SFB 340 - project B9 "Computational Semantics"

1998 - 2000

In the context of this project, we assume that computational semantics means an inference-based transformation process, whose premises involve information from at least the following sources:
  • lexical semantics
  • syntactic configuration
  • type-theoretic information
  • contextual restrictions (of various kinds)
The goal of this project is to realize an inference-based method of computational semantics without making the conventional assumption that all, or at least most, of the premises be available in an unambiguous, fully specified manner. Firstly, the basis for our theory and for the realization of a computational semantics system has been layed out by the previous projects B3 and B9. In those projects, construction and inference methods for underspecified semantic representations for various phenomena (quantifier scope, plural, tense, propositional attitudes) have been developed and implemented. Secondly, we will take into account various pilot studies and projects outside of the SFB concerning the integration of syntactic and semantic underspecificiation, anaphora resolution and presupposition handling on underspecified semantic representations.

On the basis of the existing B9 semantic fragment and the above mentioned studies, we plan to experiment with alternative software architectures in order to realize a concurrent processing of syntactic, semantic and pragmatic inference mechanisms - to a further extent than it has been proposed in the literature, so far. On the one hand, application-neutral methods from constraint logic programming shall be employed, on the other hand, application-specific rules and heuristics have to be developed in order to obtain a reasonable run-time behaviour.

In parallel to the computational issues, the semantic fragment of the earlier B9 project will be extended wrt. the description of semantic phenomena and semantic composition. Apart from the theoretical aspects of computational semantics, the project should yield a 'semantic workbench', where researchers from more theoretically oriented neighbour projects can test their ideas.


Implementation (from predecessor project):

iXi system ("Integrated Computational Semantics Implementation")

Pointers for project internal use

IMS Stuttgart, Fri Apr 16 21:15:30 1999 (