Motivation

In our work, we have built Conversational Interfaces for tutoring purposes in a variety of situations. Through this work, we have encountered situations which highlight the shortcomings of techniques developed for state of the art Conversational Interfaces, like Spoken Dialog Systems. Such situations arise due to the unique new challenges posed by the need for next generation of Conversational Agents. These agents should have the ability to:

• Indulge in conversation with multiple human users simultaneously on the same floor (e.g. in collaborative situations)
• Demonstrate socially motivated behaviors like affect, small-talk, politeness strategies, opportunism, persuasiveness/manipulativeness, ...
• Maintain relationships with users over long-term and demonstrate personality traits
• Be Ubiquitous via reach-ability through interfaces ranging from telephones to webpages to virtual worlds to generic hardware devices like mobiles and toys/robots

By identifying these shortcomings in current systems, we realize that the scope of their use extends beyond tutoring systems. Some clear examples where agents with such abilities could be useful besides tutoring includes agents as secretaries / receptionists, as shopping / sales assistants, applications for collaborative work, games/entertainment, therapy, elder care, ... among many others.

In a general case, these applications call for a new generation of conversational interfaces that would be situated deeply into people's everyday activites.

Introduction and Scope

Through this work, we intend to review existing work on building Conversational Interfaces which demand such abilities and in the process create an architectural framework that would allow researchers and engineers to

• Build the next generation of Conversation Interfaces which make use of the abilities listed above
• Achieve higher degree of flexibility in designing systems, including aspects of architecture such as communication and coordination, without needing to start from scratch
• Build such systems from Reusable Components and distribute component development among team members
• Build systems that can intelligently integrate and function in multiple environments / channels

In the pursuit of this goal, we have been working on our framework prototype - Basilica (currently standing identifier). You can refer to our Publications and Research Notes for more details as this thought evolves.

Disclaimer