On Sun, 09 Dec 2007 18:57:06 -0500, Anonymous
<anyone@[EMAIL PROTECTED]
> wrote:
>It seems to me the results don't quite look/sound natural. Any
>suggestions for tweaks? It seems to me "n" should change to "m" in
>certain situations, for example "hanbu" should be changed to "hambu."
>Also there's something unbelievably random about the number of words in
>each sentence and the sequence of vowels within longer sentences, but I
>can't quite put my finger on it.
The lack of dipthongs (ai, au, ou, etc), consonant clusters (st, sh,
ch, etc), changed word endings (ing, ed, es, etc.), and double letters
tells me that this is random generation or possibly a conlang. One
sentence, though, could have fooled me if I saw it by itself, because
Nipon means Japan, and I think I could have mistaken it for Japanese
because of the heavy use of the letter j, and the syllabic nature of
Japanese, which I do not speak but have seen on occasion in Latin
letters (I do recognize their writing systems, but do not read it).
> Nipon ka hugodujo no jija pimo.
I think if you could at least generate syllables that go on the end of
words only, and repeat some words in a longer text and add an ending,
it would look much more convincing, i.e., similar to "Our horse was
the fastest of all the horses on the track." or "Of all the horses on
the track, our horse was the fastest." or "There were many horses at
the track that day. It had rained and was muddy. Our horse beat them
all." So the repeated/changed word does not have to appear in the
same sentence or in the same order every time (i.e. "horse...horses"
or "horses...horse"). If the text was really long, having the program
limit the number of words generated and repeating more of them and
changing more of them with endings would also be more convincing than
a single long string of never-repeated randomly generated words.
Otherwise, I think single sentences only could hope to fool someone
who is familiar with languages.
Leah
|
