NeuralWOZ: Learning to Collect Task-Oriented Dialogue via Model-based Simulation (ACL-IJCNLP 2021)

Overview

NeuralWOZ

This code is official implementation of "NeuralWOZ: Learning to Collect Task-Oriented Dialogue via Model-based Simulation".

Sungdong Kim, Minsuk Chang, Sang-woo Lee
In ACL 2021.

Citation

@inproceedings{kim2021neuralwoz,
  title={NeuralWOZ: Learning to Collect Task-Oriented Dialogue via Model-based Simulation},
  author={Kim, Sungdong and Chang, Minsuk and Lee, Sang-woo},
  booktitle={ACL},
  year={2021}
}

Requirements

python3.6
torch==1.4.0
transformers==2.11.0

Please install apex for the mixed precision training.
See details in requirements.txt

Data Download and Preprocessing

1. Download dataset

Please run this script at first. It will create data repository, and save and preprocess MultiWOZ 2.1 dataset.

python3 create_data.py

2. Preprocessing

To train NeuralWOZ under various settings, you should create each training instances with running below script.

python3 neuralwoz/preprocess.py --exceptd $TARGET_DOMAIN --fewshot_ratio $FEWSHOT_RATIO
  • exceptd: Specify "target domain" to exclude from training dataset for leave-one-out scheme. It is one of the (hotel|restaurant|attraction|train|taxi).
  • fewshot_ratio: Choose proportion of examples in the target domain to include. Default is 0. which means zero-shot. It is one of the (0.|0.01|0.05|0.1). You can check the fewshot examples in the assets/fewshot_key.json.

This script will create "$TARGET_DOMAIN_$FEWSHOT_RATIO_collector_(train|dev).json" and "$TARGET_DOMAIN_$FEWSHOT_RATIO_labeler_train.h5".

Training NeuralWOZ

You should specify output_path to save the trained model.
Each output consists of the below four files after the training.

  • pytorch_model.bin
  • config.json
  • vocab.json
  • merges.txt

For each zero/few-shot settings, you should set the TRAIN_DATA and DEV_DATA from the preprocessing. For example, hotel_0.0_collector_(train|dev).json should be used for the Collector training when the target domain is hotel in the zero-shot domain transfer task.

We use N_GPU=4 and N_ACCUM=2 for Collector training and N_GPU=2 and N_ACCUM=2 for Labeler training to fit 32 for batch size based on V100 32GB GPU.

1. Collector

python3 neuralwoz/train_collector.py \
  --dataset_dir data \
  --output_path $OUTPUT_PATH \
  --model_name_or_path facebook/bart-large \
  --train_data $TRAIN_DATA \
  --dev_data $DEV_DATA \
  --n_gpu $N_GPU \
  --per_gpu_train_batch_size 4 \
  --num_train_epochs 30 \
  --learning_rate 1e-5 \
  --gradient_accumulation_steps $N_ACCUM \
  --warmup_steps 1000 \
  --fp16

2. Labeler

python3 neuralwoz/train_labeler.py \
  --dataset_dir data \
  --output_path $OUTPUT_PATH \
  --model_name_or_path roberta-base-dream \
  --train_data $TRAIN_DATA \
  --dev_data labeler_dev_data.json \
  --n_gpu $N_GPU \
  --per_gpu_train_batch_size 8 \
  --num_train_epochs 10 \
  --learning_rate 1e-5 \
  --gradient_accumulation_steps $N_ACCUM \
  --warmup_steps 1000 \
  --beta 5. \
  --fp16

Download Synthetic Dialogues from NeuralWOZ

Please download synthetic dialogues from here

  • The naming convention is nwoz_{target_domain}_{fewshot_proportion}.json
  • Each dataset contains synthesized dialogues from our NeuralWOZ
  • Specifically, It contains synthetic dialogues for the target_domain while excluding original dialogues for the target domain (leave-one-out setup)
  • You can check the i-th synthesized dialogue in each files with aug_{target_domain}_{fewshot_proprotion}_{i} for dialogue_idx key.
  • You can use the json file to directly train zero/few-shot learner for DST task
  • Please see readme for training TRADE and readme for training SUMBT using the dataset
  • If you want to synthesize your own dialogues, please see below sections.

Download Pretrained Models

Pretrained models are available in this link. The naming convention is like below

  • NEURALWOZ: (Collector|Labeler)_{target_domain}_{fewshot_proportion}.tar.gz
  • TRADE: nwoz_TRADE_{target_domain}_{fewshot_proportion}.tar.gz
  • SUMBT: nwoz_SUMBT_{target_domain}_{fewshot_proportion}.tar.gz

To synthesize your own dialogues, please download and unzip both of Collector and Labeler in same target domain and fewshot_proportion at $COLLECTOR_PATH and $LABELER_PATH, repectively.

Please use tar -zxvf MODEL.tar.gz for the unzipping.

Generate Synthetic Dialogues using NeuralWOZ

python3 neuralwoz/run_neuralwoz.py \
  --dataset_dir data \
  --output_dir data \
  --output_file_name neuralwoz-output.json \
  --target_data collector_dev_data.json \
  --include_domain $TARGET_DOMAIN \
  --collector_path $COLLECTOR_PATH \
  --labeler_path $LABELER_PATH \
  --num_dialogues $NUM_DIALOGUES \
  --batch_size 16 \
  --num_beams 1 \
  --top_k 0 \
  --top_p 0.98 \
  --temperature 0.9 \
  --include_missing_dontcare

License

Copyright 2021-present NAVER Corp.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Owner
NAVER AI
Official account of NAVER AI, Korea No.1 Industrial AI Research Group
NAVER AI
Plug-n-Play Reinforcement Learning in Python with OpenAI Gym and JAX

coax is built on top of JAX, but it doesn't have an explicit dependence on the jax python package. The reason is that your version of jaxlib will depend on your CUDA version.

128 Dec 27, 2022
Code for "ATISS: Autoregressive Transformers for Indoor Scene Synthesis", NeurIPS 2021

ATISS: Autoregressive Transformers for Indoor Scene Synthesis This repository contains the code that accompanies our paper ATISS: Autoregressive Trans

138 Dec 22, 2022
MonoRCNN is a monocular 3D object detection method for automonous driving

MonoRCNN MonoRCNN is a monocular 3D object detection method for automonous driving, published at ICCV 2021. This project is an implementation of MonoR

87 Dec 27, 2022
A (PyTorch) imbalanced dataset sampler for oversampling low frequent classes and undersampling high frequent ones.

Imbalanced Dataset Sampler Introduction In many machine learning applications, we often come across datasets where some types of data may be seen more

Ming 2k Jan 08, 2023
CTRL-C: Camera calibration TRansformer with Line-Classification

CTRL-C: Camera calibration TRansformer with Line-Classification This repository contains the official code and pretrained models for CTRL-C (Camera ca

57 Nov 14, 2022
Solving SMPL/MANO parameters from keypoint coordinates.

Minimal-IK A simple and naive inverse kinematics solver for MANO hand model, SMPL body model, and SMPL-H body+hand model. Briefly, given joint coordin

Yuxiao Zhou 305 Dec 30, 2022
Toward Spatially Unbiased Generative Models (ICCV 2021)

Toward Spatially Unbiased Generative Models Implementation of Toward Spatially Unbiased Generative Models (ICCV 2021) Overview Recent image generation

Jooyoung Choi 88 Dec 01, 2022
A template repository for submitting a job to the Slurm Cluster installed at the DISI - University of Bologna

Cluster di HPC con GPU per esperimenti di calcolo (draft version 1.0) Per poter utilizzare il cluster il primo passo รจ abilitare l'account istituziona

20 Dec 16, 2022
Customer-Transaction-Analysis - This analysis is based on a synthesised transaction dataset containing 3 months worth of transactions for 100 hypothetical customers.

Customer-Transaction-Analysis - This analysis is based on a synthesised transaction dataset containing 3 months worth of transactions for 100 hypothetical customers. It contains purchases, recurring

Ayodeji Yekeen 1 Jan 01, 2022
Implementation of Segformer, Attention + MLP neural network for segmentation, in Pytorch

Segformer - Pytorch Implementation of Segformer, Attention + MLP neural network for segmentation, in Pytorch. Install $ pip install segformer-pytorch

Phil Wang 208 Dec 25, 2022
Pytorch implementation for ACMMM2021 paper "I2V-GAN: Unpaired Infrared-to-Visible Video Translation".

I2V-GAN This repository is the official Pytorch implementation for ACMMM2021 paper "I2V-GAN: Unpaired Infrared-to-Visible Video Translation". Traffic

69 Dec 31, 2022
This is the repo for Uncertainty Quantification 360 Toolkit.

UQ360 The Uncertainty Quantification 360 (UQ360) toolkit is an open-source Python package that provides a diverse set of algorithms to quantify uncert

International Business Machines 207 Dec 30, 2022
Gradient Step Denoiser for convergent Plug-and-Play

Source code for the paper "Gradient Step Denoiser for convergent Plug-and-Play"

Samuel Hurault 11 Sep 17, 2022
Code for the Paper: Alexandra Lindt and Emiel Hoogeboom.

Discrete Denoising Flows This repository contains the code for the experiments presented in the paper Discrete Denoising Flows [1]. To give a short ov

Alexandra Lindt 3 Oct 09, 2022
Code for You Only Cut Once: Boosting Data Augmentation with a Single Cut

You Only Cut Once (YOCO) YOCO is a simple method/strategy of performing augmenta

88 Dec 28, 2022
PyTorch Implementation of Temporal Output Discrepancy for Active Learning, ICCV 2021

Temporal Output Discrepancy for Active Learning PyTorch implementation of Semi-Supervised Active Learning with Temporal Output Discrepancy, ICCV 2021.

Siyu Huang 33 Dec 06, 2022
Doosan robotic arm, simulation, control, visualization in Gazebo and ROS2 for Reinforcement Learning.

Robotic Arm Simulation in ROS2 and Gazebo General Overview This repository includes: First, how to simulate a 6DoF Robotic Arm from scratch using GAZE

David Valencia 12 Jan 02, 2023
UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning

UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning This is the official PyTorch implementation for UniMoCo pape

dddzg 49 Jan 02, 2023
Segmentation-Aware Convolutional Networks Using Local Attention Masks

Segmentation-Aware Convolutional Networks Using Local Attention Masks [Project Page] [Paper] Segmentation-aware convolution filters are invariant to b

144 Jun 29, 2022
meProp: Sparsified Back Propagation for Accelerated Deep Learning

meProp The codes were used for the paper meProp: Sparsified Back Propagation for Accelerated Deep Learning with Reduced Overfitting (ICML 2017) [pdf]

LancoPKU 107 Nov 18, 2022