时间戳

首页 > 解决方案 > 如何在 keras Sequntial 模型中使用来自 doc2vec 的句子向量进行句子情感分析？

python - 如何在 keras Sequntial 模型中使用来自 doc2vec 的句子向量进行句子情感分析？

问题描述

创建 doc2vec 模型

x：句子列表（电影评论）

x的长度=2000

doc2vec_data = []
for line in x:
temp = ''.join(str(token) for token in line.lower())
doc2vec_data.append(temp)

File = open('doc2vec_data.txt', 'w',encoding="utf-8") 
for item in doc2vec_data:
File.write("%s\n" % item)

sentences=gensim.models.doc2vec.TaggedLineDocument("doc2vec_data.txt")

d2v =gensim.models.Doc2Vec(sentences, dm=0,window = 5, 
                   size=5,
                   iter = 100, workers=32,dbow_words=1,
                   alpha=2,min_aplha=0.5)

创建向量的 Numpy 数组：因为 doc2vec 模型不能直接用于 keras 顺序模型。

vec=np.array([d2v.infer_vector(item) for item in x])

Keras 序列模型：

model=Sequential()
model.add(Embedding(2000,128,input_length=vec.shape[1]))
model.add(LSTM(200,dropout=0.2,recurrent_dropout=0.2))
model.add(Dense(1,activation='sigmoid'))

model.compile(loss='binary_crossentropy',
      optimizer='rmsprop',
      metrics=['accuracy'])

y：句子标签（0 和 1）

model.fit(vec,y,
  batch_size=32,epochs=8,
  verbose=1)

上面的代码给了我这个错误-

InvalidArgumentError                      Traceback (most recent call last)
~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in 
_do_call(self, fn, *args)
1349     try:
-> 1350       return fn(*args)
1351     except errors.OpError as e:

~\Anaconda3\lib\site-packages\tensorflow\python\client\session.py in 
_run_fn(session, feed_dict, fetch_list, target_list, options, run_metadata)
 1328                                    feed_dict, fetch_list, target_list,
-> 1329                                    status, run_metadata)
 1330 

~\Anaconda3\lib\site-packages\tensorflow\python\framework\errors_impl.py in 
__exit__(self, type_arg, value_arg, traceback_arg)
472             compat.as_text(c_api.TF_Message(self.status.status)),  
--> 473             c_api.TF_GetCode(self.status.status))
474     # Delete the underlying status object from memory otherwise it 
stays alive

InvalidArgumentError: indices[0,0] = -19 is not in [0, 2000)
 [[Node: embedding_1/Gather = Gather[Tindices=DT_INT32, Tparams=DT_FLOAT, 
validate_indices=true, 
_device="/job:localhost/replica:0/task:0/device:CPU:0"] 
(embedding_1/embeddings/read, embedding_1/Cast)]]

During handling of the above exception, another exception occurred:

InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-34-3d0fc0b22a78> in <module>()
 1 model.fit(vec,y,
  2          batch_size=32,epochs=8,
----> 3          verbose=1) 

InvalidArgumentError: indices[0,0] = -19 is not in [0, 2000)
 [[Node: embedding_1/Gather = Gather[Tindices=DT_INT32, Tparams=DT_FLOAT, 
validate_indices=true, 
_device="/job:localhost/replica:0/task:0/device:CPU:0"] 
(embedding_1/embeddings/read, embedding_1/Cast)]]

有人可以告诉我错误是什么，我该如何解决？

标签： pythonkerasdeep-learningsentiment-analysis

解决方案

您已经将句子转换为向量并使用 Keras 模型重新尝试。它抱怨您的嵌入层没有收到正确的索引，因为它已经嵌入。假设您vec.shape == (samples, doc2vec_vector_size)需要删除 Embedding 因为它已经嵌入，而 LSTM 因为现在每个句子而不是每个单词都有 1 个向量：

model = Sequential()
model.add(Dense(hidden_size, activation='relu', input_dim=doc2vec_vector_size))
model.add(Dense(1, activation='sigmoid'))

推荐阅读