A Bag of Useful Techniques for Unification-Based Finite-State Transducers

Hans-Ulrich Krieger, Witold Drozdzynski, Jakub Piskorski, Ulrich Schäfer, Feiyu Xu

In: Proceedings of 7th KONVENS. Konferenz zur Verarbeitung natürlicher Sprache (KONVENS) Vienna Seiten 105-112 9/2004.


We present several extensions to the shallow text processor sprout, viz.,
  1. a fast imperfect unifiability test,
  2. a special form of sets together with a polymorphic lazy and destructive unification operation,
  3. a cheap form of negation,
  4. a weak unidirectional form of coreferences,
  5. optional context-free stages in the shallow cascade,
  6. a compile time type check,
  7. compile time transition sorting under subsumption,
  8. several output merging techniques, and
  9. a compaction technique for lexical resources.
The extensions have been found relevant in several projects and might be of importance to other systems, even to deep processing.


sproutKONVENS2004.pdf (pdf, 180 KB)

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence