ConvLab: Multi-Domain End-to-End Dialog System Platform

TL;DR

ConvLab is an open-source platform that facilitates research in multi-domain end-to-end dialog systems by providing reusable components, datasets, and pre-trained models for easy experimentation and comparison.

Contribution

It introduces ConvLab, a comprehensive platform with annotated datasets and pre-trained models, enabling efficient development and evaluation of multi-domain dialog systems.

Findings

Extended MultiWOZ dataset with user dialog act annotations.

Demonstrated ease of conducting complex multi-domain experiments.

Supported comparison of various dialog system approaches.

Abstract

We present ConvLab, an open-source multi-domain end-to-end dialog system platform, that enables researchers to quickly set up experiments with reusable components and compare a large set of different approaches, ranging from conventional pipeline systems to end-to-end neural models, in common environments. ConvLab offers a set of fully annotated datasets and associated pre-trained reference models. As a showcase, we extend the MultiWOZ dataset with user dialog act annotations to train all component models and demonstrate how ConvLab makes it easy and effortless to conduct complicated experiments in multi-domain end-to-end dialog settings.