A word-grammar based morphological analyzer for agglutinative languages

A word-grammar based morphological analyzer for agglutinative languages Aduriz I.�, Agirre E., Aldezabal I., Alegria I., Arregi X., Arriola J. M., Artola X., Gojenola K., Maritxalar A., Sarasola K., Urkia M.� Dept. of Computer Languages and Systems, University of the Basque Country, 649 P. K., E-20080 Donostia, Basque Country �UZEI, Aldapeta 20, E-20009 Donostia, Basque Country �Universidad de Barcelona, Gran Vía de las Cortes Catalanas, 585, E-08007 Barcelona jipgogak@si.ehu.es. Abstract Agglutinative languages present rich morphology and for some applications they need deep analysis at word level. The work here presented proposes a model for designing a full morpho- logical analyzer. The model integrates the two-level formalism and a unification-based formalism. In contrast to other works, we propose to separate the treatment of sequential and non-sequential morpho- tactic constraints. Sequential constraints are applied in the segmentation phase, and non-sequential ones in the final feature-combination phase. Early appli- cation of sequential morphotactic constraints during the segmentation process makes feasible an efficient implementation of the full morpho- logical analyzer. The result of this research has been the design and implementation of a full morphosyntactic analysis procedure for each word in unrestricted Basque texts. Introduction Morphological analysis of words is a basic tool for automatic language processing, and indispensable when dealing with highly agglutinative languages like Basque (Aduriz et al., 98b). In this context, some applications, like spelling correction, do not need more than the segmentation of each word into its different component morphemes along with their morphological information. However, there are other applications such as lemmatiza- tion, tagging, phrase recognition, and determination of clause boundaries (Aduriz et al., 95), which need an additional global morphological parsing1 of the whole word. Such a complete morphological analyzer has to consider three main aspects (Ritchie et al., 92; Sproat, 92): 1) Morphographemics (also called morpho- phonology). This term covers orthographic variations that occur when linking morphemes. 2) Morphotactics. Specification of which morphemes can or cannot combine with each other to form valid words. 3) Feature-combination. Specification of how these morphemes can be grouped and how their morphosyntactic features can be combined. The system here presented adopts, on the one hand, the two-level formalism to deal with morphographemics and sequential morpho- tactics (Alegria et al., 96) and, on the other hand, a unification-based word-grammar2 to combine the grammatical information defined in morphemes and to tackle complex morphotactics. This design allowed us to develop a full coverage analyzer that processes efficiently unrestricted texts in Basque. The remainder of this paper is organized as follows. After a brief description of Basque morphology, section 2 describes the architecture for morphological processing, where the morphosyntactic component is included. Section 3 specifies the phenomena covered by the analyzer, explains its design criteria, and presents implementation and evaluation details. Section 4 compares the 1 This has also been called morphosyntactic parsing. When we use the term morphosyntax we will always refer to the hierarchical structure at word level, combining morphology and syntax. 2 The term word-grammar should not be confused with the syntactic …

Content truncated to 500 words