A corpus study on the number of true proportional analogies between chunks in two typologically different languages

Yves Lepage (1) , Julien Migeot (1) , Erwan Guillerm (1)

1 Equipe Hultech - Laboratoire GREYC - UMR6072

We measure the number of true proportional analogies between chunks in two typologically different languages on a similar corpus: a 20,000 sentence long Japanese-English bicorpus. We verify that at least 96% of analogies of form between chunks are also analogies of meaning. We conclude that analogy ought to be considered as a reliable structuring device between chunks.

Domaines: Traitement du texte et du document

Yves Lepage, Julien Migeot, Erwan Guillerm. A corpus study on the number of true proportional analogies between chunks in two typologically different languages. The seventh symposium on natural language processing, 2007, Bangkok, Thailand. pp.117-122. ⟨hal-00261018⟩