"""
This is experiment 5
bert as a service, concat, scikit mlp
"""
import numpy as np
import sklearn.neural_network as nn
from scipy.stats.stats import pearsonr
from utils.resourceManager import getEmbeddedResource
print("Getting data...")
data = getEmbeddedResource("exp5", "BertAsService", "zh", "train", subname="en-ch", MultiServerBert=("./bert/uncased_L-12_H-768_A-12","./bert/chinese_L-12_H-768_A-12"))
val_data = getEmbeddedResource("exp5", "BertAsService", "zh", "dev", subname="en-ch", MultiServerBert=("./bert/uncased_L-12_H-768_A-12","./bert/chinese_L-12_H-768_A-12"))
test_data = getEmbeddedResource("exp5", "BertAsService", "zh", "test", subname="en-ch", MultiServerBert=("./bert/uncased_L-12_H-768_A-12","./bert/chinese_L-12_H-768_A-12"))
print("Tokenized data")
data.extend(val_data)
es,cs,y =[],[],[]
for e, c, y_ in data:
es.append(e)
cs.append(c)
y.append(y_)
x = np.concatenate((es, cs), axis=1)
y = np.asarray(y)
es,cs=[],[]
for e, c in test_data:
es.append(e)
"""
This is experiment 5
bert as a service, concat, scikit mlp
"""
import numpy as np
import sklearn.neural_network as nn
from scipy.stats.stats import pearsonr
from utils.resourceManager import getEmbeddedResource
print("Getting data...")
data = getEmbeddedResource("exp5", "BertAsService", "de", "train")
val_data = getEmbeddedResource("exp5", "BertAsService", "de", "dev")
print("Tokenized data")
es, cs, y = [], [], []
for e, c, y_ in data:
es.append(e)
cs.append(c)
y.append(y_)
x = np.concatenate((es, cs), axis=1)
y = np.asarray(y)
es, cs, val_y = [], [], []
for e, c, y_ in val_data:
es.append(e)
cs.append(c)
val_y.append(y_)
from scipy.stats.stats import pearsonr
from utils.resourceManager import getEmbeddedResource
def getAverage(data):
data_averaged = []
for e, c, l in data:
e_avg = np.mean(e, axis=0)
c_avg = np.mean(c, axis=0)
data_averaged.append((np.array(list(e_avg) + list(c_avg)), l))
return data_averaged
print("Getting data...")
data = getEmbeddedResource("exp2", "FastText", "zh", "train")
val_data = getEmbeddedResource("exp2", "FastText", "zh", "dev")
print("Tokenized data")
train = getAverage(data)
model = nn.MLPRegressor(max_iter=4, hidden_layer_sizes=(5,), verbose=True)
X = [x[0] for x in train]
y = [x[1] for x in train]
model.fit(X, y)
test = getAverage(val_data)
X_test = [x[0] for x in test]
y_test = [x[1] for x in test]
my_y = model.predict(X_test)
print("PEARSON:", pearsonr(y_test, my_y))
This is experiment 5
bert as a service, concat, scikit mlp
"""
import numpy as np
import sklearn.neural_network as nn
from scipy.stats.stats import pearsonr
from utils.resourceManager import getEmbeddedResource
print("Getting data...")
data = getEmbeddedResource("exp5",
"BertAsService",
"de",
"train",
subname="uncased",
model_dir="./bert/uncased_L-12_H-768_A-12")
val_data = getEmbeddedResource("exp5",
"BertAsService",
"de",
"dev",
subname="uncased",
model_dir="./bert/uncased_L-12_H-768_A-12")
print("Tokenized data")
es, cs, y = [], [], []
for e, c, y_ in data:
es.append(e)
cs.append(c)
y.append(y_)