In this paper we will propose the use of social robots as interface between users and services in a Smart Environment. We will focus on the need for a robot to recognize the user’s feedback, in order to respond and revise its behaviour according to user’s needs. As we believe speech is a natural and immediate input channel in human-robot interaction, we will discuss the importance of recognising, besides the linguistic content of the spoken sentence, the attitude of the user towards the robot and the environment. In this way, the meaning of the user dialog will be made clear when hardly recognisable by the analysis of the utterance structure. Then, we will present the results of the application of a potential approach used for integrating the linguistic analysis with the recognition of the valence and arousal of the user’s utterance. In order to achieve this goal, we collected and analysed a corpus of data to build an interpretation model based on a Bayesian network. Then we tested the accuracy of the model using a test dataset. Results will show that the integration of the linguistic content with the recognition of some acoustic features of spoken sentences perform better in recognising the key aspects of user’s feedback.
Interpretation of User’s Feedback in Human-Robot Interaction
DE CAROLIS, Berardina;
2009-01-01
Abstract
In this paper we will propose the use of social robots as interface between users and services in a Smart Environment. We will focus on the need for a robot to recognize the user’s feedback, in order to respond and revise its behaviour according to user’s needs. As we believe speech is a natural and immediate input channel in human-robot interaction, we will discuss the importance of recognising, besides the linguistic content of the spoken sentence, the attitude of the user towards the robot and the environment. In this way, the meaning of the user dialog will be made clear when hardly recognisable by the analysis of the utterance structure. Then, we will present the results of the application of a potential approach used for integrating the linguistic analysis with the recognition of the valence and arousal of the user’s utterance. In order to achieve this goal, we collected and analysed a corpus of data to build an interpretation model based on a Bayesian network. Then we tested the accuracy of the model using a test dataset. Results will show that the integration of the linguistic content with the recognition of some acoustic features of spoken sentences perform better in recognising the key aspects of user’s feedback.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
