@Tom9358 Hello, i have this problem, too. I've tried several languages, but the outcome is still the same. The random sentences feature also doesn't work properly.

Exact same problem here.

The search is now working again. Sorry for the inconvenience!

Thanks for checking in with us.

There is a page that covers the process of making a language request:

https://en.wiki.tatoeba.org/art...nguage-request

It's part of the Tatoeba wiki, which means you can find it by using the link "Tatoeba Wiki" at the bottom of any page.

From the introduction: "A new language should be requested only if the requester is planning to add a substantial number of sentences in that language."

Ervin should set up a Tatoeba account and follow the procedure described on the page.

You could send the Tatoeba user sharptoothed (the maintainer of that page) a private message.

Thank you, this is very helpful. How did you decide which sentences to add?

Hey! Nice list! I'm sure it contains a very large portion of important aspects of English grammar. If somebody would like to tackle this in a research-grounded way at some point, perhaps the following reference might be of value. This paper comes with a collection of all sorts of English ungrammaticality:

Sprouse, Jon, Carson T. Schütze, & Diogo Almeida. 2013. A comparison of informal and formal acceptability judgments using a random sample from Linguistic Inquiry 2001-2010. Lingua 134: 219-248.

Jon Sprouse provides an xlsx with a list of example sentences here (under Journal Articles): https://www.jonsprouse.com/research.html (control+f for "formal acceptability" and download the materials spreadsheet)
Each ungrammatical sentence has a grammatical counterpart, so I think that means that all grammatical aspects listed in the spreadsheet have a sentence that is grammatical according to that aspect.

I used it for my master's thesis back in early 2023, when ChatGPT was brand new and we were wondering if it would be able to a. tell us whether a given sentence is grammatical or not, b. repair it and c. explain why it was ungrammatical.
The conclusion was that a. went quite well, b. went alright, and c. it did quite badly - just like humans.

Any opinions on this? I personally believe that separating Southern Quechua (from cusco) 🇵🇪 from Kichwa of Ecuador 🇪🇨 from all the other variations would be a great improvement. As I said, the cusco variation is the most standardized one and the one we should probably invest the most effort on

Because classifying language is hard work and the Tatoeba project has low ressources, it strictly follows the ISO 639-3 language classification operated by the SIL. This classification has a concept of macrolanguage [1] and Quechua happens to be one [2].

Actually, Tatoeba should not be treating macrolanguages as individual languages because it leads to the kind of issues you are describing. This problem also affects other language families on Tatoeba and we have already identified it [2] but it is yet to be resolved.

Tatoeba is a volunteer-based project, so nothing will happen unless people affected by/interested in this particular issue proactively help. To people who want to help: you can use tags to identify Quechuan languages, as AlanF_US suggested. As for the tag names, you can use the names or ISO codes listed in [2]. You can also open an new issue on GitHub [4] specifically for transitioning sentences in Quechuan languages [que] into each individual language. Ideally we need all Quechuan sentences to be properly tagged, or at least to know what member is using what language. We also need to decide on a language icon for each Quechuan language (maybe the current flag + iso code on the side could do).

[1] https://en.wikipedia.org/wiki/I..._macrolanguage
[2] https://iso639-3.sil.org/code/que
[3] https://github.com/Tatoeba/tatoeba2/issues/1673
[4] https://github.com/Tatoeba/tatoeba2

We're supposed to translate sentences without looking at that sentence's translations into other languages, meaning the gender in English shouldn't enter into your translation.

Your options are:
1. pick a gender to translate into
2. translate into both genders. we're allowed to translate the same sentence multiple ways

In practice, I find looking at other translations as a reference to be helpful sometimes, but they shouldn't affect your translation's choice of things like gender.

No, they will continue to be theirs indefinitely. A profile is usually only set to "inactive" (which enables adoptation of sentences by other users) if they request it, or in justified exceptional cases.

Further supporting information for adding Tawallammat Tamajaq [ttq]:

To demonstrate the digital readiness and active support for this language, please note that there is already a substantial contribution on Glosbe, with over 250,000 translated entries spanning Arabic, English, and French. Please note that these are recent contributions, so they might not be fully reflected in Glosbe's public statistics yet:
- Tamajaq to Arabic: https://glosbe.com
- Tamajaq to English: https://glosbe.com
- Tamajaq to French: https://glosbe.com

Additionally, an electronic dictionary for Tamajaq is fully available. We can actively leverage this resource as a reference to contribute, verify, and expand high-quality sentence pairs directly on Tatoeba once the language is approved.

Thank you for considering this request!

I'm hiding your comment not for the usual reasons but because you've already made the same comment on GitHub. As gillux said a few hours ago:

https://tatoeba.org/en/wall/sho...#message_42003