For this language pair we don’t have the negative effects of the errors from the machine translation system. As we can see in the figure, with this algorithm we can obtain about 20.000 translated synsets with a precision of 80%, and about 30.000 with a precision of 70% for the pair English-English. As we start the evaluation with the most frequent synset it means that the precision decreases with the decrease of frequency, as it was expected. As we can observe, the precision decreases with the number of synsets. In the figure we can observe the precision obtained regarding the number of translated synsets. In figure 1 we can observe the results of this algorithm for all the language pairs (for the translated pairs we show the results for Google Translate). For English-Catalan and English- Spanish we have used both Google Translate and Microsoft Translator. We conducted the experiments for three language pairs: For all the language pair we have tested all the algorithms. In table 2 we can observe the number of synsets and variants of these versions. These preliminary versions have been obtained from versions 1.6 applying a mapping and other techniques. The results obtained with the different algorithms have been automatically evaluated with a preliminary version of the Spanish and Catalan 3.0 WordNets. All sentences in the corpus, both source and target sentences, have been morphosyntactically tagged whith Freeling (Padró et al., 2010a). systems are statistical and allow us to translate from English to Spanish and Catalan.

Google translate in spanish and english