Left-Associative Morphological Analysis of English with Stochastic Disambiguation

Morphosyntactic Analysis assigns to every word in a text information about word class, syntactic constraints (valency and agreement) and base form. At CLUE there exist morphological analyzers for German, Italian and Korean developed with Malaga, the implementation language for Left-Associative Gramars (LAGs).
A hybrid software component for the morphological Analysis of English is to be developed. English, which differs from the languages considered so far in that it has a simpler inflectional system, but a higher average categorial membership per word.
Rule-based and stochastic methods are to be integrated in a way that unites linguistic transparency and probabilistic robustness.

Supervisors:

• Prof. Dr. phil. Roland R. Hausser, Ph.D. (Computational Linguistics)
• Prof. Dr.-Ing. Günther Görz (Computer Science VIII: Artificial Intelligence)

Tasks

• Collection of annotated corpora and MRD
  
• Conversion to ISO 8859-1 character set and Malaga format
  
• Extraction of a large lexicon from MRD and annotated corpora
  
• Formulation of LAG rules for inflectional/derivational morphology
• Development of a probabilistic tagger to disambiguate the output of the morphological analysis
• Training of the probabilistic model on annotated corpora

Download [in German]

• Leidner, Jochen (1998): Linksassoziative morphologische Analyse des Englischen mit stochastischer Disambiguierung. Master's thesis. Friedrich-Alexander-Universität Erlangen-Nürnberg, Abteilung for Computerlinguistik (CLUE).
  [PDF, 770 kB] [PS, 2 MB]

• Beutel, Björn (1997): Malaga 4.0 Documentation [PS; PDF]
• Hausser, Roland (1989): Principles of Computational Morphology (CMU Technical Report) [PS; PDF]
• Hausser, Roland (1992): Complexity in Left-Associative Grammar Theoretical Computer Science 106 [PS; PDF]