This repository contains the source code and datasets for the EMNLP 2020 paper Filtering before Iteratively Referring for Knowledge-Grounded Response Selection in Retrieval-Based Chatbots by Gu et al.

Our proposed Filtering before Iteratively REferring (FIRE) model has achieved a new state-of-the-art performance of knowledge-grounded response selection on the PERSONA-CHAT and CMU_DoG datasets.

Python 2.7
Tensorflow 1.4.0

Your can download the datasets and their corresponding embedding and vocabulary files used in our paper from the following links.

• PERSONA-CHAT and its embedding and vocabulary files.
  
• CMU_DoG and its embedding and vocabulary files.
  

Unzip the datasets to the folder of data and run the following commands. The processed files are stored in data/personachat_processed/ or data/cmudog_processed/.

cd data
python data_preprocess_pc.py
python data_preprocess_cd.py

Then, unzip their corresponding embedding and vocabulary files to the folder of data/personachat_processed/ or data/cmudog_processed/.

Take PERSONA-CHAT as an example.

cd scripts
bash train_personachat.sh

The training process is recorded in log_FIRE_train_personachat_original.txt file.

bash test_personachat.sh

The testing process is recorded in log_FIRE_test_personachat_original.txt file. And your can get a test_out_FIRE_personachat_original.txt file which records scores for each example. Run the following command and you can compute the metric of Recall.

python compute_recall.py

If you use the code and datasets, please cite the following paper: "Filtering before Iteratively Referring for Knowledge-Grounded Response Selection in Retrieval-Based Chatbots" Jia-Chen Gu, Zhen-Hua Ling, Quan Liu, Zhigang Chen, Xiaodan Zhu. EMNLP (2020)

@inproceedings{gu-etal-2020-filtering,
 title = "Filtering before Iteratively Referring for Knowledge-Grounded Response Selection in Retrieval-Based Chatbots",
 author = "Gu, Jia-Chen and
           Ling, Zhenhua and
           Liu, Quan and
           Chen, Zhigang and
           Zhu, Xiaodan",
 booktitle = "Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: Findings",
 month = nov,
 year = "2020",
 address = "Online",
 publisher = "Association for Computational Linguistics",
 url = "https://www.aclweb.org/anthology/2020.findings-emnlp.127",
 pages = "1412--1422",
}

Thank ParlAI for providing the PERSONA-CHAT dataset.
Thank Xueliang Zhao for providing the processed CMU_DoG dataset used in their paper.

Please keep an eye on this repository if you are interested in our work.
Feel free to contact us ([email protected]) or open issues.