engleski

Formalization of morphosyntactic features of flective language as exemplified by Croatian

The article is about a morphosyntactic model of Croatian language with which help the formal processing is being approached. It presents grammatical basis of the project 'Machine Understanding of Croatian Language' (MZT-RH-130440) in tagger developing for Croatian language which is based on morphologic generator included in Microsoft Word's Spelling Checker (Silić, 1996). Besides the researches on the removal of ambiguity in formal processing of natural languages, based in the first place on qualitative (grammatical) criteria, classes of morphosyntactic units obtained by formal analysis can also serve as a theoretical approach to the typology of word classes. There are 15 classes of lexical units in Croatian language, divided according to 31 morphosyntactic features of 9 formal categories. The work is divided in two parts: the first one contains a summary of formal features, their structure and methodology applied for the division into classes of units or parts of speech. The second part provides a concise explanation of each unit class, and points out problematic situations with the modes applied for their solving. It is emphasized that (i) the introduction of the full morphosyntactic constraints in the classification of units into classes and types is possible, (ii) formal procession of natural language could use features that traditional linguistics does not find descriptive, (iii) due to the morphosyntactic predetermination of syntax, parsing should start from the morphosyntactic features seen as the minimal units, (iv) every formal analysis which takes into consideration the grammatical categorial features is marked by language-specific approach.

natural language procesing; computational linguistics; flective languages; Croatian language; morphosyntax; formal features; valency features; morphosyntactic features; word classes; tagging

nije evidentirano