A deep dive into how the morphological structure of human languages shapes the design of NLP systems — from classical rule-based methods to BPE and the frontiers of morphology-aware LLMs.
#transformers
Content tagged with "transformers"
Transformer-based sentiment classification for Arabic text (MSA and dialectal), built at CERIST using BERT and AraBERT.
Intent classification for Arabic conversational text, supporting downstream applications like chatbots and voice assistants.
Sentiment classification models generalized across multiple languages, extending Arabic-focused work to a broader multilingual setting at CERIST.