G. Results and Evaluation of Hungarian Nominal WordNet v1.0.pdf

Results and Evaluation of Hungarian Nominal WordNet v1.0 Márton Miháltz and Gábor Prószéky MorphoLogic Késmárki u. 8, Budapest, 1118 Hungary Email: mihaltz@morphologic.hu, proszeky@morphologic.hu Abstract. This paper presents recent results of the ongoing project aimed at creating the nominal database of the Hungarian WordNet. We present 9 different automatic methods, developed for linking Hungarian nouns to WN 1.6 synsets. Nominal entries are obtained from two different machine-readable dictionaries, a bilingual English- Hungarian and an explanatory monolingual (Hungarian). The results are evaluated against a manually disambiguated test set. The final version of the nominal database is produced by combining the verified result sets and their intersections when confidence scores exceeded certain threshold values. 1 Introduction The project started in 2000, with the aim of creating a Hungarian nominal WordNet ontology with semi-automatic methods [6]. Our basic strategy was to attach Hungarian entries of a bilingual English-Hungarian dictionary to the nominal synsets of Princeton WordNet, version 1.6 (WN, [4]), following the so-called expand approach [7]. This way, the synsets formed by the Hungarian nouns can inherit the WN semantic relations. In order to achieve this, we used heuristic methods, developed partly by previous similar projects [1,2], and partly by us, which rely on information extracted from several machine-readable dictionaries (MRDs). This approach relies on the assumption that nominal conceptual hierarchies, which describe the world, would be similar across English and Hungarian languages to a degree which is sufficient for producing a preliminary version of our WordNet. 2 Machine-Readable Dictionaries Used We used two different MRDs to assist the heuristics which disambiguate the Hungarian nouns against Princeton WordNet synsets. The MoBiDic bilingual English-Hungarian electronic dictionary contains 17,700 Hungarian nominal entries, corresponding to