Joint Prediction of Morphosyntactic Categories for Fine-Grained Arabic Part-of-Speech Tagging Exploiting Tag Dictionary Information

Proceedings of the 21st Conference on Computational Natural Language Learning (CoNLL 2017), pages 421–431, Vancouver, Canada, August 3 - August 4, 2017. c⃝2017 Association for Computational Linguistics Joint Prediction of Morphosyntactic Categories for Fine-Grained Arabic Part-of-Speech Tagging Exploiting Tag Dictionary Information Go Inoue, Hiroyuki Shindo and Yuji Matsumoto Graduate School of Information and Science Nara Institute of Science and Technology 8916-5 Takayama, Ikoma, Nara, 630-0192, Japan {inoue.go.ib4, shindo, matsu}@is.naist.jp Abstract Part-of-speech (POS) tagging for morpho- logically rich languages such as Arabic is a challenging problem because of their enor- mous tag sets. One reason for this is that in the tagging scheme for such languages, a complete POS tag is formed by com- bining tags from multiple tag sets defined for each morphosyntactic category. Previ- ous approaches in Arabic POS tagging ap- plied one model for each morphosyntactic tagging task, without utilizing shared in- formation between the tasks. In this pa- per, we propose an approach that utilizes this information by jointly modeling mul- tiple morphosyntactic tagging tasks with a multi-task learning framework. We also propose a method of incorporating tag dic- tionary information into our neural models by combining word representations with representations of the sets of possible tags. Our experiments showed that the joint model with tag dictionary information re- sults in an accuracy of 91.38% on the Penn Arabic Treebank data set, with an abso- lute improvement of 2.11% over the cur- rent state-of-the-art tagger. 1 1 Introduction Part-of-speech (POS) tagging is a fundamental task in natural language processing. The granu- larity of the POS tag set that reflects language- specific information varies from language to lan- guage. In morphologically simple languages such as English, the size of the tag set is typically less than a hundred. On the other hand, in morpholog- ically rich languages such as Arabic, the number 1Our code is available at https://github.com/ go-inoue/FineGrainedArabicPOSTagger of theoretically possible tags can be up to 333,000, of which only 2,200 tags might appear in an actual corpus (Habash and Rambow, 2005). One reason for this is that in the tagging scheme for such lan- guages, a complete POS tag is formed by com- bining tags from multiple tag sets defined for each morphosyntactic category. For example, a com- plete POS tag for the word Hb (“love”)2 can be de- fined as the combination of a noun from the coarse POS category, a nominative (n) from the case cat- egory, “not applicable” (na) from the mood cat- egory, and so on. The enormous number of re- sulting tags causes fine-grained POS tagging for Arabic to be more challenging. In order to perform this task, it is beneficial to utilize information from other morphosyntac- tic categories when predicting a label for one cat- egory. For example, if a word is a noun, it should take one of three tags from the case category: nominative (n), accusative (a), or genitive (g), while it should take “not applicable” (na) from the mood category since mood is not defined for nom- …

Content truncated to 500 words