Cornell Movie Dialog classification with convolution layers.
It is slightly simplified implementation of Kim's Convolutional Neural Networks for Sentence Classification paper in Tensorflow.
- To (Download and) preprocess data
./data.py - To create model graph, and test data flow
./model.py - To train (existing or new) model
./train.py - To get actual options use
-hor--helpflag after./script.py - Dependency: (((data) model) train)
- Python 3
- Tensorflow > 0.12
- Numpy
- Possibly [update][update] required for Google Protobuf library if TFRecord is slow [update]: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/g3doc/get_started/os_setup.md#protobuf-library-related-issues
- Yoon Kim's github repo
- Denny's blog post on text classification
- Suriya Deepan blog post on preprocessing Cornell Data
- A Sensitivity Analysis of (and Practitioners' Guide to) Convolutional Neural Networks for Sentence Classification
- Indico blog post on Protobufs and Queues