Hello, corrected, there was a little database corruption.

Right now it's no possible, in the future I will try my best to offer an adequate API that would permit external software to access Tatoeba in order to permit third-developers to provide additional functionality or simply to access to some of the already present features without needing to open a browser.

IMPORTANT

Concerning translation mismatch and what to do
(I blame nobody as I do understand some reason are not straightforward)

1 - What do I mean by translation mismatch


it's when we have for example

sentence A and sentence B linked

sentence A itself is correct , natural etc.
sentence B itself is correct , natural etc.

but sentence A is not a translation of B and B not a translation of A


2 So what's wrong

Very often I see people asking to each other and debatting about what is the correct/original "sentence" in order to know if it's A and B that need to be corrected

3 Ok but why it's wrong ?

Because the problem are not the sentences themselves (remember we are in case that both sentences are, taken alone, correct)

it's wrong because

A - That may require to "adapt" translation in cascade

Indeed, most of the time we do not have only sentence A and sentence B, but maybe sentence C (translation of B) , sentence D ( translation of C) etc. , so if the Owner of B change the sentence B, he will need to notify the owner of the sentence C that will himself need to notify the own of the sentence D etc.

That's wrong because :

At one moment someone will not be "active" anymore, and then the sentence will not be adapted, leaving this "mismatch"

B - Maybe people are recording this sentence

For example even in the case we have sentence A and B, one may think the previous problem does not apply and then it's fine

But imagine at the same moment someone is using a dump of Tatoeba (containing sentence B) to record audio version of it, then we will finish with a audio and sentence mismatch that will be very long to find (as most of the time, especially for this case, the modification will be minor, and maybe only attentive listening, or listening by a native, may see the difference AND comment about it)

And in that case that will certainly mislead at lot of learner.


So What to do instead ?

Answer : Unlink and tell on the "unlinked sentences" that it was unlinked


That the best solution because

1 - We kept both correct sentence, solving the two problem before

2 - the owner of sentence B may add a sentence B' match sentence A and then learners, will have a sentence B and B' (and maybe a sentence A' to match B) with a slightly different meaning

This may help learners to see the difference of meaning between two slightly difference sentences

So if you see something asking for this, please post a comment that link to see, so that soon everyone will do the "best practice"

(NOTE, I'm currently developing, as part of the new version, but fortunately enough independent from it to be released before the full new version is achieved
something that will permit us to centralize all this messages about important statement/explanation)

If people feel like translating it, I will be very thankful

hmm it seems my personnal server (on which the language detector is hosted) is experiencing some problems, so the autodetection is not available yet, i hope to solve that soon.

I've added audio recording that marcelostockle send me in

spanish (more than 1000 new audios recording)
and
esperanto

a big thank you to him

hmmm indeed for freefighter the detector has only english for you , something certainly went from when the detector was re-generated

but for alexmarcelo you're marked as contributing in english, portuguese, spanish , french and italian

I'll try to set up a webpage to detect a language of any text with possibility to precise or not a given user, so that to diagnostic where the problem come from

actually yesterday i've restarted the server (added more hard disk spaces) and forgotten to restart the pinyin generator , it has been fixed for some hours, but just to let know.

as you said, certainly a spelling mistake,

After you also need to wait me to make a mistake and forget to update a table, that part is more tricky to do on purpose

after no your bug and that bug were different

we have the ":" bug that is well known

and the bug (caused by an error from me) that make possible to have sentences linked to non-existing (no more existing tag) if you manipulate directly the database

so when we arrived at the point that the bug 1 and 2 where both together, the current situation happened.

and anyway once again as I'm the only one to be able to make the bug 2 to appear (for scientific purpose you can tag a sentence with a useless tag, and i will make it an empty tag, and you will see that you will appear as the owner of that empty tag)

question 1: as said before (did you read me?) because i'm merging them from times to times and error human est , these time i've may have forgotten to also update the "tag -> sentence" table

question 2: you really didn't read me, i've said it 3 time
because there's 3 table
1 sentence
2 tag
3 tag->sentence
if a link in the third table point to a tag that is non existent, it will have such a tag.

question 3: as said the bug makes the first sentence with the lower empty tag to be displayed. the tag were introduced at a period at which this two users that you don't appreciate where active. After the first tag being on offensive sentences ... after I'm going to repeat myself , if i point out something funny people will say "oh look he point out something, and know what what he point out was something funny, that can't be a coincidence"


---- if you're to read one thing , at least read that ------

the empty tag have been created only by a mistake of me, so as CK does not control my mind, i'm the only one who have created empty tag because of a mistake of mine,

so question 1 : how can CK know that i will make a mistake on merging this specific tag ?

so if i'm to think like you that it's not a coincidence, then I'm the only one who could have done this, because I'm the only to be able to create empty tag and by so being able to exploit the bug

seems also a reasonable advice to me.

so when i was saying "reconstructing the tagging" process, it's not what jakov did, because jakov only can see the table of my previous post (well it's true it's not the username but the user_id, and what jakov script is simply to replace this user_id) , and as said that show nothing of what was this original "tag ID 42" (or whatever id it is)

What i was talking about was to look which tag have been applied/deleted, which is not possible to know directly as i have no log table of this, but it should be possible if looking to the log of the URL call on the server + looking in the dump which tag was present on a given sentence at a given time

first as scottt show, there was not one , but a lot of sentences with this empty tag, and as told you the one shown was the one with the lowest tag_id

second, in which language should i repeat that, there's no hidden feature that CK can access to, if he has been able to apply so much tags that fast it's because the URL system permit the things to be scriptable , ask Jakov

the url are always /module/action/parameter1/parameter2 etc. , like /sentences/link/sentence1/sentence2, the same things apply for tags, so it's just that instead of using the interface, he made the same things that Jakov do with some of his script.

And once again, the tag was not created empty, it was made empty because that's what happen when me, I, sysko merge a tag and i forgotten to merge the "link" , so when i merge let say the tag "offensive" (id 33) and "ofensive" (id 32) there's a "link" table that say

tag_id sentence_id added_by time
32 1 CK 12/12/12
32 42 Toto 12/12/13
33 55 Ck 12/13/42

if i merge the two tags but i forget to merge also the previous table, you'll have this "tag number 33 on sentence 55" , but then tag 33 is now non-existing, but the information about when it was added are still there, so you think that the tag was added originially empty , which is not

yep for the bible I think keeping the same

from SOURCE will permit an easier conversion in the future when I will move these tags to dedicated metadata like

KEY => VALUE
author => sacredceltic/Goethe
source => The bible/L'île mystèrieuse
etc.

moreover if that bug would have applied to a "nothing special sentence" , i doubt that Scott would have post it on the wall and would have, as he did before, report the bug to me by private message.
So for me saying this is the same as saying "oh see that's a sign of god" if i saw to someone a photo of for example a cloud making a letter, occulting the fact that i didn't show you the thousand of photos I've taken before that get nothing special.

it would have been applied to "fuck your mother" or "your sister is a communist" you would have made the same "why that precise one" comment.

If you really continue to deny the evidence i will dig into the dump of tatoeba and reconstruct the history of the tagging and merging process, that would be long, not interesting but if that's really needed ...

yes you're right about it, i used to do such merging in the past, basically i think the best way to handle it is

1 - we take one case : for example the "idiome/idiom/idiomatic expression"
2 - people discuss about which one is the best
3 - once one is chosen i merge them

Note: step 2 may be never-ending but of course with people keeping focus, we should quickly arrive to step 3 , and then quickly jump to an other "case" etc.

First the tag was not created empty, because as said that's not possible.

After why this specific sentence, it's because that empty tag (well know it's not possible to see but each "empty" tag had actually different id, the id of the 'former' tag they came from), was the one with lowest id, and as the first people to use tag was mainly CK for tagging the "non-politically correct" sentences (as you can that one contains other tags from him dating from 2010, i didn't check but from my memories that was very soon after the tag feature was introduced) that "ghost" tag was certainly one related to offensive sentences.

So it wouldn't have been this sentence it may have been any other offensive sentence.

ok c'est résolu aussi, pour "comment" les tags vides apparaissaient, il y a une explication détaillée en dessous, merci de m'avoir fourni directement les liens vers des phrases contenant ce problème.

No actually CK has not such a privilege because there's not such a feature (you can check the code of tatoeba as it's public)

After looking to the sentences containing that Scott provided in an other post of that thread, I've investigated. Here how it's happened

An empty tag appear when a sentence say it contains a tag that does not exist anymore (for those speaking sql that happens when in the table that link "sentences" and "tags" , there's a "tag_id" that point to a tag that does not exists anymore)



CK created a tag on that sentence (maybe to say it's offensive what so ever, i've no way to track this, or actually if you really want to know i've the dump of tatoeba of everyweek, so "simply" need to reload them in an empty database and check for the last dump what kind of tags this sentences had).

When merging tag because of spelling mistakes/similar spelling (Ok vs OK vs O.K. for example) I may have forgotten for some of them to also update the table that join sentences and tags, leaving some "links" to some tag_id that point now to nothing, creating by that way an empty tag

this in addition to the ":" bug created that situation.

So it's impossible that CK or any user created that on purpose because it's impossible through the interface, as you needed to be able to delete a tag without deleting the "links", which only possible by entering directly SQL command.

Case solved, i think there's one sentence which need to be deleted...