engleski

Automatic Enrichment of Croatian Morphological Lexicon Using Large Corpora and Web Search

The first version of the Croatian Morphological Lexicon (HML) was developed as early as 1994 and was utilized in the implementation of various experiments and systems dealing with Croatian. Since the HML is frequently used both as a stand-alone application and as a module in many other systems for processing Croatian, the lexicon is constantly being updated to newer versions by manual inserting unknown wordforms (i.e. the corresponding 3- tuples of lemmas, wordforms and morphosyntactic tags) in batches. Current version of HML cosists of 110.000 lemmas and more than 4.000.000 lexicon entries. Due to limitations in availability of expert human annotators and various other constraints, the process of manual inspection, lemma assingment and inflectional pattern selection for unknown wordforms is a rather slow procedure. Accordingly, in this paper, we propose a generic approach to (semi-)automatic generation of new candidate lemmas for HML, their verification, assingment of inflectional patterns and finally creation and insertion of new lexicon entries to HML in a single processing pipeline.

morphological lexicon; automatic enlargement; Croatian language

nije evidentirano