Sentences don't kill Tatoeba. Users kill Tatoeba... (when given sentences).

Would it be possible to extend the "Show All Sentences Not Directly Translated Into..." to "...Not Indirectly Translated Into..."?

This would be really, really useful.

Does that include one of racist ones by Russians about Russian minorities? Can you send that one to me, or make it into a list, because I find them to be absolutely hilarious...

It doesn't seem like one worders should be a serious issue, since you can only have so many valid ones before you run into duplicates. And what's wrong with having a one-word imperative sentence for each verb in the language?

^ö^

I vote 5 an infinite number of times! This idea is infinitely perfect.

What do you guys think of this idea?

1 - Atrocious
2 - Terrible
3 - Horrifying
4 - Bad
5 - Not good

Tatoeba with Ratings:

Okay, this is probably a really bad idea, but... If we wanted to experiment with what Tatoeba would be like with ratings, we could technically try and simulate it with tags. Simply leave a tag "good + #" or "bad + #" for a thumbs up/down system, or "1-10 + [some string to make it unique]" for a 1-10...

Just a thought.

Haha... can one person vote multiple times?

Hah! Too funny...

Transliteration seems to be up for Uzbek.

Question: do you know why е is transliterated as "ye"?

Some corrections:

*I can’t check Persian _yet_.

*I don’t know it _yet_.

*But other people can _for now_.

Nevermind, it's just a flag display error on the main page.

There's just the mystery of there being 71 sentences in one and 69 in the other... I was sure I always added them in pairs. I wonder what happened.

Exponential growth!

Except for the early ones... I think that many of those are linked by "unknown".

I've been meaning to bring up the same point, but feared I would sound repetitive. Hasn't somebody mentioned this idea somewhere at some point? I forget what was the reason for not implementing it (well, it's probably a lot of coding, again...)

But yea, I agree. Adding the third layer would be really, really helpful - especially when doing the chain translating. I also think that it's very true that this is where the majority of duplicates come from.

> I've mentioned to Trang that I think it would be useful to have "owners" of links as well as sentences. That would be one way to get the attention of the interested parties.

I completely agree with you here. There currently seems to be a duality in Tatoeba between the quality of an individual sentence and the translations it has. If ratings ever come to be realized, this will also resurface as a problem (what do you rate - the naturalness of some given sentence, or the translation?) IMO, the translation is the most important as it is clear who's the responsible party for *both* the naturalness and translation of the two sentences. The only time an owner of any given sentence would play a crucial role would be when he/she creates a sentence from scratch and leaves it untranslated (but as soon as someone does translate it, that someone also takes partial responsibility for the original sentence).

Anyway, it's all a lot of work, and probably wouldn't come for a long, long time unless sysko wins the lottery...

> As I was writing my last comment I got to thinking that it might not be too difficult to write a script to send comments to all sentences linked to a specific sentence and tag them with "@translation check".


This would still require going through and leaving comments on all the tagged sentences, since it wouldn't be clear which link the check refers to.

> It's either that or wait for a person who knows all the languages.

Exactly. No person will know all the languages, but from a probability standpoint, the tagger or the owner would be more likely to know more of them.

But yea, the problem grows very complicated very quickly as people translate/link... There needs to be something in the interface to handle this, cause the lot-of-work method won't be practical. I think you'd agree that few will take the time to do all that, and even if they do, it's just for 1 sentence out of 500,000+...

> "In your weird sentence case, just change the sentence to what you consider the most likely one (or just pick one at random), mention the others in a comment, point to it in comments on the linked sentences and tag them with "@translation check"."

Not that I'm lazy or anything, but that's a *lot* of work for an orphaned sentence... Few would be willing to do that. Especially when the sentence has a lot of direct translations.