An unsupervised approach for bootstrapping Arabic sense tagging
Mona Diab · 2004
To date, there are no WSD systems for Arabic. In this paper we present and evaluate a novel unsupervised approach, SALAAM, which exploits translational correspondences between words in a parallel Arabic English corpus to annotate Arabic text using an English WordNet taxonomy. We illustrate that our approach is highly accurate in ≤ 90.1% of the evaluated data items based on Arabic native judgement ratings and annotations. Moreover, the obtained results are competitive with state-of-the-art unsupervised English WSD systems when evaluated on English data.