Nostre datos es liberate sub licentias de Creative Commons varie.Plus information
Si tu ama iste contento, per favor considera un donation.
Menu
Is there a way to make new members read the rules before contributing?
+1 A simple line of code could be written that detects a user's first contribution on the website and automatically displays a message containing the rules.
I was thinking about something like that, yes.
Something like the following text would be short enough that they might likely read it.
https://en.wiki.tatoeba.org/art...rt-quick-start
I think it would be better, though, not to link to the wiki page which takes people off the main tatoeba.org site.
As an alternative (or in addition) to displaying information when a user first adds a sentence, we could add this information to whatever they see when they're signing up for an account.
I think most people skip the "terms" while signing up for accounts.
Terms of use are typically very long and often include information that seems obvious or irrelevant. We're talking about something much shorter, like:
- Write complete sentences, with correct capitalization and punctuation, in a language that you know well. Don't take sentences from elsewhere unless they're in the public domain, and don't use AI-generated or machine-translated sentences.
...and make sure you're adding your translation to the right sentence.
I'd like to understand this better. Are you frequently seeing people adding translations that have nothing to do with the originals? Does it seem as though they intended to add translations to other sentences that they saw recently? Do you have any examples?
No. I'm seeing people adding translations to any language they see meaning to translate the indirect English or other language they know.
They see, for instance, this Russian sentence:
https://tatoeba.org/ru/sentences/show/13788899
Russian isn't listed in their profile among the languages they know, and they obviously don't understand it. However, they understand the indirect English (I want Tom to read this.) and they add their translation to the Russian sentence instead of adding it to the English one.
Then they translate from Finnish, Marathi, Berber, Chinese etc.
See e.g.
https://tatoeba.org/ru/activiti...Tigre?page=200
or
https://tatoeba.org/ru/sentence...omment-1570172
https://tatoeba.org/ru/sentence...omment-1570171
or
https://tatoeba.org/ru/sentence...omment-1567660
I see it happen once a month or so. Sometimes, by the time I notice this, hundreds of sentences have already been added.
Ich glaube, dass es nicht viel bringt, wenn man versucht, die Leute zu zwingen, die Regeln zu lesen. Es wird genügend Leute geben, die das einfach wegklicken, ohne es richtig durchgelesen zu haben. Und andererseits besteht die Gefahr, dass man Leute, die sinnvoll beitragen wollen damit vergrault.
Was war denn der Auslöser für deine Frage? Welche Fehler machen die Leute?
Ich könnte mir vorstellen, dass eine Revert-Funktion sinnvoll wäre. Soweit ich weiß, gibt es sowas bislang nicht: Administratoren (und vielleicht auch Korpus-Maintainer) könnten sich anzeigen lassen, welche Aktionen ein anderer Benutzer vorgenommen hat und dann bei jeder anklicken, dass diese revertiert werden soll. Wichtig dabei ist, dass man viele Änderungen auf einmal rückgängig machen kann, und nicht jeden Satz einzeln anfassen muss.
Oder man könnte sich auch Kriterien überlegen, die automatisiert bestimmt werden können (kein Satzzeichen am Ende, sehr kurz, etc.) Wenn nach 5 Sätzen viele solcher Kriterien erfüllt sind, wird der Benutzer erst mal temporär gesperrt und ihm mitgeteilt, dass es so aussieht, als hätte er die Regeln nicht gelesen. Sollte man natürlich netter formulieren.
I think it's worth the risk.
Once a month a new user comes and starts adding hundreds of non-sentences without paying attention to comments nor (sometimes) to PMs. What’s worse, other users immediately start translating them, and thus the situation is snowballing.
Every other new user keeps adding sentences without proper capitalization and/or punctuation.
Now and then, a new user starts translating from every language they see, not understanding what direct and indirect translations are.
Some of them do that on purpose, using machine translation or AI.
Let alone those who add sentences in the languages they don't know.
There’s no need to mention that it ruins others' work.
So maybe we should take a risk and do something to make them read a few lines that Alan mentioned.
I'm not mean, just fair.
Regarding your suggestions, I'm not sure whether we have resources to carry out the first one, and whether it will be transparent enough. As for the second one, it’s really worth considering, if it’s not too hard to implement.
I think, the implementation shouldn't be too difficult (*). But first of all we need a list of such criteria; they need to be language specific, of course.
I try to collect some criteria which should give a penalty:
* Sentence is very short. Maybe extra penalty if only one word.
* Wrong capitalization.
* Wrong punctuation.
* Contains a symbol that's unusual in that language.
(*) Actually I'm struggeling a lot with implementing something in Tatoeba. Stuff that I thought would be implemented in a few minutes takes weeks and I already had had to give up on one project. So maybe my estimate here is wrong too.
I do not agree about the short sentences, or think they should be at least dependent on the language. E.g. in Spanish they are more common than in English, because there is more conjugation of verbs, and pronouns can therefore be dropped. New users may feel more comfortable translating shorter sentences, and they're also what comes up first when you search by relevance, so this could lead to them being unfairly blocked. Maybe in this case instead of blocking them outright, it could just flag them to be checked by a moderator.
For the other suggestions, I agree they would be helpful, but I also wonder how difficult it would be to create an accurate set of characters, punctuation, and capitalization rules for all the languages on Tatoeba, especially as some languages use multiple scripts.
I agree. Actually, a simple message for new users would already be of great help.
I meanwhile think it would be a good idea to issue a warning, whenever a user adds a sentence with some abnormalities. This would be useful for more experienced users as well.
For example, the last symbol of the sentence can be rather easily checked. I counted them for all languages with at least 10,000 sentences (top 69). This is my result:
jpn, cmn, yue: 。？！」…）：
ara, pes, ckb: .؟!")
hin, ben, asm: ।?!."
tig: ።:?.!፧፥፡
hye: ։:
ell: .;!"'”“»:…)
yid: .?!״”‟":
zhg: .?!"
jbo: aeiou.
all others: .?!"”':…)»“
I'm not familiar with all of these languages. If you are familiar with one of these languages and you think an other symbol should be added, please tell me. And if you think, some of them shouldn't be included, please also tell me (for example I'm quite unsure if the dot is considered correct in jbo).
This can be implemented in an easy algorithm, skipping all the languages that are not in the top 69. Of course, this check isn't perfect, because, for example french sentences should not end with ohne of those: "”'“ This list might be refined of course, but for a start, I think it would be sufficient, especially if we do not talk about stopping someone completely, but only about issuing a warning.
PS: I added an issue at GitHub: https://github.com/Tatoeba/tatoeba2/issues/3264
> Maybe in this case instead of blocking them outright, it could just flag them to be checked by a moderator.
I think it would be very useful to have some sort of menu or autotags that would allow moderators to easily double-check new sentences (maybe not just those written by new contributors, but by anyone; perhaps with a different tag or something). I suppose adding an autotag system wouldn't be too difficult, but it would be very beneficial since it would make it easier to double-check new sentences and show what has and hasn't been reviewed by a moderator.
It's not practical to have every sentence should go through a moderator before it is considered okay. It's not even practical to have every sentence contributed by a new user go through a moderator.
I'm not saying it should go through a moderator before appearing in the corpus or something. I meant a tag or something that would make it easier to find unreviewed sentences, like removing marked sentences from a search. I'd love to have something like this for reviewing Toki Pona sentences while excluding previously reviewed ones.
I would suggest the following additional "warning sign" criteria: if a brand-new user translates into/from many different languages within a very short time of joining the site, or translates from languages that are not added to their profile. (Users are already prevented from translating INTO languages absent from their profile, but a user must have some skill level in a language in order to translate FROM that language as well.)
As far as how easy these are to implement - that's another question entirely. I've had the same experience as kumakyoo with developing on Tatoeba. (I have not done anything recently, but I added some things a few months ago.) Gillux is a great help when it comes to getting familiar with the code base and workflow, but it still takes a long time to implement very simple things.
You make a good point about the asymmetry in the fact that a user cannot add a sentence in a language they don't know, but they're free to translate a sentence in a language they don't know. I think that's an oversight, and we could get a lot of advantage, without obvious disadvantages, by not just warning users who do this, but preventing them from doing it in the first place. I wrote an issue ticket (3266) about this on GitHub.
Thanks, Alan!
(BTW, I think the link should be: https://github.com/Tatoeba/tatoeba2/issues/3266 )
When I wrote "#3266", I wasn't intending it to be a link, and it didn't occur to me that Tatoeba would treat it as a sentence link. Thanks for providing the actual link. I couldn't access it on a computer where I wasn't logged into GitHub, but hopefully the people who want to see it will be able to.:)