A Finite-State Approach to Shallow Parsing and Grammatical Functions Annotation of German
Frank Henrik Müller · 2005
Chapter 2 gives an introduction to the nature and architecture of our parser. Section 2.1 presents finite-state automata (FSAs) as a parsing model and illustrates the motivation for using this very model for the parser at hand. Section 2.2 explains the motivation for and the main features of the parsing approach used in the dissertation at hand, namely the annotation of constituent structures and grammatical functions with hand-crafted rules. Section 2.3 introduces and illustrates principles spanning the annotation of all linguistic phenomena (i.e. robustness, modularity and underspecification) and, thus, all components of the parser. Section 2.4 shows how the FSAs are implemented with the help of the tool fsgmatch by giving illustrative examples of typical linguistic phenomena. 2.1 FSAs as a Parsing Model A parser is the implementation of an abstract automaton. In the thesis at hand, we use finite-state automata (FSAs) as the abstract automaton which serves as the basis of our parser because FSAs are very effective formalisms for parsing. The reason for this is that FSAs can be described by regular grammars, which are the most restricted and, thus, least powerful type of grammar in the Chomsky Hierarchy (Chomsky, 1956). Since the power of the formalism is restricted, for example it cannot deal with recursive structures, it is considerably faster than other approaches as Abney (1996c) has shown.