System and method for conversational agent via adaptive caching of dialogue tree

Machine readable and non-transitory medium having information recorded thereon for managing a user machine dialogue, wherein the information, when read by the machine, causes the machine to perform:

• receiving, at a device, sensor data including an utterance representing a speech of a user engaged in a dialogue with the device and information related to a current state of the dialogue;
• determining the speech of the user based on the utterance;
• searching, by a local dialogue manager residing on the device, a sub-dialogue tree stored on the device for an acoustic response answering the utterance from the user, wherein the sub-dialogue tree is previously created by a server by carving out from a dialogue tree generated prior to the dialogue to predict content of dialogues, the local dialogue manager is previously generated by the server in accordance with the sub-dialogue tree for driving the dialogue on the device, without involving the server, within a scope defined by the sub-dialogue tree;
• rendering the acoustic response answering the utterance from the user in response to the speech, if the response is identified from the sub-dialogue tree;
• and
• sending, if the response is not available in the sub-dialogue tree, a request to the server for the response, an updated sub-dialogue tree, and an updated local dialogue manager, wherein the updated sub-dialogue tree is to be carved out by the server from the dialogue tree based on the response and the current state of the dialogue to predict future content of the dialogue, the updated local dialogue manager is created according to the updated sub-dialogue tree for driving, on the device without involving the server, the dialogue based on the predicted future content of the dialogue represented by the updated local sub-dialogue tree.

A system for managing a user machine dialogue, the method comprising:

• a sensor analyzer in a device implemented on a processor and configured for receiving, at a device, sensor data including an utterance representing a speech of a user engaged in a dialogue with the device and information related to a current state of the dialogue;
• a surrounding information understanding unit configured for determining the speech of the user based on the utterance;
• a local dialogue manager residing on the device and configured for searching a sub-dialogue tree stored on the device for an acoustic response answering the utterance from the user, wherein the sub-dialogue tree is previously created by a server by carving out from a dialogue tree generated prior to the dialogue to predict content of dialogues, the local dialogue manager is previously generated by the server in accordance with the sub-dialogue tree for driving the dialogue on the device, without involving the server, within a scope defined by the sub-dialogue tree;
• a response rendering unit configured for rendering the acoustic response answering the utterance from the user in response to the speech, if the response is identified from the sub-dialogue tree;
• and
• a device/server coordinator configured for sending, if the response is not available in the sub-dialogue tree, a request to the server for the response, an updated sub-dialogue tree, and an updated local dialogue manager, wherein the updated sub-dialogue tree is to be carved out by the server from the dialogue tree based on the response and the current state of the dialogue to predict future content of the dialogue, the updated local dialogue manager is created according to the updated sub-dialogue tree for driving, on the device without involving the server, the dialogue based on the predicted future content of the dialogue represented by the updated local sub-dialogue tree.