A New Representation Model for the Automatic Recognition and Translation of Arabic Named Entities with NooJ

Proceedings of Recent Advances in Natural Language Processing, pages 545–550, Hissar, Bulgaria, 12-14 September 2011. A New Representation Model for the Automatic Recognition and Translation of Arabic Named Entities with NooJ Héla Fehri Laboratory MIRACL, University of Sfax hela.fehri@fss.rnu.tn Kais Haddar Laboratory MIRACL, University of Sfax kais.haddar@fss.rnu.tn Abdelmajid Ben hamadou Laboratory MIRACL, University of Sfax abdelmajid.benhamadou@isimsf.rnu.tn Abstract Recognition and translation of named entities (NEs) are two current research topics with regard to the proliferation of electronic documents exchanged through the Internet. The need to assimilate these documents through NLP tools has become necessary and interesting. Moreover, the formal or semi- formal modeling of these NEs may intervene in both processes of recognition and translation. Indeed, the modeling makes more reliable the constitution of linguistic resources, limits the impact of linguistic specificities and facilitates transformations from one representation to another. In this context, we propose an approach of recognition and translation based on a representation model of Arabic NEs and a set of transducers resolving morphological and syntactical phenomena. 1 Introduction The formal or semi-formal modeling of NEs can be involved in recognition and translation process. It enables to constitute more reliable linguistic resources. Indeed, such a modeling can represent all the constituents of a NE in a standard manner and limit the impact of linguistic specificities. In fact, a formal representation of Arabic NEs can help, firstly, in the identification of dictionaries and grammars required for a given application and, secondly, in the use of advanced linguistic methods of translation (i.e., transfer or pivot method). This abstraction level favors the reuse of certain linguistic resources. The elaboration of a formal and generic representation of an NE is not an easy task because, on the one hand, we have to find a representation that takes into consideration the concept of recursion and length of NE. In fact, a NE can be formed by other NEs. So, its length is not known in advance. On the other hand, the representation to be proposed should also contain a sufficient number of features that can represent any NE independently of the domain and grammatical category. It is in this context that the present work is situated. In fact, the main objective is to propose an approach of recognition and translation of Arabic NEs based on a representation model, a set of bilingual dictionaries and a set of transducers resolving morphological and syntactical phenomena related to the Arabic NEs and implemented with the linguistic platform NooJ (Silberztein, 2005). In this paper, we present, firstly, a brief overview of the state-of the art. Next, we describe the hierarchy type of Arabic NEs and the identified problems in recognition and translation processes. Then, we detail our proposed representation model. After that, we give a general idea of our resources construction and their implementation in the linguistic platform NooJ. Finally, the paper concludes with some perspectives. 2 Related work Research on NEs revolves around two complementary axes: the first involves the typing of NEs while the second concerns …

Content truncated to 500 words