Automatic extraction and evaluation of MWE

Leonardo ZilioLuiz SvobodaLuiz Henrique Longhi RossiRafael Martins Feitosa

This short paper aims at presenting a method for automatically extracting and evaluating MWE in the Europarl corpus. For this purpose we make use of mwetoolkit and utilize its output to find rules for the automatic evaluation of MWE. We then developed an XML parser to evaluate MWE candidates against those rules and also against online dictionaries. A sample of the results was manually evaluated by linguists and we had 87% of precision.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: