The Implementation of Arabic Subject Markers in the LKB System

Arabic Subject Markers are interface phenomena (specifically between morphology and syntax). In this paper, I describe them briefly, I give my linguistic analysis within the framework of the Head-Driven Phrase Structure Grammar and I show how I implement them in the LKB system. I show that this system, despite its strength, does not allow for a proper implementation of these units. Standard Arabic (henceforth Arabic) Subject Markers (henceforth SMs) are morphemes that convey information on the gender, number and the person of subject or topic1. They are attached to perfective and imperfective verbs and respect well-defined morphological patterns. They are suffixes when attached to perfective verbs and prefixes and / or suffixes when attached to imperfective verbs. Two examples of these units are provided in (1) and (2). In (1) the SM -ta, which encodes the features 2MS, is attached to the perfective stem katab "wrote". In (2), the SM ti : , which encodes the features 2FS, is attached to the imperfective indicative stem ktub "write". (1) katab wrote -ta -2MS maqa :l paper -a -ACC -n -INDE “You (masculine) wrote a paper” (2) t2aktub write -i : -FS -na -IND maqa :l paper -a -ACC -n -INDE “You (feminine) write a paper” Introduction The study of these linguistic markers go back to the first Arabic grammarians, such as Sibawayhi in the thirteenth century. They have also a privileged place in contemporary linguistic research, although they do not constitute an object of research on their own right. Thus, we find that they are studied by addressing issues affecting the agreement, such as the well-known agreement asymmetries (see, inter alia, (Aoun, Benmamoun, and Sportiche 1994), (Harbert and Bahloul 2002) and (Benmamoun and Lorimor 2006)). Copyright c © 2009, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. See (Li and Thompson 1976). (Jebali 2008) is the first work entirely dedicated to the modeling of these markers, a work that includes the description, the linguistic analysis and the implementation. The problems raised by SMs are mainly morphosyntactic. The question that arises is : are they better treated as arguments or as agreement markers ? The terminology used to designate them in the literature hesitates between disparate labels such as pronouns (even pronominal affixes), agreement markers and clitics. These labels are most often not based on independently motivated criteria and authors who use them do not wonder about the impact of their classification on the analysis of other phenomena encountered in Arabic. I propose a linguistic analysis of SMs based on independent criteria , I assess the impact of this analysis on other phenomena, I formalize this analysis in the framework of the Head-Driven Phrase Structure Grammar (HPSG) (Ginzburg and Sag 2000), (Pollard and Sag 1994), (Pollard and Sag 1987) and (Sag, Wasow, and Bender 2003). I implement the analysis in the Linguistic Knowledge Building (LKB) system (Copestake 2002) to test its validity and I show some limitations of this system. Describing SMs Perfective SMs When attached to perfective verbs, SMs are suffixes. They precede any other morpheme that could be as well attached to the verb. In the example (3) we have such a case where the SM -ta (2MS) is attached to the verbal stem katab and the object marker -ha : (it) is attached to the left of the SM, any other order being impossible. (3) katab wrote -ta -2MS -ha : -it