StringTools

Christopher2222 — Sat, 21 Oct 2017 23:28:17 Z

Any thoughts as to why StringTools(SyllableLength) has not been fixed in Maple yet?

Well this is a can of worms!

tomleslie — Mon, 23 Oct 2017 01:51:18 Z

The conclusion I have come to is that accurate syllable counting cannot be achieved

I think that Maple's "get-out" on this one, is (from the help page)

The morphological syllable length is defined to be the number of vowel-consonant changes in word. (Note that this may not agree with the lexical syllable length found in dictionaries.)

Before I came to the conclusion above I thought it might be fun to try and work out what "rules" are actually being applied by Maple's SyllableLength() command, and whether ti might be possible to do better

A moment's thought will convinece you that a reasonable stab at the basic algorithm would be to "telescope" all adjacent consonants to a single consonant, and all adjacent vowels to a single vowel. With 'V' representing a vowel and 'C' representing a consonant, this would result in a string of alternating "V", "C" characters - for example the word "straight" would be transformed to "CCCVVCCC", which would then be "telescoped" to "CVC". Then just count the number of occurrences of the "V" character in the resulting string. So I wrote a little procedure to do this -it's pretty trivial. (I apologise for parts of this - it's a long time since I used any regexp stuff, and I was never that great with 'regexps' anyway)

My experimentation suggest that this is pretty much what the SyllableLength() command in Maple actually does. However it does apply a few extras: for example the above algorithm will render "bake" as "CVCV" and hence count two syllables. The Maple command appears to "get around" this by ignoring all trailing vowels, so SyllableLength("bake") returns 1, but so does SyllableLength("llama") or SyllableLength("guru") - not good.

However a trivial piece of regexp magic will reduce the syllable count by one only if the word ends in consonant+"e"+(zero or one)"s". This works, but has the unfortunate side-effect or reducing words such as "he", "the", "due", etc to zero syllables. Another no-brainer: jonly apply this step if the input syllable length is greater than 1.

Then I got to your last example - ie "radio", which the basic algorithm describe above would collapse to "CVCV" and so the syllable count would be off by 1. Again. another little piece of regexp magic will fix this - any word ending in "io" should have the syllable count increased by 1. However what happens if the "io" is not at the end of the word? consonant+"io"+consonant can be one or two syllables depending on whether the leading consonant is "t", "g", or "s". So OK another little piece or regexp.

Probably because of this last example, I then started thinking about the "ious" problem - think the difference between "specious" and "serious". It would seem that "ious" is one syllable when preceded by "c", "g" "t" or "x" (think "specious", "facetious", "egregious" or "noxious") but otherwise is two syllables ( think "serious", "odious", "pious", "harmonious"). Another regexp and this is solved

At this point, I'm cooking with gas: my routine now seems to be way more accurate than the the inbuilt Maple version, and I'm thinking that there can't be many(?) more "exceptional" cases to handle - until I thought about the "re+vowel" combination. This was when I really realised the magnitude of the problem, reinforced by contemplating reindeer. I simply cannot think of any way to deal with this issue algorithmically (or its relation the "de+vowel" combination,- think dealmaker and deactivate). My conclusion - we're screwed - as in there is no algothmic way to count "dictionary" syllables:-(

FWIW, the routine I came up with is attached

getSyll.mw

tough one

Christopher2222 — Mon, 23 Oct 2017 14:13:43 Z

@tomleslie thanks. I appreciate your work on the subject. Another tough word is "doable".

Your example of dealmaker and deactivate are examples of two different types of words, one is a combination of two words "deal-maker" while the other is a word with a prefix "de-activate".

I think if we create enough rules we can crack the syllable code, but perhaps there are too many one off rules that make the code impossible to complete?

MaplePrimes - answers and comments on Question, StringTools[SyllableLength] missing rules

StringTools

Well this is a can of worms!

tough one