Assyriology sitrep

[pozorvlak]

Here's the current state of play with the Assyriology:

• I was conflating stress and vowel length. Some languages (like English) use metres based on stress, and some use metres based on vowel length (Latin, Greek, Turkish...). Turkish vowels, incidentally, come in three strengths rather than two.


• the ECTSL note on the language says that the Sumerians probably used long and short versions of a,i,e and u (IIRC), but that the differences between the long and short versions of a given vowel aren't included in the transliteration. I don't know if the difference is marked in the cuneiform - that would be a good thing to find out.


• Stress tends to be very regular in agglutinative languages: first syllable (Hungarian), or last syllable, or last syllable unless you're negating the word when it's the second syllable (Turkish), or something like that. Not terribly interesting from the point of view of poetry.


• Of the 38,000 or so words used in the corpus, 27,000 are used exactly once. This is far too many for any approach based on assigning syllable-patterns to words to work. However, if we make the assumption that there's some regular rule a la Turkish, we might be able to make progress. Or if we assume that it's based on vowel length, and that vowel length is a constant for each use of a given cuneiform symbol, we could make progress then. The other option is to give up on Sumerian and try Akkadian, which is inflected (isn't it?) and thus doesn't suffer from this problem.


• I got one rather good suggestion: instead of looking at entropy or autocorrelation or anything like that, I could try looking at the frequencies of short (two-or-three syllable) patterns of shorts and longs, in the hope of finding feet. Given the frequencies of shorts and longs in the lexicon, we could work out what frequencies we'd expect for eg "long short short" under the assumption of total randomness, and compare them with the frequencies our hypothesised stress pattern produces. We should then be able to see how far apart the two are - is a χ-squared test appropriate here?


• But there's another problem: counting syllables for lines of a few poems produces results that look totally random:
  
  
  
  I thought I'd found a repeating pattern in one, but it turned out that the poet had just repeated the same block of 30 lines five times. I don't know how to check that it is random (secret_screams? benparker?), but it doesn't look very hopeful. This, I would have thought, would rule out Japanese-style syllable-counting poetry and more-or-less rules out Latin-style metric poetry (though Latin metres allowed for some variation, so it might be worth proceeding anyway).


• One thing it doesn't rule out is Old Germanic-style stress-counting poetry, where you have the same number of stresses per line but ignore the unstressed syllables. Or long/short vowels, possibly. Beowulf's an example, but so is Baa Baa Black Sheep:

  BAA baa BLACK sheep
  HAVE you ANy wool?
  YES sir, YES sir,
  THREE bags FULL.
  ONE for the MASTer and
  ONE for the DAME,
  and ONE for the LITTle boy
  who LIVES down the LANE.

  We could search for this using basically the same strategy, only now it gets easier: just take the distribution of stresses/ long syllables per line and take its kurtosis, which measures the tendency of a distribution to cluster around the mean. (Can you tell that I've been reading the stats pages on Wikipedia recently? :-) )

Comments

[half-of-monty.livejournal.com]

Of the 38,000 or so words used in the corpus, 27,000 are used exactly once.

However did they work out what these words mean then? Good job assyriologists!

(Also, in the second line of Baa baa black sheep, the second stress is on wool. Sing it while tapping the table.)

[pozorvlak.livejournal.com]

It's agglutinative (http://en.wikipedia.org/wiki/Agglutinative_language), like Hungarian or Japanese: words are composed of strings of shorter morphemes which glue together. Morphemes can then be decoded individually. But the decipherment still took decades and extreme cleverness. More here (http://en.wikipedia.org/wiki/Sumerian_language#Decipherment).

[half-of-monty.livejournal.com]

In that case, can't you start with a hypothesis that the stress on the whole is the same as it would be on each part? Not true for English, granted, but is true-ish for German?

[pozorvlak.livejournal.com]

German isn't actually very agglutinative: it's an inflected language with some agglutination. Pure agglutinative languages tend to have very regular stress - see point 3. But I could make the guess that Sumerian poetry worked on vowel length instead of stress and then do exactly what you suggest for vowel length (which mathematically makes no difference). I'll certainly try this when I get some time - paper submission deadline for conference proceedings on Wednesday, and I'm away at an EPSRC GRADschool from tomorrow morning until Sunday night.