Frequently Asked Questions

Building the Companion

Is it a robot?

The Companions project is not concerned with robotics and we do not envisage to use a Companion like a robot in the immediate future. The difference is that a robot is usually designed to perform one or more tasks (paint a car, cut a machined piece of metal, clean a floor), while the Companion is designed to interact with the user, develop a persistent representation of the information provided, and allow the user access to various kinds of information.

However, there is no reason why a future project could not attempt to combine Companions and robots into a more complex product.

What will it look like?

The Companion has been intentionally designed to be able to be 'embodied' in a variety of objects. It could be an avatar on your computer screen or mobile phone, it could be embodied in an object like Nabaztag, or it might even be embodied in something like a handbag that you can carry around.

The most appropriate form of embodiment will depend on the intended user and purpose. We expect to develop a number of embodiments during the course of the project.

Why is making a Companion difficult?

First, because we are using different technologies which until now have not worked together. Secondly, each technology currently has serious limitations (e.g. speech recognition is known to work less well if there is surrounding noise, or with large vocabularies): part of the challenge is using the strengths of one technology to compensate for the weaknesses in another. Finally, if making a Companion was easy, it would not be a research project.

What is 'natural language'?

This is the kind of language that ordinary people use when they write or speak with each other. The term is used by computer scientists to distinguish this from programming languages, which are used to write computer programs. Natural Language Processing, a core research area of Companions, concerns making software which can process written language by extracting information, finding answers, automatically summarising etc.

What is an 'Embodied Conversational Agent' (ECA)?

An ECA is a cartoon-like character, or avatar, controlled by software which is able to communicate with the user using speech, gestures, facial expressions, etc. They provide a sophisticated user interface for accessing information, and generally allowing the user to interact with a computer.

Historically, most ECA research has focused on the visual design and gesture of the agent and has taken for granted that speech and dialogue systems would be available off the shelf. In Companions, we take an integrated approach where the design of the dialogue and speech components is intended to complement and interact with the ECA in a systematic manner.

Why is the European Commission funding this research project?

The European Commission has a series of research funding 'Frameworks' which aim to increase the competitiveness of European research and development, facilitate and increase collaboration between researchers in the many different EU countries and support the transfer of technology from research organisation such as universities and institutes to industry.

The Companions Project fits into these objectives as part of the Information Society Technologies thematic area, and is one of the EC's largest multimedia multimodal interface projects.