alexyangli / absa_keras Goto Github PK

Thank you for sharing this wonderful tutorial. I downloaded the glove.42B.300d file and run the preprocess.py code，but occurd an error: ValueError: invalid literal for int() with base 10: ',' have you ever come across this error? how can i slove it ？I am looking forward for your reply. Thank you .

run preprocess.sh to generate word_glove.npy

Originally posted by @AlexYangLi in #3 (comment)

Running the preprocess.sh file does not generate word_glove.npy
Could you please upload that file?

Hi,

I tried to run this code but I got this error

could someone help plz

Thank u

Dear author(s)

What great work! I was cloning your codes, and it was running very well. I plan to run these codes using my own datasets according to my research. However, I have some questions:

1. May I directly replace the new dataset in folder data?
2. The data must be in XML format, isn't it?
3. Which part command should I change when the new dataset is used?

Thank you in advance.

regards,
Novi

Hi,

I ran train.py of the previous commit, but only 'cabasc' got an error as follows:

tensorflow.python.framework.errors_impl.InvalidArgumentError: Matrix size-incompatible: In[0]: [2336,300], In[1]: [1324,1324] [[{{node model_1/content_attention_1/MatMul_1}}]] [[{{node metrics/acc/Mean}}]]

At that time, the code toggle comments on most of train_model(...) and just ran fixed word/aspect embedding part, i.e.

config.word_embed_trainable = False
config.aspect_embed_trainable = False
train_model('laptop/term', 'laptop', 'word', 'td_lstm')
train_model('laptop/term', 'laptop', 'word', 'tc_lstm')
train_model('laptop/term', 'laptop', 'word', 'ae_lstm')
train_model('laptop/term', 'laptop', 'word', 'at_lstm')
train_model('laptop/term', 'laptop', 'word', 'atae_lstm')
train_model('laptop/term', 'laptop', 'word', 'memnet')
train_model('laptop/term', 'laptop', 'word', 'ram')
train_model('laptop/term', 'laptop', 'word', 'ian')
# train_model('laptop/term', 'laptop', 'word', 'cabasc')

Only 'cabasc' was toggled comment and did not plot on the results of combining ELMo Embedding as well.

So I guess that 'cabasc' can not run in this condition?

When I am trying to run train.py It is showing me the error

File "train.py", line 183, in <module>
    train_model('twitter', 'twitter', 'word', 'atae_lstm')
  File "train.py", line 46, in train_model
    model = SentimentModel(config)
  File "C:\Users\papa\Documents\ABSA_Keras\models.py", line 75, in __init__
    self.config.word_embed_type))
  File "C:\ProgramData\Anaconda3\lib\site-packages\numpy\lib\npyio.py", line 415, in load
    fid = open(os_fspath(file), "rb")
FileNotFoundError: [Errno 2] No such file or directory: './data/twitter/word_glove.npy'

results in the [] are the performances of models with noth word embeddings and aspect embeddings fixed

您好，我用您的代码跑了这几个数据集，发现实现结果跟Paper上的低了1%或者2%。请问是什么问题？感谢回复