Menu
I've got a message sent by "Anylove", titled "Hello, my dear", which I deleted unread. Such messages are written with criminal intent, and you shouldn't expect that in their content is anything true.
The same stuff appears in ''his/her'' comments to the sentences!
That Anita keeps sending and repeating her messages, please /block user ANILOVE.
Ankaŭ mi ricevis ŝian/lian mesaĝon.
Le problème est que Tatoeba n'a pas de système de filtrage des inscriptions comme des captchas, qui permettent de filtrer les robots. C'est pourquoi, parmi les 8000 utilisateurs de Tatoeba, près de 6000 ont été créés par des robots.
Comme vous pouvez voir ici, leurs noms d'utilisateurs ont été créés par un algorithme : http://tatoeba.org/epo/users/all/page:408
I also received the very same message, I back the idea of block this bot. It'd be a problem if this kind of "users" start to pop out of their lairs...
Is it difficult to add a captcha in the register? I hope it doesn't.
The problem with captchas is that they are not always well understood and deter people from becoming users...
Ok but where I said captchas you can say alternative methods as «sum this two numbers»,«write the last 5letters of the word xxxxxxxxx»,systems.
That is indeed better acorrding to me, as long as questions are random, otherwise hackers will find a way to automate this as well...
A good Tatoeba-in-house trick could be to pick a sentence at random and ask in which language it is written (with a limited list of choices)...
Or even simpler: show a flag, and ask which language it matches...
Of course, the questions have to be in the language of the speaker...that's the difficult part, since otherwise you bias the service, excluding people who don't understand the questions...
I think I have a much better idea:
How about, if you don't get the possibility of sending PMs until you contribute with, I don't know, at least 15 sentences, or something like that?
We kill two birds with one shot ;)
That's a brilliant idea. Tio estas brila ideo.
Boracasli already created scripts to produce as many sentences as he wants, with simple dots...
The process of filtering must be automation-proof. So far, only captchas or questions (non repetitive, non predictable) cannot be automated.
On the other handm it's the new users that need help the most! They couldnt contact other users for help then.
This was also my concern. There might be users who join the website mainly to indicate a mistake in a sentence and only then begin to like the work and get working on translations.
Also, if someone joins, while he's getting his 15 sentences done, he might still be doing many mistakes (forgettings periods, not translating directly, including annotations or the-like) and while people could contact him, he could not answer until he reaches 15 sentences.
So I don't really like the idea with the 15 sentences...
How about captchas only for a user's first ten or so messages? That should keep spammers at bay and it wouldn't affect long-time users.
Captchas just for messages, eh?
Hm, that might be a good idea. On the other hand, might spammers then simple drop the idea of adressing users and instead put their spam messages in comments below sentences or as sentences themselves?