Central Kurdish (Sorani) Sentiment Analysis investigates opinion mining, emotion detection, and subjective language classification for social media and digital texts written in Kurdish. This research explores both machine learning classifiers and modern deep learning transformer architectures adapted to low-resource agglutinative languages.
Research Objectives
The project addresses the unique challenges of Kurdish social media texts, including informal slang, code-switching, script variations, and the lack of standardized annotated polarity corpora.
Key Features
- Sentiment classification and polarity detection across Sorani texts
- Explainable machine learning models for low-resource NLP evaluation
- Curated Kurdish sentiment lexicons and domain-specific benchmark datasets
- Comparative analysis of transformer-based contextual language models
Methodology
The methodology combines rule-based lexicon processing with fine-tuned multilingual and Kurdish-adapted transformer architectures, evaluated across public opinion and social commentary corpora.
Impact
This project advances Kurdish computational linguistics, providing automated tools for market analysis, public sentiment monitoring, and sociolinguistic research in Kurdish digital spaces.