At Plotweaver, I transcribe and annotate complex West African audio, including recordings with background noise, overlapping speakers, and regional accents. My work has maintained 98%+ text accuracy across more than 150 hours of audio.
I segment recordings into millisecond-timestamped text and attribute speech to speakers. I also correct automated transcription errors caused by localized phrasing and switching between English, Pidgin, and Yoruba.
At AEZAR, I reviewed AI-generated transcripts and transcribed interviews, focus groups, and field research voice notes. I preserved cultural context and formatted deliverables for machine learning pipelines.
On a West African ASR model training dataset project, I tagged local slang and transcribed phonetic details that automated tools missed, helping reduce the target model’s Word Error Rate.
