[Article] Decoding Anagrammed Texts Written in an Unknown Language and Script

[DonaldFisk]

1) the statement:

So when the pair say that Hebrew was the highest scoring match for the manuscript without rating the likelihood, this is a bit of a meaningless boast. “Someone has to have the highest score,” says Argamon.

is doing insufficient justice to the close match given in Figure 4. At first I also considered that it could just be the end of the probability curve, but it is justified to call it an outlier and it is not understood why this happens. This has to be investigated further.

I think it is just the end of a probability curve.   Below 0.05 on the x-axis, the probability of finding a language drops below 1 (on the y-axis).   This means that in practice you'll find zero or one value at each point there, which is exactly what we have: zero at 0.01, 0.03 and 0.04 and one at 0.02.

We also don't know how the histogram would look if early 15th century languages were used instead, as they should have been.  For all we know, Middle High German or Middle Georgian might have matched much better than their modern counterparts and taken the top spot instead, with Mediaeval Hebrew being a much poorer match.

[ReneZ]

Donald,

the distance metric is a sum of a series of positive figures. In order for the sum to be small, every individual item in the series has to be small. For this point, the sum is less than half the sum of the second best case, and from there, a continuum of values starts.
The authors also say that Hebrew is a closer match to the Voynich text than to any other language.

This is exactly the kind of behaviour that one would expect if it were a correct match.

That of course doesn't confirm that it is indeed a correct match, but at least it is surprising.
I do believe that it could be coincidental, but a closer look is warranted.

A bigger problem is that the Voynich text used is entirely Currier-B language. The statistics for Currier-A are probably so different, that there will be another match.
This is a prediction, not based on any analysis.

[ReneZ]

the distance metric is a sum of a series of positive figures [...]

actually, that is a bit of an over-simplification, but it amounts to this.

[DonaldFisk]

This is exactly the kind of behaviour that one would expect if it were a correct match.

It's the kind of behaviour you'd expect in general.   You have a distribution with a peak.   Far enough away from the peak, the probability will drop below 1, and you'll get gaps in the observed distribution, so what looks like an outlier isn't really one.   There has to be a best match and it's that particular one.

[ReneZ]

Well, I'm not going to engage in a "yes it is" / "no it isn't" discussion ...

[-JKP-]

Rene, yes, thank you, I had intended to include two links so readers could link to the news site (to see new entries) and to the Verge article directly. I forgot to add the second one (it was late).

Thanks for clarifying it.

[DonaldFisk]

Well, I'm not going to engage in a "yes it is" / "no it isn't" discussion ...

I understand your reasons for thinking the match might be significant, but it's still what I'd expect given the distribution of the other points for the reason I explained, and I think we should be looking at the whole distribution and not just the values close to x=0.0.

[szvladimir]

I have not found something extraordinary in the article. The only small Hebrew passage looking ugly and it is not enough to believe in deciphering. I have seen no AI, only experiments with statistics.

Maybe they publish some more, and people having good native feeling of Hebrew and Kabbalah, can review the Hebrew transliterations (with anagrams or wthout) and make feedback.

Sorry, I had more expectations

[-JKP-]

Times of Israel response:

You are not allowed to view links. Register or Login to view.

[-JKP-]

Here's how the PaleoHebrew site reacted to the Hauer and Kondrak translation:

You are not allowed to view links. Register or Login to view.