Addressing Machine Translation of Legal Terminology in Arabic Legislation: From Error Evaluation to Systems Optimisation

Legal translatability is a space of possibilities, an autonomous realm of ‘cross-cultural events’, within which the ‘system-bound’ of legal concepts and notions deeply rooted in language, history, culture, and societal evolution of one country are transformed and integrated into the language of another. Arabic-English / Arabic-French Legal translation is considered an arduous task because of the wide gap between the source and target languages linguistic and legal systems. Thus, translators from and into Arabic face difficulties on different linguistic levels, be they syntactic, semantic, or terminological (i.e., Sharī'ah Law vs Common and Civil Law terms). The forementioned legal systems are strikingly different and each of them is embedded in the cultural background of each system. Accordingly, the main complexity in legal translatability from Arabic will be to gather terminology of multiple origins and to transfer it into another linguistic and cultural framework. These frameworks are originally a binary code and comprises two semantic spaces: the source space Arabic and the target space English and French. To translate Arabic legislation that holds technical yet complex legal terminology, translators need access to extensive and reliable resources that serve as an ashgate guide to avoid literal translation, which may lead to serious legal consequences due to the misinterpretation of a legal term in a given legal context. Unfortunately, unlike for the Indo-European languages, there is a significant lack in the literature of Arabic legal terminology in terms of efforts to develop legal terminology resources and other resources that can be adopted to the needs of the legal translators. This absence has a direct effect on many Natural Language Processing applications namely the optimisation of Machine Translation systems to meet the accuracy requirement in the translation of Arabic legislation. Therefore, this thesis work comes as a reaction to fill this research gap by fostering collaborative initiatives between legal experts to develop a robust infrastructure of multilingual terminology for Arabic as a reference for translators, and as a foundational knowledge-base necessary for enhancing the accuracy criteria of Machine Translation systems required in the legal domain for Arabic. To do this, the study attempts to answer three main relevant questions: i) why is the translation of Arabic terminology in the legal domain challenging? ii) how accurate is the automatic translation of Arabic legal terms? iii) how can we reinforce the Machine Translation systems performance in the legal domain for Arabic?