Dialogue may seem natural to you, but it involves a series of distinct, complex tasks in order for two or more people to interact smoothly. To recreate that for a conversational AI, such as the one to be featured in the ARI-SPRING robot, you need to emulate a dialogue system. A symplified workflow includes at least an Automated Speech Recognizer (to recognize words), followed by a Natural Language Understanding component (to decode the words and attach semantic meaning to them), a dialogue manager (to infer the intent of the speaker and prepare a response), a Natural Language Generator (selecting words with appropriate semantic meaning for the response), and finally a synthesizer (to make the response intelligible). That’s how complex that is! An introduction to these concepts is given in this SPRING Technical Seminar #3, entitled “Conversational AI for Human-Robot Interaction: an introduction”, by Prof. Oliver Lemon from Heriot Watt University on 10 June 2020 (part 1) and by Prof. Christian Dondrup from Heriot Watt University on 16 June 2020 (part 2).

For privacy reasons YouTube needs your permission to be loaded. For more details, please see our Privacy policy.
I Accept
For privacy reasons YouTube needs your permission to be loaded. For more details, please see our Privacy policy.
I Accept