This repository contains the source code and datasets for the CIKM 2019 paper Interactive Matching Network for Multi-Turn Response Selection in Retrieval-Based Chatbots by Gu et al.
Our proposed Interactive Matching Network (IMN) has achieved a new state-of-the-art performance on four large-scale datasets that are publicly available for research on multi-turn conversation.
Python 2.7
Tensorflow 1.4.0
Your can download the processed datasets used in our paper here and unzip it to the folder of data
.
Ubuntu_V1
Ubuntu_V2
Douban
Ecommerce
Take Ubuntu_V1 as an example.
cd scripts
bash ubuntu_train.sh
The training process is recorded in log_train_IMN_UbuntuV1.txt
file.
bash ubuntu_test.sh
The testing process is recorded in log_test_IMN_UbuntuV1.txt
file. And your can get a ubuntu_test_out.txt
file which records scores for each context-response pair. 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: "Interactive Matching Network for Multi-Turn Response Selection in Retrieval-Based Chatbots" Jia-Chen Gu, Zhen-Hua Ling, Quan Liu. CIKM (2019)
@inproceedings{Gu:2019:IMN:3357384.3358140,
author = {Gu, Jia-Chen and
Ling, Zhen-Hua and
Liu, Quan},
title = {Interactive Matching Network for Multi-Turn Response Selection in Retrieval-Based Chatbots},
booktitle = {Proceedings of the 28th ACM International Conference on Information and Knowledge Management},
series = {CIKM '19},
year = {2019},
isbn = {978-1-4503-6976-3},
location = {Beijing, China},
pages = {2321--2324},
url = {http://doi.acm.org/10.1145/3357384.3358140},
doi = {10.1145/3357384.3358140},
acmid = {3358140},
publisher = {ACM},
}