UAE scientists develop AI system to identify Arabic dialects in 22 countries

Researchers at the University of Sharjah in the UAE have created an artificial intelligence (AI) system capable of accurately identifying different Arabic dialects, reports Techxplore. Their work, published in IEEE Xplore, addresses the complexities of the Arabic language, which has numerous dialects across 22 countries.

“Arabic is a rich language with many regional dialects, each with its own unique vocabulary, expressions and pronunciation,” Professor Ashraf Elnagar, lead researcher was quoted as saying. “Our system can automatically identify which Arabic dialect someone is speaking.”

The AI model was trained on over 3,000 hours of audio from YouTube, covering 19 dialects from across the Middle East, North Africa and the Arabian Peninsula. The system has shown a remarkable 97.29 per cent accuracy for regional dialects and 94.92 per cent for country-specific dialects.

READ: UAE launches ‘world’s highest quality’ Arabic AI software

“What is remarkable is that we achieved this using only 29 per cent of the training data typically required by other researchers,” said Elnagar, adding that the system is publicly available for further development.

The AI system has potential applications in virtual assistants, translation services and customer support, bridging communication gaps between Arabic speakers. According to student researcher Amr Barakat, the project enables “more inclusive and accurate communication for Arabic speakers worldwide.”

Another student researcher, Abdulla Aldhaheri, noted the industry-wide interest in the project, as it “holds the potential for widespread adoption, offering numerous benefits and improvements to various AI-driven language applications and services.”

In May it was reported that a generative AI programme being developed in Saudi collaboration with IBM is set to work with several Arabic dialects. The Saudi Data and Artificial Intelligence Authority (SDAIA) announced that its Arabic large language model (LLM) for text generation will be integrated into IBM’s AI and data platform, watsonx.

READ: UAE-made AI models could position Middle East as global tech hub, says AWS executive