Language Engineering in Grammatical Framework (GF)

Abstract: This thesis describes a number of practical experiments rather than theoretical investigations in the area of natural language processing. The basis for the work presented is Grammatical Framework (GF). It is a very complex system, which comprises among other things a grammar formalism based on type theory and its implementation written in Haskell. GF is intended for high-quality machine translation (of INTERLINGUA type) in the restricted language domains. The primary concern of this thesis is however limited to the usage of GF as a piece of software. The main results are: ? Implementing a syntax editor, which provides a graphical user interface (GUI) for the command-line GF core. ? Writing a part of code for automatic generation of gramlets?pure Java programs with limited (compared to GF) functionality that can be run on PDA (Portable Device Assistants) and as applets in a browser. ? Writing the Russian resource grammar that takes care of the most basic morphological and syntactic rules and serves as a standard library for building application grammars (describing restricted language domains) in Russian. These results contribute to language engineering in GF on two different levels: ? Author level (end-user) ? constructing sentences in natural languages. ? Grammarian level ? building a grammar description, which is later used on the author level. The last part of the thesis deals with a non-linguistic domain. In that experiment we try to apply functional parsing technique to the well-known problem of protein secondary structure prediction (bioinformatics).

  CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)