> I wonder if machine learning can be used to reconstruct the language behind Linear A. Similar symbols should represent similar phonemes.

This is an insight well within the means of any humans working on Linear A. What would machine learning add?

Identify words and grammatical structure. It’s not like there are spaces, and inflected languages change the ending of words. Even with a small set documents this must be incredibly difficult when you don’t know what the words are. Similarities with Semitic cuneiform phoneme order would probably be easier to recognize by machine

According to Wikipedia:

"The extant corpus, comprising some 1,427 specimens totalling 7,362 to 7,396 signs, if scaled to standard type, would fit easily on two sheets of paper."

What's the point of using ML to process two pages of text? It's well within human scale, and humans can infer things better.

> and humans can infer things better.

While that is true, it's not necessarily an advantage; you do see the claim that the decipherment of Linear B is entirely spurious, a product of the decipherers' imaginations.

Previous HN discussion: https://news.ycombinator.com/item?id=27191899

There are ancient forms found in Linear B that have to be preserved in some sections of the Iliad to preserve the dactylic hexameter, like wanax. This gives more credence to a continuity of the Greek language from the Mycenaean Bronze Age

Those are forms we knew about before we knew anything about Linear B; they are not independently evidence that we are reading the Linear B correctly. When you hallucinate material, it's not surprising that you hallucinate something you're already familiar with.

It's even less surprising in a context like this, where a proposal that "this is Bronze Age Greek!" is more or less required to include reconstructed Bronze Age forms. Who's going to take it seriously otherwise?

I’ve never seen this take. I was following that dutch philologist’s attempt to decode Linear A thinking it was luwian. Woudhuizen or something?

Is there no evidence of similar place names in Linear A and B?

The latest research suggests Linear A and Linear B is the same script, but used for different languages analogous to how the Latin script is both used for say English and Polish, but with language specific adjustments.

I saw a video not long ago from some research using ML to gain new insights into Linear A. I cannot find it atm.