Chidinma Okoro

Developed and implemented comprehensive annotation guidelines for Igbo speech and text data, increasing data quality by 25% and reducing annotation discrepancies by 15%.

Trained and mentored a team of 10+ Igbo linguists and annotators, improving team productivity by 20% and ensuring consistent application of complex linguistic rules across projects.

Collaborated with AI engineers to identify and resolve model biases specific to Igbo dialects, enhancing model accuracy for diverse regional speech patterns by 18%.

Managed the full lifecycle of Igbo dataset creation, from raw data collection to final validation, supporting the launch of two new Igbo-language AI features.

Annotated over 50,000 utterances of Igbo speech and text data for a major speech recognition project, achieving a consistent accuracy rate of 98%.

Assisted in the onboarding and initial training of 5 new data annotators, effectively reducing their ramp-up time by 30% through structured guidance and feedback.

Identified and categorized over 200 unique Igbo linguistic phenomena (e.g., tone, compounding, idiomatic expressions) to improve annotation guideline specificity.

Participated in weekly quality assurance reviews, providing detailed feedback that led to a 10% improvement in overall dataset consistency.

Analyzed over 1,000 hours of recorded Igbo speech data, identifying key tonal patterns and their grammatical functions for a research publication.

Developed a preliminary digital corpus of Igbo proverbs and idioms, comprising over 500 entries, to support future NLP research.

Collaborated with senior researchers on a project investigating cross-dialectal variation in Igbo, presenting findings at two national linguistic conferences.

Assisted in data collection and transcription for various linguistic studies, ensuring high fidelity and adherence to phonetic standards.