DIGITAL PRESERVATION OF ARABIC EUPHEMISTIC SYNONYMY: A POST-HUMAN PEDAGOGICAL APPROACH TO CULTURAL HERITAGE THROUGH LARGE LANGUAGE MODELS

The preservation of Arabic linguistic heritage faces unprecedented challenges in the digital age, particularly regarding euphemistic synonymy—a fundamental aspect of Arabic cultural identity. This study examines how large language models (LLMs) can serve as digital preservation tools for Arabic euphemistic traditions while maintaining cultural authenticity through post-human pedagogical frameworks. We evaluated GPT-4o's capacity to identify, contextualize, and preserve euphemistic synonym pairs that satisfy classical Arabic semantic criteria, focusing on cultural transmission rather than mere computational accuracy. Six heritage-rich textbook units were processed through 150 culturally-sensitive prompts, generating 7,300 model tokens. Two senior Arabic cultural specialists created a 26-pair gold standard including three historically contested pairs from the ithbāt/inkār debate, rating system outputs on cultural preservation adequacy. GPT-4o achieved F₁ = 0.81 overall (F₁ = 0.67 on contested pairs), with 78% of generated cultural annotations deemed preservation-adequate. Implementing sustainable AI practices through local 8-bit inference reduced energy consumption by 32%, aligning with ethical cultural preservation principles. Results demonstrate that LLMs, when guided by cultural heritage preservation frameworks and post-human pedagogical approaches, can effectively support the digital transmission of Arabic linguistic traditions while maintaining cultural integrity and environmental responsibility.