MACA: A Modular Architecture for Conversational Agents

TL;DR

MACA is a flexible, modular software architecture that simplifies building and extending dialogue systems, enabling rapid prototyping, multi-domain support, and easy data collection for conversational agents.

Contribution

This paper introduces MACA, a plug-and-play architecture that separates conversation domain from strategy, facilitating quick development and reproduction of dialogue systems across multiple domains.

Findings

Supports multiple domains with existing dialogue strategies

Enables rapid prototyping of dialogue systems

Provides tools for data collection on Amazon Mechanical Turk

Abstract

We propose a software architecture designed to ease the implementation of dialogue systems. The Modular Architecture for Conversational Agents (MACA) uses a plug-n-play style that allows quick prototyping, thereby facilitating the development of new techniques and the reproduction of previous work. The architecture separates the domain of the conversation from the agent's dialogue strategy, and as such can be easily extended to multiple domains. MACA provides tools to host dialogue agents on Amazon Mechanical Turk (mTurk) for data collection and allows processing of other sources of training data. The current version of the framework already incorporates several domains and existing dialogue strategies from the recent literature.