Virtual assistant acquisitions and training
Techniques and architectures for implementing a team of virtual assistants are described herein. The team may include multiple virtual assistants that are configured with different characteristics, such as different functionality, base language models, levels of training, visual appearances, personalities, and so on. The characteristics of the virtual assistants may be configured by trainers, end-users, and/or a virtual assistant service. The virtual assistants may be presented to end-users in conversation user interfaces to perform different tasks for the users in a conversational manner. The different virtual assistants may adapt to different contexts. The virtual assistants may additionally, or alternatively, interact with each other to carry out tasks for the users, which may be illustrated in conversation user interfaces.
1. A method comprising:
presenting a virtual assistant trainer interface, the virtual assistant trainer interface including information regarding multiple virtual assistants;
receiving a user selection of one of the multiple virtual assistants;
based at least in part on the user selection, identifying, by a first computing device, a particular virtual assistant of the multiple virtual assistants for a first user to train;
associating the particular virtual assistant with an account of the first user;
obtaining, by the first computing device, a trained version of the virtual assistant from the first user by:
causing the particular virtual assistant to be output via a second computing device that is associated with the account to facilitate a conversation between the particular virtual assistant and the first user;
receiving first audio input during the conversation;
determining a concept of the first audio input;
determining a response to the first audio input based at least in part on the concept and a base language model that is associated with the particular virtual assistant;
causing the particular virtual assistant to output the response;
receiving second audio input during the conversation;
based at least in part on the second audio input, determining that the response is not an accurate response for the first audio input; and
based at least in part on the second audio input, modifying the base language model that is associated with the particular virtual assistant;
evaluating, by the first computing device, the trained version of the particular virtual assistant upon completion of the updating, the evaluating being based at least in part on subjecting the trained version of the particular virtual assistant to one or more tests; and
offering, by the first computing device, the trained version of the particular virtual assistant for acquisition at a value to a second user to be implemented within a conversation user interface.
2. The method of claim 1 , wherein evaluating the trained version of the particular virtual assistant further comprises:
determining a rating for the first user based at least in part on the evaluation; and
associating the rating with a user profile of the first user.
3. The method of claim 1 , further comprising:
providing compensation to the first user based at least in part on the evaluation of the trained version of the particular virtual assistant.
4. The method of claim 1 , further comprising:
determining a number of acquisitions of the trained version of the particular virtual assistant by users; and
providing compensation to the first user based at least in part on the number of acquisitions of the trained version of the particular virtual assistant by users.
5. The method of claim 1 , wherein the base language model comprises related units of inputs mapped to responses, tasks, or a combination thereof, the relation of the units in the base language model being based at least in part on the concept.
6. The method of claim 1 , wherein the subjecting the trained version of the particular virtual assistant to one or more tests being performed via an automated manner, the one or more tests comprising tests that ensure the particular virtual assistant satisfies one or more criteria.
7. A system comprising:
one or more processors; and
memory communicatively coupled to the one or more processors and storing executable instructions that, when executed by the one or more processors, cause the one or more processors to perform acts comprising:
causing display of a virtual assistant trainer interface to enable one or more trainers to train a virtual assistant, the virtual assistant trainer interface displaying a set of concepts utilized by the virtual assistant to interpret user input;
receiving, from the one or more trainers via the virtual assistant trainer interface, input regarding configuration of the set of concepts;
based at least in part on the input from the one or more trainers, causing the set of concepts to be configured to obtain a trained version of the virtual assistant, the causing the set of concepts to be configured including at least one of modifying the base language model; and
causing display of a virtual assistant agency interface to offer the trained version of the virtual assistant for acquisition for a value, the virtual assistant agency interface displaying a level of training associated with the trained version of the virtual assistant, the level of training being based at least in part on information identifying qualifications of the one or more trainers and a count of how many of the one or more trainers were associated with the training of the trained virtual assistant.
8. The system of claim 7 , wherein the virtual assistant trainer interface displays a previous conversation of the virtual assistant with a user; and the acts further comprise:
receiving a rating from the one or more trainers regarding a response of the virtual assistant that was provided during the previous conversation; and
causing a characteristic of the virtual assistant to be configured based at least in part on the rating for the response of the virtual assistant that was provided during the previous conversation.
9. The system of claim 7 , wherein the acts further comprise enabling the one or more trainers to configure at least one of a task mapping associated with the virtual assistant, a base language model associated with the virtual assistant, functionality associated with the virtual assistant, a visual appearance of the virtual assistant, an audible manner of output of the virtual assistant, a level of security associated with the virtual assistant, or a language in which the virtual assistant communicates.
10. The system of claim 7 , wherein:
the virtual assistant trainer interface displays a base language model of the virtual assistant; and the acts further comprise:
receiving input regarding training of the virtual assistant, the input indicating how to configure the base language model of the virtual assistant; and
causing the base language model to be configured.
11. The system of claim 10 , wherein the base language model maps a response to a set of user inputs.
12. The system of claim 7 , wherein the acts further comprise:
receiving a user selection of the virtual assistant;
associating the virtual assistant with an account that is associated with the one or more trainers;
causing the virtual assistant to be output via a computing device that is associated with the account to facilitate a conversation between the virtual assistant and the one or more trainers;
receiving first audio input during the conversation;
determining a concept of the first audio input;
determining a response to the first audio input based at least in part on the concept;
causing the virtual assistant to output the response;
receiving second audio input during the conversation;
based at least in part on the second audio input:
determining that the response is not an accurate response for the first audio input; and
determining a correction to the response;
updating a base language model that is associated with the virtual assistant based at least in part on the correction to the response;
receiving an indication that the virtual assistant is trained; and
offering the virtual assistant for acquisition.
13. The system of claim 12 , wherein the acts further comprise:
before receiving the indication that the virtual assistant is trained, receiving third audio input during another conversation;
determining another response to the third audio input;
causing the virtual assistant to output the other response;
receiving fourth audio input during the other conversation;
based at least in part on the fourth audio input:
determining that the other response is not an accurate response for the third audio input; and
determining a correction to the other response;
updating the base language model associated with the virtual assistant based at least in part on the correction to the other response.
14. The system of claim 7 , wherein the level of training of the trained version of the virtual assistant is based at least in part on an amount of training performed by the one or more qualified trainers.
15. They system of claim 7 , wherein the level of training is based at least in part on an evaluation of the trained version of the virtual assistant, the evaluation being performed by one or more automated tests ensuring the particular virtual assistant satisfies one or more criteria.
16. One or more non-transitory computer storage media storing computer-readable instructions that, when executed, instruct one or more processors to perform operations comprising:
causing display of a virtual assistant trainer interface to enable one or more trainers to train a virtual assistant, the virtual assistant trainer interface displaying a set of concepts utilized by the virtual assistant to interpret user input;
receiving, from the one or more trainers via the virtual assistant trainer interface, input regarding configuration of the set of concepts;
based at least in part on the input, causing the base language model to be configured to obtain a trained version of the virtual assistant;
evaluating the trained version of the virtual assistant based at least in part by subjecting the trained version of the virtual assistant to one or more tests; and
causing display of a virtual assistant agency interface to offer the trained version of the virtual assistant for acquisition, the virtual assistant agency interface displaying a level of training associated with the trained version of the virtual assistant and information identifying a rating of the one or more trainers that trained the virtual assistant, the level of training of the trained version of the virtual assistant being based at least in part on an amount of training that is received from the one or more trainers and the rating of the one or more trainers.
17. The one or more non-transitory computer storage media of claim 16 , wherein the virtual assistant trainer interface displays a previous conversation of the virtual assistant with a user; and the operations further comprise:
receiving a rating from the one or more trainers regarding a response of the virtual assistant that was provided during the previous conversation; and
causing a characteristic of the virtual assistant to be configured based at least in part on the rating for the response of the virtual assistant that was provided during the previous conversation.
18. The one or more non-transitory computer storage media of claim 16 , wherein the operations further comprise enabling the one or more trainers to configure at least one of a task mapping associated with the virtual assistant, a base language model associated with the virtual assistant, functionality associated with the virtual assistant, a visual appearance of the virtual assistant, an audible manner of output of the virtual assistant, a level of security associated with the virtual assistant, or a language in which the virtual assistant communicates.
19. The one or more non-transitory computer storage media of claim 16 , wherein:
the virtual assistant trainer interface displays a base language model of the virtual assistant; and the operations further comprise:
receiving input to configure the base language model of the virtual assistant; and
causing the base language model to be configured.
20. The one or more non-transitory computer storage media of claim 19 , wherein the base language model maps a response to a set of user inputs.
21. The one or more non-transitory computer storage media of claim 17 , wherein the operations further comprise receiving a reason for the rating for the response of the virtual assistant that was provided during the previous conversation.