Mapping actions and objects to tasks
View Patent ↗Techniques for mapping actions and objects to tasks may include identifying a task to be performed by a virtual assistant for an action and/or object. The task may be identified based on a task map of the virtual assistant. In some examples, the task may be identified based on contextual information of a user, such as a conversation history, content output history, user preferences, and so on. The techniques may also include customizing a task map for a particular context, such as a particular user, industry, platform, device type, and so on. The customization may include assigning an action, object, and/or variable value to a particular task.
1. One or more computer-readable storage media storing computer-readable instructions that, when executed, instruct one or more processors to perform operations comprising:
obtaining user input that is received at a smart device during a conversation between a user and a virtual assistant;
identifying context associated with at least one of the user or the smart device;
selecting, based at least in part on the context, a task map from a plurality of task maps that map a plurality of actions or a plurality of objects to tasks of the virtual assistant;
analyzing the user input to determine an action from the plurality of actions and an object from the plurality of objects;
determining, based at least in part on the task map, that at least one of the action or the object is mapped to a first task;
causing the first task to be at least partly performed by the virtual assistant;
determining that a second task was initiated by the user subsequent to the first task being performed; and
based at least in part on the second task being initiated by the user subsequent to the first task being performed, updating the task map by mapping at least one of the action or the object to the second task in the task map.
2. The one or more computer-readable storage media of claim 1 , wherein the action comprises a verb and the object comprises a noun.
3. The one or more computer-readable storage media of claim 1 , wherein the operations further comprise:
upon determining that the at least one of the action or the object is mapped to the first task, identifying a variable for performing the task; and
obtaining a value for the variable by at least one of causing the user to be prompted for further user input or analyzing the context.
4. The one or more computer-readable storage media of claim 1 , wherein the operations further comprise:
learning information to be associated with the first task based at least in part on the conversation between the user and the virtual assistant, the information comprising at least one of the action, the object, or a value of a variable for the first task.
5. The one or more computer-readable storage media of claim 1 , wherein the context associated with at least one of the user or the smart device comprises at least one of:
a particular industry associated with at least one of the user or the virtual assistant;
a platform of the smart device;
a type of the smart device;
a type of the user; or
a location associated with the user.
6. The one or more computer-readable storage media of claim 1 , the operations further comprising:
determining that the first task is associated with a confidence score higher than a threshold confidence score;
and wherein the causing the first task to be at least partly performed by the virtual assistant is based at least in part on determining that the first task is associated with the confidence score higher than the threshold confidence score.
7. A method comprising:
under control of one or more computing devices configured with executable instructions,
causing a conversation user interface to be output on a smart device to enable a conversation between a user of the smart device and a virtual assistant;
learning information to be associated with a first task to be performed by the virtual assistant based at least in part on the conversation of the virtual assistant with the user, the information comprising at least one of an action, an object, or a value of a task variable, wherein the learning comprises:
identifying input that is received from the user during the conversation;
analyzing the input to identify the action and the object;
identifying, based at least in part on a task map that maps a plurality of actions or a plurality of objects to tasks of the virtual assistant, a second task that is mapped to at least one of the action or the object;
performing the second task;
identifying that the user initiated the first task after the second task was performed;
based at least in part on identifying that the first task was initiated by the user, identifying at least one of the action or the object of the input to be associated with the first task that was initiated by the user; and
associating at least one of the action, the object, or the value of the task variable with the first task by mapping at least one of the action, the object, or the value of the task variable to the first task in the task map.
8. The method of claim 7 , wherein the learning further comprises:
identifying a request that is received from the user during the conversation, the request requesting to associate a particular term or phrase with the first task.
9. The method of claim 7 , wherein the input is received through a user interface that enables the user to associate a task with an action-object pair.
10. One or more computer-readable storage media storing computer-readable instructions that, when executed, instruct one or more processors to perform operations comprising:
identifying a context for configuring a task map of a virtual assistant, the task map mapping tasks to be performed by the virtual assistant to action-object pairs, the context comprising an industry to which the virtual assistant is to be deployed;
obtaining information related to the context, the information comprising one or more terms or phases that are used within the industry;
configuring the task map for the context by assigning at least one task of the tasks to a particular action-object pair based at least in part on the information related to the context;
receiving input from a user and via a conversation interface output on a device;
analyzing the input to identify an action-object pair;
identifying, based at least in part on the task map, a first task to which the action-object pair is mapped;
causing the virtual assistant to perform the first task;
determining that a second task was initiated by the user subsequent to the first task being performed; and
based at least in part on determining that the second task was initiated by the user subsequent to the first task being performed, updating the task map by mapping the object-action pair to the second task in the task map.
11. The one or more computer-readable storage media of claim 10 , wherein:
the context comprises a platform for which the virtual assistant is to be deployed; and
the information related to the context comprises one or more terms or phrases that are used in association with the platform.
12. The one or more computer-readable storage media of claim 10 , wherein:
the context comprises a device type for which the virtual assistant is to be deployed; and
the information related to the context comprises one or more terms or phrases that are used in association with the device type.
13. The one or more computer-readable storage media of claim 10 , wherein:
the context comprises a user for which the virtual assistant is to be deployed; and
the information related to the context comprises at least one of a current conversation history of the user with the virtual assistant, a previous conversation history of the user with the virtual assistant, input of the user that requests that the at least one task be associated with the particular action-object pair, content output history of the user that identifies content that has been output to the user, user preference information indicating one or more preferences of the user, or device information indicating a type of device that is used by the user.
14. A method comprising:
under control of one or more computing devices configured with executable instructions,
causing a conversation user interface to be output on a smart device associated with a user to enable a conversation between the user of the smart device and a virtual assistant;
receiving, via the user interface on the smart device, a phrase containing an action and an object;
identifying, based at least in part on a task map that maps a plurality of actions or a plurality of objects to tasks of the virtual assistant, a first task that is mapped to at least one of the action or the object;
performing the first task;
determining that the user initiated a second task after the first task was performed; and
based at least in part on determining that the user initiated the second task after the first task was performed, updating the task map by mapping at least one of the action or the object to the second task in the task map.
15. The method of claim 14 , further comprising:
identifying a variable for performance of the first task;
obtaining a value for the variable by at least one of (i) analyzing contextual information that is related to at least one of the conversation or the user or (ii) causing the user to be prompted for the value of the variable; and
updating the task map by mapping the value of the variable to the second task.
16. The method of claim 15 , wherein the obtaining the value for the variable comprises determining that a term or phrase in the contextual information corresponds to the value for the variable based on at least one of a word type or category of the term or phrase.
17. The method of claim 15 , wherein obtaining the value for the variable includes utilizing a goal-based dialog.
18. The method of claim 14 , wherein the virtual assistant is configured for at least one of multi-modal input/out, multi-language communication, or multi-channel communication.