Spoken dialogue system, a spoken dialogue method and a method of adapting a spoken dialogue system

A spoken dialogue system comprising:

• an input configured to receive data relating to a speech signal originating from a user, where the speech signal forms a part of a dialogue;
• a processor configured to:
• extract one or more acoustic features from the input speech signal;
• update a system state based on the input speech signal using a state tracker model;
• determine a system action using a policy model, wherein the updated system state is the input to the policy model, and a reward function depends on the system action and the system state;
• generate a success measure using one or more classifiers inputting feature vectors including the acoustic features, wherein the success measure as an input is included in the reward function;
• calculate a reward value R of the reward function using the success measure, the reward function being calculated by formula:
• R={Y×(success measure)}−N,
• wherein N is a total number of dialog turns that have actually occurred and Y is a pre-determined factor;
• and
• update the policy model using the reward value;
• and
• an output configured to output information specified by the system action;
• wherein a feature vector is generated using the acoustic features, and inputted into one or more classifiers, and
• the one or more classifiers are configured to output the success measure.