Audio Data Annotation for Speech Recognition AI
Worked on a large-scale audio data labeling project for a speech recognition and natural language processing system. Tasks included transcribing audio files with accurate time stamps, tagging speaker emotions, identifying background noise, and performing speaker diarization. The dataset included voice assistant commands, call center conversations, and general human speech. Used tools like Label Studio and Audacity for accurate segmentation and labeling. Applied classification tags for speech type (command, query, conversation), emotion (neutral, happy, angry, etc.), and sound events (e.g., door closing, typing, traffic). Followed strict quality assurance guidelines, peer-reviewed annotations, and met a 98% accuracy benchmark across thousands of labeled samples.