def dev_tuner(full_csv, split_type):
"""Tuning n_components on dev set.
Note: This is a very basic tunning for nn based affinity.
This is work in progress till we find a better way.
"""
DER_list = []
pval = None
for n_lambdas in range(1, params["max_num_spkrs"] + 1):
# Process whole dataset for value of n_lambdas
concate_rttm_file = diarize_dataset(full_csv, split_type, n_lambdas,
pval)
ref_rttm = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm = concate_rttm_file
[MS, FA, SER, DER_] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
)
DER_list.append(DER_)
# Take n_lambdas with minmum DER
tuned_n_lambdas = DER_list.index(min(DER_list)) + 1
return tuned_n_lambdas
def dev_p_tuner(full_csv, split_type):
"""Tuning p_value affinity matrix
"""
DER_list = []
prange = np.arange(0.002, 0.015, 0.001)
n_lambdas = None
for p_v in prange:
# Process whole dataset for value of p_v
concate_rttm_file = diarize_dataset(full_csv, split_type, n_lambdas,
p_v)
ref_rttm = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm = concate_rttm_file
[MS, FA, SER, DER_] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
)
DER_list.append(DER_)
# Take p_val that gave minmum DER on Dev dataset
tuned_p_val = prange[DER_list.index(min(DER_list))]
return tuned_p_val
def dev_nn_tuner(full_csv, split_type):
"""Tuning n_neighbors on dev set.
Assuming oracle num of speakers.
"""
DER_list = []
pval = None
for nn in range(5, 15):
# Fix this later. Now assumming oracle num of speakers
n_lambdas = 4
# Process whole dataset for value of n_lambdas
concate_rttm_file = diarize_dataset(full_csv, split_type, n_lambdas,
pval, nn)
ref_rttm = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm = concate_rttm_file
[MS, FA, SER, DER_] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
)
DER_list.append([nn, DER_])
DER_list.sort(key=lambda x: x[1])
tunned_nn = DER_list[0]
return tunned_nn[0]
def dev_nn_tuner(full_meta, split_type):
"""Tuning n_neighbors on dev set. Assuming oracle num of speakers.
This is used when nn based affinity is selected.
"""
DER_list = []
pval = None
# Now assumming oracle num of speakers.
n_lambdas = 4
for nn in range(5, 15):
# Process whole dataset for value of n_lambdas.
concate_rttm_file = diarize_dataset(full_meta, split_type, n_lambdas,
pval, nn)
ref_rttm = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm = concate_rttm_file
[MS, FA, SER, DER_] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
)
DER_list.append([nn, DER_])
if params["oracle_n_spkrs"] is True and params["backend"] == "kmeans":
break
DER_list.sort(key=lambda x: x[1])
tunned_nn = DER_list[0]
return tunned_nn[0]
def dev_ahc_threshold_tuner(full_meta, split_type):
"""Tuning threshold for affinity matrix. This function is called when AHC is used as backend.
"""
DER_list = []
prange = np.arange(0.0, 1.0, 0.1)
n_lambdas = None # using it as flag later.
# Note: p_val is threshold in case of AHC.
for p_v in prange:
# Process whole dataset for value of p_v.
concate_rttm_file = diarize_dataset(full_meta, split_type, n_lambdas,
p_v)
ref_rttm = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm = concate_rttm_file
[MS, FA, SER, DER_] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
)
DER_list.append(DER_)
if params["oracle_n_spkrs"] is True:
break # no need of threshold search.
# Take p_val that gave minmum DER on Dev dataset.
tuned_p_val = prange[DER_list.index(min(DER_list))]
return tuned_p_val
def dev_pval_tuner(full_meta, split_type):
"""Tuning p_value for affinity matrix.
The p_value used so that only p% of the values in each row is retained.
"""
DER_list = []
prange = np.arange(0.002, 0.015, 0.001)
n_lambdas = None # using it as flag later.
for p_v in prange:
# Process whole dataset for value of p_v.
concate_rttm_file = diarize_dataset(full_meta, split_type, n_lambdas,
p_v)
ref_rttm = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm = concate_rttm_file
[MS, FA, SER, DER_] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
)
DER_list.append(DER_)
if params["oracle_n_spkrs"] is True and params["backend"] == "kmeans":
# no need of p_val search. Note p_val is needed for SC for both oracle and est num of speakers.
# p_val is needed in oracle_n_spkr=False when using kmeans backend.
break
# Take p_val that gave minmum DER on Dev dataset.
tuned_p_val = prange[DER_list.index(min(DER_list))]
return tuned_p_val
out_boundaries = diarize_dataset(
full_csv,
"dev",
n_lambdas=n_lambdas,
pval=best_pval,
n_neighbors=best_nn,
)
# Evaluating on DEV set
logger.info("Evaluating for AMI Dev. set")
ref_rttm_dev = os.path.join(params["ref_rttm_dir"], "fullref_ami_dev.rttm")
sys_rttm_dev = out_boundaries
[MS_dev, FA_dev, SER_dev, DER_dev] = DER(
ref_rttm_dev,
sys_rttm_dev,
params["ignore_overlap"],
params["forgiveness_collar"],
individual_file_scores=True,
)
msg = "AMI Dev set: Diarization Error Rate = %s %%\n" % (str(
round(DER_dev[-1], 2)))
logger.info(msg)
# AMI Eval Set
full_csv = []
with open(params["csv_diary_eval"], "r") as csv_file:
reader = csv.reader(csv_file, delimiter=",")
for row in reader:
full_csv.append(row)
out_boundaries = diarize_dataset(
split_type,
n_lambdas=n_lambdas,
pval=best_pval,
n_neighbors=best_nn,
)
# Computing DER.
msg = "Computing DERs for " + split_type + " set"
logger.info(msg)
ref_rttm = os.path.join(params["ref_rttm_dir"],
"fullref_ami_" + split_type + ".rttm")
sys_rttm = out_boundaries
[MS, FA, SER, DER_vals] = DER(
ref_rttm,
sys_rttm,
params["ignore_overlap"],
params["forgiveness_collar"],
individual_file_scores=True,
)
# Writing DER values to a file. Append tag.
der_file_name = split_type + "_DER_" + tag
out_der_file = os.path.join(params["der_dir"], der_file_name)
msg = "Writing DER file to: " + out_der_file
logger.info(msg)
diar.write_ders_file(ref_rttm, DER_vals, out_der_file)
msg = ("AMI " + split_type + " set DER = %s %%\n" %
(str(round(DER_vals[-1], 2))))
logger.info(msg)
final_DERs[split_type] = round(DER_vals[-1], 2)
