Character Level CNNs in Keras
This repository contains Keras implementations for Character-level Convolutional Neural Networks for text classification on AG's News Topic Classification Dataset.
The following models have been implemented:
- Xiang Zhang, Junbo Zhao, Yann LeCun. Character-level Convolutional Networks for Text Classification. NIPS 2015
- Yoon Kim, Yacine Jernite, David Sontag, Alexander M. Rush. Character-Aware Neural Language Models. AAAI 2016
- Shaojie Bai, J. Zico Kolter, Vladlen Koltun. An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling. ArXiv preprint (2018)
Kim's CharCNN was originally part of an end-to-end trained pipeline for language modelling, but has been adapted for text classification.
- Install dependencies (Tensorflow 1.3 and Keras 2.1.3):
$ pip install -r requirements.txt
Specify the training and testing data sources and model hyperparameters in the
Run the main.py file as below:
$ python main.py --model [model_name]
[model_name] with either
kim to run the desired model.