Language-Based Environment for Natural Language Parsing

LANGUAGE-BASED ENVIRONMENT FOR NATURAL LANGUAGE PARSING Lehtola, A., J~ppinen, H., Nelimarkka, E. sirra Foundation (*) and Helsinki University of Technology Helsinki, Finland ABSTRACT This paper introduces a special programming environment for the definition of grammars and for the implementation of corresponding parsers. In natural language processing systems it is advantageous to have linguistic knowledge and processing mechanisms separated. Our environment accepts grammars consisting of binary dependency relations and grammatical functions. Well-formed expressions of functions and relations provide constituent surroundings for syntactic categories in the form of two-way automata. These relations, functions, and automata are described in a special definition language. In focusing on high level descriptions a linguist may ignore computational details of the parsing process. He writes the grammar into a DPL-description and a compiler translates it into efficient LISP-code. The environment has also a tracing facility for the parsing process, grammar-sensitive lexical maintenance programs, and routines for the interactive graphic display of parse trees and grammar definitions. Translator routines are also available for the transport of compiled code between various LISP-dialects. The environment itself exists currently in INTERLISP and FRANZLISP. This paper focuses on knowledge engineering issues and does not enter linguistic argumentation. INTRODUCTION Our objective has been to build a parser for Finnish to work as a practical tool in real production applications. In the beginning of our work we were faced with two major problems. First, so far there was no formal description of the Finnish grammar. Second difficulty was that Finnish differs by its structure greatly from the Indoeuropean languages. Finnish has relatively free word order and syntactico-semantic knowledge in a sentence is often expressed in the inflections of the words. Therefore existing parsing methods for Indoeuropean languages (eg. ATN, DCG, LFG etc.) did not seem to grasp the idiosyncracies of Finnish. The parser system we have developed is based on functional dependency. Grammar is specified by a family of two-way finite automata and by dependency function and relation definitions. Each automaton expresses the valid dependency context of one constituent type. In abstract sense the working storage of the parser consists of two constituent stacks and of a register which holds the current constituent (Figure I). The register of the current constituent LI L2 L3 RI R2 R3 The left The righ constituent constituent stack stack Figure I. The working storage of DPL-parsers (*) SITRA Foundation P.O. Box 329, SF-00121 Helsinki, Finland 98 <-Phrase Adverbial )<+Phrase Adverbial IILD PHRASE ON RIGHT ~*Phrase Subject~ ~ophrase Phrase ] I L Adverbial ! *Phrase IAdverbial IILO PHRASE ON RIGHT ~Phrase Phrase Sublet1 ILO PHRASE ON RIGHT • - -Nomina empty left- hand side BUILD PXRA: ON RIGHT = ,Nominal - +Nominal ~nd of inpul @ FIND REGENT ON RIGHT Notations: On the left is I On the left is a state transition the state node ?X with priority, conditions for of the automaton {cond$ . . . . the dependent candidate (if not Toncllon) otherwised stated) and k The question mark I indicates the direction 4, connection function …

Content truncated to 500 words