Exploiting Hidden Morphophonemic Constraints for Finding the Underlying Forms of ‘weak’ Arabic Verbs

Proceedings of Recent Advances in Natural Language Processing, pages 448–454, Hissar, Bulgaria, 12-14 September 2011. Exploiting Hidden Morphophonemic Constraints for Finding the Underlying Forms of ‘weak’ Arabic Verbs Allan Ramsay School of Computer Science University of Manchester Manchester M13 9PL, UK Allan.Ramsay@manchester.ac.uk Hanady Mansour Department of Arabic, College of Arts and Sciences Qatar University, State of Qatar hanadyma@qu.edu.qa Abstract We present a treatment of Arabic morphol- ogy which allows us to deal with ‘weak’ verbs by paying attention to the underly- ing phonological process. This provides us with a very clean way of thinking about such verbs, and also makes maintenance of the lexicon very straightforward. 1 Introduction It is well known that Arabic morphology is com- plex: the language uses a combination of concate- native and discontinuous processes, and the effects of these are obscured by the fact that many phono- logically significant items (short vowels, gemina- tions) are not written in modern Arabic. We present a treatment of Arabic morphology which covers the standard cases, but which has two significant advantages. (i) We delay making decisions about the underlying form until we have the information that is necessary for getting the de- cision right. Unlike most attempts at diacriticisa- tion, we do not enumerate all the possible forms and then try to choose between them. Instead we leave decisions on specific diacritics until we are in a position to get them right–e.g. we delay choos- ing between declarative and interrogative present tense prefixes for a verb until we know whether it is being used in a statement or a question. This enables us to weave morphological and syntactic processing together very efficiently, as described below. (ii) We can take account of the phonolog- ical processes that produce the varying forms of ‘weak’ verbs without having to declare these verbs as belonging to a special class. Weak verbs are in fact regular verbs whose spelling reflects a small set of phonological contractions. Our analysis al- lows us to obtain ‘underlying forms’ for the sur- face forms of weak verbs which show how they are related to their roots. 2 Basic Mechanisms The basic problems of Arabic morphology are well known. A single word may have numerous forms, marking various syntactic features, where a form may have a combination of prefixes and af- fixes and the vowels at the heart of the word may vary. Thus ��� ����� (kataba) . . . are all forms of a single verb, with a variety of prefixes and suffixes mark- ing such things as tense, agreement and mode, and with each form involving different vowels be- tween the consonants ��� ������� (k?t?b). The situation is made worse by the fact that the short vowels, and a number of other significant items, are not generally written. This means that the full forms ��� ����� (kataba), ��� ��� �� (kutiba), . . . are all written as �� ��� (ktb). To make things even worse, Arabic gener- ally forms families of words around a single …

Content truncated to 500 words