-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdata_classify.py
182 lines (156 loc) · 7.65 KB
/
data_classify.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
import os
import sys
import pickle
import itertools
import pandas as pd
import numpy as np
import graphviz
import settings as s
from sklearn import svm, tree
from sklearn.utils import class_weight, shuffle
from sklearn.metrics import confusion_matrix, classification_report
from sklearn.preprocessing import OneHotEncoder
from sklearn.model_selection import GridSearchCV
import keras
from keras import layers
import matplotlib.pyplot as plt
MINIMUM_CONF = -1000000.0
OVERSAMPLING = True
TEST_SPLIT = 0.8
def plot_history(history):
plt.plot(history.history['categorical_accuracy'])
plt.plot(history.history['val_categorical_accuracy'])
plt.title('Model accuracy')
plt.ylabel('Accuracy')
plt.xlabel('Epoch')
plt.legend(['Train', 'Test'], loc='upper left')
plt.show()
def get_length(path):
path = os.path.join(s.OUTPUT_DIR, "autovot_files", path)
with open(path) as f:
length = int(f.readline().split(" ")[0])
return length
def get_values(path):
path = os.path.join(s.OUTPUT_DIR, "autovot_files", path)
matrix = np.nan_to_num(np.genfromtxt(path, skip_header=1, filling_values=0))
matrix = np.concatenate([np.mean(x,axis=0) for x in np.array_split(matrix, 20)])
return matrix
def evaluate_model(truth, predictions, classes):
conf_mat = confusion_matrix(truth, predictions)
plt.imshow(conf_mat, interpolation='nearest', cmap=plt.cm.Blues)
plt.title(f"Confusion matrix for /{stop}/")
plt.colorbar()
tick_marks = np.arange(len(classes))
plt.xticks(tick_marks, classes, rotation=45)
plt.yticks(tick_marks, classes)
thresh = conf_mat.max() / 2.
for i, j in itertools.product(range(conf_mat.shape[0]), range(conf_mat.shape[1])):
plt.text(j, i, "{0:d}".format(conf_mat[i, j]),
horizontalalignment="center",
color="white" if conf_mat[i, j] > thresh else "black")
plt.ylabel('True label')
plt.xlabel('Predicted label')
plt.tight_layout()
plt.show()
prediction_path = os.path.join(s.OUTPUT_DIR, "real_pred")
feature_slice = [f"feat-{x}" for x in range(63*20)]
z_scores = {}
if not os.path.isfile(os.path.join(s.OUTPUT_DIR, "dataframe.pkl")):
data = pd.read_csv(prediction_path, sep=" ", names=["phoneme", "allophone", "word_pos", "label", "id", "conf", "begin", "end"])
data["path"] = data['phoneme'].astype(str)+"-"+data['allophone']+"-"+data['word_pos']+"-"+data['label']+"-"+data['id' ].astype(str)
data["VOT"] = data["end"] - data["begin"]
temp = list(zip(*data["path"].map(get_values)))
for i, c in enumerate(feature_slice):
data[c] = temp[i]
z_scores[c] = (data[c].mean(), data[c].std(ddof=0))
data[c] =(data[c] - data[c].mean())/data[c].std(ddof=0)
data["length"] = data["path"].apply(get_length)
data.to_pickle(os.path.join(s.OUTPUT_DIR, "dataframe.pkl"))
with open(os.path.join(s.OUTPUT_DIR, "z_scores.pkl"), 'wb') as f:
pickle.dump(z_scores, f)
else:
data = pd.read_pickle(os.path.join(s.OUTPUT_DIR, "dataframe.pkl"))
with open(os.path.join(s.OUTPUT_DIR, "z_scores.pkl"), 'rb') as f:
z_scores = pickle.load(f)
sotc_prediction_path = os.path.join(s.OUTPUT_DIR, "real_pred_sotc")
if not os.path.isfile(os.path.join(s.OUTPUT_DIR, "sotc_dataframe.pkl")):
sotc_data = pd.read_csv(sotc_prediction_path, sep=" ", names=["phoneme", "measurable", "id", "conf", "begin", "end"])
sotc_data["path"] = sotc_data['phoneme'].astype(str)+"-"+sotc_data['measurable'].astype(str)+"-"+sotc_data['id'].astype(str)
sotc_data["VOT"] = sotc_data["end"] - sotc_data["begin"]
temp = list(zip(*sotc_data["path"].map(get_values)))
for i, c in enumerate(feature_slice):
sotc_data[c] = temp[i]
sotc_datac[c] = (sotc_data[c] - z_scores[c][0])/z_scores[c][1]
sotc_data["length"] = sotc_data["path"].apply(get_length)
sotc_data.to_pickle(os.path.join(s.OUTPUT_DIR, "sotc_dataframe.pkl"))
else:
sotc_data = pd.read_pickle(os.path.join(s.OUTPUT_DIR, "sotc_dataframe.pkl"))
for col in ["conf", "VOT", "length"]:
data.loc[:, col] = (data[col] - data[col].mean())/data[col].std(ddof=0)
sotc_data.loc[:, col] = (sotc_data[col] - data[col].mean())/data[col].std(ddof=0)
#data.loc[:, "allophone"] = data["allophone"].replace([f"{stop}rl" for stop in s.STOPS]+s.STOPS, "yes")
#data.loc[:, "allophone"] = data["allophone"].replace(["dx", "q"]+[f"{stop}cl" for stop in s.STOPS], "no")
data.loc[:, "allophone"] = data["allophone"].replace([f"{stop}rl" for stop in s.STOPS]+s.STOPS+["dx", "q"]+[f"{stop}cl" for stop in s.STOPS], "yes")
for stop in s.STOPS:
stop_data = data[(data["phoneme"] == stop)]
#stop_data = data[(data["allophone"] == "yes") | (data["allophone"] == "no")]
o_data = data[(data["phoneme"] == "vowel") | (data["phoneme"] == "nasal") | (data["phoneme"] == "fric")]
o_data.loc[:, "allophone"] = "no"
stop_data = stop_data.append(o_data[np.random.rand(len(o_data)) < len(stop_data)/2.5/len(o_data)])
mask = np.random.rand(len(stop_data)) < 0.8
y, classes = pd.factorize(stop_data.loc[:, "allophone"])
X = stop_data[feature_slice]
X = X.values.reshape((-1, 20, 63))
X = np.append(X, np.repeat(stop_data[["conf", "VOT", "length"]].values.reshape(-1, 1, 3), 20, axis=1), axis=2)
train_y = y[mask]
test_y = y[~mask]
train_X = X[mask]
test_X = X[~mask]
if stop == "t":
n_data = data[(data["phoneme"] == "t+")]
n_y = n_data.loc[:, "allophone"].apply(lambda x: np.where(classes==x)[0][0])
n_X = n_data[feature_slice]
n_X = n_X.values.reshape((-1, 20, 63))
n_X = np.append(n_X, np.repeat(n_data[["conf", "VOT", "length"]].values.reshape(-1, 1, 3), 20, axis=1), axis=2)
train_y = np.concatenate((train_y, n_y))
train_X = np.vstack((train_X, n_X))
train_X, train_y = shuffle(train_X, train_y, random_state=0)
#Resample
if len(classes) == 2:
uniques, counts = np.unique(train_y, return_counts=True)
print(counts)
if len(classes) != 2:
train_y = keras.utils.to_categorical(train_y)
model = keras.Sequential()
model.add(layers.Conv1D(64, 3, input_shape=(20, 66), activation="relu"))
model.add(layers.Conv1D(64, 3, activation="relu"))
model.add(layers.MaxPooling1D(3))
model.add(layers.Conv1D(128, 3, activation="relu"))
model.add(layers.GlobalAveragePooling1D())
model.add(layers.Dropout(rate=0.5))
adam = keras.optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=True)
if len(classes) == 2:
model.add(layers.Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer=adam, metrics=["accuracy"])
else:
model.add(layers.Dense(len(classes), activation="softmax"))
model.compile(loss='categorical_crossentropy', optimizer=adam, metrics=["categorical_accuracy"])
history = model.fit(train_X, train_y, epochs=10, batch_size=64, validation_split=0.05)
if len(classes) == 2:
pred = model.predict(test_X)
else:
pred = np.argmax(model.predict(test_X), axis=1)
sotc_stop_data = sotc_data[sotc_data["phoneme"] == stop]
sotc_X = sotc_stop_data[feature_slice]
sotc_X = sotc_X.values.reshape((-1, 20, 63))
sotc_X = np.append(sotc_X, np.repeat(sotc_stop_data[["conf", "VOT", "length"]].values.reshape(-1, 1, 3), 20, axis=1), axis=2)
sotc_y = sotc_stop_data.loc[:, "measurable"].apply(lambda x: np.where(classes==x)[0][0])
pred = model.predict(sotc_X)
for x in [0.25, 0.5, 0.75, 0.8, 0.85, 0.9, 0.95]:
print(x)
pred_x = np.copy(pred)
pred_x[pred_x > x] = 1
pred_x[pred_x <= x] = 0
print(confusion_matrix(sotc_y, pred_x))
#plot_history(history)
#evaluate_model(test_y, pred, classes)