fixed array split and empty response errors in TranscriptionService

Joll59 / d-ser-t

d-ser-t quantifies speech recognition accuracy of the MSFT speech service and/or user created MSFT custom speech service models.

2 stars 2 forks source link

Closed ovishesh closed 5 years ago

ovishesh commented 5 years ago

fixed a bug where the arrays were being wrongly split. Replaced Math.floor with Math.ceil and added a check to have the most balanced spilt.
fixed a bug where empty utterance responses were being associated with expected responses.

Note that this does not fix the partial transcription bug as described on #4

HobbyProjects commented 5 years ago

Please consider adding a test. We just had an instances where a refactor broke another part and we didn't know until late.