Unification: Computational Issues

Unification grammars have become the most common framework in computational linguistics for expressing grammatical theories, as well as a means for implementing linguistically informed natural language processing systems. The basic building blocks for unification grammars are feature structures, which are used to represent linguistic information at all levels. The major operation on feature structures is unification, combining the information encoded by two compatible feature structures. This article defines the two concepts formally, laying the ground work for a more informal definition of unification grammars. It then discusses computational issues concerned with unification, including the computational representation of feature structures and parsing with unification grammars.