Emotion tone analysis in dialectal Arabic detects fine-grained emotions in informal text but is difficult due to rich morphology, non-standard spelling, Arabizi, and many dialects. Research still focuses mostly on binary sentiment or MSA, while emotion categories and dialects such as Iraqi, Algerian, Moroccan, Egyptian, and Gulf lack labeled data and consistent annotation. Recent work uses ML and DL: traditional ML gives strong baselines but struggles with nuanced emotions, while hybrid and transformer models perform better with good preprocessing and embeddings. The main challenge is building robust, data-efficient pipelines that generalize across dialects and emotions with limited high-quality data.

Закзук Ш., Каража Н. (науч. рук. Авксентьева Е.Ю.) Machine learning and deep learning for emotion tone analysis in dialectal arabic texts // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2026]. URL: https://kmu.itmo.ru/digests/article/16920