ITEE Innovation Expo 2008 : Project Details

Language and space: bio-inspired robot languages

As robots are increasingly integrated into everyday life, it is becoming more important for ordinary users to be able to communicate with them, without needing to learn specialised vocabulary or signals. Ideally, users would be able to use natural language (such as English) and be understood by the robots.

Through the work of the RatChat project, robots are able to develop a limited artificial form of language, dealing with place names and prepositions and grounded in their exploration of the real world. Adding the ability for the robots to use verbs in their language increases the possible communications that can be accomplished. This thesis project looks at the nature of verbs and previous efforts to develop the use of verbs in robots, in order to design a suitable 'mental' representation for verb concepts. This representation is based on the 'most information' lexical strategy already used in RatChat. To test the viability of the design, a language game was run which relied on the linguistic communication of information in order for the robots to complete the activity.

The final representation design builds heavily on work done by Ruth Schulz and Steven Pinker, and resulted in the robots achieving approximately 100% accuracy in guessing the meaning of the verb spoken by another robot. Further work to be done includes the addition of a greater number of verb concepts, dealt with at varying levels of detail.