rockt/hierarchical-neural-autoencoder

A Hierarchical Neural Autoencoder for Paragraphs and Documents

Implementations of the three models presented in the paper "A Hierarchical Neural Autoencoder for Paragraphs and Documents" by Jiwei Li, Minh-Thang Luong and Dan Jurafsky, ACL 2015

Requirements:

GPU

matlab >= 2014b

memory >= 4GB

Folders

Standard_LSTM: Standard LSTM Autoencoder

hier_LSTM: Hierarchical LSTM Autoencoder

hier_LSTM_Attention: Hierarchical LSTM Autoencoder with Attention

DownLoad Data

• dictionary: vocabulary
• train_permute.txt: training data for standard Model. Each line corresponds to one document/paragraph
• train_source_permute_segment.txt: source training data for hierarchical Models. Each line corresponds to one sentence. An empty line starts a new document/sentence. Documents are reversed.
• test_source_permute_segment.txt: target training data for hierarchical Model.

Training roughly takes 2-3 weeks for standard models and 4-6 weeks for hierarchical models on a K40 GPU machine.

For any question or bug with the code, feel free to contact [email protected]

@article{li2015hierarchical,
    title={A Hierarchical Neural Autoencoder for Paragraphs and Documents},
    author={Li, Jiwei and Luong, Minh-Thang and Jurafsky, Dan},
    journal={arXiv preprint arXiv:1506.01057},
    year={2015}
}