Wordnets: State of the Art and Perspectives. Case Study: the Romanian Wordnet

During a quarter of a century of existence and in spite of much criticism, wordnets have thoroughly proved their appropriateness as repositories of linguistic knowledge and their usefulness in various applications. In this paper we present the methodology of creating the Romanian wordnet (RoWN), with special emphasis on the strategies adopted during ten years of ceaseless implementation and which highlight the efforts invested, the way we dealt with the alignment of the RoWN (previously aligned to PWN 2.0) to the PWN 3.0, as well as the future work we envisage for enriching and extending this resource. 1 Generalities on wordnets Language is a system of signs. This structuralist perspective on language serves extremely well the description of natural languages by both theoretical linguists and specialists in the formal representation of language. Notions like paradigm (i.e. class of similar elements), syntagm (i.e. a linguistic environment in which the elements of a paradigm can occur), value (the distinguishable functional role of an element in a syntagm) are modeled to serve the formal representation of language as a whole. Among the different knowledge representation formalisms, we focus here on wordnets, a special kind of semantic networks. While semantic networks have words in nodes and the arcs are semantic relations, wordnets are definitely more than that: they are:  monolingual dictionaries: they contain words with definitions for each of their senses;  multilingual dictionaries: via the InterLingual Index, ILI, access from one language-specific network to all the others is facilitated; thus, it is possible to compare the organization of the lexical material of various languages, to find examples supporting the thesis of semantic specificity of languages, to introduce the multilingual dimension in various applications relying on wordnets;  thesauri: lexical information is organized in terms of word meanings, not word forms;  lexical ontologies: wordnets contain concepts lexicalizations from various domains and the relations between these concepts lexicalizations. There have been more projects enriching wordnet with ontological information: WordNet Domains (Pianta et al. 2002), SUMO (Niles and Pease 2003).