Work

Roving Mind: a balancing act between open–domain and engaging dialogue systems

Year: 2017
Type: conference-paper
Abstract: Open–domain dialogue systems should be able to cover a very large set of domains \nand at the same time keep the user engaged in the interaction. Current approaches \nto dialogue modeling are divided between domain–independent, non–modular \napproaches using sequence–to–sequence models and the domain–specific modular \nsystems developed for task–based dialogue. Furthermore, user engagement in \ndialogue, addressed in this bot challenge, is a rather new research area. In this \nwork we describe Roving Mind, a dialogue system which combines domain– \nindependence with a modular architecture for open–domain spoken conversation, \nwith a specific module for user engagement. Our architecture takes a balanced \napproach between human expert design and data–driven approaches, adapting the \ntraditional task–based architecture relying on slots and intents to open–domain \nconversation using entities and domain-independent dialogue acts. The system was \ntested extensively during the Alexa Prize semifinals and received ratings from a large number of users, which allowed us to draw statistically relevant facts on how users interact with an open domain, non task-based social bot. Our experiments \nshow that on average users give higher ratings to task-driven conversations (using user–entertaining strategies) compared to completely open–topic conversations. We also find a correlation between cumulative sentiment (using sentiment and \ndialogue acts) and user ratings and argue that this could be investigated to estimate an error–signal for strategy computation.
Language: English
Open Access status: green
OpenAlex ID: W7014930978
FWCI: 0.198
Cites: 16
Cited by: 3
Related to: 10