Synthetic speech clips for 101 rare drug terms, intended for evaluation only — measuring how
well an ASR system recognises rare / out-of-vocabulary medical vocabulary (target-word WER / CER /
recall). Each clip reads a real DailyMed sentence containing one target drug name, synthesised with
Google Gemini TTS across multiple voices. This is the frozen evaluation set from the master's thesis
"Audio-free lexical adaptation of Whisper's decoder"… See the full description on the dataset page:
https://huggingface.co/datasets/aharalambieva/target-words-geminitts.
