Identifying Broken Plurals, Irregular Gender, and Rationality in Arabic Text

Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics, pages 675–685, Avignon, France, April 23 - 27 2012. c⃝2012 Association for Computational Linguistics Identifying Broken Plurals, Irregular Gender, and Rationality in Arabic Text Sarah Alkuhlani and Nizar Habash Center for Computational Learning Systems Columbia University {sma2149,nh2142}@columbia.edu Abstract Arabic morphology is complex, partly be- cause of its richness, and partly because of common irregular word forms, such as broken plurals (which resemble singular nouns), and nouns with irregular gender (feminine nouns that look masculine and vice versa). In addition, Arabic morpho- syntactic agreement interacts with the lex- ical semantic feature of rationality, which has no morphological realization. In this paper, we present a series of experiments on the automatic prediction of the latent linguistic features of functional gender and number, and rationality in Arabic. We com- pare two techniques, using simple maxi- mum likelihood (MLE) with back-off and a support vector machine based sequence tagger (Yamcha). We study a number of orthographic, morphological and syntactic learning features. Our results show that the MLE technique is preferred for words seen in the training data, while the Yam- cha technique is optimal for unseen words, which are our real target. Furthermore, we show that for unseen words, morphological features help beyond orthographic features and that syntactic features help even more. A combination of the two techniques im- proves overall performance even further. 1 Introduction Arabic morphology is complex, partly because of its richness, and partly because of its com- plex morpho-syntactic agreement rules which de- pend on functional features not necessarily ex- pressed in word forms. Particularly challeng- ing are broken plurals (which resemble singu- lar nouns), nouns with irregular gender (mascu- line nouns that look feminine and feminine nouns that look masculine), and the semantic feature of rationality, which has no morphological re- alization (Smrž, 2007b; Alkuhlani and Habash, 2011). These features heavily participate in Ara- bic morpho-syntactic agreement. Alkuhlani and Habash (2011) show that without proper model- ing, Arabic agreement cannot be accounted for in about a third of all noun-adjective pairs and a quarter of verb-subject pairs. They also report that over half of all plurals in Arabic are irregular, 8% of nominals have irregular gender and almost half of all proper nouns and 5% of all nouns are rational. In this paper, we present results on the task of automatic identification of functional gender, number and rationality of Arabic words in con- text. We consider two supervised learning tech- niques: a simple maximum-likelihood model with back-off (MLE) and a support-vector-machine- based sequence tagger, Yamcha (Kudo and Mat- sumoto, 2003). We consider a large number of orthographic, morphological and syntactic learn- ing features. Our results show that the MLE tech- nique is preferred for words seen in the training data, while the Yamcha technique is optimal for unseen words, which are our real target. Further- more, we show that for unseen words, morpho- logical features help beyond orthographic features and that syntactic features help …

Content truncated to 500 words