In this paper, we report the results of an experiment aimed at automatically mapping corpus-derived Semantic Types to WordNet synsets. The algorithm for the automatic alignment of Semantic Types with WordNet synsets relies on lexical correspondence, i.e. it performs an automatic alignment of Semantic Types labels with the corresponding WordNet entry nouns, when present (for example, the Semantic Type [[Activity]] is mapped to synsets containing the entry noun "activity". In this way, 150 Types out of 180 are mapped automatically, while 30 gaps have to be resolved manually. Automatic mapping based on lexical correspondence, however, does not guarantee that the mapping is good, i.e. that the items which make up the extension of a certain Semantic Types match the set of hyponyms of the corresponding synset(s). An evaluation of 43 Semantic Types against a gold standard reveals that, for 30% of them, a manual revision is needed.

Mapping Semantic Types onto WordNet Synsets

Jezek, Elisabetta;Feltracco, Anna;Gatti, Lorenzo;Magnolini, Simone;Magnini, Bernardo
2016

Abstract

In this paper, we report the results of an experiment aimed at automatically mapping corpus-derived Semantic Types to WordNet synsets. The algorithm for the automatic alignment of Semantic Types with WordNet synsets relies on lexical correspondence, i.e. it performs an automatic alignment of Semantic Types labels with the corresponding WordNet entry nouns, when present (for example, the Semantic Type [[Activity]] is mapped to synsets containing the entry noun "activity". In this way, 150 Types out of 180 are mapped automatically, while 30 gaps have to be resolved manually. Automatic mapping based on lexical correspondence, however, does not guarantee that the mapping is good, i.e. that the items which make up the extension of a certain Semantic Types match the set of hyponyms of the corresponding synset(s). An evaluation of 43 Semantic Types against a gold standard reveals that, for 30% of them, a manual revision is needed.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11582/305535
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
social impact