def smooth(p, gendrop=False, hierarchy=False): ''' Error smoothing function Takes the vector output by the neural network Partitions it into vectors each representing a potential phoneme Converts those partitions to the closest phoneme vectors ''' # Partition the input vector into individual phonemes chunked_list = list(chunks(p, 12)) # Get list of phoneme tuples phoneme_tuples = phon_to_feat.values() output_tuple = () # For potential phoneme in tuple, find closest phoneme to it using dictionary keying distance to phoneme for phoneme in chunked_list: dist_from_realphon = { dist(phoneme, phoneme_tuples[i]): phoneme_tuples[i] for i in range(len(phoneme_tuples)) } smoothed_vector = min(dist_from_realphon.keys()) output_tuple += dist_from_realphon[smoothed_vector] return output_tuple
def smooth(p): ''' Error smoothing function Takes the vector output by the neural network Takes dictionary that converts suffix to relevant tuple Converts those partitions to the closest phoneme vectors ''' # Partition the input vector into individual phonemes chunked_list = list(chunks(p, n_feat)) # Get list of phoneme tuples phoneme_tuples = phon_to_feat.values() output_tuple = () # For potential phoneme in tuple, find closest phoneme to it using dictionary keying distance to phoneme for phoneme in chunked_list: dist_from_realphon = { dist(phoneme, phoneme_tuples[i]): phoneme_tuples[i] for i in range(len(phoneme_tuples)) } smoothed_vector = min(dist_from_realphon.keys()) output_tuple += dist_from_realphon[smoothed_vector] return output_tuple
def smooth(p, gendrop=False, hierarchy=False): ''' Error smoothing function Takes the vector output by the neural network Partitions it into vectors each representing a potential phoneme Converts those partitions to the closest phoneme vectors ''' # Partition the input vector into individual phonemes chunked_list = list(chunks(p, 12)) # Get list of phoneme tuples phoneme_tuples = phon_to_feat.values() output_tuple = () # For potential phoneme in tuple, find closest phoneme to it using dictionary keying distance to phoneme for phoneme in chunked_list: dist_from_realphon = {dist(phoneme, phoneme_tuples[i]): phoneme_tuples[i] for i in range(len(phoneme_tuples))} smoothed_vector = min(dist_from_realphon.keys()) output_tuple += dist_from_realphon[smoothed_vector] return output_tuple
def smooth(p): ''' Error smoothing function Takes the vector output by the neural network Takes dictionary that converts suffix to relevant tuple Converts those partitions to the closest phoneme vectors ''' # Partition the input vector into individual phonemes chunked_list = list(chunks(p, n_feat)) # Get list of phoneme tuples phoneme_tuples = phon_to_feat.values() output_tuple = () # For potential phoneme in tuple, find closest phoneme to it using dictionary keying distance to phoneme for phoneme in chunked_list: dist_from_realphon = {dist(phoneme, phoneme_tuples[i]): phoneme_tuples[i] for i in range(len(phoneme_tuples))} smoothed_vector = min(dist_from_realphon.keys()) output_tuple += dist_from_realphon[smoothed_vector] return output_tuple