Creation and Development of the Romanian Lexical Resources

Proceedings of Recent Advances in Natural Language Processing, pages 678–685, Hissar, Bulgaria, 12-14 September 2011. Creation and Development of the Romanian Lexical Resources Elena Boian, Constantin Ciubotaru, Svetlana Cojocaru, Alexandru Colesnicov, Ludmila Malahov, Mircea Petic Institute of Mathematics and Computer Science, Chis¸in˘au, Moldova {lena,chebotar,svetlana.cojocaru,kae,mal,mirsha}@math.md Abstract This article describes the Romanian lexical resources containing morpholog- ical data and dictionaries: synonyms, Romanian-English, and Romanian- Russian. The inflection process at the creation of morphological resources based on the functional grammar with scattered context is considered. An arbitrary word is in- flected knowing only its part of speech, and the gender for nouns. New words were obtained also by using prefixing and suffixing. The research in automated prefixing and suffixing permit- ted us to determine some word classes for which this method is applicable, and to im- plement the corresponding algorithms. We describe the database structure, and the DB population programming tools. The article describes an approach to the checking of integrity and correctness of the morphological resources presented as a database mapping Romanian words to their morphological derivatives. 1 Introduction Creation and development of the lexical resources are important parts of Natural Language Process- ing (NLP). One of such resources for the Romanian lan- guage are Reusable Resources for the Romanian Language (RRRL).1 This article describes how these lexical re- sources were created and developed, the database structure, and the corresponding programming tools. The morphological and, more specifically, the inflectional aspects are pointed out. 1http://www.math.md/elrr/ In Sec. 2, the implemented programs are de- scribed that populate the resources by automatic inflection (Boian and Cojocaru, 1996). The start- ing point for this approach was the book (Lom- bard and Gˆadei, 1981), where most of Romanian productive classes of words (nouns, adjectives and verbs) were classified according to their inflection groups. The classification was made from the lin- guistic point of view, and, for example, the accents were taken into account. In this case, it is possible to operate only with the graphical representation of the word that equally simplifies and complicates the problem. Nevertheless, this classification was useful and led to the idea to formalize word-forms producing with the special grammar that is pre- sented into Sec. 3. Other parts of speech (numer- als, pronouns, articles, conjunctions, prepositions, interjections) were entered into the Database (DB) manually as being not so numerous. It is shown also how the inflectional model for an arbitrary word can be determined (Sec. 4). Knowing this information it is possible to perform the inflection automatically (Sec. 5). The proposed solution is not restricted by the Romanian language but could be also applied to other natural languages with inflectional mecha- nisms similar to these of Romanian. Another way to populate the DB is affixing. New words were generated by prefixing and suf- fixing (Cojocaru et al., 2009). The research in the possibilities of automated prefixing and suffix- ing permitted to determine some word classes for which this method is applicable, and to implement the corresponding algorithms. This led to consid- erable …

Content truncated to 500 words