def main():
model, encoder_model, decoder_model = create_models(300, 512, 300)
model.load_weights(sys.argv[1])
ft_en = FastText('embeddings/wiki.en.bin')
ft_tl = FastText('embeddings/wiki.tl.bin')
start_seq = ft_en.get_numpy_vector(SOS, normalized=True).reshape(1, 1, -1)
chars = '.,?!()'
while True:
input_sentence = input('Input Tagalog: ').lower() #'kamusta ka ?'
for c in chars:
input_sentence = input_sentence.replace(c, ' ' + c + ' ')
print('Embedding...')
input_seq = input_sentence.lower().split()
aaa = np.zeros((1, 15, 300), dtype='float32')
for i, w in enumerate(input_seq):
aaa[0, i] = ft_tl.get_numpy_vector(w, normalized=True)
#input_seq = [ft_tl.get_numpy_vector(i, normalized=True) for i in input_seq]
#input_seq = np.stack(input_seq).reshape(1, -1, 300)
input_seq = aaa
print(input_seq)
print('Translating...')
decoded_sentence = decode_sequence(input_seq, encoder_model,
decoder_model, ft_en, start_seq)
print('-')
print('Input sentence:', input_sentence)
print('Decoded sentence:', decoded_sentence)
def main():
model = FastText('model_text8.bin')
target_words = [
'granada', 'python', 'harmony', 'mafia', 'yoga', 'goth', 'cyberpunk',
'nasa', 'japan', 'boolean', 'foodball', 'algorithm', 'china', 'usa',
'internet', 'harvard', 'earth', 'horse', 'angel', 'rock'
]
for t_word in target_words:
# get embedding
target_word_embedding = model.get_numpy_vector(t_word)
print('Target word:', t_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(t_word, k=15)
# init array
nn_word_embedding = np.zeros(shape=(15, 128))
i = 0
for word, similarity in closest_words:
# get each word embedding
nn_word_embedding[i] = model.get_numpy_vector(word)
#print('Word:', word, 'Vec:', nn_word_embedding[i])
i = i + 1
# kmeans
#print(nn_word_embedding.shape)
#print(closest_words)
cluster_model = KMeans(n_clusters=3, init='k-means++')
prediction = cluster_model.fit_predict(nn_word_embedding)
print(prediction)
j = 0
for word in closest_words:
print('Word:', word[0], '- Cluster #%d' % (prediction[j] + 1))
j = j + 1
class FeatureGenerator:
def __init__(self, fastext_path):
self.fasttext = FastText(fastext_path)
def generate_record(self, tuple):
tr = self.fasttext.get_numpy_vector(tuple[0])
si = self.fasttext.get_numpy_vector(tuple[1])
lm = self.fasttext.get_numpy_vector(tuple[2])
#return numpy.concatenate((tr, lm))
#return numpy.concatenate((tr, si, lm))
return numpy.concatenate((tr, si, lm, lm - tr))
#return numpy.concatenate((si, lm - tr, tr - lm))
def generate(self, values):
return numpy.array([self.generate_record(value) for value in values])
def use_pyfasttext_model():
# OK
# 训练模型可以使用fasttext命令行工具进行(../doc/fastText_train.png)，也可以使用本文件使用的pyfasttext包训练。
"""
# OK: 1. pyfasttext包训练的模型的导入
model = FastText("../data/lxw_model_sg_pyfasttext.bin")
print(model["先生"]) # type(model["先生"]): <class 'array.array'>
print(model.get_numpy_vector("先生")) # type: <class 'numpy.ndarray'>
print(model["刘晓伟"]) # OOV
print(model.get_numpy_vector("刘晓伟"))
print(model["陈贺"]) # OOV
print(model.get_numpy_vector("陈贺"))
model = FastText("../data/lxw_model_cbow_pyfasttext.bin")
print(model["先生"])
print(model.get_numpy_vector("先生")) # type: <class 'numpy.ndarray'>
print(model["刘晓伟"]) # OOV
print(model.get_numpy_vector("刘晓伟"))
print(model["陈贺"]) # OOV
print(model.get_numpy_vector("陈贺"))
# NOTE: 简单的测试发现, 两个不同的模型针对同一个OOV计算得到的向量是一样的(与fasttext包的情况相同，详情可参见NO_2_use_fasttext_model), 非OOV的向量是不一样的。
"""
# OK: 2. fasttext命令行工具训练出来的模型的导入
model = FastText("../data/880w_fasttext_skip_gram.bin")
print(model["先生"]) # type(model["先生"]): <class 'array.array'>
print(model.get_numpy_vector("先生"))
# print(model["刘晓伟"]) # OK. OOV
# print(model["陈贺"]) # OK. OOV
# Sentence and text vectors.
sentence_vec = model.get_numpy_sentence_vector("刘晓伟 是 个 好人")
print(sentence_vec)
"""
def make_embedding_matrix(word_index, fname):
model = FastText(os.path.join('embeddings', fname))
embedding_matrix = np.zeros((len(word_index) + 1, EMBEDDING_DIM),
dtype='float32')
for word, i in word_index.items():
embedding_matrix[i] = model.get_numpy_vector(word, normalized=True)
return embedding_matrix
class FastTextEmbedding(Embedding):
def __init__(self, binfile, normalize = False):
self.file = binfile
self.vdim = -1
self.normalize = normalize
def load(self):
print('Loading fasttext model.')
self.ftmodel = FastText()
self.ftmodel.load_model(self.file)
self.vdim = len(self.ftmodel['is'])
print('Finished loading fasttext model.')
return self
def getVector(self, word):
return self.ftmodel.get_numpy_vector(word, normalized = self.normalize)
def search(self, q, topk = 4):
raise NotImplementedError()
def wordForVec(self, v):
word, sim = self.ftmodel.words_for_vector(v)[0]
return word, sim
def containsWord(self, word):
return True
def vocabulary(self):
return self.ftmodel.words
def dim(self):
return self.vdim
class FastTextEmbeddings:
def __init__(self, path):
self.fasttext = FastText(path)
def generate(self, sentence):
return [self.fasttext.get_numpy_vector(word) for word in sentence]
def size(self):
return 300
def build_w2v(relevant_tokens, model_file='wiki.cy.bin'):
# using this library because it's more memory friendly for python :)
from pyfasttext import FastText
model = FastText(model_file)
w2v = {}
for token in relevant_tokens:
vec = model.get_numpy_vector(token)
w2v[token] = vec
return w2v
def main():
model = FastText('model_text8.bin')
target_word = 'dog'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
print('Embedding shape:', target_word_embedding.shape)
print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
def __init__(self, doc_catgy, n_vocab, emb_dim, out_channels, filter_size, word2index, pre_trained_embedding, multi_label):
self.in_channels = 1
self.out_channels = out_channels
self.row_dim = emb_dim
self.hidden_dim = 512 ## fixed
self.doc_catgy = doc_catgy
self.n_classes = len(doc_catgy)
self.n_vocab = n_vocab
self.filter_size = filter_size
self.word2index = word2index
self.mutli_label = multi_label
self.le = None
if self.mutli_label == 1:
self.le = MultiLabelBinarizer(classes=[i[0] for i in sorted(self.doc_catgy.items(), key=lambda x: x[1])],sparse_output=False)
elif self.mutli_label == 0:
self.le = LabelEncoder()
self.le.fit([i[0] for i in sorted(self.doc_catgy.items(), key=lambda x: x[1])])
self.look_up_table = None
self.pre_trained_embedding = pre_trained_embedding
super(XMLCnn, self).__init__()
self.to_gpu()
if not self.pre_trained_embedding is None:
model = FastText(self.pre_trained_embedding)
dim = len(model['a'])
n_vocab = len(self.word2index.keys())
self.look_up_table = self.xp.zeros((n_vocab, dim),dtype=np.float32)
for word,index in tqdm(self.word2index.items()):
try:
self.look_up_table[index] = chainer.cuda.to_gpu(model.get_numpy_vector(word))
except:
self.xp.random.seed(index)
self.look_up_table[index][:] = self.xp.random.uniform(-0.25, 0.25, dim)
self.set_seed_random(123)
with self.init_scope():
if self.look_up_table is None:
self.embedding=L.EmbedID(n_vocab, self.row_dim, ignore_label=-1,initialW=linear_init)
else:
self.embedding=L.EmbedID(n_vocab, self.row_dim, ignore_label=-1,initialW=self.look_up_table)
self.conv1 = L.Convolution2D(self.in_channels,self.out_channels,(filter_size[0],self.row_dim), stride=2,initialW=linear_init)
self.conv2 = L.Convolution2D(self.in_channels,self.out_channels,(filter_size[1],self.row_dim), stride=2,initialW=linear_init)
self.conv3 = L.Convolution2D(self.in_channels,self.out_channels,(filter_size[2],self.row_dim), stride=2,initialW=linear_init)
self.l1=L.Linear(in_size = None, out_size = self.hidden_dim, initialW=linear_init)
self.l2=L.Linear(in_size = self.hidden_dim, out_size = self.n_classes,initialW=linear_init)
self.to_gpu()
class SVM:
def __init__(
self,
TFIDF): # TODO: plz give me the model or using global singleton
self.model = FastText('/home/yee0/Atos/wiki.en.bin')
self.TFIDF = TFIDF
self.tfidf_dict, self.words = TFIDF.get_tfidf()
def set_tfidf(self, TFIDF):
self.TFIDF = TFIDF
self.tfidf_dict, self.words = TFIDF.get_tfidf()
def to_feature(self, X):
return [self.vec(self.TFIDF.predict_field(post), post) for post in X]
def vec(self, field_index, post):
v = np.zeros(300)
post = set(post) # make unique
for pl in post:
if pl != '' and pl in self.words:
v += self.model.get_numpy_vector(
pl) * self.tfidf_dict[field_index][pl]
return v
def train(self, X, y):
# 建立 SVC 模型
self.svc = svm.SVC()
svc_fit = self.svc.fit(self.to_feature(X), y)
def predict(self, post):
return self.svc.predict(self.to_feature([post]))
def save_model(self):
with open('steevebot/save/svm.pickle', 'wb') as f:
pickle.dump(self.svc, f)
def restore_model(self):
with open('steevebot/save/svm.pickle', 'rb') as f:
self.svc = pickle.load(f)
# In[ ]:
# In[ ]:
class FastTextWrapper(EmbeddingWrapper):
"""
Contains the KeyedVectors object, name of the file from which the embeddings were loaded as well as its md5.
"""
def __init__(self, fasttext_path):
super(FastTextWrapper, self).__init__(fasttext_path)
self._fasttext = FastText(fasttext_path)
self._word_output_len = self.get_numpy_vector("checklen").shape[0]
def __len__(self):
return self._word_output_len
def get_numpy_vector(self, word):
return self._fasttext.get_numpy_vector(word)
def emb_obj(self):
"""
Returns FastText object
"""
return self._fasttext
class FastTextEmbedding(Embedding):
def __init__(self, binfile, normalize=False):
self.file = binfile
self.vdim = -1
self.normalize = normalize
def load(self):
print('Loading fasttext model.')
self.ftmodel = FastText()
self.ftmodel.load_model(self.file)
self.vdim = len(self.ftmodel['is'])
print('Finished loading fasttext model.')
return self
def getVector(self, word):
return self.ftmodel.get_numpy_vector(word, normalized=self.normalize)
def wordForVec(self, v):
word, sim = self.ftmodel.words_for_vector(v)[0]
return word, sim
def nearest_neighbors(self, word, n=200):
tuples = ftmodel.nearest_neighbors(word, n)
return tuples
def nearest_neighbors_by_vector(self, word, n=200):
tuples = self.ftmodel.words_for_vector(v, n)
return tuples
def containsWord(self, word, explicit=False):
if explicit:
return word in vocabulary()
return True
def vocabulary(self):
return self.ftmodel.words
def dim(self):
return self.vdim
def main():
texts_tl, texts_en = data.parse_corpora('corpus')
word_index_tl, word_index_en, encoder_input_data, decoder_input_data, decoder_target_data = data.preprocess(
texts_en, texts_tl)
embedding_dim = 300
latent_dim = 512
model, encoder_model, decoder_model = create_models(
embedding_dim, latent_dim, embedding_dim)
model.load_weights(sys.argv[1])
indexes = np.random.randint(0, len(texts_tl), 100)
ft_model = FastText(os.path.join('embeddings', 'wiki.en.bin'))
start_seq = ft_model.get_numpy_vector(data.SOS,
normalized=True).reshape(1, 1, -1)
embedding_weights = np.load('embedding-weights.npz')
e_tl = embedding_weights['tl'].astype('float32')
for seq_index in indexes:
# Take one sequence (part of the training set)
# for trying out decoding.
sentence = texts_tl[seq_index]
input_seq = encoder_input_data[seq_index]
input_seq = np.take(e_tl, input_seq, axis=0).reshape(1, -1, 300)
print(input_seq)
#input_seq = sentence.split()[1:-1]
#print(input_seq)
#input_seq = np.stack(list(ft_model.get_numpy_vector(i, normalized=True) for i in input_seq)).reshape(1, -1, 300)
#input_seq = np.stack(list(map(ft_model.get_numpy_vector, input_seq))).reshape(1, -1, 300)
#print(input_seq)
#print(input_seq.shape)
decoded_sentence = decode_sequence(input_seq, encoder_model,
decoder_model, ft_model, start_seq)
print('-')
print('Input sentence:', texts_tl[seq_index])
print('Decoded sentence:', decoded_sentence)
from pyfasttext import FastText
ft = FastText('model.bin')
print ft.get_numpy_vector(u'you')
def main():
model = FastText('model_text8.bin')
target_word = 'deep'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
print('Embedding shape:', target_word_embedding.shape)
print('Embedding:', target_word_embedding[0:15], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'president'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'self'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'insult'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'general'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'inclined'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'property'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'international'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'many'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'imprisoned'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'branches'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'communist'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'france'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'strict'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'earthly'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
terget_word = "zero"
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'feminism'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'ideas'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'theory'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
target_word = 'writings'
# get embedding
target_word_embedding = model.get_numpy_vector(target_word)
print('Target word:', target_word)
#print('Embedding shape:', target_word_embedding.shape)
#print('Embedding:', target_word_embedding[0:10], '...')
# find closest words
closest_words = model.nearest_neighbors(target_word, k=15)
closest_word_embeddings = []
numw = 0
for word, similarity in closest_words:
print('Word:', word, 'similarity:', similarity)
closest_word_embeddings.append(model.get_numpy_vector(word))
kmeans = cluster.KMeans(n_clusters=3)
kmeans.fit(closest_word_embeddings)
labels = kmeans.labels_
print ('Cluster id labels for inputted data')
print (labels)
cluster1 = []
cluster2 = []
cluster3 = []
for i in range(0,15):
if labels[i] == 0:
cluster1.append(closest_words[i][0])
if labels[i] == 1:
cluster2.append(closest_words[i][0])
if labels[i] == 2:
cluster3.append(closest_words[i][0])
print("cluster #1 : ", cluster1)
print("cluster #2 : ", cluster2)
print("cluster #3 : ", cluster3)
class NB_Implement():
def __init__(self):
start_time = time.time()
# self.model = FastText("../data/input/models/sg_pyfasttext.bin") # DEBUG
self.model = FastText(
"../data/input/models/880w_fasttext_skip_gram.bin")
end_time = time.time()
print(f"Loading word vector model cost: {end_time - start_time:.2f}s")
# self.vocab_size, self.vector_size = self.model.numpy_normalized_vectors.shape # OK
self.vocab_size = self.model.nwords
self.vector_size = self.model.args.get("dim")
print(
f"self.vector_size:{self.vector_size}, self.vocab_size: {self.vocab_size}"
) # self.vector_size:200, self.vocab_size: 925242
# 句子的表示形式: {"avg": 向量和的平均, "fasttext": get_numpy_sentence_vector, "matrix": matrix}
self.sentence_vec_type = "avg"
def set_sent_vec_type(self, sentence_vec_type):
assert self.sentence_vec_type in [
"avg", "matrix", "fasttext"
], "self.sentence_vec_type must be in ['avg', 'fasttext', 'matrix']"
self.sentence_vec_type = sentence_vec_type
def gen_sentence_vec(self, sentence):
"""
:param sentence:
:return:
"""
sentence = sentence.strip()
if self.sentence_vec_type == "fasttext":
return self.model.get_numpy_sentence_vector(sentence)
word_list = [word for word in sentence.split(" ")]
word_len = len(word_list)
if self.sentence_vec_type == "matrix":
sentence_matrix = np.empty(word_len, dtype=list)
for idx, word in enumerate(word_list):
sentence_matrix[idx] = self.model.get_numpy_vector(word)
return sentence_matrix
else: # self.sentence_vec_type == "avg":
sentence_vector = np.zeros(self.vector_size) # <ndarray>
# print(f"type(sentence_vector): {type(sentence_vector)}")
for idx, word in enumerate(word_list):
# print(f"type(self.model.get_numpy_vector(word)): {type(self.model.get_numpy_vector(word))}") # <ndarry>
sentence_vector += self.model.get_numpy_vector(word)
return sentence_vector / len(word_list)
def gen_train_val_data(self):
"""
构造训练, 验证数据
"""
X_train = list()
y_train = list()
for line in open("../data/input/training_set.txt"):
line = line.strip().split("\t")
sent_vector = self.gen_sentence_vec(line[-1])
X_train.append(sent_vector)
y_train.append(int(line[0]))
X_val = list()
y_val = list()
for line in open("../data/input/validation_set.txt"):
line = line.strip().split("\t")
sent_vector = self.gen_sentence_vec(line[-1])
X_val.append(sent_vector)
y_val.append(int(line[0]))
return np.array(X_train), np.array(y_train), np.array(X_val), np.array(
y_val),
def train_bayes(self, X_train, y_train):
"""
基于Naive Bayes的分类算法
"""
from sklearn.naive_bayes import GaussianNB
model = GaussianNB()
model.fit(X_train, y_train)
joblib.dump(model, "../data/output/models/bayes_model")
def evaluate_bayes(self, model_path, X_val, y_val):
"""
基于Naive Bayes分类器的预测
"""
model = joblib.load(model_path)
y_val = list(y_val)
correct = 0
"""
y_predict = list()
for sent_vec in X_val: # sent_vec.shape: (self.vector_size,)
predicted = model.predict(sent_vec.reshape(1, -1)) # sent_vec.reshape(1, -1).shape: (1, self.vector_size)
y_predict.append(predicted[0])
"""
y_predict = model.predict(X_val)
print(f"len(y_predict): {len(y_predict)}, len(y_val): {len(y_val)}")
assert len(y_predict) == len(
y_val
), "Unexpected Error: len(y_predict) != len(y_val), but it should be"
for idx in range(len(y_predict)):
if int(y_predict[idx]) == int(y_val[idx]):
correct += 1
score = correct / len(y_predict)
print(f"Bayes Classification Accuray:{score}")
return score
def predict_bayes(self, model_path):
"""
实际应用测试
"""
model = joblib.load(model_path)
sentence = "这件 衣服 真的 太 好看 了 ！ 好想 买 啊 "
sent_vec = np.array(self.gen_sentence_vec(sentence)).reshape(1, -1)
print(f"'{sentence}': {model.predict(sent_vec)}") # 1: 负向
sentence = "这个 电视 真 尼玛 垃圾 ， 老子 再也 不买 了"
sent_vec = np.array(self.gen_sentence_vec(sentence)).reshape(1, -1)
print(f"'{sentence}': {model.predict(sent_vec)}") # 1: 负向
# Update states
states_value = [h, c]
return decoded_sentence
#indexes = np.random.randint(0, len(input_texts), 100)
indexes = np.random.randint(0, len(valid_input_texts), 100)
for seq_index in indexes:
# Take one sequence (part of the training set)
# for trying out decoding.
#text = input_texts[seq_index]
text = valid_input_texts[seq_index]
words = text.split()
encoder_input_data = np.zeros((1, max_encoder_seq_length, embedding_dims),
dtype='float32')
for t, word in enumerate(words):
encoder_input_data[0,
t, :] = filModel.get_numpy_vector(word,
normalized=True)
#print("decodeding",word)
input_seq = encoder_input_data
decoded_sentence = decode_sequence(input_seq)
print('-')
print('Input sentence:', input_texts[seq_index])
print('Decoded sentence:', decoded_sentence)
# Update the target sequence (of length 1).
target_seq = np.zeros((1, 1, embedding_dims))
target_seq[0, 0, :] = engModel.get_numpy_vector(sampled_word, normalized=True)
# Update states
states_value = [h, c]
return decoded_sentence
#indexes = np.random.randint(0, len(input_texts), 100)
indexes = np.random.randint(0, len(valid_input_texts), 100)
for seq_index in indexes:
# Take one sequence (part of the training set)
# for trying out decoding.
#text = input_texts[seq_index]
text = valid_input_texts[seq_index]
words = text.split()
encoder_input_data = np.zeros((1, max_encoder_seq_length, embedding_dims),dtype='float32')
for t, word in enumerate(words):
encoder_input_data[0,t,:] = filModel.get_numpy_vector(word, normalized=True)
#print("decodeding",word)
input_seq = encoder_input_data
decoded_sentence = decode_sequence(input_seq)
print('-')
print('Input sentence:', input_texts[seq_index])
print('Decoded sentence:', decoded_sentence)
class Preprocessing:
def __init__(self):
start_time = time.time()
# self.model = FastText("../data/input/models/sg_pyfasttext.bin") # DEBUG
self.model = FastText(
"../data/input/models/880w_fasttext_skip_gram.bin")
end_time = time.time()
print(f"Loading word vector model cost: {end_time - start_time:.2f}s")
# self.vocab_size, self.vector_size = self.model.numpy_normalized_vectors.shape # OK
self.vocab_size = self.model.nwords
self.vector_size = self.model.args.get("dim")
# self.vector_size:200, self.vocab_size: 925242
print(
f"self.vector_size:{self.vector_size}, self.vocab_size: {self.vocab_size}"
)
# 句子的表示形式:
# {"avg": 向量和的平均, "fasttext": get_numpy_sentence_vector, "concatenate": 向量拼接和补齐, "matrix": 矩阵}
self.sentence_vec_type = "matrix"
self.MAX_SENT_LEN = 70 # DEBUG: 超参数. self.get_sent_max_length()
# 对于"concatenate": self.MAX_SENT_LEN = 30, 取其他不同值的结果: 100: 50.22%, 80: 50.23%, 70: 50.33%, 60: 55.92%, 50: 69.11%, 40: 68.91%, 36: 69.34%, 30: 69.22%, 20: 69.17%, 10: 67.07%
# 对于"matrix": self.MAX_SENT_LEN = 70, 取其他不同值的结果: TODO:
@classmethod
def data_analysis(cls):
train_df = pd.read_csv("../data/input/training_set.txt",
sep="\t",
header=None,
names=["label", "sentence"])
val_df = pd.read_csv("../data/input/validation_set.txt",
sep="\t",
header=None,
names=["label", "sentence"])
y_train = train_df["label"]
y_val = val_df["label"]
sns.set(style="white", context="notebook", palette="deep")
# 查看样本数据分布情况(各个label数据是否均匀分布)
sns.countplot(y_train)
plt.show()
sns.countplot(y_val)
plt.show()
print(y_train.value_counts())
print(y_val.value_counts())
def set_sent_vec_type(self, sentence_vec_type):
assert sentence_vec_type in ["avg", "concatenate", "fasttext", "matrix"], \
"sentence_vec_type must be in ['avg', 'fasttext', 'concatenate', 'matrix']"
self.sentence_vec_type = sentence_vec_type
def get_sent_max_length(self): # NOT_USED
sent_len_counter = Counter()
max_length = 0
with open("../data/input/training_set.txt") as f:
for line in f:
content = line.strip().split("\t")[1]
content_list = content.split()
length = len(content_list)
sent_len_counter[length] += 1
if max_length <= length:
max_length = length
sent_len_counter = sorted(list(sent_len_counter.items()),
key=lambda x: x[0])
print(sent_len_counter)
# [(31, 1145), (32, 1105), (33, 1017), (34, 938), (35, 839), (36, 830), (37, 775), (38, 737), (39, 720), (40, 643), (41, 575), (42, 584), (43, 517), (44, 547), (45, 514), (46, 514), (47, 480), (48, 460), (49, 470), (50, 444), (51, 484), (52, 432), (53, 462), (54, 495), (55, 487), (56, 500), (57, 496), (58, 489), (59, 419), (60, 387), (61, 348), (62, 265), (63, 222), (64, 153), (65, 127), (66, 103), (67, 67), (68, 34), (69, 21), (70, 22), (71, 8), (72, 6), (73, 4), (74, 10), (75, 2), (76, 4), (77, 2), (78, 1), (79, 2), (80, 4), (81, 2), (82, 3), (83, 1), (84, 5), (86, 4), (87, 3), (88, 3), (89, 2), (90, 2), (91, 3), (92, 5), (93, 2), (94, 4), (96, 1), (97, 5), (98, 1), (99, 2), (100, 2), (101, 2), (102, 1), (103, 2), (104, 2), (105, 2), (106, 5), (107, 3), (108, 2), (109, 3), (110, 4), (111, 1), (112, 2), (113, 3), (114, 1), (116, 1), (119, 3), (679, 1)]
return max_length
def gen_sentence_vec(self, sentence):
"""
:param sentence:
:return:
"""
sentence = sentence.strip()
if self.sentence_vec_type == "fasttext":
return self.model.get_numpy_sentence_vector(sentence)
word_list = sentence.split(" ")
if self.sentence_vec_type == "concatenate":
sentence_vector = self.model.get_numpy_vector(word_list[0])
for word in word_list[1:]:
sentence_vector = np.hstack(
(sentence_vector, self.model.get_numpy_vector(word)))
return sentence_vector # NOTE: 对于concatenate情况, 每个句子的sentence_vector是不一样长的
if self.sentence_vec_type == "matrix": # for Deep Learning.
sentence_matrix = []
for word in word_list[
-self.
MAX_SENT_LEN:]: # NOTE: 截取后面的应该是要好些(参考https://github.com/lxw0109/SentimentClassification_UMICH_SI650/blob/master/src/LSTM_wo_pretrained_vector.py#L86)
sentence_matrix.append(self.model.get_numpy_vector(word))
length = len(sentence_matrix)
# 一定成立，因为上面做了切片截取
assert length <= self.MAX_SENT_LEN, "CRITICAL ERROR: len(sentence_matrix) > self.MAX_SENT_LEN."
# 参数中的matrix类型为list of ndarray, 返回值的matrix是ndarray of ndarray
sentence_matrix = np.pad(sentence_matrix,
pad_width=((0,
self.MAX_SENT_LEN - length),
(0, 0)),
mode="constant",
constant_values=-1)
return sentence_matrix
else: # self.sentence_vec_type == "avg":
sentence_vector = np.zeros(self.vector_size) # <ndarray>
# print(f"type(sentence_vector): {type(sentence_vector)}")
for idx, word in enumerate(word_list):
# print(f"type(self.model.get_numpy_vector(word)): {type(self.model.get_numpy_vector(word))}") # <ndarray>
sentence_vector += self.model.get_numpy_vector(word)
return sentence_vector / len(word_list)
def gen_train_val_data(self):
# 构造训练数据 & 验证数据
train_df = pd.read_csv("../data/input/training_set.txt",
sep="\t",
header=None,
names=["label", "sentence"])
val_df = pd.read_csv("../data/input/validation_set.txt",
sep="\t",
header=None,
names=["label", "sentence"])
# 打乱训练集的顺序. TODO: 不打乱感觉训练出来的模型是有问题的?(好看那句总是预测结果是1？)
train_df = train_df.sample(frac=1, random_state=1)
# val_df = val_df.sample(frac=1, random_state=1) # 验证集不用打乱
X_train = train_df["sentence"]
X_train_vec = list()
for sentence in X_train:
sent_vector = self.gen_sentence_vec(sentence)
X_train_vec.append(sent_vector)
y_train = train_df["label"] # <Series>
X_val = val_df["sentence"]
X_val_vec = list()
for sentence in X_val:
sent_vector = self.gen_sentence_vec(sentence)
X_val_vec.append(sent_vector)
y_val = val_df["label"] # <Series>
if self.sentence_vec_type == "concatenate":
# NOTE: 注意，这里的dtype是必须的，否则dtype默认值是"int32", 词向量所有的数值会被全部转换为0
X_train_vec = sequence.pad_sequences(X_train_vec,
maxlen=self.MAX_SENT_LEN *
self.vector_size,
value=0,
dtype=np.float)
X_val_vec = sequence.pad_sequences(X_val_vec,
maxlen=self.MAX_SENT_LEN *
self.vector_size,
value=0,
dtype=np.float)
return np.array(X_train_vec), np.array(X_val_vec), np.array(
y_train), np.array(y_val)
def main():
#enc_word2vec = FastText('wiki.tl/wiki.tl.bin')
#dec_word2vec = FastText('wiki.en/wiki.en.bin')
dec_word2vec = FastText('wiki.en/wiki.en.bin')
enc_word2vec = FastText('wiki.tl/wiki.tl.bin')
#data_path = 'tgl-eng/tgl.txt'
#test_path = 'valid_split'
test_path = 'train_split'
#data_path = 'health_shortened.tsv'
eos = "eos"
sos = "sos"
#savemodel_filename = 's2s_fasttextloader_batch64_twodata.h5'
#training parameters
#batch_size = 64 # Batch size for training.
batch_size = 64
epochs = 500 # Number of epochs to train for.
latent_dim = 512 # Latent dimensionality of the encoding space.
word_vec_size = 300
#chkpt_path="checkpoints/weights-improvement-twodata-{epoch:05d}.hdf5"
#checkpoint = ModelCheckpoint(chkpt_path, verbose=1)
# Define the model that will turn
# `encoder_input_data` & `decoder_input_data` into `decoder_target_data`
#so far checkpoints/weights_cosine_proximity_combined2-00063.hdf5 seems to work
weights_path = 'checkpoints/weights_cosine_proximity_combined2-00063.hdf5'
model, encoder_model, decoder_model = build_model(word_vec_size,
latent_dim, weights_path)
model.summary()
# Compile & run training
#model.compile(optimizer='rmsprop', loss='mean_squared_error')
#model.compile(optimizer='rmsprop', loss='cosine_proximity')
#model.compile(optimizer='rmsprop', loss='mean_squared_error')
# Note that `decoder_target_data` needs to be one-hot encoded,
# rather than sequences of integers like `decoder_input_data`!
#num_sentence = 77990
#steps_per_epoch = int(num_sentence//batch_size)
decoder_model.summary()
def decode_sequence(input_seq, sos, eos):
# Encode the input as state vectors.
states_value = encoder_model.predict(input_seq)
# Generate empty target sequence of length 1.
target_seq = np.zeros((1, 1, word_vec_size))
# Populate the first character of target sequence with the start character.
#target_seq[0, 0, target_dict[sos]] = 1.
''' create vector for sos '''
target_seq[0, 0, :] = dec_word2vec.get_numpy_vector(sos,
normalized=True)
# Sampling loop for a batch of sequences
# (to simplify, here we assume a batch of size 1).
stop_condition = False
decoded_sentence = ''
while not stop_condition:
output_tokens, h, c = decoder_model.predict([target_seq] +
states_value)
# Sample a token
#sampled_token_index = np.argmax(output_tokens[0, -1, :])
#sampled_word = target_rev_dict[sampled_token_index]
sampled_word = dec_word2vec.words_for_vector(
output_tokens[0, -1, :])[0][0]
decoded_sentence += sampled_word + " "
# Exit condition: either hit max length
# or find stop character.
# if sampled_word in [".", "?", "!"] or
if (sampled_word == eos
or len(decoded_sentence) > max_decoder_seq_length):
stop_condition = True
if (decoded_sentence.endswith(eos + ' ')):
decoded_sentence = decoded_sentence[:-len(eos + ' ')]
# Update the target sequence (of length 1).
target_seq = np.zeros((1, 1, word_vec_size))
target_seq[0,
0, :] = dec_word2vec.get_numpy_vector(sampled_word,
normalized=True)
# Update states
states_value = [h, c]
return decoded_sentence
input_texts, target_texts = input2target(test_path, sos, eos)
indexes = np.random.randint(0, len(input_texts), 50)
#max_encoder_seq_length = max([len(words.split()) for words in input_texts])
#max_decoder_seq_length = max([len(words.split()) for words in target_texts])
max_encoder_seq_length = 130
max_decoder_seq_length = 100
encoder_input_data = np.zeros(
(len(input_texts), max_encoder_seq_length, word_vec_size),
dtype='float32')
'''
for i, text, in enumerate(input_texts):
words = text.split()
#words.reverse()
for t, word in enumerate(words):
encoder_input_data[i, t, :] = enc_word2vec.get_numpy_vector(word, normalized=True)
'''
while True:
input_sentence = input('Enter Filipino sentence: ')
print('Input:', input_sentence)
input_sentence = input_sentence.replace(",", " ,")
input_sentence = input_sentence.replace(".", " .")
input_sentence = input_sentence.replace("!", " !")
input_sentence = input_sentence.replace("?", " ?")
input_sentence = input_sentence.lower()
input_words = input_sentence.split()
for t, word in enumerate(input_words):
encoder_input_data[0, t, :] = enc_word2vec.get_numpy_vector(
word, normalized=True)
#for seq_index in indexes:
# Take one sequence (part of the training set)
# for trying out decoding.
input_seq = encoder_input_data
decoded_sentence = decode_sequence(input_seq, sos, eos)
print('-')
#print('Input sentence:', input_texts[seq_index])
print('Decoded sentence:', decoded_sentence)
class Embeddings(object):
def __init__(self,
name,
path='./embedding-registry.json',
lang='en',
extension='vec',
use_ELMo=False):
self.name = name
self.embed_size = 0
self.static_embed_size = 0
self.vocab_size = 0
self.model = {}
self.registry = self._load_embedding_registry(path)
self.lang = lang
self.extension = extension
self.embedding_lmdb_path = None
if self.registry is not None:
self.embedding_lmdb_path = self.registry["embedding-lmdb-path"]
self.env = None
self.make_embeddings_simple(name)
self.static_embed_size = self.embed_size
self.bilm = None
# below init for using ELMo embeddings
self.use_ELMo = use_ELMo
if use_ELMo:
self.make_ELMo()
self.embed_size = ELMo_embed_size + self.embed_size
description = self._get_description('elmo-en')
self.env_ELMo = None
if description:
self.embedding_ELMo_cache = os.path.join(
description["path-dump"], "cache")
# clean possible remaining cache
self.clean_ELMo_cache()
# create and load a cache in write mode, it will be used only for training
self.env_ELMo = lmdb.open(self.embedding_ELMo_cache,
map_size=map_size)
def __getattr__(self, name):
return getattr(self.model, name)
def _load_embedding_registry(self, path='./embedding-registry.json'):
"""
Load the description of available embeddings. Each description provides a name,
a file path (used only if necessary) and a embeddings type (to take into account
small variation of format)
"""
registry_json = open(path).read()
return json.loads(registry_json)
def make_embeddings_simple_in_memory(self,
name="fasttext-crawl",
hasHeader=True):
nbWords = 0
print('loading embeddings...')
begin = True
description = self._get_description(name)
if description is not None:
embeddings_path = description["path"]
embeddings_type = description["type"]
self.lang = description["lang"]
print("path:", embeddings_path)
if self.extension == 'bin':
self.model = FastText(embeddings_path)
nbWords = self.model.nwords
self.embed_size = 300
else:
if embeddings_type == "glove":
hasHeader = False
with open(embeddings_path) as f:
for line in f:
line = line.strip()
line = line.split(' ')
if begin:
if hasHeader:
# first line gives the nb of words and the embedding size
nbWords = int(line[0])
self.embed_size = int(line[1].replace(
"\n", ""))
begin = False
continue
else:
begin = False
word = line[0]
#if embeddings_type == 'glove':
vector = np.array(
[float(val) for val in line[1:len(line)]],
dtype='float32')
#else:
# vector = np.array([float(val) for val in line[1:len(line)-1]], dtype='float32')
if self.embed_size == 0:
self.embed_size = len(vector)
self.model[word] = vector
if nbWords == 0:
nbWords = len(self.model)
print('embeddings loaded for', nbWords, "words and",
self.embed_size, "dimensions")
'''
def make_embeddings_fasttext_bin(self, name="wiki.en.bin"):
nbWords = 0
print('loading embeddings...')
description = self._get_description(name)
if description is not None:
embeddings_path = description["path"]
print("path:", embeddings_path)
self.model = load_fasttext_format(embeddings_path)
'''
def make_embeddings_lmdb(self, name="fasttext-crawl", hasHeader=True):
nbWords = 0
print(
'\nCompiling embeddings... (this is done only one time per embeddings at first launch)'
)
begin = True
description = self._get_description(name)
if description is not None:
embeddings_path = description["path"]
embeddings_type = description["type"]
self.lang = description["lang"]
print("path:", embeddings_path)
if embeddings_type == "glove":
hasHeader = False
txn = self.env.begin(write=True)
batch_size = 1024
i = 0
nb_lines = 0
with open(embeddings_path) as f:
for line in f:
nb_lines += 1
with open(embeddings_path) as f:
#for line in f:
for line in tqdm(f, total=nb_lines):
line = line.split(' ')
if begin:
if hasHeader:
# first line gives the nb of words and the embedding size
nbWords = int(line[0])
self.embed_size = int(line[1].replace("\n", ""))
begin = False
continue
else:
begin = False
word = line[0]
#if embeddings_type == 'glove':
vector = np.array(
[float(val) for val in line[1:len(line)]],
dtype='float32')
#else:
# vector = np.array([float(val) for val in line[1:len(line)-1]], dtype='float32')
if self.embed_size == 0:
self.embed_size = len(vector)
if len(word.encode(
encoding='UTF-8')) < self.env.max_key_size():
txn.put(word.encode(encoding='UTF-8'),
_serialize_pickle(vector))
#txn.put(word.encode(encoding='UTF-8'), _serialize_byteio(vector))
i += 1
# commit batch
if i % batch_size == 0:
txn.commit()
txn = self.env.begin(write=True)
#if i % batch_size != 0:
txn.commit()
if nbWords == 0:
nbWords = i
self.vocab_size = nbWords
print('embeddings loaded for', nbWords, "words and",
self.embed_size, "dimensions")
def make_embeddings_simple(self, name="fasttext-crawl", hasHeader=True):
description = self._get_description(name)
if description is not None:
self.extension = description["format"]
if self.embedding_lmdb_path is None or self.embedding_lmdb_path == "None":
print(
"embedding_lmdb_path is not specified in the embeddings registry, so the embeddings will be loaded in memory..."
)
self.make_embeddings_simple_in_memory(name, hasHeader)
elif self.extension == "bin":
print(
"embedding is of format .bin, so it will be loaded in memory..."
)
self.make_embeddings_simple_in_memory(name, hasHeader)
else:
# check if the lmdb database exists
envFilePath = os.path.join(self.embedding_lmdb_path, name)
if os.path.isdir(envFilePath):
description = self._get_description(name)
if description is not None:
self.lang = description["lang"]
# open the database in read mode
self.env = lmdb.open(envFilePath,
readonly=True,
max_readers=2048,
max_spare_txns=4)
# we need to set self.embed_size and self.vocab_size
with self.env.begin() as txn:
stats = txn.stat()
size = stats['entries']
self.vocab_size = size
with self.env.begin() as txn:
cursor = txn.cursor()
for key, value in cursor:
vector = _deserialize_pickle(value)
self.embed_size = vector.shape[0]
break
cursor.close()
# no idea why, but we need to close and reopen the environment to avoid
# mdb_txn_begin: MDB_BAD_RSLOT: Invalid reuse of reader locktable slot
# when opening new transaction !
self.env.close()
self.env = lmdb.open(envFilePath,
readonly=True,
max_readers=2048,
max_spare_txns=2)
else:
# create and load the database in write mode
self.env = lmdb.open(envFilePath, map_size=map_size)
self.make_embeddings_lmdb(name, hasHeader)
def make_ELMo(self):
# Location of pretrained BiLM for the specified language
# TBD check if ELMo language resources are present
description = self._get_description('elmo-en')
if description is not None:
self.lang = description["lang"]
vocab_file = description["path-vocab"]
options_file = description["path-config"]
weight_file = description["path_weights"]
print('init ELMo')
# Create a Batcher to map text to character ids
self.batcher = Batcher(vocab_file, 50)
# Build the biLM graph.
self.bilm = BidirectionalLanguageModel(options_file, weight_file)
# Input placeholders to the biLM.
self.character_ids = tf.placeholder('int32',
shape=(None, None, 50))
self.embeddings_op = self.bilm(self.character_ids)
with tf.variable_scope('', reuse=tf.AUTO_REUSE):
# the reuse=True scope reuses weights from the whole context
self.elmo_input = weight_layers('input',
self.embeddings_op,
l2_coef=0.0)
def dump_ELMo_token_embeddings(self, x_train):
if not self.use_ELMo:
print(
"Warning: ELMo embeddings dump requested but embeddings object wrongly initialised"
)
return
description = self._get_description('elmo-en')
if description is not None:
print("Building ELMo token dump")
self.lang = description["lang"]
options_file = description["path-config"]
weight_file = description["path_weights"]
working_path = description["path-dump"]
all_tokens = set(['<S>', '</S>'])
for i in range(0, len(x_train)):
# as it is training, it is already tokenized
tokens = x_train[i]
for token in tokens:
if token not in all_tokens:
all_tokens.add(token)
vocab_file = os.path.join(working_path, 'vocab_small.txt')
with open(vocab_file, 'w') as fout:
fout.write('\n'.join(all_tokens))
tf.reset_default_graph()
token_embedding_file = os.path.join(working_path,
'elmo_token_embeddings.hdf5')
dump_token_embeddings(vocab_file, options_file, weight_file,
token_embedding_file)
tf.reset_default_graph()
self.batcher_token_dump = TokenBatcher(vocab_file)
self.bilm_token_dump = BidirectionalLanguageModel(
options_file,
weight_file,
use_character_inputs=False,
embedding_weight_file=token_embedding_file)
self.token_ids = tf.placeholder('int32', shape=(None, None))
self.embeddings_op_token_dump = self.bilm_token_dump(
self.token_ids)
"""
with tf.variable_scope('', reuse=tf.AUTO_REUSE):
# the reuse=True scope reuses weights from the whole context
self.elmo_input_token_dump = weight_layers('input', self.embeddings_op_token_dump, l2_coef=0.0)
"""
print("ELMo token dump completed")
def get_sentence_vector_only_ELMo(self, token_list):
"""
Return the ELMo embeddings only for a full sentence
"""
if not self.use_ELMo:
print(
"Warning: ELMo embeddings requested but embeddings object wrongly initialised"
)
return
# Create batches of data
local_token_ids = self.batcher.batch_sentences(token_list)
max_size_sentence = local_token_ids[0].shape[0]
# check lmdb cache
elmo_result = self.get_ELMo_lmdb_vector(token_list, max_size_sentence)
if elmo_result is not None:
return elmo_result
with tf.Session() as sess:
# weird, for this cpu is faster than gpu (1080Ti !)
with tf.device("/cpu:0"):
# It is necessary to initialize variables once before running inference
sess.run(tf.global_variables_initializer())
# Compute ELMo representations (2 times as a heavy warm-up)
elmo_result = sess.run(
self.elmo_input['weighted_op'],
feed_dict={self.character_ids: local_token_ids})
elmo_result = sess.run(
self.elmo_input['weighted_op'],
feed_dict={self.character_ids: local_token_ids})
#cache computation
self.cache_ELMo_lmdb_vector(token_list, elmo_result)
return elmo_result
def get_sentence_vector_with_ELMo(self, token_list):
"""
Return a concatenation of standard embeddings (e.g. Glove) and ELMo embeddings
for a full sentence
"""
if not self.use_ELMo:
print(
"Warning: ELMo embeddings requested but embeddings object wrongly initialised"
)
return
"""
# trick to extend the context for short sentences
token_list_extended = token_list.copy()
#print("token_list_extended before: ", token_list_extended)
for i in range(0, len(token_list_extended)):
local_list = token_list_extended[i]
j = i
while len(local_list) <= 5:
#print(j, local_list)
if j < len(token_list_extended)-1:
local_list = local_list + token_list_extended[j+1]
else:
break
j = j + 1
token_list_extended[i] = local_list
#print("token_list_extended after: ", token_list_extended)
max_size_sentence = 0
for i in range(0, len(token_list)):
local_length = len(token_list[i])
if local_length > max_size_sentence:
max_size_sentence = local_length
"""
# Create batches of data
local_token_ids = self.batcher.batch_sentences(token_list)
max_size_sentence = local_token_ids[0].shape[0]
# check lmdb cache
elmo_result = self.get_ELMo_lmdb_vector(token_list, max_size_sentence)
if elmo_result is None:
with tf.Session() as sess:
# weird, for this cpu is faster than gpu (1080Ti !)
with tf.device("/cpu:0"):
# It is necessary to initialize variables once before running inference
sess.run(tf.global_variables_initializer())
# Compute ELMo representations (2 times as a heavy warm-up)
elmo_result = sess.run(
self.elmo_input['weighted_op'],
feed_dict={self.character_ids: local_token_ids})
elmo_result = sess.run(
self.elmo_input['weighted_op'],
feed_dict={self.character_ids: local_token_ids})
#cache computation
self.cache_ELMo_lmdb_vector(token_list, elmo_result)
concatenated_result = np.zeros(
(elmo_result.shape[0], max_size_sentence - 2, self.embed_size),
dtype=np.float32)
for i in range(0, elmo_result.shape[0]):
for j in range(0, len(token_list[i])):
#if is_int(token_list[i][j]) or is_float(token_list[i][j]):
# dummy_result = np.zeros((elmo_result.shape[2]), dtype=np.float32)
# concatenated_result[i][j] = np.concatenate((dummy_result, self.get_word_vector(token_list[i][j])), )
#else:
concatenated_result[i][j] = np.concatenate(
(elmo_result[i][j], self.get_word_vector(
token_list[i][j])), )
return concatenated_result
def get_sentence_vector_ELMo_with_token_dump(self, token_list):
if not self.use_ELMo:
print(
"Warning: ELMo embeddings requested but embeddings object wrongly initialised"
)
return
with tf.variable_scope('', reuse=tf.AUTO_REUSE):
# the reuse=True scope reuses weights from the whole context
self.elmo_input_token_dump = weight_layers(
'input', self.embeddings_op_token_dump, l2_coef=0.0)
# Create batches of data
local_token_ids = self.batcher_token_dump.batch_sentences(token_list)
with tf.Session() as sess:
# weird, for this cpu is faster than gpu (1080Ti !)
with tf.device("/cpu:0"):
# It is necessary to initialize variables once before running inference
sess.run(tf.global_variables_initializer())
# Compute ELMo representations
elmo_result = sess.run(
self.elmo_input_token_dump['weighted_op'],
feed_dict={self.token_ids: local_token_ids})
return elmo_result
def _get_description(self, name):
for emb in self.registry["embeddings"]:
if emb["name"] == name:
return emb
for emb in self.registry["embeddings-contextualized"]:
if emb["name"] == name:
return emb
return None
def get_word_vector(self, word):
"""
Get static embeddings (e.g. glove) for a given token
"""
if (self.name == 'wiki.fr') or (self.name == 'wiki.fr.bin'):
# the pre-trained embeddings are not cased
word = word.lower()
if self.env is None:
# db not available, the embeddings should be available in memory (normally!)
return self.get_word_vector_in_memory(word)
try:
with self.env.begin() as txn:
txn = self.env.begin()
vector = txn.get(word.encode(encoding='UTF-8'))
if vector:
word_vector = _deserialize_pickle(vector)
vector = None
else:
word_vector = np.zeros((self.static_embed_size, ),
dtype=np.float32)
# alternatively, initialize with random negative values
#word_vector = np.random.uniform(low=-0.5, high=0.0, size=(self.embed_size,))
# alternatively use fasttext OOV ngram possibilities (if ngram available)
except lmdb.Error:
# no idea why, but we need to close and reopen the environment to avoid
# mdb_txn_begin: MDB_BAD_RSLOT: Invalid reuse of reader locktable slot
# when opening new transaction !
self.env.close()
envFilePath = os.path.join(self.embedding_lmdb_path, self.name)
self.env = lmdb.open(envFilePath,
readonly=True,
max_readers=2048,
max_spare_txns=2,
lock=False)
return self.get_word_vector(word)
return word_vector
def get_ELMo_lmdb_vector(self, token_list, max_size_sentence):
"""
Try to get the ELMo embeddings for a sequence cached in LMDB
"""
if self.env_ELMo is None:
# db cache not available, we don't cache ELMo stuff
return None
try:
ELMo_vector = np.zeros(
(len(token_list), max_size_sentence - 2, ELMo_embed_size),
dtype='float32')
with self.env_ELMo.begin() as txn:
for i in range(0, len(token_list)):
txn = self.env_ELMo.begin()
# get a hash for the token_list
the_hash = list_digest(token_list[i])
vector = txn.get(the_hash.encode(encoding='UTF-8'))
if vector:
# adapt expected shape/padding
local_embeddings = _deserialize_pickle(vector)
if local_embeddings.shape[0] > max_size_sentence - 2:
# squeeze the extra padding space
ELMo_vector[
i] = local_embeddings[:max_size_sentence - 2, ]
elif local_embeddings.shape[
0] == max_size_sentence - 2:
# bingo~!
ELMo_vector[i] = local_embeddings
else:
# fill the missing space with padding
filler = np.zeros((max_size_sentence -
(local_embeddings.shape[0] + 2),
ELMo_embed_size),
dtype='float32')
ELMo_vector[i] = np.concatenate(
(local_embeddings, filler))
vector = None
else:
return None
except lmdb.Error:
# no idea why, but we need to close and reopen the environment to avoid
# mdb_txn_begin: MDB_BAD_RSLOT: Invalid reuse of reader locktable slot
# when opening new transaction !
self.env_ELMo.close()
self.env_ELMo = lmdb.open(embedding_ELMo_cache,
readonly=True,
max_readers=2048,
max_spare_txns=2,
lock=False)
return self.get_ELMo_lmdb_vector(token_list)
return ELMo_vector
def cache_ELMo_lmdb_vector(self, token_list, ELMo_vector):
"""
Cache in LMDB the ELMo embeddings for a given sequence
"""
if self.env_ELMo is None:
# db cache not available, we don't cache ELMo stuff
return None
txn = self.env_ELMo.begin(write=True)
for i in range(0, len(token_list)):
# get a hash for the token_list
the_hash = list_digest(token_list[i])
txn.put(the_hash.encode(encoding='UTF-8'),
_serialize_pickle(ELMo_vector[i]))
txn.commit()
def clean_ELMo_cache(self):
"""
Delete ELMo embeddings cache, this takes place normally after the completion of a training
"""
if self.env_ELMo is None:
# db cache not available, nothing to clean
return
else:
for file in os.listdir(self.embedding_ELMo_cache):
file_path = os.path.join(self.embedding_ELMo_cache, file)
if os.path.isfile(file_path):
os.remove(file_path)
os.rmdir(self.embedding_ELMo_cache)
def get_word_vector_in_memory(self, word):
if (self.name == 'wiki.fr') or (self.name == 'wiki.fr.bin'):
# the pre-trained embeddings are not cased
word = word.lower()
if self.extension == 'bin':
return self.model.get_numpy_vector(word)
if word in self.model:
return self.model[word]
else:
# for unknown word, we use a vector filled with 0.0
return np.zeros((self.static_embed_size, ), dtype=np.float32)
def __init__(self,
n_layers,
n_source_vocab,
n_units,
catgy,
doc_catgy,
senseid2netout,
word2index,
pre_trained_embedding,
model_type,
multi_label,
wsd_epoch=0,
h=8,
dropout=0.1,
max_length=500,
use_label_smoothing=False,
embed_position=False,
wsd_model=None):
super(Transformer, self).__init__()
self.to_gpu()
self.set_random_seed(123)
self.word2index = word2index
self.pre_trained_embedding = pre_trained_embedding
self.model_type = model_type
self.wsd_model = wsd_model
self.multi_label = multi_label
with self.init_scope():
if not self.pre_trained_embedding is None:
model = FastText(self.pre_trained_embedding)
dim = len(model['a'])
n_vocab = len(self.word2index.keys())
self.look_up_table = self.xp.zeros((n_vocab, dim),
dtype=np.float32)
for word, index in self.word2index.items():
try:
self.look_up_table[index] = chainer.cuda.to_gpu(
model.get_numpy_vector(word))
except:
self.xp.random.seed(index)
self.look_up_table[index][:] = self.xp.random.uniform(
-0.25, 0.25, dim)
self.embed_x = L.EmbedID(n_source_vocab,
n_units,
ignore_label=-1,
initialW=self.look_up_table)
else:
self.embed_x = L.EmbedID(n_source_vocab,
n_units,
ignore_label=-1,
initialW=linear_init)
self.encoder = Encoder(n_layers, n_units, h, dropout)
self.fc2 = L.Linear(in_size=n_units,
out_size=len(doc_catgy),
initialW=linear_init)
self.fc2_wsd = L.Linear(in_size=n_units,
out_size=len(catgy),
initialW=linear_init)
self.lookup_table_sense = L.EmbedID(in_size=len(catgy),
out_size=n_units,
ignore_label=-1,
initialW=linear_init)
self.lookup_table_sense_fixed = self.lookup_table_sense.W.data
self.senseid2netout = senseid2netout
self.senseid2netout['<PAD>'] = [-1]
self.wsd_epoch = wsd_epoch
if embed_position:
self.embed_pos = L.EmbedID(max_length,
n_units,
ignore_label=-1)
self.n_layers = n_layers
self.n_units = n_units
self.dropout = dropout
self.use_label_smoothing = use_label_smoothing
self.initialize_position_encoding(max_length, n_units)
self.scale_emb = self.n_units**0.5 ## origin 0.5
self.doc_catgy = doc_catgy
self.catgy = catgy
self.inverse_catgy = {v: k for k, v in self.catgy.items()}
self.wsd_netout2wordindex = {
k: self.word2index[v]
for k, v in self.inverse_catgy.items()
}
self.wsd_netout2wordindex[-1] = -1
self.max_len = max_length
self.le = None
if self.multi_label == 1:
self.le = MultiLabelBinarizer(classes=[
i[0]
for i in sorted(self.doc_catgy.items(), key=lambda x: x[1])
],
sparse_output=False)
elif self.multi_label == 0:
self.le = LabelEncoder()
self.le.fit([
i[0]
for i in sorted(self.doc_catgy.items(), key=lambda x: x[1])
])
self.to_gpu()
# 加载一个已经存在的模型
# model = FastText()
# model.load_model('./path/to/model.bin')
# 使用skip-gram模型训练
skip_gram_model = FastText()
skip_gram_model.skipgram(input='./train.txt',
output='skip_gram_model',
epoch=100,
lr=0.7)
print(skip_gram_model['贷款'])
# print(skip_gram_model.get_numpy_vector('贷款'))
# print(skip_gram_model.get_numpy_vector('贷款', normalized=True))
var1 = skip_gram_model.get_numpy_vector('人民币')
var2 = skip_gram_model.get_numpy_vector('贷款')
var3 = skip_gram_model.get_numpy_vector('外币')
skip_gram_model.words_for_vector(var1 + var2 - var3, k=1)
# for word in skip_gram_model.words:
# print(word, skip_gram_model[word])
print(skip_gram_model.nearest_neighbors('贷款', k=2))
# test data is stored inside a file, use this:
# skip_gram_model.predict_proba_file('./test.txt', k=2)
print("\n")
##################
