Overview of phenomena and proposed tests

 

Tests run in this exercise (for details, see appendix A)

• Disambiguation of proper names and common nouns. (NN/NE)
  Proper names and common nouns have very similar distribution, their distinction is mainly semantic. Furthermore in German (unlike in other European languages), there is no orthographical distinction between proper and common nouns.

  There is also interaction/ambiguity with foreign language material.
  (Type Ia)

• Separable verb prefixes (PTKVZ)
  Verbal prefixes may occur as separated units and are then ambiguous with ADV, ADJD, APPR and APPO. We are most interested in tests with adjectival and adverbial prefixes as they have a similar distribution as predicative adjectives or adverbs.
  (Type II, Ic?)
• Verb infinitives (VINF)
  All infinitives are ambiguous with finite 1st and 3rd Plural verbs. Sequences of infinitives are only possible with auxiliaries or modals. Therefore we expect external impact from verb type disambiguation (lexical distinction VM, VA, VV).
  Modal auxiliaries: different distribution (perfect constructions with infinitive forms).
  (Type Ia, Ic?)
• Disambiguation of reflexive and personal pronous (PRF/PPER)
  Ambiguous forms are mich, dich, dir, mir, euch, uns. reflexive and personal pronouns have identical distribution; the decision about reflectiveness depends on (verbal) subcategorization, but automatic taggers have no access to criteria used by humans (reflexitivity of verb; case, person, number of subject and pronoun). Are unambiguous 3rd person personal pronouns (er, sie, es) and reflexives (sich) enough to train taggers to distinguish ambiguous forms (mich, dich, ...)?
  (Type Ib)

Phenomena suggested for further testing (cf. appendix B):

• Functional vs. adverbially used participles (VVPP/ADJD)
  Difficulty: copula constructions and stative passive have identical formation. Stepwise change to definitions in guidelines, by syntactic constructions.
  (Type II)
• Indefinite vs. adjective-like determiners (PD-/PI-/ADJ-)
  Merging of demonstrative and indefinite pronouns/determiners? Membership to demonstr./indef. is lexically determined!
  (Type Ic?)
• Indefinite determiners with or without articles (PIDAT/PIAT)
  Context dependent or lemma dependent?
  (Type Ia, Ic?)
• adverbially used adjectives vs. non-adjectival adverbs ADJD/ADV
  Differences in semantics.
  (Type II)
• wordforms doch, aber, denn (KON/ADV)
  This word forms are ambiguous between coordinating conjunctions and adverbs, depending on position within sentence
  (Type Ib)
• wordforms wie, als (KOUS/PWAV/KOKOM/KON)
  These word forms have a variety of functions and therefore tags. Which functions can be merged into one tag easiest? (Type I, Ic?)
• proper names vs. foreign material (FM/NE)
  Very problematic: dependent on semantics. Interaction with NE/NN test.
  (Type I)
• Punctuation $(, $., $,
  Is there external impact of behavour of comma on relative pronouns, articles, subordinating conjunctions?
  (Type Ic?)