Re: OFF: Dissociated Press

At 13:52 08.4.2000 -0300, FFlores wrote:

>Ah yes, there's a letter-based variant too. If you just take a space-
>separated list of words and preserve the spaces, you have your BOW/
>EOW mark. It seems to work better with English, since words are short,
>than in Spanish, for example, and better yet if you allow many sound
>clusters. The main problem is that the algorithm ignores context --
>it just knows what came immediately before. For example, see what
>happens to this paragraph (before the last "For"):
What about setting a limit on word length?  I've also seen similar programs
that look for letter pairs or triplets.  If "_" is any
punctuation/space/newline/tab and "*" is any letter, then you first look up
all instances of _*, then if you got _a you look for a* and append the
found * to _a, you get _ab, look for b* and so on.  If you get to the
word-length limit -1 -- say the limit is 6 and you got _abcde, then you
start looking for e*_.

You can also start with _ab, look for ab*, add *, look for bc*, add *, look
for cd*, and so on.  Trying different combinations of word length and
length of matching pattern you should be able to fine-tune the degree of
resemblance between source and output.




/BP

 B.Philip Jonsson  <mailto:bpj@...>bpj@netg.se
<mailto:melroch@...>melroch@my-deja.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~__
            Anant' avanaute quettalmar!               \ \
   __  ____ ____    _____________  ___ __   __ __     / /
   \ \/___ \\__ \  /___  _____/\ \\__ \\ \  \ \\ \   / /
   / /   / /  /  \    / /Melroch\ \_/ // /  / // /  / /
  / /___/ /_ / /\ \  / /Melarocco\_  // /__/ // /__/ /
 /_________//_/  \_\/ /Eowine__   / / \___/\_\\___/\_\
I neer Pityancalimeo\ \_____/ /ar/ /_atar Mercasso naan
~~~~~~~~~Cuinondil~~~\_______/~~~\__/~~~Noolendur~~~~~~
|| Lenda lenda pellalenda pellatellenda cuivie aiya! ||
"A coincidence, as we say in Middle-Earth" (JRR Tolkien)