in voxpopuli/text/wer_tools.py [0:0]
def load_word_align_file(path_file: Path) -> List[WordAlignFile]:
with open(path_file, "r") as file:
data = json.load(file)
out = []
for file_data in data:
align_path = get_align_index_path(
file_data["target"], file_data["word_prediction_no_lm"]
)
if len(align_path) == 0:
continue
out.append(
WordAlignFile(
file_id=correct_name_fbcluster_output(file_data["sample_id"]),
target=file_data["target"],
decoded=file_data["word_prediction_no_lm"],
wer=file_data["wer"],
ler=file_data["ler"],
align_path=align_path,
)
)
return out