engleski

Towards Parsing Croatian Complex Sentences: Dependent Noun Clauses

In this paper, authors will present methods for parsing Croatian complex sentences in which a dependent clause serves as a direct object to the main verb. This research is based on the resources that have already been developed for parsing simple Croatian sentences. So far, sentences that we were able to parse using these resources are of the basic structure consisting of a subject, verb, direct and indirect object, adverbial of time and place. Methods we shall present in this paper will extend this structure to the following sentence structure <main clause <dependent clause>> and, although quite rare and stylistically marked, to the structure <<dependent clause> main clause>. Our primary indicator for this type of sentence will be the absence of the required direct object in the main clause as well as the presence of one of the subordinating conjunctions (‘da’, ‘kako’) or complementizers (relative pronoun, adverb of place, time, cause or manner). Since this type of complex sentences is very common in Croatian language, we believe that this research will be a valuable contribution to Croatian module for NooJ. At the end of the paper, we will evaluate the adequacy of the model through precision, recall and f-measure.

parse trees; partial parser; main clause; dependent clause; Croatian language; NooJ.

nije evidentirano