Parsing with principles and classes of information (Q1375383)

The book investigates the complex task of natural language (NL) parsing and proposes an efficient interplay between computational efficiency and linguistic adequacy. The proposed parsing technology makes use of two key observations considered by the author: (1) Linguistic information belongs to five main classes (configurations; lexical features; syntactic features; locality conditions; referential indices), each one defininq information ``modules'' in the parser. (2) The linguistic principles are complex constraints which can be factored into simple ones and pre-compiled off-line. Thus a key solution to the principle-based parsing would be an efficient trade-off between modularity of the linguistic theory and partial compilation of its comprised principles. The book has five chapters and two consistent appendices: Chapter 1 debates the relationship between grammars and parsers, considers a particular type of partial precompilation of linguistic principles, and analyzes the parser modularity based on linguistic information content. Chapter 2 gives an overall view of the parser organization and a sample of its parsing capabilities, comparing them to other parser performances. Chapter 3 illustrates comparative results of the \(\text{LR}(k)\) parsing engine with various grammar compilations. Chapter 4 discusses the assignment of features and presents algorithms to perform local feature annotation and chain computing. The issue of \(\text{LR}(k)\) incremental parsing is addressed. Chapter 5 presents the implementation of the routines computing restrictions on wh-movement, bringing arguments in favour of \(\text{LR}(k)\) parsing for cross-linguistic variation on locality restrictions, with evidences from Italian and English. Appendix A contains a substantial overview of the computational and linguistic framework (levels of representation, the linguistic modules, and the parsing algorithms). Appendix B shows results of the parsing examples for several grammatical constructions. Merlo's book represents a new step and fruitful experience made at the frontier of two important research and very worked fields: formal linguistic theories and parsing technologies.