DaoD / ReBoost

IR Journal 2019: ReBoost: A Retrieval-Boosted Sequence-to-Sequence Model for Neural Response Generation

ReBoost: A Retrieval-Boosted Sequence-to-Sequence Model for Neural Response Generation

This repository contains the source code and datasets for the IR Journal paper ReBoost: A Retrieval-Boosted Sequence-to-Sequence Model for Neural Response Generation by Zhu et al.

Model overview

Results

Dependencies

Python 3.5
Tensorflow 1.1.2

Datasets

Your can download the processed datasets used in our paper here and unzip it to the folder of data.
Weibo (STC)
OpenSubtitle

Train and test a model

python3 runModel.py

Cite

If you use the code and datasets, please cite the following paper:
"ReBoost: A Retrieval-Boosted Sequence-to-Sequence Model for Neural Response Generation"
Yutao Zhu, Zhicheng Dou, Jian-Yun Nie and Ji-Rong Wen. IR Journal (2019)

@article{DBLP:journals/ir/ZhuDNW20,
  author    = {Yutao Zhu and
               Zhicheng Dou and
               Jian{-}Yun Nie and
               Ji{-}Rong Wen},
  title     = {ReBoost: a retrieval-boosted sequence-to-sequence model for neural
               response generation},
  journal   = {Inf. Retr. Journal},
  volume    = {23},
  number    = {1},
  pages     = {27--48},
  year      = {2020},
  url       = {https://doi.org/10.1007/s10791-019-09364-x},
  doi       = {10.1007/s10791-019-09364-x}
}