Coordinating electronic personal assistants
A method includes receiving a representation of a spoken utterance, processing the representation of the spoken utterance to identify, from a number of candidate domains, a request and a serving domain, and routing the request to a personal assistant based on the request and the serving domain. Identification of the serving domain is based on one or more of a contextual state, a behavior profile of a speaker of the utterance, and a semantic content of the utterance.
1. A method implemented using one or more processors, the method comprising:
receiving a representation of a spoken utterance;
processing the representation of the spoken utterance to identify, from a plurality of candidate domains, a request and a serving domain based on one or more of:
a contextual state,
a behavior profile of a speaker of the utterance, and
a semantic content of the utterance;
identifying a plurality of candidate personal assistants based on the request and the serving domain and computing a match score of the request and each candidate personal assistant; and
routing the request to a personal assistant based on the request and the serving domain, in which the request is routed to the personal assistant having the maximum match score from among the candidates;
wherein a match score of a particular personal assistant is reduced if routing the request to the particular personal assistant would result in a pre-determined dangerous condition.
2. The method of claim 1 , in which the contextual state includes a location of the speaker.
3. The method of claim 1 , in which the contextual state includes a time of day.
4. The method of claim 1 , in which the contextual state includes a current weather condition.
5. The method of claim 1 , in which a contextual state includes calendar information of the speaker.
6. The method of claim 1 , in which a match score of a particular personal assistant is reduced if routing the request to the particular personal assistant would result in no action.
7. The method of claim 1 , in which one of the candidate personal assistants is operable to control functionality of a vehicle.
8. The method of claim 1 further comprising monitoring an interaction between the personal assistant and a speaker of the request during servicing of the request by the personal assistant.
9. The method of claim 8 wherein the monitoring includes determining when the personal assistant has finished servicing the request.
10. The method of claim 8 wherein the monitoring includes processing the interaction between the personal assistant and the speaker according to a user interaction model.
11. The method of claim 8 wherein routing the request to a personal assistant based on the request and the serving domain includes determining the personal assistant for servicing the request including processing the request according to a user interaction model.
12. The method of claim 8 wherein at least one personal assistant of the plurality of personal assistants is native to a vehicle and at least one personal assistant of the plurality of personal assistants is non-native to the vehicle.
13. The method of claim 12 wherein at least one of the personal assistants is associated with a smart phone or a smart speaker.
14. A system comprising:
an input for receiving a representation of a spoken utterance;
one or more processors for
processing the representation of the spoken utterance to identify, from a plurality of candidate domains, a request and a serving domain based on one or more of:
a contextual state,
a behavior profile of a speaker of the utterance, and
a semantic content of the utterance; and
identifying a plurality of candidate personal assistants based on the request and the serving domain and computing a match score of the request and each candidate personal assistant; and
an output for providing the request to one of a plurality of personal assistants based on the request and the serving domain, in which the request is routed to the personal assistant having the maximum match score from among the candidates,
wherein a match score of a particular personal assistant is reduced if routing the request to the particular personal assistant would result in a pre-determined dangerous condition.