Is anyone working on comprehensible input using ChatGPT?

Textkit Community Open Board
1

Hi all, just a general question below on “comprehensible input” and ChatGPT. Has anyone on Textkit started using it to produce “comprehensible input” (or at least a first draft of comprehensible input for review and correction by a classicist)?

Background

I have very much been of the “composition” approach to classics, and still am, toiling away over the smallest elements of a text to make it as close to an ancient text as possible (and thereby to learn more about ancient texts as the final end). You can see my approach here (for which I was kindly given a cash prize: which I waived of course, since as someone who is not a fee-paying university student, I didn’t want to accept cash from academics who need it most to do their great work on which I heavily rely as an outsider; in return I was very kindly sent one of the Antigone editor’s personal books): https://antigonejournal.com/wp-content/uploads/2023/02/Socrates-vs-Hawking-Bochan.pdf

However, I’m now curious as to where the “other” approach has reached: active learning, comprehensible input, that whole movement, and will delve into this for a year or so in parallel with my traditional “composition” approach, as a bit of a sandbox.

I remember in a previous Textkit conversation calculating that, based on an article by Krashen, the comprehensible input approach would need somewhere between 30k and 5.85m words for learners, and I guessed that machine learning in the future might be able to assist with this (if properly edited by a human classicist: there’s my paranoid “composition” approach shining through). See that thread here: http://discourse.textkit.com/t/krashen-and-ancient-greek-teaching-methods/15043/1

With ChatGPT I just tested the first sentence of Lingua Latina in this way:

Using this Latin sentence as a model: “Roma in Italia est”, can you produce 10 more examples in Latin which only feature four words: (1) a city or country as the subject in the nominative, (2) a prepositional phrase in Latin made up of “in” plus a country or continent in the ablative, and (3) the copula “est” or “sunt” (as appropriate, depending on the number of the subject)?

It then responded:

Certainly! Here are ten examples in Latin that follow the specified model:

Athenae in Graeciā sunt. (Athens is in Greece.)
Carthāgō in Āfricā est. (Carthage is in Africa.)
Londinium in Britanniā est. (London is in Britain.)
Corduba in Hispāniā est. (Cordoba is in Hispania.)
Byzantium in Thrāciā est. (Byzantium is in Thrace.)
Tarentum in Italiā est. (Tarentum is in Italy.)
Massilia in Galliā est. (Massilia is in Gaul.)
Tyrus in Phoeniciā est. (Tyre is in Phoenicia.)
Numantia in Celtiberiā est. (Numantia is in Celtiberia.)
Byzantium in Europā est. (Byzantium is in Europe.)

I’m guessing that someone (not me: I’m looking into something else for the moment) could easily do this for each new sentence structure in Lingua Latina and produce a large set of comprehensible input pretty quickly, which could then be reviewed by people on this forum or elsewhere.

Cheers, Chad

How well does chatgpt translate Latin?
2

…actually, if anyone wants to collaborate on a fun project on the side along these lines, please jump in here. The aim would be to produce together a Lingua Latina + version, free online, as follows:

  1. Each model sentence (in the way I described above) in the two main Lingua Latina books would be used to generate approx. 10x more examples. You could probably produce 1 million words of comprehensible input in this way at a guess (I counted one page of FR and it had a little over 200 words, with approx. 300 pages in the textbook, so somewhere between 50–60k words in the book at a very rough guess; times by 10 for the new content generation, and then 2 again as there are two main books I understand in the series, and you get a decent amount of material around the 1 million words mark).

  2. Checking and correcting of the new content here on Textkit by forum members with strong Latin skills.

  3. Turning the content into more interesting formats by forum members with strong Latin pronciation and media editing skills (e.g. making a video of the new content by reading it out and adding graphics, e.g. zooming out from Rome to Italy, from Athens to Greece, etc.) .

  4. Gamifying the new content for active practice by forum members with good app development skills (e.g. making a Duolingo-like / Education Perfect-style app / online platform where the new content is taken apart and reassembled by users to complete quizzes).

(3 and 4 would be the “+” part, trying to improve the Lingua Latina approach, which has been recommended by teachers who use the book, e.g.:

https://www.youtube.com/watch?v=adv_lbelCJk

Any thoughts on this? It would democratise online the approach I understand they use at active learning courses. I could help with 1 and 2 but it would need a few people to jump in, in order to get a critical mass.

EDIT: Here is a quick go at making a video, using public source pictures and AI voices (I went Italian for the closest “ring”). Doing this project would also give us the chance to improve the model sentences where Oerberg uses rarer word orders or syntactical structures - I’ve called this out at the end on the position of “est” (“non est” often goes at clause end, but “est” in locatival sentences does not). Any thoughts on this quick prototype gratefully received!

https://www.youtube.com/watch?v=2EYlWVoxhdw

Cheers, Chad

3

…please ignore the video link above (Textkit is not letting me edit anymore). I will start a new playlist in Youtube having thought a little more about what I want to do. Basically, I see errors in Lingua Latina and want to start off by creating new comprehensible input using word order resources, frequency lists (e.g. relative and interrogative pronouns are some of the highest frequency words in Latin, and should really be naturalised in comprehensible input from the very beginning), common expressions and spaced interval repetition. Here’s a mission statement intro - not actually my voice, the wonders of AI (I’ve then uploaded a first video and have more in the works - it will be a LONG-term project):

https://www.youtube.com/watch?v=Kg9IlGKeFOc

Cheers, Chad

4

This seems like a very interesting and promising idea. I would love to help out on this project!