Building an Arabic Multiword Expressions Repository

We introduce a list of Arabic multiword expressions (MWE) collected from various dictionaries. The MWEs are grouped based on their syntactic type. Every constituent word in the expressions is manually annotated with its full context-sensitive morphological analysis. Some of the expressions contain semantic variables as place holders for words that play the same semantic role. In addition, we have automatically annotated a large corpus of Arabic text using a pattern-matching algorithm that considers some morphosyntactic features as expressed by a highly inflected language, such as Arabic. A sample part of the corpus is manually evaluated and the results are reported in this paper.