In brief
The International Conference on Artificial Intelligence and its technologies (MCAIT 2026) finished on November 6, 2026, at Taibah University, Medina, after three days of sessions that highlighted advances in AI applications across health, education, cybersecurity and smart cities. Researchers presented findings on the reliability of Arabic language models, a lip‑reading speech‑recognition system, medical question triage, historical manuscript digitization and automatic legal‑compliance checking.
The sixth International Conference on Artificial Intelligence and its technologies (MCAIT 2026) concluded on Friday, November 6, 2026, at Taibah University in Medina after three days of in‑person and virtual presentations. The event, organized by the university’s Institute for Advanced Computing, featured session themes that spanned healthcare AI, education, cybersecurity, and smart city technologies.
The conference emphasized collaborative research, noting ties with Malaysia’s national university that had contributed to several parallel sessions. A panel of speakers highlighted the pair’s joint efforts in expanding Arabic‑language AI resources across the region.
One of the most cited studies was presented by the researcher ريم التميمي, who conducted 17 separate analyses to gauge the reliability and dialectal sensitivity of large Arabic language models. Her team found that models varied markedly in their ability to preserve meaning across different dialects, with the most reliable models maintaining contextual integrity but still showing gaps in regional expressions. The research calls for a framework that expands evaluation metrics beyond standard benchmarks to better reflect linguistic diversity.
In a parallel session, researchers demonstrated a lip‑reading speech‑recognition prototype that decoded four Arabic words—"نعم، لا، آسف، شكرًا"—by tracking mouth movements. The system achieved an 87.02% accuracy rate, while the inclusion of an unknown‑word detection layer lowered accuracy to 78.01%. The experiment used deep‑learning neural networks trained on a dataset of characterized lip gestures.
The conference also featured a medical domain study by الباحثتان إيناس الرحيلي و أريج الهذلي, who classified roughly 170,000 Arabic questions on medical queries into three severity tiers. Their AraBERT model reached a 98.29% F1‑score, surpassing an alternative CNN model that scored 95.59%. The results suggest an AI‑driven triage tool could streamline patient guidance and prioritize urgent care needs.
Another session presented by الباحثة بشاير الصيدلانية examined Arabic optical character recognition applied to 500 lines of historic manuscripts. Results indicated a 27.92% error rate for one OCR engine compared with 43.66% for another, and a specialized restoration system achieved a markedly lower 4.06% error rate. The study underlines the importance of tailored OCR solutions for digitising fragile historical texts.
A separate research group featuring نورة الراجح و ميادة المصري explored the consistency of AI‑generated legal answers. They curated 90 bilingual legal questions regarding Saudi personal data protection and the European GDPR. The evaluation found 87% consistency across 254 paired replies from three language models, underscoring the potential for automated legal compliance checks while also highlighting the need for additional oversight.
Conference officials stated that the event underscored the university’s commitment to advancing AI capabilities in Arabic‑speaking contexts. They noted that the demonstrated research outcomes would inform future developments in health, education, cyber‑security and public‑services infrastructure, paving the way for broader adoption of AI solutions that respect cultural and regulatory nuances.
