Neural Generation Meets Real People: Building a Social, Informative Open-Domain Dialogue Agent

TL;DR

This paper introduces Chirpy Cardinal, a social chatbot that combines neural generation with hand-crafted dialogue to create engaging, emotionally intelligent conversations, achieving high user ratings in a competitive setting.

Contribution

It presents a novel hybrid approach integrating neural and rule-based methods for open-domain social dialogue systems.

Findings

Handled thousands of conversations daily

Achieved second place in Alexa Prize Socialbot Challenge

Received an average user rating of 3.58 out of 5

Abstract

We present Chirpy Cardinal, an open-domain social chatbot. Aiming to be both informative and conversational, our bot chats with users in an authentic, emotionally intelligent way. By integrating controlled neural generation with scaffolded, hand-written dialogue, we let both the user and bot take turns driving the conversation, producing an engaging and socially fluent experience. Deployed in the fourth iteration of the Alexa Prize Socialbot Grand Challenge, Chirpy Cardinal handled thousands of conversations per day, placing second out of nine bots with an average user rating of 3.58/5.