Enlarging the Croatian WordNet with WN-Toolkit and Cro-Deriv

Proceedings of Recent Advances in Natural Language Processing, pages 480–487, Hissar, Bulgaria, Sep 7–9 2015. Enlarging the Croatian Wordnet with WN-Toolkit and CroDeriV Antoni Oliver Universitat Oberta de Catalunya aoliverg@uoc.edu Kreˇsimir ˇSojat Sveuˇciliˇste u Zagrebu ksojat@ffzg.hr Matea Srebaˇci´c Sveuˇciliˇste u Zagrebu msrebaci@unizg.hr Abstract Wordnet is a standard semantic resource for several Natural Language Processing tasks and it is available for an increas- ing number of languages. The Croatian Wordnet (CroWN) was a relatively small resource with 10.026 synsets and 31.367 synset-variant pairs covering only 45.91% of the so-called Core WordNet. Com- paring these figures with the size of the Princeton WordNet for English version 3.0, that has 117,659 synsets and 206,975 synset-variant pairs, it is clear that the CroWN should be expanded. First exper- iments for the expansion of the CroWN were performed using the WN-Toolkit, a set of Python programs for wordnet cre- ation and expansion using dictionary, Ba- belnet and parallel-corpora based strate- gies. The WN-Toolkit was previously suc- cessfully applied to other languages as Spanish, Catalan and Galician. After this first expansion, CroWN reached 70.63% of the core wordnet. In the second step we used CroDeriv, a derivational database for Croatian and the manual creation of 1,457 synset-variant pairs until reaching 100% of the Core WordNet. After second step was completed, CroWN reached 23,137 synsets and 47,931 synset-lemma pairs. 1 Introduction In this paper we explain the methodology and re- sults of the experiments for the enlargement of the Croatian Wordnet using the WN-Toolkit and the derivational database CroDeriV. The paper is or- ganised as follows: first, we will explain the de- velopment of the previous version of CroWN and we will present some figures about the size of this wordnet before and after the expansion. Then we will present the WN-Toolkit and its main fea- tures. After that, in section 4, we will present the CroDeriV, morphological database of Croatian verbs which was used in one of our experiments. Next, the experimental methodology is presented followed by the results of the experiments. After that the main sources of errors are presented and analyzed. Finally, the conclusions and future work are presented. 2 CroWN and wordnets for other languages The Croatian Wordnet has been developed under the Central and South-East European Resources (CESAR) project, funded by the European Com- mission (50%) and the Ministry of Science, Ed- ucation and Sports of the Republic of Croatia (50%). The first version of the Croatian Word- net had 10.026 synsets and 31.252 synset-variant pairs. The synset ID’s are those of the Princeton WordNet for English v 3.0. The Princeton WordNet for English version 3.0 has 117,659 synsets and 206,975 synset-variant pairs. In table 1 we can observe the number of synset variant pairs both in old and new versions of CroWN. In table 2 the number of synsets in both versions is shown. The starting version of the Croatian Wordnet covered 45.91% of the so- called Core WordNet (Boyd-Graber et al., 2006), that is, approximately the 5,000 most frequently used …

Content truncated to 500 words