A Morphological Parser for Afrikaans

A MORPHOLOGICAL PARSER FOR AFRIKAANS L G de Stadler and M W Coetzer University of Stellenbosch South Africa The parser has been developed as a component in a text-to-speech system. The system, which includes a syntactic parser and a set of stress rules, is based on a morpheme dictionary and a set of phonological, morpho- phonological and orthographical rules; a design assuming a full lexicon of words with their phonemic transcriptions would not permit the system to convert unrestricted text. The morpheme dictionary consists of approximately 12 000 monomorphemes and a number of morphological complexes. Each item is marked for its syntactic and morphological category. In the presentation attention will be drawn to similarities and differences between this system and other systems such as MITalk. 85 AN INTEGRATED SYSTEM FOR MORPHOLOGICAL ANALYSIS OF THE SLOVENE LANGUAGE Toma~ Erjavec, Peter Tancig NLU Lab., Department of Computer Science and Informatics Jo~ef Stefan Institute Jamova 39, 61000 Ljubljana Yugoslavia ABSTRACT: The paper presents an integrated environment for morphological analysis of word-forms of the Slovene language. The system consists of a lexicon input and maintenance module, a lexicon output module for accessing lexical word forms, a two-level rule compiler and a two-level morphological analysis/synthesis unit. The basic paradigms and lexical alternations of word forms are handled by the lexicon system, while the two-level component takes care of phonologically induced alternations. 1. Introduction We present an integrated environment for morphological analysis of (written) word-forms of the Slovene language. The language belongs to the Slavic family of languages, but exhibits some very idiosyncratic properties (e.g. having also a "dual" number and very rich inflection). Our project of writing a morphological analyzer and synthesizer (MAS) for the Slovene language has had primarily two aims. First, to write a useful MAS, which could serve as a front-end to other Slovene language processing systems, and second, to implement a model general enough to allow us to facilitate the study of Slovene morphology. The work on the project itself is split into two parts much along the same lines. First of the task of selecting and implementing a model versatile enough to cover the quirks of Slovene morphology, and second, the task of writing down the rules of Slovene morphology (Toporigi~ 84) in the formalism of the chosen model. The two-level model of Kimmo Koskeniemmi (Karttunen 88, Koskenniemi 84,85,86) was selected as the basic scheme for our MAS, our choice being influenced - among other things - by its prevalence in current (computer) morphological studies. This makes the system well documented and thus easy to implement, as well as simplifying the task of writing the rules for (phonologically) induced alternations of Slovenian word-forms (Erjavec 89). 2. Structure The system was implemented on VAX/VMS in Quintus Prolog and consists of the following parts: (1) The compiler, which takes as its input two-level rules and produces final state automata (transducers). (2) The lexicon module which provides a user interface for the creation and updating of the lexicon - the lexicon input module. …

Content truncated to 500 words