2005/039 - Literality based sample sorting for syntax projection
- Bruno Cavestro,Nicola Cancedda
Cross-Language knowledge induction workshop, "Babes-Bolyai" University, Cluj-Napoca, Romania, 25 July - 6 August, 2005.
In the present paper we face the problem of projecting syntax trees over different sides of a parallel corpus, without using any language dependent feature. To achieve this task we introduce a literality score, and accordingly, sort the bi-sentences of the parallel corpus in different classes. We will then show how it is possible to iteratively train a parser over those classes