inferring hierarchies from sequences
  


Full source of Sequitur with many options

An ObjectPascal implementation by Michalis Kamburelis. Michalis' server is a little slow, so I've mirrored his tarball.

SEQUITUR is a method for inferring compositional hierarchies from strings. It detects repetition and factors it out of the string by forming rules in a grammar. The rules can be composed of non-terminals, giving rise to a hierarchy. It is useful for recognizing lexical structure in strings, and excels at very long sequences.

Craig Nevill-Manning, Rutgers University,
Ian Witten, University of Waikato, New Zealand