A GBML Approach to DFA Construction and its Application to Turkish Word Inflection

Meltem Turhan

April 1996

Abstract:

The construction of the grammar that accepts a given set of strings is a Machine Learning problem. In this work it has been shown that genetic-based machine learning (GBML) can successfuly be used in the generation of some grammars. A genetic algorithm (GA) engine is developed and Turkish word inflection which has an agglutinative nature, is used as the testbed for the proposed technique. Results for a subset of all word inflections namely the noun inflections is presented.

• Introduction
• The Problem
• GA Representation
  • An Overview of GA
  • GA Representation of DFA
  • Genetic Operations
    • Reproduction
    • Crossover
    • Mutation
    • Swap
    • Exchange
  • The Engine
• Application of the Technique to Turkish Word Inflection
  • Overview of Turkish Word Inflection
  • Generation of the Testbet
• Result
• Conclusion
• References