Dates: 2023
Principal Investigator: Dr. Irfan Ahmad
Funded by: SDAIA-KFUPM Joint Research Center for Artificial Intelligence

Description: This research explores new text representations to improve NLP for Arabic, focusing on dotless Arabic text, which could reduce vocabulary size, create more compact models, and speed up training without sacrificing performance. The work studies its impact on linguistic properties and downstream tasks like translation, speech recognition, and question answering, with implications for other languages using Arabic script. It also investigates dynamic embeddings to handle Arabic’s rich morphology more efficiently. Applied work includes figurative speech detection beyond sarcasm, mental health analysis from social media, and historical text classification, aiming to enhance broader NLP tasks. Beyond Arabic, the research also addresses imbalanced datasets with tailored methods and applies AI to fields like civil engineering and epilepsy prediction.