Using AI for (alternative) translations and context

[nvandorp]

Hi there!
(this is my first post after my short introduction)

I was curious as to how AI would deal with the original transcript of the Oera Linda. To test this I gave it a section and asked it to give a translation. To my surprise the translation was pretty accurate when compared to translations on the wiki. Or maybe it is not a surprise, and it's just something LLM's (large language models) are good at.

I made a custom GPT, that has the entire transcript (NOT the translation) as 'knowledge' that it refers to when you ask it questions about the Oera Linda. This model/data right now is private to me, as I'm unsure if people here would want this to be publicly available. However, I can share the link here if people want to use it/play around with it. It is pretty good at giving translations and putting things into historical and linguistic context (though of course it's still AI; it can make mistakes/make things up...) I think it could be a nice educational tool perhaps, or give slightly different translations that can lead to new ideas/insights about the text. I provided a screenshot of an interaction with the gpt.

I understand AI can be a divisive subject, so I don't want to step on anyone's toes. Please let me know if you'd like me to share a link of the gpt to try it out, or if we can improve the model by providing it with more context!

Cheers,

Nick

[Helgiteut]

Hello nvandorp. I wonder how translation is done, if the language of the Oera Linda is only from one book? Where does the data come from for the translation?

[nvandorp]

Thanks for your reply Coco. I agree that thinking for oneself, and not relying on these kind of tools is probably better and is also more fun! Though I can imagine when someone gets stuck with a translation or is unsure, this could be a helpful second opinion.

[nvandorp]

Hello nvandorp. I wonder how translation is done, if the language of the Oera Linda is only from one book? Where does the data come from for the translation?

Hi Helgiteut, great question! I asked the GPT to explain how it does this, I provided the answer below.
This does immediately expose an important flaw, where it translates WÉIFÁREN with 'wandering or travelling', even though 'sailing' seems more fitting in the context (though a Wayfarer is someone who travels by foot...but that aside). These kind of errors, as Coco pointed out, make it unreliable as a primary method of translating, though in my opinion it can still be helpful in some situations, or to get more context.

GPT answer:

The language of the Oera Linda is a distinctive form of Old Frisian, but it incorporates elements that are unlike any other Frisian text, blending archaic Dutch, Germanic languages, and invented words. Here's how I approach translating this text:

1. Understand the Structure and Vocabulary
The text uses an older, more formal structure that can seem foreign due to its unique spelling conventions and syntax. By recognizing roots from Germanic languages (Old Frisian, Dutch, and German) and comparing them to modern equivalents, I can decipher the words.

For example:

"THA.K ALTHUS WÉIFÁREN WAS" – This translates to "As I had wandered." Here, "THA.K" can be broken down as "when I," and "WÉIFÁREN" comes from a verb meaning "to wander" or "to travel."
"NÉN ORLOCH" – "Nén" means "no," and "Orloch" is a recognizable word for "war" or "conflict," akin to the German "Krieg" or the Dutch "oorlog."
2. Identify Archaic and Invented Words
Some terms, like "HAVESMVDA" (harbor) or "KRÉTA" (Crete), are used creatively or in a non-standardized form. I compare them to similar words in neighboring languages or seek contextual meaning from their use in the passage. For instance:

"HAVESMVDA" could be parsed as "harbor-mouth," similar to how ancient words would describe geographical features (e.g., fjords, river mouths).
"KRÉTA" clearly refers to Crete, but the phrase "thrvch min ljud kréta héton" gives it a local meaning where the author’s people gave the island this name, blending legend and linguistic creativity.
3. Use of Contextual Meaning
Understanding the passage’s context is crucial. Even if the exact form of a word is unclear, the context—like interactions between people, references to war, peace, and trade—helps clarify meaning. For example:

When the narrator describes a peaceful landing, followed by trade ("BOTA MITH ÍSER.ARK"), it's clear the context involves a form of barter or exchange.

[ott]

I was curious as to how AI would deal with the original transcript of the Oera Linda.

I think this kind of curiousity is good. Why not experiment with new technology and see if it is helpful. I understand the resistance people feel towards what is called AI, specially if the advanced computer program answers questions as if it were human. People can easily forget it's just a program. It may make them lazy and lose creativity. I myself know little about it still, have experimented slightly with creating images (curiosity), but was not very succesful. Yesterday I was moved by a song someone had created, inspired by the new Dutch interview. later I realized that voice and chords were created artificially. The lyrics will have been genuine and are good. Overall, I did enjoy listening to it, but an acoustic version would be preferable.

I made a custom GPT, that has the entire transcript (NOT the translation) as 'knowledge' that it refers to when you ask it questions about the Oera Linda.

This can be interesting indeed, in particular for comparing texts.

Did you use the edited transcript (a work in progress)? For some purposes the raw transcript (with all its errors and variety in spelling and grammar) will be better. We aim to provide that too (as we did earlier).

There have been experiments with so-called stylometry software some years ago. Have you heard of this or similar techniques?