Naive_sentiment_analysis

• Here is the lab of CS438 in SJTU. We realize a simple classifier to judge if a comment to a goods is positive or not.

• Used model:

  • CNN+LSTM for feature extraction
  • A trained word2vec model. (By Stanford GloVe project.)

• Some module need(may):

  Tensorflow1.12.0, jieba0.39, numpy1.15.2, gensim3.6.0, nltk3.4, scipy1.1.0, tqdm4.28.1
  

• To train the model:

  • Type python3 train_dl.py (if you want to change some parameters, go to config.py)

• To evaluate and output a file:

  • Type python3 evaluate.py(you may firstly refer our form.)