aeglaeca wrote:EDIT2: Actually, just wondering, how many of the actual Haikus are flagged by your program? It would be interesting to know what accent the program uses.
Definitely not all of them, since it only looks for single sentences (where a 'sentence' is determined by a method in the framework, using whatever settings it has by default) with 17 syllables, and some of the real haiku have more than one sentence (I'm thinking about adding the capability to recognise multi-sentence haiku.) Also, because of the way I do it (first finding 17-syllable sentences, then looking at the syllable counts in the individual words to find the line breaks), it won't find any which have punctuation which is both pronounced and considered to be separating words rather than part of them, but that shouldn't make much difference unless you have URLs in your haiku (an early bug, before I doublechecked the length of the last line, was that it thought the following was a haiku:
As for the accent, I don't specify anything in the program (maybe I'll add an interface for that later), so I think it uses whichever accent the current system voice has; I'm currently using a US English voice but there are also ones for English, Irish, Scottish, Australian and South African English (as well as other languages) so I'll have to try those and see if they split the syllables up any differently. I'm not sure how much of the accent is simply in the sounds of the voices and how much is in the way text is converted to phonemes. Maybe I should try the English English voices to see if they have triphthongs.
Anyway, I ought to tear myself away from this thread tidy up the latest version of it and release that, since it's nice and multi-needle-pulled-thinged now.