Parsing Schemata

Abstract

Parsing schemata provide a general framework for specication, analysis and comparison of (sequential and/or parallel) parsing algorithms. A grammar specifies implicitly what the valid parses of a sentence are; a parsing algorithm specifies explicitly how to compute these. Parsing schemata form a well-defined level of abstraction in between grammars and parsing algorithms. A parsing schema specifies the types of intermediate results that can be computed by a parser, and the rules that allow to expand a given set of such results with new results. A parsing schema does not specify the data structures, control structures, and (in case of parallel processing) communication structures that are to be used by a parser. Part I, Exposition, gives a general introduction to the ideas that are worked out in the following parts. Part II, Foundation, unfolds a mathematical theory of parsing schemata. Different kinds of relations between parsing schemata are formally introduced and illustrated with examples drawn from the parsing literature. Part III, Application, discusses a series of applications of parsing schemata. - Feature percolation in unification grammar parsing can be described in an elegant, legible notation. - Because of the absence of algorithmic detail, parsing schemata can be used to get a formal grip on highly complicated algorithms. We give substance to this claim by means of a thorough analysis of Left-Corner and Head-Corner chart parsing. - As an example of structural similarity of parsers, despite differences in form and appearance, we show that the underlying parsing schemata of Earley's algorithm and Tomita's algorithm are virtually identical. Using this structural correspondence we can obtain a novel parallel parser by cross-fertilizing a parallel Earley parser with Tomita's graph-structured stack. - Parsing schemata can be implemented straightforwardly by boolean circuits. This means that, in principle, parsing schemata can be coded directly into hardware. Part IV, Perspective, discusses the prospects for natural language parsing applications and draws some conclusions. An important observation is that the theoretical and practical part of the book reinforce each other. The proposed framework is abstract enough to allow a thorough mathematical treatment and practical enough to allow rewriting a variety of real parsing algorithms (i.e. seriously proposed in the literature, not toy examples) in a clear and coherent way.

Related Papers

• → Recursive ascent parsing: from Earley to Marcus(1992)16 cited
• Parsing Schemata - a framework for specification and analysis of parsing algorithms(1997)
• → Top-Down Parsing with Parsing Contexts(2015)4 cited
• Top-Down Parsing with Parsing Contexts A Simple Approach to Context-Sensitive Parsing(2014)
• → A Kind of Syntax Parsing Algorithm Based on The Predictive Parsing Table(2022)