Comprehensive Performance and Qualitative Analysis of a Speech Recognition Model
Conducted an in-depth technical analysis to benchmark and evaluate a leading speech recognition model (Whisper) on modern Apple Silicon hardware. This involved a significant qualitative assessment of transcription outputs, where I systematically compared model-generated text against a ground-truth dataset. I identified and categorized specific error patterns related to capitalization, spacing, and brand name recognition, creating a structured rubric to provide actionable feedback for developers.