Comparison of the Final Wordnets Dutch, Spanish and Italian
Piek T. J. M. Vossen, Salvador Climent, Maria Antònia Martí · Data Archiving and Networked Services (DANS) · 1999
This deliverable describes the comparison of the final wordnets for Dutch, Spanish and Italian.The wordnets contain between 25-44K synsets and 50-70K word meanings.This is between 30-50% the size of WordNet1.5. The comparison of the wordnets is done on the basis of the ILI-records to which the synsets refer.There is not a one-to-one mapping of synsets to ILI-records and therefore the comparison is only a rough approximation of the compatibility.Three types of comparisons have been done:1. intersection of the associated ILI-records: this indicates the possible translatability of concepts across the languages.2. the clustering of the associated ILI-records over the EuroWordNet top-ontology: this gives an indication of the conceptual coverage and balancing of the wordnets.3. the compatibility of hyponymy relations in the wordnets, projected on the associated ILIrecords: this gives a rough indication of the similarity in classification structure across the wordnets.The overall statistics is useful for users of the database to get an idea of the global coverage and matching of the data. LE2-4003 EuroWordNetLE2-4003 EuroWordNet Table 9: Dutch, Spanish and Italian Verbs clustered over the WordNet1.5Lexicographer's file codes WN NL ES IT Lexicographer's file code TC-Tokens % of wn TC-Tokens % wn TC-Tokens % wn TC-Tokens % wn 29 verb.body1095 0.39% 719 65.66% 505 46.12% 484 44.20% 30 verb.change