This paper describes a morphological parser that is capable of handling more than 2,000 auxiliary forms in standard Basque, as arrayed in Euskaltzaindia’s Aditz laguntzaile batua (1973). The modular designed parser, implemented in Edinburgh Prolog, focuses on consistent patterns conveyed by inflectional morphemes. Feature-value pairs use a unification-based grammar to encode the relevant information (for example, number, person, valency, mood, or tense).
The main morphotactic generalizations are derived from a partition of the finite verb into the following components: initial preverbal affix (ba-); bound morphemes for first and second person and third-person markers, including split ergatives; various affixes; stems; bound pronominal morphemes with a dative argument; a bound pronominal morpheme column indicating the person for the ergative argument; and a postverbal suffix (-n).
The parser was developed in two person-weeks, thanks to the computational tools available in Manchester--f-structure facilities, grammar rule compilers, and a customized debugger. The authors briefly describe feature-value pairs based on LFG [1], a rewrite grammar, spelling rules, feature specifications and co-occurrences restrictions (FCRs), and rewrite rules. They emphasize a feature called choosetense, which invokes the co-occurrence pattern specified by its FCR and thus ensures that the f-structure has appropriate values for tense and mood. The authors hope to apply their parsing strategies to the development of a Turkish grammar.