jonatasgrosman / huggingsound

HuggingSound: A toolkit for speech-related tasks based on Hugging Face's tools
MIT License
430 stars 42 forks source link

Skipped Reference when using evaluation #89

Open Shaobo-Z opened 1 year ago

Shaobo-Z commented 1 year ago

I received a warning message: WARNING:huggingsound.speech_recognition.model:6 references skipped because they were empty after text normalization

Should I worry about this message? Does it mean some of the data were not used for evaluation? (6 Skipped, 2 Used?)

Here are my sample data:

{"path": "sample-002469.mp3", "transcription": "we speak of them only to children once before"},
{"path": "sample-003480.mp3", "transcription": "why can't you be serious"},
{"path": "sample-002944.mp3", "transcription": "finally a young woman approached who was not dressed in black"},
{"path": "sample-001877.mp3", "transcription": "the woman was silent for some time"},
{"path": "sample-002230.mp3", "transcription": "the boy told him then that he needed to get to the pyramids"},
{"path": "sample-003314.mp3", "transcription": "you brought a new feeling into my crystal shop"},
{"path": "sample-003743.mp3", "transcription": "even though the sheep didn't teach me to speak arabic"},
{"path": "sample-004448.mp3", "transcription": "before this i always looked to the desert with longing said fatima"},