Evaluation

evaluate_samples

def evaluate_samples(lang_samples)

Calculates word and phoneme error rates per language and their mean across languages

Args

• lang_samples (Dict): Data to evaluate. Contains languages as keys and list of result samples as values. Prediction samples is given as a List of Tuples, where each Tuple is a tokenized representation of (text, result, target).

Returns

• Dict: Evaluation result carrying word and phoneme error rates per language.