Towards-Implicit-Content-Introducing-for-Generative-Short-Text-Conversation-Systems

This repository is the implementation of the EMNLP '17 paper Towards Implicit Content-Introducing for Generative Short-Text Conversation Systems. If you use this code (based on dl4mt) or our results in your research, please cite as appropriate:

@inproceedings{yao2017towards,
    title={Towards implicit content-introducing for generative short-text conversation systems},
    author={Yao, Lili and Zhang, Yaoyuan and Feng, Yansong and Zhao, Dongyan and Yan, Rui},
    booktitle={Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing},
    pages={2190--2199},
    year={2017}
}

Usage

• Step1: Prepare your query, reply, cue words file; one instance per line.
• Step2: Build dictionary using build_dictionary.py
• Step3: Update the absolute path of data in training and testing script.
• Step4: You can train and inference after that.

$ sh train.sh
$ sh test.sh

It's a preliminary version. Please feel free to contact with me if there are any bugs. My email: yaolili12235 AT gmail DOT com