engleski

Improved Parser for Simple Croatian Sentences

In this paper, authors will present the work that has been done to improve the existing syntactic parser. This work is a continuation of the work presented at the NooJ 2009 conference. We will show and explain the grammar for detecting nominal predicate in a simple sentence. The nominal predicate in Croatian language is made of the auxiliary verb ‘to be’ and an <NP> in Nominative case. The <NP> can be a complex <NP> made of a single noun and any number of adjectives, pronouns and numbers proceeding that noun and agreeing with it in number, gender and case, but also a single noun, a single pronoun, a single adjective or even an adverb. A problem of coordination of two or more <NP> nodes of different gender and its agreement with the main verb in the cases where coordination is a subject of a sentence will be discussed. The work will further enlight and discuss other important properties of Croatian sentence complexity. At the end of the paper, the results will be evaluated through precision, recall and f-measure to show the adequacy of the model.

parse trees; partial parser; simple sentences; Croatian language; NooJ; coordination; nominal predicate

nije evidentirano