Developing CereChat and Collaborating with the University of Nottingham

Last week, Paola and Ewan, from the CereProc team, traveled to Nottingham for a one-day workshop on our latest project: CereChat. This innovative technology facilitates unscripted conversations with an agent, enabling fast, human-like responses. The team met up with researchers from the University of Nottingham, our collaborators on this AI project at the Cobot Maker Space, a great environment for team meetings. I had the chance to interview CereProc’s own Paola and Ewan to learn more about their trip.

Can you tell us a little more about this project?

Absolutely! We have been developing a system that actively listens to a speaker and facilitates unscripted conversations. The goal is to create a system where a robot such as CereBot, a robot based on the MiRo robot, understands what the user is saying and generates a response before the user has even finished speaking, allowing for faster turn taking and a more natural sounding response.

The teams have been working on a demo using our conversational system, CereChat, for a major Japanese corporation, that has partnered with CereProc on numerous occasions. For the demo, CereChat has been integrated with CereBot, to show a full range of dynamic interactions.

What has the collaboration with the University of Nottingham involved?

Thanks to their amazing work, we’ve integrated movements into the CereBot robot, enhancing the dynamic conversational experience. For example, our CereBot robot can blink, rotate its head, and move around while conversing, making interactions more human-like.

What was the best part of the workshop?

“The trip was a great work experience and it is nice to meet face to face with colleagues you have been communicating with on a project”, Ewan and Paola said in unison. The workshop’s purpose was to facilitate an in-person meeting to review our progress and plan the next steps. Both teams presented their work and engaged in informative discussions about our progress.

Receiving positive feedback about our progress was encouraging, and we left the workshop with a clear understanding of the future direction of the project.

What have been the main challenges of working on CereChat?

Ewan: “It was my first time working with robots, and learning a new Python-based language, ROS, was particularly challenging.”

Paola: “The biggest challenge has been implementing a system where users can interrupt the bot, or vice versa. This is a complex task for any conversational system, and developing this capability has been quite challenging.”

What are the next steps?

Moving forward, we will merge the work from both teams into our existing demo and focus on making incremental improvements. We aim to enhance the response time and expand the range of topics the CereBot robot can discuss. Overall, we strive to make our demo as modular and dynamic as possible.