Reducing driver distraction in spoken dialogue
View Patent ↗Methods and systems for reducing driver distraction and situation non-awareness related to a dialog of an automated dialog system in a vehicle. For a dialog policy learning session, driver distraction is introduced as an input into a penalty assigner that assesses dialog quality, and dialog acts are extended to include dialogs and dialog act presentation styles which reduce driver workload related to dialogs. The automated dialog system policy is developed during the learning process by optimizing the penalties, so that automated dialog workload is reduced in response to increased workload or anticipated workload on the driver. Methods and systems are presented for responding to both actual workload in regular vehicles as well as anticipated workload in autonomous vehicles.
1. A method for reducing user distraction associated with interaction with an automated dialogue system, the method comprising:
receiving, by a processor, a user workload parameter;
responsively to the user workload parameter, controlling the automated dialogue system to perform a system dialogue turn that reduces the user workload associated with interacting with the automated dialogue system, wherein the system dialogue turn includes a dialogue act selected from a group consisting of:
a workload-reducing dialogue act; and
a regular dialogue act modified by a workload-reducing dialogue modification;
detecting a user distraction level for a plurality of dialogues;
responsive to detecting the user distraction level, computing a penalty value for assessing a dialogue quality according to at least the user distraction level for a dialogue of the plurality of dialogues;
storing the penalty value in a database which includes the plurality of dialogues; and
computing a dialogue policy for a controller of the automated dialogue system according to the database.
2. The method of claim 1 , wherein the workload-reducing dialogue act comprises an act selected from a group consisting of:
a dialogue pause;
a suggestion of a dialogue pause;
a dialogue termination;
a suggestion of a dialogue termination.
3. The method of claim 2 , further comprising a spectrum of pause-handling and termination handling choices ranging by degree of user involvement.
4. The method of claim 1 , wherein the workload-reducing dialogue modification comprises a dialogue modification selected from a group consisting of:
a switch to voice modality;
breaking up compound prompts into items;
a change to a yes/no prompt; and
speeding up of a dialogue.
5. The method of claim 4 , wherein the speeding up of a dialogue act is in response to a predicted workload increase.
6. The method of claim 1 , wherein the user workload parameter includes a hazard warning of a hazard ahead of the vehicle, and wherein at least one of the workload reducing act or the workload reducing dialogue modification includes an immediate suspension of the dialogue.
7. The method of claim 6 , wherein the hazard warning includes a warning of a braking condition of another vehicle ahead of the vehicle.
8. A method for reducing user situation non-awareness associated with an automated dialogue system, the method comprising:
receiving, by a processor, a user anticipated workload parameter;
responsively to the user anticipated workload parameter, controlling the automated dialogue system to perform a system dialogue turn that reduces the user workload associated with interacting with the automated dialogue system, wherein the system dialogue turn includes a dialogue act selected from a group consisting of:
a workload-reducing dialogue act; and
a regular dialogue act modified by a workload-reducing modification;
detecting a user situation non-awareness level for a plurality of dialogues stored in a database;
responsive to detecting the user situation non-awareness level, computing a penalty value for assessing a dialogue quality according to at least the user situation non-awareness; and
computing a dialogue policy for a controller of the automated dialogue system according to the database.
9. The method of claim 8 , wherein the workload-reducing dialogue act comprises an act selected from a group consisting of:
a dialogue pause;
a suggestion of a dialogue pause;
a dialogue termination;
a suggestion of a dialogue termination.
10. The method of claim 9 , further comprising a spectrum of pause-handling and termination handling choices ranging by degree of user involvement.
11. The method of claim 6 , wherein the workload-reducing dialogue modification comprises a dialogue modification selected from a group consisting of:
a switch to voice modality;
breaking up compound prompts into items;
a change to a yes/no prompt; and
speeding up of a dialogue.
12. The method of claim 11 , wherein the speeding up of a dialogue act is in response to a predicted anticipated workload increase.
13. A dialogue system for reducing user distraction associated with interaction with automated dialogue, the system comprising:
a dialogue control unit;
a storage device containing a dialogue policy;
a workload estimation unit operative to:
receive a workload parameter indicative of a user workload;
compute a workload estimate;
input the workload estimate into the dialogue control unit;
perform a system dialogue turn that reduces the user workload associated with interacting with the automated dialogue system, wherein the system dialogue turn includes a dialogue act selected from a group consisting of:
a workload-reducing dialogue act; and
a regular dialogue act modified by a workload-reducing dialogue modification;
detect a user distraction level for a plurality of dialogues;
responsive to detecting the user distraction level, compute a penalty value for assessing a dialogue quality according to at least the user distraction level for a dialogue of the plurality of dialogues;
store the penalty value in a database which includes the plurality of dialogues; and
compute a dialogue policy for a controller of the automated dialogue system according to the database.
14. The dialogue system of claim 13 , wherein the workload parameter includes a hazard warning of a hazard ahead of the vehicle, and wherein the workload estimate includes the hazard warning.
15. The dialogue system of claim 14 , wherein the hazard warning includes a warning of a braking condition of another vehicle ahead of the vehicle.
16. A dialogue system for reducing user situation non-awareness associated with interaction with automated dialogue, the system comprising:
a dialogue control unit;
a storage device containing a dialogue policy;
an anticipated workload estimation unit operative to:
receive an anticipated workload parameter indicative of an anticipated user workload;
compute an anticipated workload estimate;
input the anticipated workload estimate into the dialogue control unit;
perform a system dialogue turn that reduces the user workload associated with interacting with the automated dialogue system, wherein the system dialogue turn includes a dialogue act selected from a group consisting of:
a workload-reducing dialogue act; and
a regular dialogue act modified by a workload-reducing modification; detect the user situation non-awareness level for a plurality of dialogues stored in a database;
responsive to detecting the user situation non-awareness level, compute a penalty value for assessing a dialogue quality according to at least the user situation non-awareness; and
compute a dialogue policy for the dialogue control unit according to the database.
17. The dialogue system of claim 16 , wherein the anticipated workload parameter includes a hazard warning of a hazard ahead of the vehicle, and wherein the anticipated workload estimate includes the hazard warning.
18. The dialogue system of claim 17 , wherein the hazard warning includes a warning of a braking condition of another vehicle ahead of the vehicle.