A web app to read Latin texts with inline translations

  vaguely Latin-shaped set of gibberish that ChatGPT nondeterministically generated

OP hadn't used AI to generate Latin, but to generate English.

You're severely overstating the case here, as if it just produces random text like a markov generator or something.

I do know Latin (quite well), and gpt-4o does an extremely good job at translating from English to Latin and vice versa.

> It's a book that starts off with very simple sentences and gradually introduces more complex topics like tenses and declensions as the story goes along.

I'm not sure this is actually an approach I'd recommend. I was recently asked to give some supplemental English tutoring to a Chinese brother and sister, 9 and 5 years old. The 5-year-old could already use and understand 'simple' sentences such as "what do you see?" and "where is your brother?", though I'll note that the subject-auxiliary inversion required by a question of that form isn't exactly a simple concept.

I got them a copy of The Cat in the Hat, and their mother objected that it was too advanced for either of them, because most of the verbs in The Cat in the Hat are in the past tense, which apparently isn't covered within the first four years of Chinese English instruction.

You can't learn what you're not exposed to, but you can learn a lot of what you are exposed to in a language.

Meh, when you say Duolingo isn't great, you should compare it to something in its same class. I don't think Duolingo competes with sitting down with a grammar book. The whole reason I used Duolingo, which got me up to speed with Spanish enough to read books, is that I could do it anywhere and any time rather than bust out a book for serious study. I was never going to do that, so Duolingo was strictly better for learning a language than a book.

People lose track of that every time they dunk on Duolingo.

I just wish Duolingo would move faster.

I've been working on a tool to display the structure (and glosses) of a text. All the structural and definitional annotation is provided by me based on personal knowledge. So far I have most of a single short fable in Mandarin Chinese done. ;D

Does this sound like something you'd be interested in checking out?

> The choice of the word “fer” isn’t the most obvious choice for “bring”, though.

Well, if you asked me how to say "bring" in Latin, that would be my first choice, and the irregularity of the verb tells us that it's very common in general, though not necessarily for this.

Lewis and Short has "In general, to bear, carry, bring"; "In particular, to move, bring, lead, conduct, drive, raise", which seems to hit the concept of "bringing help" squarely in the center. https://www.perseus.tufts.edu/hopper/morph?l=fero&la=la#lexi...

I am still working on the app part of the website (https://app.adhyeta.org.in/). Will enable registration/login in a week or two.

This is how the interface looks like/works, when logged in.

https://www.adhyeta.org.in/a/images/z85.png (ruby annotations can be seen floating above the source text)

https://www.adhyeta.org.in/a/images/z88.png

The "app" is laid out exactly the same way as the www version is. The only difference is that it provides tools to manipulate the status of words. Login is required because vocabulary progress has to be tracked on a per-user basis.

Ruby annotations?[1] They were originally meant to provide pronunciation hints for Asian languages. The hints naturally float above or below the corresponding word. You do not have to play around with tables/grids/flex to make this happen.

[1] https://developer.mozilla.org/en-US/docs/Web/HTML/Element/ru...

It does seem under resourced. Version 5 never appeared and I don’t know what 6 is supposed to look like. It’s still a great online resource, but having my own library, and access to the Hellenic and Roman Library in London, I’m happy to take things slow these days.

The accuracy of the results might depend on the language you’re reading, the LLM you’re using, the nature of the text, and the amount of context you give to the LLM. When I’ve tested the best models with more or less standard texts, such as excerpts from novels or news articles, in English, Japanese, and Russian, the results have been extremely good. The latest versions of ChatGPT, Claude, and Gemini are able to explain the meetings of words quite well, and they also get the grammar correct. (I say this as a long-time language teacher and lexicographer. I have written and edited many textbooks and dictionaries for learners of English and Japanese; LLMs come close to my ability and maybe exceed it sometimes.)

They are not always so good, however, with more granular aspects of language, particularly the way words are written or pronounced—the problem the models have with the word “strawberry” is well known. I’ve also seen them struggle with the meanings of words and sentences in isolation, as a lack of context can confuse them (as it can confuse people).

In the case of emails or transcripts, the text might contain mistakes or non-standard language that might trip up the LLMs as well.

In any case, at least for major languages and non-critical applications, I think LLM’s are a great way to understand what is written in another language.

It is very good, that's how I learned Latin. For those wanting to speak it, there are online chat rooms I used to use.

This is also quite common in some more synthetic languages, like most slavic ones for example.

And the latin case is made muuuch more extreme due to the fact that we read poetry and stuff written in the yambic hexameter, which would have also been quite alien sounding to an average roman citizen on the street.

It’s like passing data around in a list or tuple, versus key-value objects or dictionaries.

I remember in school one trick was always to start finding the verb and then use that to build a broader context of the phrase

People might find it interesting to know that English still has remnants of its case system in its personal pronouns, for example I/me/my/mine, he/him/his, she/her/her, who/whom/whose.

Some people struggle with who/whom but in fact it’s the same as “he” versus “him”: if you replace “who” with “he” in the sentence and it sounds wrong then you should be using “whom”.

I know various e-readers have the ability to look up unfamiliar words or phrases in a dictionary by tapping on them, so you can focus on only looking up things that you don't know, or look for another definition for some word if the definition you have doesn't quite fit. For instance, I would probably need to revisit my parse in that first sentence, "Arion, since he (or perhaps some unspecified deed? but why not the ablative absolute here?) was accomplished abroad, ..." using proficio as "has been accomplished" instead of proficiscor for "departed"); looking up profectus in a dictionary would yield the other base word.

I also wonder about jotting down translation notes while reading -- when seriously trying to translate a text, writing down notes in the margins or between lines (if there's appropriate spacing) helps a ton. At least until you're familiar with the grammar and the constructions, laboring through the translation is a huge help.

For someone like me who is reasonably familiar with the grammar, but might be rusty on the actual vocabulary (having last needed to use it in any semi-serious context more than a decade ago), I could see myself referring to the existing app's translation on occasion to give me some ideas regarding the actual words. But I don't know how helpful it is for someone newer to the language, who'd need more than just vocab.

Another commenter also mentions the possibility of pairing the Latin text with a well-established English translation. That might also be interesting; I could certainly muddle my way through a translation without knowing the grammar, but I would make plenty of mistakes if I didn't stop to think about cases and verb forms. What I think would help most is to perform a surface reading, followed by a refinement to make the text more idiomatic.

For instance, taking one of the sentences in the second paragraph: "Pecunia omni nautis oblata, vitam deprecatus est." My surface reading would look something like... "With all money offered to sailors, he begged his life." I might then refine that to "Having offered the sailors all of his money, he begged for his life."