Datasets:
refactor_loader_script
The goal of this work is to refactor MOCKS.py and allow for direct use of this testset in the Python scripts. Currently the testset is loaded with three columns: audio_id, audio and transcription. It was assumed that the user would prepare the pairs using */*/test/[offline,online]/[all,subset].pair.[positive,similar,different].tsv files. Such solution is not very efficient and does not use the datasets package. Hence the goal of this refactoring is to implement some better functionality.
The testset will be loaded with the following columns: keyword_id, keyword_transcription, test_id, test_transcription, test_audio, label. The goal is to allow for easy evaluation of the QbyT KWS. Unfortunately there is no easy and fast way to generate test cases for QbyE KWS with current implementation of the datasets package.
The work on this refactoring is finished and tested.
LGTM