Corpora based Approach for Arabic/English Word Translation Disambiguation
Farag Ahmed, Andreas Nürnberger · 2009
We are presenting a word sense disambiguation method applied in automatic translation of a query from Arabic into English. The developed machine learning approach is based on statistical models, that can learn from parallel corpora by analysing the relations between the items included in this corpora in order to use them in the word sense disambiguation task. The relations between items in this corpora are obtained by using and developing a purely statistical method from corpora in order to avoid the use of structured linguistic resources like ontology which are not yet available for Arabic in an appropriate quality. The results of this analysis should provide us with some useful semantic information that can help to find the best translation equivalents of the polysemous items.