Representing Meaning of Arabic Sentence Dynamically and more Smoothly

This paper presents work in progress on the formal representation of the meanings of Arabic sentences. It, specifically, deals with two limitations in the current approaches to Arabic sentence representation, namely, object identification and variable naming. Generally, the amount of work and research devoted to this area is very limited due to its high complexity. These limited attempts have adopted, either a Montagovian formalization, or a unification-based formalism. However, the two approaches have drawbacks. Montague’s framework has been criticized for its inadequacy to handle the anaphoric pronouns. The discourse representation theory (DRT) has provided a solution for this problem by the systematic use of free variables. However, the choice of names of the variables has remained problematic. That is, taking the set union on the set of markers, the idea of Henk Zeevat, means that the discourse representation structures (DRSs), containing the same names of variables, are taken to refer to the same object. Consequently, one should be extremely careful, whenever choosing these names. Instead, we adopt the so-called referent system, as proposed by Kees Vermeulen, in which the sets of variables are made disjoint, and the names of the variables are assigned locally, instead of assigning them globally in Zeevat’s approach. This original form of referent system fits well with a specific word order, namely SVO, because the renaming process is based on syntax. Arabic is characterized by a flexible word order, which restricts the role played by directionality in the renaming process. Consequently, we make some modifications to referent systems to associate variables in the DRSs with morphosyntactic information. This means that variables are identified if and only if morphosyntactic information matches.