Orchid XP v8 wrote:
> http://blog.plover.com/lang/finnpar.html
> 
> I've never actually seen any text written in Finnish, so... does this 
> look even remotely plausible?
> 
Scientific American had an article on this in the late 1980's.  The 
author first tried making random text by selecting letters, based on 
their frequency in the English language.  The longest English word found 
in the entire result set was the word "rare".

The author then tried to generate text based on letter pairs; the 
algorithm, having one letter, searched for that letter and randomly 
selected the letter to follow based on the frequency of letters that 
followed in the source text.  This produced results that looked a lot 
like English.

He then tried using longer strings of letters, by matching pairs of 
letters and choosing the next letter based on the letters following the 
pair to be matched.  The results were much better this time; many of the 
resulting words were actual English words, and the ones that weren't 
could be taken for words with which the reader was not unfamiliar.

Setting the match string to a longer length made it more likely that the 
  random text would consists of lengthy passages from the source text.

He then tried a different tack; instead of matching letters, he decided 
to match whole words.  A word to follow the current word was randomly 
chosen from all of the words that follow the current word.  This 
produced interesting results as well.

I had fun using AmigaBASIC to test out these ideas.

Regards,
John

Post a reply to this message