Towards developing an Arabic word alignment annotation tool with some Arabic alignment guidelines

Word Alignment is an important supporting task for different NLP applications like training of machine translation systems, translation lexicon induction, word sense discovery, word sense disambiguation, information extraction and the cross-lingual projection of linguistic information. In this paper we study the main rules and guidelines required to build an aligner tool for Arabic language which should help in correcting most of alignment errors. These errors are identified by considering the outputs of some already existing aligner annotation tools.