siat-nlp/ddmn

Code and data for the paper "Dual Dynamic Memory Network for End-to-End Multi-turn Task-oriented Dialog Systems".

DDMN

This repository contains data and code for the COLING'2020 paper "Dual Dynamic Memory Network for End-to-End Multi-turn Task-oriented Dialog Systems".

In this paper, we propose a Dual Dynamic Memory Network (DDMN) for multi-turn dialog generation, which maintains two core components: dialog memory manager and KB memory manager. The dialog memory manager dynamically expands the dialog memory turn by turn and keeps track of dialog history with an updating mechanism, which encourages the model to filter irrelevant dialog history and memorize important newly coming information. The KB memory manager shares the structural KB triples throughout the whole conversation, and dynamically extracts KB information with a memory pointer at each turn.

Requirements

The implementation is based on Python 3.x. To install the dependencies used in this project, please run:

pip install -r requirements.txt

Quickstart

Step 1: Training

For different datasets, please first set up the following parameters in the script run_train.sh:

data_dir=[xxx]        # directory of the specific dataset
save_dir=[xxx]        # directory to store trained models

and then run:

sh run_train.sh

Note that more arguments for training can be found in the main.py.

For self-critical sequence training, please set up num_epochs larger than pre_epochs but no larger than 3 epochs, since self-critical sequence training for a long time might be unstable sometimes.

Step 2: Testing

For different datasets, please first set up the following parameters in the script run_test.sh:

data_dir=[xxx]        # directory of the specific dataset
save_dir=[xxx]        # directory of the rrtrained models
output_dir=[xxx]      # directory to store generation results

and then run:

sh run_test.sh

Note that more arguments for testing can be found in the main.py.

Step 3: Evaluation

For different datasets, please first set up the following parameters in the script run_eval.sh:

data_name=[xxx]      # ['kvr', 'camrest', 'multiwoz']
data_dir=[xxx]       # directory of the specific dataset
eval_dir=[xxx]       # directory of the generation output

and then run:

sh run_eval.sh

Citation

If you use our code in your work, please cite as:

@inproceedings{wang-etal-2020-dual,
    title = "Dual Dynamic Memory Network for End-to-End Multi-turn Task-oriented Dialog Systems",
    author = "Wang, Jian  and
      Liu, Junhao  and
      Bi, Wei  and
      Liu, Xiaojiang  and
      He, Kejing  and
      Xu, Ruifeng  and
      Yang, Min",
    booktitle = "Proceedings of the 28th International Conference on Computational Linguistics",
    year = "2020",
    address = "Barcelona, Spain (Online)",
    publisher = "International Committee on Computational Linguistics",
    pages = "4100--4110"
}