AI Research Consultant (Sentiment Analysis Project)
Provided AI research consulting support for text sentiment annotation of Nigerian e-service feedback from Google Play Store reviews. Applied aspect-based sentiment labeling and used inter-annotator agreement (Cohen's kappa) to ensure annotation reliability. Evaluated predictions from transformer models (AfriBERTa and mBERT) and reviewed labeling errors and model failure modes. • Labeled 5,000+ reviews into positive/negative/neutral (and aspect-based sentiment categories) • Measured labeling consistency using Cohen's kappa and QA procedures • Addressed class imbalance with SMOTE to improve minority sentiment learning • Documented collection-to-annotation-to-evaluation pipeline with audit trails