https://github.com/shagunsodhani/CNN-Sentence-Classifier

Implementation of "Convolutional Neural Networks for Sentence Classification" paper
https://github.com/shagunsodhani/CNN-Sentence-Classifier

Last synced: about 2 months ago
JSON representation

Implementation of "Convolutional Neural Networks for Sentence Classification" paper

• Host: GitHub
• URL: https://github.com/shagunsodhani/CNN-Sentence-Classifier
• Owner: shagunsodhani
• License: mit
• Created: 2016-10-22T19:19:45.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2017-09-23T16:15:42.000Z (almost 7 years ago)
• Last Synced: 2024-05-03T06:21:19.856Z (5 months ago)
• Language: Python
• Homepage: http://shagunsodhani.in/CNN-Sentence-Classifier/
• Size: 477 KB
• Stars: 144
• Watchers: 9
• Forks: 29
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# CNN-Sentence-Classifier

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

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.971868.svg)](https://doi.org/10.5281/zenodo.971868)

## Usage

* Install [Keras](https://keras.io/#installation)

* 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](http://nlp.stanford.edu/projects/glove/) to train or download Glove embeddings and [this](https://code.google.com/archive/p/word2vec/) 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](app/model/model.py) and invoke from `model_selector` function (in [model.py](app/model/model.py)).

* All supported arguments can be seen in [here](app/utils/argumentparser.py)

## References

* [Kim, Yoon. "Convolutional neural networks for sentence classification." arXiv preprint arXiv:1408.5882 (2014)](https://arxiv.org/abs/1408.5882)

* [Summary of paper](https://gist.github.com/shagunsodhani/9ae6d2364c278c97b1b2f4ec53255c56)