Recognition and translation Arabic-French of Named Entities: case of the Sport places

TL;DR

This paper presents a method combining translation and transliteration for Arabic-French Named Entity recognition and translation in the sports domain, improving translation accuracy especially for personal names.

Contribution

It introduces an integrated translation-transliteration approach using the NooJ platform, tailored for Arabic sports place names, with a refined typological model and evaluation results.

Findings

Improved translation accuracy for Arabic sports place names

Effective integration of transliteration into translation system

Positive evaluation results demonstrating method's effectiveness

Abstract

The recognition of Arabic Named Entities (NE) is a problem in different domains of Natural Language Processing (NLP) like automatic translation. Indeed, NE translation allows the access to multilingual in-formation. This translation doesn't always lead to expected result especially when NE contains a person name. For this reason and in order to ameliorate translation, we can transliterate some part of NE. In this context, we propose a method that integrates translation and transliteration together. We used the linguis-tic NooJ platform that is based on local grammars and transducers. In this paper, we focus on sport domain. We will firstly suggest a refinement of the typological model presented at the MUC Conferences we will describe the integration of an Arabic transliteration module into translation system. Finally, we will detail our method and give the results of the evaluation.