CNN-Sentence-Classifier

Simplified implementation of "Convolutional Neural Networks for Sentence Classification" paper

Usage

• Install Keras
• Repository contains "Movie reviews with one sentence per review" (Pang and Lee, 2005) dataset in sample_dataset.
• Alternatively, to use some other dataset, make two files
  • input.txt where each line is a sentence to be classified
  • label.txt where each line is the label for corresponding line in input.txt
• Make model folder by running mkdir model
• Refer this to train or download Glove embeddings and this for Word2Vec embeddings.
• Run python3 app/train.py --data_dir=path_to_folder_containing_input.txt_and_label.txt --embedding_file_path=path_to_embedding_vectors_file --model_name=name_of_model_from_the_paper
• For example, if data is in data folder, embedding file is vectors.txt and model is cnn_static, run python3 app/train.py --data_dir=data --embedding_file_path=vectors.txt --model_name=cnn_static
• To define your own model, pass model_name as self, define your model in app/model/model.py and invoke from model_selector function (in model.py).
• All supported arguments can be seen in here

References

• Kim, Yoon. "Convolutional neural networks for sentence classification." arXiv preprint arXiv:1408.5882 (2014)
• Summary of paper