analyzo per lando: ne konata 19
germanujo 16
francujo 6
ukrainio 6
belgujo 5
brazilo 4
hispanujo 4
aŭstrio 3
britujo 3
hungario 3
nederlando 3
rusujo 3
usono 3
argentinio 2
italujo 2
kanado 2
polujo 2
turkio 2
barato 1
belorusio 1
bulgario 1
ĉinujo 1
danujo 1
finlando 1
islando 1
israelo 1
japanujo 1
koreo 1
meksikio 1
norvegio 1
vjetnamio 1
Analyzo per indikita gepatra lingvo: ne konata 1
germana 19
angla 9
franca 9
rusa 7
hispana 6
nederlanda 5
portugala 5
ukraina 5
hungara 4
itala 3
persa 3
pola 3
araba 2
finna 2
turka 2
belorusa 1
bulgara 1
ĉina 1
dana 1
flandra 1
hebrea 1
hindia 1
islanda 1
kantona 1
malaltsaksa 1
norvegia 1
ŝanhaja 1
ŝvaba 1
ukrajna 1
vjetnama 1
mi precizigas ke pri la mezaĝo, mi ignoris la profiloj kiuj indikas fantaziajn aĝojn tiel Dejo aŭ Eldad (6 aŭ 7 jaroj...)
bonjour! je m'appelle max et j'étudie la langue française. je viens des philippines. je parle l'anglais et bien sur la langue maternelle de mon pays - le filipino/tagalog.
Bienvenue Max ! Tu es à la bonne adresse pour améliorer ton français et n'hésite pas à traduire toutes ces phrases françaises en tagalog !
En Tagalog ..
Magandang umaga! Ako si Max. Nagaaral ako ng Pranses. Nagmula ako ng Pilipinas. Marunong ako ng Ingles at siempre ng aking inang wika - Filipino/Tagalog.
Bienvenu parmi nous max :) sinon sur ton profil il y a une petite faute (le reste est correct, donc ce serait dommage de la laisser)
"ma langue maternelle est *LE philippin/tagalog"
ah oui. merci mille! ;p