indiejoseph / cnn-text-classification-tf-chinese

CNN for Chinese Text Classification in Tensorflow

CNN for Chinese Text Classification in Tensorflow

Sentiment classification forked from dennybritz/cnn-text-classification-tf, make the data helper supports Chinese language and modified the embedding from word-level to character-level, though that increased vocabulary size, and also i've implemented the Character-Aware Neural Language Models network structure which CNN + Highway network to improve the performance, this version can achieve an accuracy of 98% with the Chinese corpus

This code belongs to the "Implementing a CNN for Text Classification in Tensorflow" blog post.

It is slightly simplified implementation of Kim's Convolutional Neural Networks for Sentence Classification paper in Tensorflow.

Requirements

• Python 2.7
• Tensorflow 0.9.0
• Numpy

Running

Print parameters:

./train.py --help

optional arguments:
  -h, --help            show this help message and exit
  --embedding_dim EMBEDDING_DIM
                        Dimensionality of character embedding (default: 128)
  --filter_sizes FILTER_SIZES
                        Comma-separated filter sizes (default: '1,2,3,4,5,6,8')
  --num_filters NUM_FILTERS
                        Number of filters per filter size (default: '50,100,150,150,200,200,200')
  --l2_reg_lambda L2_REG_LAMBDA
                        L2 regularizaion lambda (default: 0.0)                        
  --dropout_keep_prob DROPOUT_KEEP_PROB
                        Dropout keep probability (default: 0.5)
  --batch_size BATCH_SIZE
                        Batch Size (default: 32)
  --num_epochs NUM_EPOCHS
                        Number of training epochs (default: 100)
  --evaluate_every EVALUATE_EVERY
                        Evaluate model on dev set after this many steps
                        (default: 100)
  --checkpoint_every CHECKPOINT_EVERY
                        Save model after this many steps (default: 100)
  --allow_soft_placement ALLOW_SOFT_PLACEMENT
                        Allow device soft device placement
  --noallow_soft_placement
  --log_device_placement LOG_DEVICE_PLACEMENT
                        Log placement of ops on devices
  --nolog_device_placement

Train:

./train.py

References

• Convolutional Neural Networks for Sentence Classification
• A Sensitivity Analysis of (and Practitioners' Guide to) Convolutional Neural Networks for Sentence Classification

cnn-text-classification-tf-chinese's Issues

你好，可以提供对应的eval.py么？

你好，我参考train.py的参数和cnn-text-classification-tf的train.py,eval.py 尝试写了一个eval.py,但是测试的Accuracy只有0.503125。
刚接触tensorflow，而且在测试集不同时，最大sequence_length不同，模型shape对不上，想改成placeholder失败，后来用整个训练集作为测试，得出的准确率50.3%和训练最后的98%左右差异很大。
可以提供一份对应的eval.py么？

out of memory on GPU gtx970(4M Mem)

请问如果有多个句子怎么办呢？

您好，我看你写的是中文的文本分类，所以就用中文问了
这个demo好像做的是句子级别的分类，但是我想请问下，如果我有多个句子应该怎么办呢？
把channel数改成相应的句子数吗？相应的batch应该选择哪个呢？

希望您有空能回复下，感激不尽

可以提供兼容Python3.5的版本吗

刚入门，在Linux系统上安装驱动太麻烦，就在Windows上搞了，可是TensorFlow对Windows系统只支持python3.5版本的。。

where is the data?

this is totaly mis-report!

sorry!

你好 这个中文情感分析 数据 是从哪里得到的？

谢谢

请问你的中文输入数据格式是什么样的？每行多个词，词之间空格间隔么？

InternalError: Dst tensor is not initialized.

提示内存不够，但是内存很明显够呀剩下100g呢

I tensorflow/core/common_runtime/bfc_allocator.cc:692] 1 Chunks of size 6652672 totalling 6.34MiB
I tensorflow/core/common_runtime/bfc_allocator.cc:696] Sum Total of in-use chunks: 42.12MiB
I tensorflow/core/common_runtime/bfc_allocator.cc:698] Stats: 
Limit:                    91684864
InUse:                    44171776
MaxInUse:                 48052736
NumAllocs:                     138
MaxAllocSize:              6652672

W tensorflow/core/common_runtime/bfc_allocator.cc:270] ******************xx**************************____******____________________________________________
W tensorflow/core/common_runtime/bfc_allocator.cc:271] Ran out of memory trying to allocate 44.94MiB.  See logs for memory state.
Traceback (most recent call last):

Name: tensorflow
Version: 0.11.0rc1
Location: /usr/local/lib/python2.7/dist-packages
Requires: mock, protobuf, numpy, wheel, six

请问想做多分类该怎么修改呢？

你好，

我最近在做中文文本多分类，tensorflow刚开始学，还不太懂，如果我没搞错这个是做二分类的，请问要做多分类的话怎么修改呢？谢谢！

请问这个代码在测试集上的准确率和召回率分别是多少呢

如题，谢谢！

你好，请问尝试过将数据分词后作为输入和用pretrained的词向量的做法吗？

如题，如尝试过，不知效果如何？是否会提高准召率？谢谢。

The model seems not converging?

The training accuracy is fluctuating between 55% -75% even after 800 steps.
I wonder in your case, how many steps did you take to make it converge to 98%?

There is no _linear function in TensorFlow 1.0.1

I managed to run the code in TF 1.0.1 using the conversion script. However, I still cannot found the _linear function in TF 1.0.1 in text_cnn.py line 14. Could you give some suggestions? Thanks.

直接运行时候, 报错

你好啊, 我这边直接 ./train.py 就报错了
会不会是版本不对? 0.9.0 >> 1.0.1

我的 tensorflow 版本为 1.0.1
python 为 2.7.13

Traceback (most recent call last): File "./train.py", line 76, in <module> l2_reg_lambda=FLAGS.l2_reg_lambda) File "/Users/edwardchan/projects/classification/cnn-text-classification-tf-chinese/text_cnn.py", line 75, in __init__ self.h_pool = tf.concat(3, pooled_outputs) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py", line 1029, in concat dtype=dtypes.int32).get_shape( File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 637, in convert_to_tensor as_ref=False) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 702, in internal_convert_to_tensor ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/constant_op.py", line 110, in _constant_tensor_conversion_function return constant(v, dtype=dtype, name=name) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/constant_op.py", line 99, in constant tensor_util.make_tensor_proto(value, dtype=dtype, shape=shape, verify_shape=verify_shape)) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/tensor_util.py", line 367, in make_tensor_proto _AssertCompatible(values, dtype) File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/tensor_util.py", line 302, in _AssertCompatible (dtype.name, repr(mismatch), type(mismatch).__name__)) TypeError: Expected int32, got list containing Tensors of type '_Message' instead.

你好，请教个问题。能解释下你的代码意思和主要做什么用的，分词？

你好。不好意思。初学tensorflow,copy了你的代码，在python 3.5.2上train了一下。

看了下准确率只有70%，是因为train时间的原因吗，正常要train多久？还有我看了下，这是个英文文章，你的cnn是做分词的 还是什么？请稍微花点时间帮忙解答下 。
然后我想实现，将一个照片上不同部分的defect识别出来，该用什么的tensorflow架构 cnn?有什么架构推荐下