Newsgroups: alt.fan.hofstadter,comp.ai,comp.ai.nat-lang,comp.lang.c++,comp.lang.lisp,comp.la
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!purdue!lerc.nasa.gov!magnus.acs.ohio-state.edu!math.ohio-state.edu!howland.reston.ans.net!swrinde!tank.news.pipex.net!pipex!uknet!newsfeed.ed.ac.uk!edcogsci!jeff
From: jeff@cogsci.ed.ac.uk (Jeff Dalton)
Subject: Re: Looking for Computer Poetry Generators
Message-ID: <DEuHKo.II0@cogsci.ed.ac.uk>
Organization: Centre for Cognitive Science, Edinburgh, UK
References: <42tdts$ruk@umt.umt.edu> <Vs78aD5w200w@alcyone.darkside.com> <432d49$r7n@gol1.gol.com>
Date: Wed, 13 Sep 1995 13:02:44 GMT
Lines: 27
Xref: glinda.oz.cs.cmu.edu comp.ai:33404 comp.ai.nat-lang:3862 comp.lang.c++:148679 comp.lang.lisp:19075

stuart@gol.com (Stuart Woodward) writes:

>>One sort of peripheral way of automatically generating poetry is to 
>>use a Markov chain generator.  Train it on some poetry, especially of 
>>one given author or a specific genre.  The output sometimes sound 
>>intentional, since poetry isn't as grammatically restricted as prose.

>This sounds interesting where can I find out more about "Markov chain
>generators"?

The basic idea is pretty simple.  You have some base text -- e.g. some
poetry -- and you generate new text by finding overlapping sequences
in the base text.  For instance, if the seq length is 2 then every
pair of words in the output appeared somewhere in the input.  So
from

  I am a zebra.
  Is 6 am better?
  Are you better now?
  Now quack, duck.
  The duck went quack quack.
  
you might get

  I am better now quack quack.

-- jeff
