Towards a methodology for filtering out gaps and mismatches across wordnets: the case of plWordNet and Princeton WordNet
Ewa Katarzyna Rudnicka, Wojciech Witkowski, Łukasz Grabowski · 2016
This paper presents the results of large-scale noun synset mapping between plWordNet, the wordnet of Polish, and Princeton Word-Net, the wordnet of English, which have shown high predominance of inter-lingual hyponymy relation over inter-synonymy relation.Two main sources of such effect are identified in the paper: differences in the methodologies of construction of plWN and PWN and cross-linguistic differences in lexicalization of concepts and grammatical categories between English and Polish.Next, we propose a typology of specific gaps and mismatches across wordnets and a rule-based system of filters developed specifically to scan all I(inter-lingual)-hyponymy links between plWN and PWN.The proposed system, it should be stressed, also enables one to pinpoint the frequencies of the identified gaps and mismatches.