WO2019083602A1 - Crowdsourced training for commands matching - Google Patents

Crowdsourced training for commands matching

Info

Publication number
WO2019083602A1
WO2019083602A1 PCT/US2018/048064 US2018048064W WO2019083602A1 WO 2019083602 A1 WO2019083602 A1 WO 2019083602A1 US 2018048064 W US2018048064 W US 2018048064W WO 2019083602 A1 WO2019083602 A1 WO 2019083602A1
Authority
WO
WIPO (PCT)
Prior art keywords
digital assistant
action
command
assistant device
determined
Prior art date
Application number
PCT/US2018/048064
Other languages
French (fr)
Inventor
Rajat Mukherjee
Sunil Patil
Mark Robinson
Original Assignee
Aiqudo, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to US201762576766P priority Critical
Priority to US62/576,766 priority
Priority to US15/984,122 priority
Priority to US15/984,122 priority patent/US20180336049A1/en
Priority to US16/110,103 priority patent/US20180366108A1/en
Priority to US16/110,103 priority
Application filed by Aiqudo, Inc. filed Critical Aiqudo, Inc.
Publication of WO2019083602A1 publication Critical patent/WO2019083602A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/16Sound input; Sound output

Abstract

Embodiments described herein are generally directed towards systems and methods relating to a crowd-sourced digital assistant system and related methods. In particular, embodiments facilitate the intuitive creation, maintenance, and distribution of action datasets that include computing events or tasks that can be reproduced when an associated command is determined received by a digital assistant device. In various implementations, multiple action datasets may be determined associated with a received command and, as such, the digital assistant device or the digital assistant server can determine one or more action datasets that are most relevant to a particular user of the digital assistant device based on contextual data collected by the digital assistant device. In further implementations, the collected contextual data can be maintained by the digital assistant device, the digital assistant server, or a combination thereof.

Description

CROWDSOURCED TRAINING FOR COMMANDS MATCHING

BACKGROUND OF THE INVENTION

Digital assistants have become ubiquitously integrated into a variety of consumer electronic devices. Modern day digital assistants employ speech recognition technologies to provide a conversational interface between users and electronic devices. These digital assistants can employ various algorithms, such as natural language processing, to improve interpretations of commands received from a user. Consumers have expressed various frustrations with conventional digital assistants due to privacy concerns, constant misinterpretations of spoken commands, unavailability of services due to weak signals or a lack of signal, and the general requirement that the consumer must structure their spoken command in a dialect that is uncomfortable for them.

SUMMARY OF THE INVENTION

This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used in isolation as an aid in determining the scope of the claimed subject matter.

Embodiments described in the present disclosure are generally directed towards systems and methods relating to a crowd-sourced digital assistant for computing devices. In particular, embodiments facilitate the intuitive creation and distribution of action datasets that include reproducible computing events and associated commands that can be employed to invoke a reproduction of the computing events on computing devices having the crowd-sourced digital assistant installed and/or executing thereon. In some further embodiments, action datasets created by and/or distributed to such computing devices can be intelligently selected for execution based on a received command (e.g., a voice command). In some aspects, a digital assistant server can intelligently select an appropriate action dataset for distribution based on a received command. In some other aspects, a digital assistant device can intelligently select an appropriate action dataset for execution based on a received command. The described embodiments describe a digital assistant system that can perform any operation on a computing device by way of a received command, the operations being limited only by the various operations executable on the computing device.

In accordance with embodiments described herein, the described digital assistant and corresponding system provides an ever-growing and evolving library of dialects that enables the digital assistant to learn from its users, in contrast to the frustrating and limited interpretation features provided by conventional digital assistants. Further, because the digital assistant and corresponding system is configured with a framework for distributing improvements to its collection of actionable operations and understandable commands, and because the digital assistant utilizes applications existing on the computing device of each user, privacy concerns typically associated with conventional digital assistants is significantly reduced.

BRIEF DESCRIPTION OF THE DRAWING

The present invention is described in detail below with reference to the attached drawing figures, wherein:

FIG. 1 is a block diagram of an exemplary computing environment for a crowd-sourced digital assistant system, in accordance with embodiments of the present invention;

FIG. 2 is a block diagram of an exemplary digital assistant device, in accordance with an embodiment of the present disclosure;

FIG. 3 is a block diagram of an exemplary digital assistant server, in accordance with an embodiment of the present disclosure;

FIG. 4 is an exemplary data structure of an action dataset, in accordance with an embodiment of the present disclosure;

FIG. 5 is a flow diagram showing a method for performing actionable digital assistant operations retrieved via a crowd-sourced digital assistant network, according to various embodiments of the present invention;

FIG. 6 is another flow diagram showing a method for performing actionable digital assistant operations retrieved via a crowd-sourced digital assistant network, according to various embodiments of the present invention; and

FIG. 7 is a block diagram of an exemplary computing environment suitable for use in implementing embodiments of the present invention. DETAILED DESCRIPTION OF THE INVENTION

The subject matter of the present invention is described with specificity herein to meet statutory requirements. However, the description itself is not intended to limit the scope of this patent. Rather, the inventors have contemplated that the claimed subject matter might also be embodied in other ways, to include different steps or combinations of steps similar to the ones described in this document, in conjunction with other present or future technologies. Moreover, although the terms "step" and/or "block" may be used herein to connote different elements of methods employed, the terms should not be interpreted as implying any particular order among or between various steps herein disclosed unless and except when the order of individual steps is explicitly described.

Aspects of the technology described herein are generally directed towards systems and methods for crowdsourcing actionable operations of a digital assistant. The described embodiments facilitate the creation, on-boarding, and distribution of action datasets to any number of computing devices having an instance of the digital assistant installed and/or executing thereon (hereinafter referenced as a "digital assistant device"). In accordance with the present disclosure, an "operation" can correspond to a final result, output, or computing operation that is generated, executed, or performed by a digital assistant device based on one or more action datasets selected and interpreted for execution by the digital assistant device, each action dataset being comprised of one or more reproducible computing events that can be invoked in response to a received command determined to correspond to the action dataset. In accordance with embodiments described herein, an "action" is described in reference to an operation that is performed in response to an action dataset selected and interpreted for execution. In this regard, an action can be performed, invoked, initiated, or executed, among other things, and any reference to the foregoing can imply that a corresponding action dataset is selected and interpreted for execution by the digital assistant device to perform the corresponding operation.

In some embodiments, actions (or the action datasets corresponding thereto) can be created, by the digital assistant device, which can record a series of detected events (e.g., inputs) that are typically provided by a user of the digital assistant device when manually invoking the desired operation (e.g., with manual inputs via a touchscreen or other input method of the digital assistant device). That is, to create a new action dataset, the digital assistant device can invoke a recording mode where a user can simply perform a series of computing operations (e.g., manual touches, click inputs) within one or more applications to achieve a desired result or operation. After the recording is stopped by the user, via a terminating input, the action dataset can store and be associated with a set of command templates corresponding to commands that the user would preferably announce to the digital assistant device when an invocation of the operation is desired. In various embodiments, a command representation can be received as speech data and converted to text (e.g., by a speech engine of the digital assistant device), or received as text input data. In accordance with embodiments described herein, a "command" is referenced herein to describe data, received as speech data or as text data. A "command representation," on the other hand is referenced to describe text data that is received, based on inputs (e.g., keyboard), received speech data converted to text data, or received text data communicated from another computing device. A "command template" is referenced herein to describe a portion of a command representation having defined parameter fields in place of variable terms.

In more detail, one or more terms or keywords in the received command can be defined as a parameter based on input(s) received from the user. A parameter, in accordance with the present disclosure, can be referenced as corresponding to one of a plurality of predefined parameter types, such as but not limited to, genre, artist, title, location, name or contact, phone number, address, city, state, country, day, week, month, year, and more. It is also contemplated that the digital assistant device can access from a memory, or retrieve (e.g., from a server), a set of predefined parameter types that are known or determined to correspond to the application or applications for which an action dataset is being created. In some embodiments, the set of predefined parameter types can be determined based at least in part on corresponding application identifying information. The digital assistant device can extract, based on the defined parameters, the corresponding keywords and generate a command template based on the remaining terms and the defined parameters. By way of example only, if the command was originally received as "play music by Coldplay," and the term "Coldplay" is defined as a parameter of type "artist," a resulting command template generated by the digital assistant device may appear as "play music by <artist>". In this regard, a command template may include the originally received command terms if no parameters are defined, or may include a portion of the originally received command terms with parameter fields defined therein, the defined parameters corresponding to variable terms of a command. The digital assistant device can receive, among other things, application identifying information, a recorded series of events, and a set command templates, among other things, to generate a new action dataset that can be retrieved, interpreted and/or invoked by the digital assistant device, simply based on a determination, by the digital assistant device, that a received command or command representation is associated with the action dataset. When an action is invoked based on a determination that a received command or command representation corresponds to an action dataset, the digital assistant device can reproduce (e.g., emulate, invoke, execute, perform) the recorded series of events associated with the corresponding action dataset, thereby performing the desired operation. Moreover, in circumstances where a received command or command representation includes a parameter term, and a determination is made that the received command or command representation corresponds to an action dataset having a parameter field that also corresponds to the parameter term, the parameter term can be employed, by the digital assistant device, to perform custom operations while performing the action. For instance, the digital assistant device can input the parameter term as a text input into a field of the application.

In some further embodiments, an action dataset, once created by the digital assistant device, can be uploaded (hereinafter also referenced as "on-boarded") to a remote server for storage thereby. The action dataset can be on-boarded automatically upon its generation or on-boarded manually based on a received instruction, by the digital assistant device. It is contemplated that individuals may want to keep their actions or command templates private, and so an option to keep an action dataset limited to locally-storage may be provided to the user (e.g., via a GUI element). The server, upon receiving an on-boarded action dataset, can analyze the action dataset and generate an associated action signature based on the characteristics and/or contents of the action dataset. Contents of an action dataset can include, among other things, application identifying information, corresponding command templates and parameters, and a recorded series of events. The action signature can be generated by various operations, such as hashing the on-boarded action dataset with a hashing algorithm, by way of example. It is also contemplated that the action signature can be generated by the on-boarding digital assistant device, the generated action signature then being stored in or appended to the action dataset before it is uploaded to the server.

In one aspect, the server can determine that the on-boarded action dataset already exists on the server, based on a determination that the action signature corresponds to the action signature of another action dataset already stored on the server. The server can either dispose of the on-boarded action dataset or merge the on-boarded action dataset (or determined differing portion(s) thereof) with an existing action dataset stored thereby, preventing redundancy and saving storage space. In another aspect, the server can analyze the on-boarded action dataset to determine if its contents (e.g., the recorded events, command templates, metadata) comply with one or more defined policies (e.g., inappropriate language, misdirected operations, incomplete actions) associated with general usage of the digital assistant system. In another aspect, the server can employ machine learning algorithms, among other things, to perform a variety of tasks, such as determining relevant parameter types, generating additional command templates for association with an on-boarded or stored action dataset, comparing similarity of events between on-boarded action datasets to identify and select more efficient routes for invoking an operation, and more.

In some further embodiments, the server can distribute one or more stored actions datasets to a plurality of digital assistant devices in communication with the server. In this way, each digital assistant device can receive action datasets or portions thereof (e.g., command templates) from the server. The action datasets can be distributed to the digital assistant devices in a variety of ways. For instance, in an embodiment, the server can freely distribute any or all determined relevant action datasets to digital assistant devices. In an embodiment, an application profile including a list of applications installed on a digital assistant device can be communicated to the server. Based on the application profile for the digital assistant device, the server can distribute any or all determined relevant action datasets to the digital assistant device. As digital assistant devices can include a variety of operating systems, and versions of applications installed thereon can also vary, it is contemplated that the application profile communicated by a digital assistant device to the server may include operating system and application version information, among other things, so that appropriate and relevant action datasets are identified by the server for distribution to the digital assistant device. For a more granular implementation, an action dataset profile including a list of action datasets or action signatures stored on the digital assistant device can be communicated to the server. In this way, only missing or updated action datasets can be distributed to the digital assistant device. In a further embodiment, a device profile including contextual information (e.g., location history, application usage history, personal data) detected and stored by a digital assistant device can be included in a determination of a command's relevance to one or more action datasets. In various embodiments, the application profile, action dataset profile, and/or device profile can be employed by either a digital assistant device or a digital assistant server described in accordance with the present disclosure.

In some embodiments, a user can simply announce a command to the digital assistant device, and in some instances, if a corresponding action dataset is determined not to be stored on the digital assistant device, the digital assistant device can send the command (e.g., the command representation) to the server for determination and selection of a set of relevant action datasets, which can then be communicated to the digital assistant device. In some other embodiments, every command can be communicated from the digital assistant device to the server for determination and selection of a set of relevant action datasets, which can then be communicated to the digital assistant device. Provided that the digital assistant device has the corresponding application installed thereon, the digital assistant device can retrieve, from the server, a set of determined most relevant action datasets, without additional configuration or interaction by the user, also reducing server load and saving bandwidth by inhibiting extraneous transfer of irrelevant action datasets. A retrieved set of relevant action datasets can be received from the server for invocation by the digital assistant device.

In some aspects, if the server determines that two or more action datasets are determined equally relevant to a received command, each action dataset may be retrieved from the server, and the digital assistant device can provide for display a listing of the determined relevant action datasets for selection and execution. In some further aspects, the server can employ, among other things, one or more of the application profile, action dataset profile, and/or device profile previously received from the digital assistant device to provide weight(s) to one or more of the determined equally relevant action datasets. In this way, the server can independently select a most relevant action dataset for communication to the digital assistant device.

In some other aspects, the digital assistant device may have stored in a local memory a plurality of action datasets locally created or received from the server. In this regard, the digital assistant device can enable features described herein when connectivity to the server is unavailable (e.g., Internet access is down). If the digital assistant device determines that two or more action datasets are equally relevant to a received command, the digital assistant device can similarly employ, among other things, one or more of the locally- stored application profile, action dataset profile, and/or device profile locally maintained by the digital assistant device to provide weight(s) to one or more of the determined equally relevant action datasets. In this way, the digital assistant device can independently select a most relevant action dataset to be interpreted, so that the included set of operations can be performed.

In some embodiments, a user of a digital assistant device can customize command templates associated with an action dataset corresponding to an application installed on their digital assistant device. Put simply, a user can employ the digital assistant (or a GUI thereof) to select an action dataset from a list of action datasets stored on the computing device, select an option to add a new command to the action dataset, and define a new command and any associated parameters for storage in the action dataset. In this regard, the user can add any custom command and parameter that can later be understood by the digital assistant device to invoke the action. In some aspects, the custom command and/or modified action can be on-boarded to the server for analysis and storage, as noted above. In some further aspects, based on the analysis, the server can distribute the custom command and/or at least a portion of the modified action dataset to a plurality of other digital assistant devices. In this regard, the list of understandable commands and corresponding actions can continue to grow and evolve, and be automatically provided to any other digital assistant device.

Accordingly, at a high level and with reference to FIG. 1, an example operating environment 100 in which some embodiments of the present disclosure may be employed is depicted. It should be understood that this and other arrangements and/or features described by the enclosed document are set forth only as examples. Other arrangements and elements (e.g., machines, interfaces, functions, orders, and groupings of functions, etc.) or features can be used in addition to or instead of those described, and some elements or features may be omitted altogether for the sake of clarity. Further, many of the elements or features described in the enclosed document may be implemented in one or more components, or as discrete or distributed components or in conjunction with other components, and in any suitable combination and location. Various functions described herein as being performed by one or more entities may be carried out by hardware, firmware, and/or software. For instance, some functions may be carried out by a processor executing instructions stored in memory.

The system in FIG. 1 includes one or more digital assistant devices 110, 115a,

115b, 115c, ... 115n, in communication with a server 120 via a network 130 (e.g., the Internet). In this example, the server 120, also in communication with the network 130, is in communication with each of the digital assistant devices 110, 115a-115n, and can also be in communication with a database 140. The database 140 can be directly coupled to the server 120 or coupled to the server 120 via the network 130. The digital assistant device 110, representative of other digital assistant devices 115a-115n, is a computing device comprising one or more applications 112 and a digital assistant module 114 installed and/or executing thereon.

The one or more applications 112 includes any application that is executable on the digital assistant device 110, and can include applications installed via an application marketplace, custom applications, web applications, side-loaded applications, applications included in the operating system of the digital assistant device 110, or any other application that can be reasonably considered to fit the general definition of an application or mobile application. On the other hand, the digital assistant module 114 can provide digital assistant services installed on the digital assistant device 110 or provided by the server 120 via the network 130, or can be implemented at least partially into an operating system of the digital assistant device 110. In accordance with embodiments described herein, the digital assistant module 114 provides an interface between a digital assistant device 110 and an associated user (not shown), generally via a speech-based exchanged, although any other method of exchange between user and digital assistant device 110 (e.g., keyboard input, communication from another digital assistant device or computing device) remains within the purview of the present disclosure.

When voice commands are received by the digital assistant device 110, the digital assistant module 114 can convert the speech command to text utilizing a speech-to- text engine (not shown) to extract identified terms and generate a command representation. The digital assistant module 114 can receive the command representation, and determine that the command representation corresponds to at least one command template of at least one action dataset stored on the digital assistant device. In some embodiments, the digital assistant module can generate an index of all command templates stored on the digital assistant device 110 for faster searching and comparison of the received command representation to identify a corresponding command template, and thereby a corresponding action dataset. Each indexed command template can be mapped to a corresponding action dataset, which can be interpreted for execution in response to a determination of a confirmed match with the received command representation.

By way of brief overview, a command template can include one or more keywords and/or one or more parameters that each have a corresponding parameter type. Each command template generally corresponds to an operation that can be performed on one or more applications 112 installed on a digital assistant device 110. Moreover, a plurality of command templates can correspond to a single operation, such that there are multiple equivalent commands that can invoke the same operation. By way of example only, commands such as "check in," check into flight," "please check in," "check into flight now," "check in to flight 12345," and the like, can all invoke the same operation that, by way of example only, directs the digital assistant module 114 to execute an appropriate airline application on the digital assistant device 110 and perform a predefined set of events or computer operations to achieve the same result.

The aforementioned commands, however, may lack appropriate information

(e.g., the specific airline). As one of ordinary skill may appreciate, a user may have multiple applications 112 from various vendors (e.g., airlines) associated with a similar service (e.g., checking into flights). A digital assistant device 110 in accordance with embodiments described herein can provide features that can determine contextual information associated with the digital assistant device 110, or its associated user, based on historical use of the digital assistant device 110, profile information stored on the digital assistant device 110 or server 120, stored parameters from previous interactions or received commands, indexed messages (e.g., email, text messages) stored on the digital assistant device, and a variety of other types of data stored locally or remotely on a server, such as server 120, to identify a most relevant parameter and supplement a command to select a most relevant action dataset. More specific commands, such as "check into Friendly Airline flight," or "Friendly Airline check in," and the like, where a parameter is specifically defined in the command, can be recognized by the digital assistant module 114.

One or more recognizable commands and corresponding action datasets can be received by the digital assistant device 110 from the server 120 at any time, including upon installation, initialization, or invocation of the digital assistant module 114, after or upon receipt of a speech command by the digital assistant module 114, after or upon installation of a new application 112, periodically (e.g., once a day), when pushed to the digital assistant device 110 from the server 120, among many other configurations. It is contemplated that the action datasets received by the digital assistant device 110 from the server 120 can be limited based at least in part on the applications 112 installed on the digital assistant device 110, although configurations where a larger or smaller set of action datasets received are contemplated. In the event an action dataset is determined not available for a particular application 112 installed on the digital assistant device 110, digital assistant module 114 can either redirect the user to a marketplace (e.g., launch an app marketplace application) to install the appropriate application determined by the server 120 based on the received command, or can invoke an action training program that prompts a user to manually perform tasks on one or more applications to achieve the desired result, the tasks being recorded and stored into a new action dataset by the digital assistant device 110. The digital assistant module 114 can also receive one or more commands from the user (e.g., via speech or text) to associate with the action dataset being generated. If the command includes variable parameters (e.g., optional fields), the action training program can facilitate a definition of such parameters and corresponding parameter types to generate command templates for inclusion in the action dataset being generated. In this way, a command template(s) is associated with at least the particular application designated by the user and also corresponds to the one or more tasks manually performed by the user, associating the generated command template to the task(s) and thus the desired resulting operation.

In some instances, the server 120 can provide a determined most-relevant action dataset to the digital assistant device 110 based on the received command. The server 120 can store and index a constantly-growing and evolving plurality of crowd-sourced action datasets submitted by or received from digital assistant devices 115a-115n also independently having a digital assistant module 114 and any number of applications 112 installed thereon. The digital assistant devices 115a-115n may have any combination of applications 112 installed thereon, and any generation of action datasets performed on any digital assistant device 110, 115-115n can be communicated to the server 120 to be stored and indexed for mass or selective deployment, among other things. In some aspects, the server 120 can include various machine-learned algorithms to provide a level of quality assurance on command templates included in on-boarded action datasets and/or the tasks and operations performed before they are distributed to other digital assistant devices via the network 130.

When the digital assistant module 114 determines an appropriate action dataset (e.g., one or more tasks to achieve a desired result) having one or more command templates that corresponds to the received command, the digital assistant module 114 can generate an overlay interface that can mask any or all visual outputs associated with the determined action or the computing device generally. The generation of the overlay interface can include a selection, by the digital assistant module 114, of one or more user interface elements that are stored in a memory of the digital assistant device 110 or server 120, and/or include a dynamic generation of the user interface element(s) by the digital assistant module 114 or server 120 based on one or more portions of the received command and/or obtained contextual data (e.g., determined location data, user profile associated with the digital assistant device 110 or digital assistant module 114, historical data associated with the user profile, etc.) obtained by the digital assistant device 110, digital assistant module 114, and/or server 120. The selected or generated one or more user interface elements can each include content that is relevant to one or more portions (e.g., terms, keywords) of the received command. In the event of dynamic generation of user interface elements, such elements can be saved locally on the digital assistant device 110 or remotely on the server 120 for subsequent retrieval by the digital assistant device 110, or can be discarded and dynamically regenerated at any time.

Example operating environment depicted in FIG. 1 is suitable for use in implementing embodiments of the invention. Generally, environment 100 is suitable for creating, on-boarding, storing, indexing, crowd-sourcing (e.g., distributing), and invoking actions or action datasets, among other things. Environment 100 includes digital assistant device 110, action cloud server 120 (hereinafter also referenced as "server") and network 130. Digital assistant device 110 can be any kind of computing device having a digital assistant module installed and/or executing thereon, the digital assistant module being implemented in accordance with at least some of the described embodiments. For example, in an embodiment, digital assistant device 110 can be a computing device such as computing device 800, as described below with reference to FIG. 7. In embodiments, digital assistant device 110 can be a personal computer (PC), a laptop computer, a workstation, a mobile computing device, a PDA, a cell phone, a smart watch or wearable, or the like. Any digital assistant device described in accordance with the present disclosure can include features described with respect to the digital assistant device 110. In this regard, a digital assistant device can include one or more applications 112 installed and executable thereon. The one or more applications 112 includes any application that is executable on the digital assistant device, and can include applications installed via an application marketplace, custom applications, web applications, side-loaded applications, applications included in the operating system of the digital assistant device, or any other application that can be reasonably considered to fit the general definition of an application. On the other hand, the digital assistant module can be an application, a service accessible via an application installed on the digital assistant device or via the network 130, or implemented into a layer of an operating system of the digital assistant device 110. In accordance with embodiments described herein, the digital assistant module 114 can provide an interface between a digital assistant device 110 and a user (not shown), generally via a speech-based exchange, although any other method of exchange between user and digital assistant device may be considered within the purview of the present disclosure.

Similarly, action cloud server 120 ("server") can be any kind of computing device capable of facilitating the on-boarding, storage, management, and distribution of crowd-sourced action datasets. For example, in an embodiment, action cloud server 120 can be a computing device such as computing device 700, as described below with reference to FIG. 7. In some embodiments, action cloud server 120 comprises one or more server computers, whether distributed or otherwise. Generally, any of the components of environment 100 may communicate with each other via a network 130, which may include, without limitation, one or more local area networks (LANs) and/or wide area networks (WANs). Such networking environments are commonplace in offices, enterprise-wide computer networks, intranets, and the Internet. The server 120 can include or be in communication with a data source 140, which may comprise data sources and/or data systems, configured to make data available to any of the various constituents of the operating environment 100. Data sources 140 may be discrete from the illustrated components, or any combination thereof, or may be incorporated and/or integrated into at least one of those components.

Referring now to FIG. 2, a block diagram 200 of an exemplary digital assistant device 210 suitable for use in implementing embodiments of the invention is shown. Generally, digital assistant device 210 (also depicted as digital assistant device 110 of FIG. 1) is suitable for receiving commands or command representations, selecting action datasets to execute by matching received commands to command templates of action datasets, or determining that no action datasets correspond to received commands, interpreting a selected action dataset to execute the associated operation, generating new action datasets, and sending action datasets to or receiving action datasets from a digital assistant server, such as server 120.

Digital assistant device 210 can include, among other things, a command receiving component 220, an action matching component 230, an action executing component 240, a training component 250, a server interfacing component 260, a contextual data detecting component 270, and a relevance component 280. The command receiving component 220 can receive a command, either in the form of speech data or text data. The speech data can be received via a microphone of the digital assistant device 210, or another computing device paired to or in communication with the digital assistant device 210. The command receiving component 220, after receiving the speech data, can employ a speech-to- text engine of the digital assistant device 210 to generate a command representation (e.g., a text string of the command). Text data received by command receiving component 220, on the other hand, can be received via a virtual keyboard or other input method of the digital assistant device 210, and similarly, can be received from another computing device paired to or in communication with the digital assistant device 210. Received text data is already in the form of a command representation, and is treated as such. In various embodiments, command receiving component 210 can be invoked manually by a user (e.g., via an input to begin listening for or receiving the command), or can be in an always-listening mode.

Based on a command representation being received, action matching component 230 can determine whether one or more action datasets stored on the digital assistant device 210 includes a command template that corresponds to or substantially corresponds (e.g., at least 90% similar) to the received command representation. In some aspects, a corresponding command template can be identified, and the action dataset of which the corresponding command template is stored in can be selected for interpretation by action executing component 240. In some other aspects, more than one corresponding command template can be identified, and at least one most relevant action dataset can be determined by relevance component 280, which will be described herein, so that the at least one most relevant action dataset can be presented for selection by a user and/or executed by the digital assistant device 210. In some other aspects, a corresponding command template cannot be identified, and either the training component 250 can be invoked, or the received command is communicated to the digital assistant server (depicted as server 120 of FIG. 1 and digital assistant server 310 of FIG. 3) via the server interfacing component 260.

The action executing component 240 can receive a selected action dataset, either selected by digital assistant device 210 from local storage, by the digital assistant server from storage accessible thereto, or selected from a list presented by digital assistant device 210. The action executing component 240 can, from the received action dataset, interpret event data, which may include executable code, links, deep links, references to GUI elements, references to screen coordinates, field names, or other pieces of data that can correspond to one or more tasks or events stored in the selected action dataset. When the event data is interpreted, the action executing component 240 can reproduce the events that were recorded when the action dataset was initially generated, by any digital assistant device such as digital assistant device 210. In some aspects, the event data can include time delays, URLs, deep links to application operations, or any other operation that can be accessed, processed, emulated, or executed by the action executing component 240. In some aspects, events like click or touch inputs, can be reproduced on the digital assistant device 210, based on the interpreted event data stored in an action dataset.

The training component 250 can facilitate the generation of an action dataset, among other things. When the training component 250 is invoked, an indication, such as a GUI element, indicating that an action recording session has begun may be presented for display. A prompt to provide the tasks or events required to perform the desired operation can also be presented for display. In this regard, a user can begin by first launching an application for which the operation is associated with, and proceed with providing inputs to the application (i.e., (performing the requisite tasks). The inputs can be recorded by the digital assistant device 210, and the training component 250 can listen for, parse, identify, and record a variety of attributes of the received inputs, such as long or short presses, time delays between inputs, references to GUI elements interacted with, field identifiers, application links activated based on received inputs (e.g., deep links), and the like. The recorded inputs and attributes (e.g., event data) can be stored, sequentially, in an event sequence, and stored into a new action dataset. The application launched is also identified, and any application identifying information, such as operating system, operating system version, application version, paid or free version status, and more, can be determined from associated metadata and also stored into the new action dataset. When the desired operation is completed (i.e., all requisite tasks/events performed), a user can activate a training termination button, which can be presented as a floating button or other input mechanism that is preferably positioned away from an active portion of the display. Other termination methods are also contemplated, such as voice activated termination, or motion activated termination, without limitation.

The training component 250 can further request that the user provide a set of commands that correspond to the desired operation. A command can be received via speech data and converted to a command representation by a speech to text engine, or received via text input as a command representation, among other ways. When the set of commands is provided and stored as command representations, the training component 250 can further prompt the user to define any relevant parameters or variables in the command representations, which can correspond to keywords or values that may change whenever the command is spoken. In this regard, a user may select one or more terms included in the received command representations, and define them with a corresponding parameter type selected from a list of custom, predefined, or determined parameter times, as described herein. The training component 250 can then extract the selected one or more terms from a command representation defined as parameter(s), replacing them with parameter field identifier(s) of a corresponding parameter type, and store the resulting data as a command template. The training component 250 can then generate the action dataset from the recorded event sequence, the application identifying information, and the one or more defined command templates. In some embodiments, the training component 250 can generate an action signature or unique hash based on the generated action dataset or one or more portions of data included therein. The action signature can be employed by the digital assistant server to determine whether the action dataset or data included therein is redundant, among other things.

In some embodiments, a contextual data detecting component 270 can obtain various types of contextual data associated with the digital assistant device 210. By way of non-limiting example, contextual data detecting component 270 can detect and store location data of the digital assistant device 210. Location data can include GPS coordinates received from a GPS module (not shown) of the digital assistant device 210, Wi-Fi signals and corresponding signal strengths received from a Wi-Fi module (not shown) of the digital assistant device 210, telecommunication tower signals and strengths detected by the digital assistant device 210, and the like. In another non- limiting example, contextual data can include application usage data, such as a number or percentage of times each application has been opened (e.g., executed) on the digital assistant device 210, a number or percentage of times certain operations are executed within each application on the digital assistant device 210, a number or percentage of times contacts of a contact list are contacted (e.g., via phone call, text messages, emails) via the digital assistant device 210, a number or percentage of times each parameter value is input into parameter fields throughout applications of the digital assistant device 210, and more. Other non- limiting examples of contextual can include content of messages (e.g., email, text, messenger applications, social media posts), a number or percentage of times each action dataset has been selected for interpretation and/or execution by the digital assistant device 210, and/or web browser history, among other things. Contextual data can be detected and/or received by digital assistant device 210 through various application APIs, permissioned access to application data, and/or via a background service of the digital assistant device 210 that can intercept and/or record contextual data, such as any of those described above in a non-limiting fashion.

In some further embodiments, contextual data detecting component 270 can employ server interfacing component 260 to communicate the obtained contextual data to the server for storage thereon. Likewise, server interfacing component 260 can independently communicate such data to the server any one or more times it is in communication with the server. In some aspects, the contextual data can be communicated with a unique device and/or user identifier (e.g., a user account registered on the digital assistant device 210) so that the server can maintain a record of contextual data associated with the digital assistant device 210 and/or user identifier associated therewith. In this way, as will be described, the server can determine relevance of one or more action datasets to command(s) received from the digital assistant device 210.

Relevance component 280 can employ contextual data stored in a memory

(not shown) of digital assistant device 210 to determine one or more most relevant action datasets, particularly in situations where multiple action datasets are determined to correspond to a received command. In some aspects, relevance component 280 can be employed by digital assistant device 210 for action datasets that are stored locally on the digital assistant device 210. It is contemplated that digital assistant device 210 may be operational to select and execute action datasets when connectivity to the server is unavailable, and as such, relevance component 280 can determine which action dataset is most relevant when the server is unavailable or not implemented to serve this role. Relevance component 280 can, in some embodiments, analyze the contextual data based on a received command, to determine whether one action dataset is more relevant to other action datasets. By way of example, if a user provides a command to digital assistant device 210 stating "open my boarding pass," the command may be relevant to a number of applications (e.g., travel applications) and/or corresponding action datasets stored on the digital assistant device 210. In this regard, it would be helpful if contextual information associated with the user of digital assistant device 210 could be employed as factors that weigh its decision of relevance to the received command. In some embodiments, relevance component 280 can parse through the obtained contextual data to identify potential keywords or terms that are relevant to the received command. In the aforementioned example, the term "boarding pass" may invoke a search of contextual data including recently-opened applications (e.g., travel applications), emails or other messages relating to travel, web browser history, and more. In some aspects, the search can be limited to a predefined set of keywords that are associated with the provided term (e.g., boarding pass). As an example, the term "boarding pass" can be associated with terms such as travel, flight, train, book, ticket, itinerary, and the like. In various embodiments, the predefined set of keywords can be included in associated action datasets, or retrieved by the digital assistant device 210 regularly or periodically from the server, among other things. In some other embodiments, the predefined set of keywords can be maintained by the server, such that if and when commands are communicated thereto, the server can similarly factor in the determined associated keywords to determine which action datasets are more relevant to the received command.

Looking now to FIG. 3, a block diagram 300 of an exemplary digital assistant server 310 suitable for use in implementing embodiments of the invention is shown. Generally, digital assistant server 310 (also depicted as server 120 of FIG. 1) is suitable for establishing connections with digital assistant device(s) 210, receiving generated action datasets, maintaining or indexing received action datasets, receiving commands or command representations to determine one or more most relevant action datasets for selection and communication to a sending digital assistant device of the command or command representation, and distributing action datasets to other digital assistant devices, such as digital assistant device 210. Digital assistant server 310 can include, among other things, an on-boarding component 320, an indexing component 330, a maintenance component 340, a relevant component 350, and a distribution component 360, among other things.

The on-boarding component 320 can receive action datasets generated by one or more digital assistant devices 210 in communication therewith. In some aspects, the on- boarding component can generate an action signature for a received action dataset, similar to how a digital assistant device may, as described herein above. Before storing the received action dataset, the action signature can be searched utilizing the indexing component 330, which maintains an index of all action datasets stored by the digital assistant server 310.

The indexing component 330 can, among other things, facilitate quick determination of uniqueness of received action datasets, and can reduce redundancy and processing load of the digital assistant server 310. By way of example, the indexing component 330 can determine whether the action signature (e.g., hash of operations) of the received action dataset corresponds to any other action signature associated with another action dataset stored by the server 310. If the action signature is determined to match another already-indexed action signature, the received action dataset can either be disregarded, or additional steps can be employed to augment the action dataset already stored by the server 310. For instance, command templates included in the already-stored action data set can be compared to those included in the received action dataset, and augmented with new or determined different command templates included in the received action dataset.

On a similar note, the maintenance component 340 can determine whether any portion of a received action dataset is different than action datasets already stored on or by the server (e.g., in a database), and extract such portions for merging into the existing corresponding action datasets. Such portions may be identified in circumstances where only command templates are hashed in the action signature, or where each portion of the action dataset (e.g., application identifying information, command template(s), event sequence) is independently hashed either by training component 240 of FIG. 2 or on-boarding component 310 of FIG. 3, to more easily identify changes or differences between action datasets. By way of example, in some embodiments, a received action dataset can include separate hashes for its application identifying information, event sequence, and command template(s). In this regard, the digital assistant server 310 can employ the indexing component 330 and maintenance component 340 to quickly identify, for instance, that the received action data corresponds to a particular application and operation, or that the command template(s) are different than those stored in the stored action dataset by virtue of the command template hashes being different. Similarly, the independent hash signatures for each portion of data included in an action dataset can facilitate efficient determination of changes or differences between any combination of data portions in a received action dataset and a stored action dataset.

Relevance component 350 can determine, based on commands or command representations received by a digital assistant device 210, a likelihood that a particular command template corresponds to the received command or command representation. While a variety of relevance determining methods may be employed, a machine learning implementation may be preferable, though a ranking of determined most similar command templates to a command or command representation received from a digital assistant device 210 can also facilitate a determination of relevance and therefore one or more most relevant command templates. Determined most-relevant command templates can thereby facilitate the selection of a most relevant action dataset to be distributed to the command- sending digital assistant device 210.

Like relevance component 280 of FIG. 2, relevance component 350 can employ contextual data associated with a digital assistant device 210 to determine one or more most relevant action datasets, particularly in situations where multiple action datasets are determined to correspond to a received command. In various aspects, relevance component 350 can be employed by digital assistant server 310 for action datasets that are stored by the server 310 and/or on the digital assistant device 210. For instance, relevance component 350 can determine which application datasets stored on the digital assistant device 210 should be invoked based on a received command. The server can communicate, to the digital assistant device 210, one of a reference to a determined most relevant action dataset stored on the digital assistant device 210 or the determined most action dataset stored by the server 310.

In another aspect, relevance component 350 can determine which application datasets stored by the server 310 should be distributed to a digital assistant device 210 based on the received command received from the digital assistant device 210. It is contemplated that digital assistant device 210 may require connectivity to the server 310, and as such, relevance component 350 can determine which action dataset is most relevant to be invoked by the digital assistant device 210 or communicated to the digital assistant device 210 for invocation thereon. Relevance component 350 can, in some embodiments, analyze contextual data stored by the server 310 and associated with the digital assistant device 210 or user account registered therewith based on a received command, to determine whether one action dataset is more relevant to other action datasets. By way of example, if a user provides a command to digital assistant device 210 stating "open my boarding pass," the command may be relevant to a number of applications (e.g., travel applications) and/or corresponding action datasets stored by the server 310. In this regard, it would be helpful if contextual information associated with the user of digital assistant device 210 could be employed as factors that weigh the server's 310 decision of relevance to the received command. In some embodiments, relevance component 350 can parse through the stored contextual data to identify potential keywords or terms that are relevant to the received command. In the aforementioned example, the term "boarding pass" may invoke a search of contextual data including recently-opened applications (e.g., travel applications), emails or other messages relating to travel, web browser history, and more. In some aspects, the search can be limited to a predefined set of keywords that are associated with the provided term(s) (e.g., boarding pass). As an example, the term "boarding pass" can be associated with terms such as "travel," "flight," "train," "book," "ticket," "itinerary," and the like. In various embodiments, the predefined set of keywords can be included in associated action datasets, or stored by the server locally or on a database that the server can access for periodic distribution of the keywords to digital assistant devices, among other things. For instance, the predefined set of keywords can be maintained by the server, such that if and when commands are communicated thereto, the server can similarly factor in the determined associated keywords to determine which action datasets are most relevant to the received command.

In some aspects, the contextual data can be maintained and stored solely by the digital assistant device 210. Particularly useful for maintenance of privacy, the server 310 can select a plurality of action datasets determined to correspond to the received command, and send each determined corresponding action dataset to the digital assistant device 210. In this regard, the digital assistant device 210 can employ the locally-maintained contextual data to determine a most relevant action dataset and select the determined most relevant action dataset for interpretation and invocation of the corresponding set of operations. The foregoing are merely exemplary implementations and combinations of the various features disclosed herein, it is contemplated that any combination of the foregoing features can be employed to facilitate the distribution, selection, relevance determination, and ultimate invocation of action datasets in accordance with the present disclosure.

The distribution component 360 can distribute or communicate to one or more digital assistant devices 210, determined relevant or most relevant action datasets, determined new action datasets, determined updated action datasets, any portion and/or combination of the foregoing, or generated notifications corresponding to any portion and/or combination of the foregoing, among other things, based on a variety of factors. For instance, the distribution component 360 can include features that determine, among other things, which applications are installed on a digital assistant device 210. Such features can enable the digital assistant server 310 to determine which action datasets or portions thereof are relevant to the digital assistant device 210, and should be distributed to the digital assistant device 210. For instance, a digital assistant device 210 profile (not shown) describing all applications currently installed or executable by a digital assistant device 210, can be maintained (e.g., stored, updated) by the digital assistant server 310. The profile can be updated periodically, manually, or dynamically by a server interfacing component 260 of the digital assistant device 210 (e.g., whenever the digital assistant is in communication with and sends a command to the digital assistant server 310, or whenever an application is installed or updated on the digital assistant device 210). The distribution component 360 can distribute or communicate notifications, action datasets, or portions thereof, in a variety of ways, such as pushing, sending in response to received requests for updates, sending in response to established communications with a digital assistant device 210, or by automatic wide scale (e.g., all digital assistant devices) or selective scale (e.g., region, location, app type, app name, app version) distribution, among other things.

Turning now to FIG. 4, a data structure 400 of an exemplary action dataset 410 in accordance with some of the described embodiments is illustrated. The depicted data structure is not intended to be limiting in any way, and any configuration of the depicted data portions of information may be within the purview of the present disclosure. Further, additional data portions or less data portions may be included in an action dataset 410 also remaining within the purview of the present disclosure.

In the depicted data structure 400, the action dataset 410 includes application identifying information 420, recorded event sequence data 430, and command templates 440. In some embodiments, the action dataset 410 further includes hash(es) 450, which can include a hash value generated based on the entire action dataset 410, or hash values generated based on any portion of the aforementioned data portions 420, 430, 440, among other things. The action dataset 410 can be generated by training component 250 of digital assistant device 210 of FIG. 2 and/or received from distribution component 360 of digital assistant server 310 of FIG. 3.

The application identifying information 420 can include information about a particular application that is required for execution to perform a particular operation for which the action dataset 410 was created. Exemplary pieces of application identifying information 420 are depicted in identifying information 425, which can include any one or more of an operating system (OS) name for which the particular application is executed on, an OS version of the aforementioned OS, a defined native language of the aforementioned OS, a name of the particular application, a version of the particular application, and the like. It is contemplated that the application identifying information 420 is required and checked (e.g., by the digital assistant server 310 of FIG. 3), before an action dataset 410 is distributed to a digital assistant device (e.g., digital assistant device 210 of FIG. 2) and employed by the digital assistant device to ensure that the action dataset 410 is compatible with, or can be correctly interpreted by action executing component 240 of FIG. 2, so that the corresponding and desired operation is performed by the digital assistant device 210.

The recorded event sequence data 430 can include any or all task or event- related data that was obtained, received, or determined by the digital assistant device (e.g., via training component 250 of FIG. 2) responsible for generating the action dataset 410. As noted herein, the recorded event sequence data can include timing attributes of received inputs (e.g., delays before or in between successive inputs, duration of inputs, GUI elements interacted with, relative positions of GUI elements, labels or metadata of GUI elements, scroll inputs and distances, links or URLs accessed activated, detected activation of application deep links activated in response to received inputs, and more). In some instances, the recorded event sequence data 430 may include conditions that require actual user intervention before subsequent events or tasks are resumed. For instance, secured login screens may require that a user input username and password information before an application is executed. In this regard, the recorded event sequence data 430 may include a condition corresponding to when user authentication has occurred, and instructions (e.g., interpretable by action executing component 240) to proceed with the tasks or events in the recorded event sequence data 430 based upon an occurrence of the condition. In various implementations, it is contemplated that the action executing component 240 of FIG. 2 can parse metadata, GUI elements, or other information from an executing application to determine when certain events occur or conditions are met. In this regard, additional conditions may be included in the recorded event sequence data 430 that require prior events or tasks to be completed, or certain GUI elements be displayed or interacted with, or any other conditions to be met, before subsequent events or tasks are performed by the action executing component 240 of FIG. 2.

Turning now to FIG. 5, an embodiment of a process flow or method 500 for performing actionable digital assistant operations (e.g., via action datasets) retrieved via a crowd-sourced digital assistant network is described. At step 510, a digital assistant device, such as digital assistant device 210 of FIG. 2, can receive a command or command representation. As described herein, a command can include an audible voice command or instruction that a user of the digital assistant device speaks aloud. The digital assistant device can receive the audio data of the command (e.g., via a microphone), and can convert the audio data to text (e.g., via a speech-to-text engine). The command can, in some embodiments, include alphanumeric text that was generated by the digital assistant device (e.g., after conversion from audio data) or was received by the digital assistant device from another computing device (e.g., via a network). The command, once converted to or received as alphanumeric text, can also be referenced herein as a command representation (e.g., text that represents a command or a requested set of operations to be performed by the digital assistant device), however the term "command" can be employed to reference either the audio data form or the alphanumeric text form of the received request.

At step 520, the digital assistant device can obtain one or more action datasets, such as action dataset 400 of FIG. 4, that are each selected based on a determined relevance of the action dataset to the received command. In some aspects, the relevance of an action dataset to a received command can be performed by a digital assistant server device, such as digital assistant server 310 of FIG. 3. In this regard, the digital assistant device can send the received command to the digital assistant server so that the command is analyzed and one or more indexed action datasets stored by the server device are evaluated against the command. After a determination is made that one or more of the action datasets correspond to or are relevant to the received command, the one or more corresponding or determined relevant action datasets can be further evaluated to determine a most relevant action dataset. In some embodiments, the one or more corresponding or determined relevant action datasets can be communicated to the digital assistant device so that the digital assistant device can determine a most relevant action dataset. In various embodiments, if the digital assistant server device is tasked with determining a most relevant action dataset, the digital assistant server device can search stored contextual data that corresponds to the digital assistant device. In this way, the contextual data can be searched for keywords or terms that are determined associated with keywords included in the received command. As described herein, associated keywords and terms can be predefined and stored in a memory to facilitate a determination of relevance or the generation of a weight or "rank" of relevance between various action datasets. As each action dataset can correspond to a particular application, a higher ranked action dataset can provide an indication that its corresponding application is more relevant than another application of another action dataset.

In some other aspects, the relevance of an action dataset to a received command can be performed by the digital assistant device, such as digital assistant device 210 of FIG. 2. In this regard, the digital assistant device can similarly analyze the received command and evaluate one or more action datasets stored locally by the digital assistant device. In is contemplated that in some further aspects, any combination of the foregoing implementations can be employed. For instance, a command can be analyzed locally by the digital assistant device and/or remotely by the digital assistant server device, action datasets can be stored locally by the digital assistant device and/or remotely by the digital assistant server device, and the respectively stored action datasets can be evaluated by either device or both devices in step, in part, in combination, or simultaneously, among other variations.

As described herein, an action dataset can include one or more command templates, such as command template 440 of FIG. 4. A command template can include, among other things, one or more keywords that correspond to a command that can be spoken or communicated to the digital assistant device to invoke a corresponding set of operations. In some aspects, the command template can also include one or more parameter fields that each corresponds to a variable keyword. That is, a parameter field can include a variable that does not need to strictly correspond to the keywords defined by the command template. As such, if a command template includes keywords "Open my <travel app> boarding pass," the third term of the command corresponding to the <travel app> parameter can be employed to determine which action dataset should be selected due to its determined relevance to the received command. In other words, if the command includes "ABC Airline" in the <travel app> parameter field, the action dataset that opens the boarding pass within the "ABC Airline" application installed on the digital assistant device can be selected due to its determined relevance to the received command.

At step 530, the digital assistant device can interpret a set of instructions included in one of the selected plurality of action datasets that were determined most relevant to the received command. That is, in various embodiments including those where either the digital assistant server device or digital assistant device determine the most relevant action dataset, a particular action dataset is selected based on a determination that it is most relevant to the received command. As relevance includes relevance of the action dataset or corresponding application to a particular user of the digital assistant device, the most relevant action dataset can be determined based on stored contextual data corresponding to the digital assistant device or a user account of the particular user associated with the digital assistant device. In some aspects, timestamps associated with contextual data, frequency of contextual data types (e.g., location, app history, recently-opened applications or documents, document type, etc.), content of various data types, or any combination thereof not being limited to the foregoing can be employed and considered in determining rank and/or relevance of an action dataset to a received command. When the digital assistant device interprets the set of instructions included in the determined most-relevant action dataset, the set of instructions included in the determined most-relevant action dataset is interpreted to replicate, reproduce, and/or execute a corresponding set of operations.

Turning now to FIG. 6, an embodiment of a process flow or method 600 for performing actionable digital assistant operations (e.g., via action datasets) retrieved via a crowd-sourced digital assistant network is described. At step 610, a digital assistant device, such as digital assistant device 210 of FIG. 2, can receive a command or command representation. As described herein, a command can include an audible voice command or instruction that a user of the digital assistant device speaks aloud. The digital assistant device can receive the audio data of the command (e.g., via a microphone), and can convert the audio data to text (e.g., via a speech-to-text engine). The command can, in some embodiments, include alphanumeric text that was generated by the digital assistant device (e.g., after conversion from audio data) or was received by the digital assistant device from another computing device (e.g., via a network). The command, once converted to or received as alphanumeric text, can also be referenced herein as a command representation (e.g., text that represents a command or a requested set of operations to be performed by the digital assistant device), however the term "command" can be employed to reference either the audio data form or the alphanumeric text form of the received request.

At step 620, the digital assistant device can obtain one or more action datasets, such as action dataset 400 of FIG. 4, that are each selected based on a determined relevance of the action dataset to the received command. In various embodiments, any number of action datasets obtained by the digital assistant device can be generated by another digital assistant device. That is, as each digital assistant device in accordance with the present disclosure is in communication with the digital assistant server device described in accordance with FIG. 3, it is contemplated that any number of digital assistant devices can generate action datasets and communicate them to the digital assistant server device for maintenance and crowd-sourced distribution in accordance with the described embodiments. In some aspects, the relevance of an action dataset to a received command can be performed by a digital assistant server device, such as digital assistant server 310 of FIG. 3. In this regard, the digital assistant device can send the received command to the digital assistant server so that the command is analyzed and one or more indexed action datasets stored by the server device are evaluated against the command. After a determination is made that one or more of the action datasets correspond to or are relevant to the received command, the one or more corresponding or determined relevant action datasets can be further evaluated to determine a most relevant action dataset. In some embodiments, the one or more corresponding or determined relevant action datasets can be communicated to the digital assistant device so that the digital assistant device can determine a most relevant action dataset. In various embodiments, if the digital assistant server device is tasked with determining a most relevant action dataset, the digital assistant server device can search stored contextual data that corresponds to the digital assistant device. In this way, the contextual data can be searched for keywords or terms that are determined associated with keywords included in the received command. As described herein, associated keywords and terms can be predefined and stored in a memory to facilitate a determination of relevance or the generation of a weight or "rank" of relevance between various action datasets. As each action dataset can correspond to a particular application, a higher ranked action dataset can provide an indication that its corresponding application is more relevant than another application of another action dataset.

In some other aspects, the relevance of an action dataset to a received command can be performed by the digital assistant device, such as digital assistant device 210 of FIG. 2. In this regard, the digital assistant device can similarly analyze the received command and evaluate one or more action datasets stored locally by the digital assistant device. In is contemplated that in some further aspects, any combination of the foregoing implementations can be employed. For instance, a command can be analyzed locally by the digital assistant device and/or remotely by the digital assistant server device, action datasets can be stored locally by the digital assistant device and/or remotely by the digital assistant server device, and the respectively stored action datasets can be evaluated by either device or both devices in step, in part, in combination, or simultaneously, among other variations.

As described herein, an action dataset can include one or more command templates, such as command template 440 of FIG. 4. A command template can include, among other things, one or more keywords that correspond to a command that can be spoken or communicated to the digital assistant device to invoke a corresponding set of operations. In some aspects, the command template can also include one or more parameter fields that each corresponds to a variable keyword. That is, a parameter field can include a variable that does not need to strictly correspond to the keywords defined by the command template. As such, if a command template includes keywords "Open my <travel app> boarding pass," the third term of the command corresponding to the <travel app> parameter can be employed to determine which action dataset should be selected due to its determined relevance to the received command. In other words, if the command includes "ABC Airline" in the <travel app> parameter field, the action dataset that opens the boarding pass within the "ABC Airline" application installed on the digital assistant device can be selected due to its determined relevance to the received command.

At step 630, the digital assistant device can interpret a set of instructions included in one of the selected plurality of action datasets that were determined most relevant to the received command. That is, in various embodiments including those where either the digital assistant server device or digital assistant device determine the most relevant action dataset, a particular action dataset is selected based on a determination that it is most relevant to the received command. As relevance includes relevance of the action dataset or corresponding application to a particular user of the digital assistant device, the most relevant action dataset can be determined based on stored contextual data corresponding to the digital assistant device or a user account of the particular user associated with the digital assistant device. In some aspects, timestamps associated with contextual data, frequency of contextual data types (e.g., location, app history, recently-opened applications or documents, document type, etc.), content of various data types, or any combination thereof not being limited to the foregoing can be employed and considered in determining rank and/or relevance of an action dataset to a received command. When the digital assistant device interprets the set of instructions included in the determined most-relevant action dataset, the set of instructions included in the determined most-relevant action dataset is interpreted to replicate, reproduce, and/or execute a corresponding set of operations.

Having described various embodiments of the invention, an exemplary computing environment suitable for implementing embodiments of the invention is now described. With reference to FIG. 7, an exemplary computing device is provided and referred to generally as computing device 700. The computing device 700 is but one example of a suitable computing environment and is not intended to suggest any limitation as to the scope of use or functionality of the invention. Neither should the computing device 700 be interpreted as having any dependency or requirement relating to any one or combination of components illustrated.

Embodiments of the invention may be described in the general context of computer code or machine-useable instructions, including computer-useable or computer- executable instructions, such as program modules, being executed by a computer or other machine, such as a personal data assistant, a smartphone, a tablet PC, or other handheld device. Generally, program modules, including routines, programs, objects, components, data structures, and the like, refer to code that performs particular tasks or implements particular abstract data types. Embodiments of the invention may be practiced in a variety of system configurations, including handheld devices, consumer electronics, general-purpose computers, more specialty computing devices, etc. Embodiments of the invention may also be practiced in distributed computing environments where tasks are performed by remote- processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

With reference to FIG. 7, computing device 700 includes a bus 710 that directly or indirectly couples the following devices: memory 712, one or more processors 714, one or more presentation components 716, one or more input/output (I/O) ports 718, one or more I/O components 720, and an illustrative power supply 722. Bus 710 represents what may be one or more busses (such as an address bus, data bus, or combination thereof). Although the various blocks of FIG. 7 are shown with lines for the sake of clarity, in reality, these blocks represent logical, not necessarily actual, components. For example, one may consider a presentation component such as a display device to be an I/O component. Also, processors have memory. The inventors hereof recognize that such is the nature of the art and reiterate that the diagram of FIG. 7 is merely illustrative of an exemplary computing device that can be used in connection with one or more embodiments of the present invention. Distinction is not made between such categories as "workstation," "server," "laptop," "handheld device," etc., as all are contemplated within the scope of FIG. 7 and with reference to "computing device."

Computing device 700 typically includes a variety of computer-readable media. Computer-readable media can be any available media that can be accessed by computing device 700 and includes both volatile and nonvolatile media, removable and nonremovable media. By way of example, and not limitation, computer-readable media may comprise computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVDs) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by computing device 700. Computer storage media does not comprise signals per se. Communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media. The term "modulated data signal" means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media includes wired media, such as a wired network or direct-wired connection, and wireless media, such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above should also be included within the scope of computer-readable media.

Memory 712 includes computer storage media in the form of volatile and/or nonvolatile memory. The memory may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard drives, optical-disc drives, etc. Computing device 700 includes one or more processors 714 that read data from various entities such as memory 712 or I/O components 720. Presentation component(s) 716 presents data indications to a user or other device. Exemplary presentation components include a display device, speaker, printing component, vibrating component, and the like.

The I/O ports 718 allow computing device 700 to be logically coupled to other devices, including I/O components 720, some of which may be built in. Illustrative components include a microphone, joystick, game pad, satellite dish, scanner, printer, wireless device, etc. The I/O components 720 may provide a natural user interface (NUI) that processes air gestures, voice, or other physiological inputs generated by a user. In some instances, inputs may be transmitted to an appropriate network element for further processing. An NUI may implement any combination of speech recognition, touch and stylus recognition, facial recognition, biometric recognition, gesture recognition both on screen and adjacent to the screen, air gestures, head and eye tracking, and touch recognition associated with displays on the computing device 700. The computing device 700 may be equipped with depth cameras, such as stereoscopic camera systems, infrared camera systems, RGB camera systems, and combinations of these, for gesture detection and recognition. Additionally, the computing device 700 may be equipped with accelerometers or gyroscopes that enable detection of motion. The output of the accelerometers or gyroscopes may be provided to the display of the computing device 700 to render immersive augmented reality or virtual reality. Some embodiments of computing device 700 may include one or more radio(s) 724 (or similar wireless communication components). The radio 724 transmits and receives radio or wireless communications. The computing device 700 may be a wireless terminal adapted to receive communications and media over various wireless networks. Computing device 700 may communicate via wireless protocols, such as code division multiple access ("CDMA"), global system for mobiles ("GSM"), or time division multiple access ("TDMA"), as well as others, to communicate with other devices. The radio communications may be a short-range connection, a long-range connection, or a combination of both a short-range and a long-range wireless telecommunications connection. When we refer to "short" and "long" types of connections, we do not mean to refer to the spatial relation between two devices. Instead, we are generally referring to short range and long range as different categories, or types, of connections (i.e., a primary connection and a secondary connection). A short-range connection may include, by way of example and not limitation, a Wi-Fi® connection to a device (e.g., mobile hotspot) that provides access to a wireless communications network, such as a WLAN connection using the 802.11 protocol; a Bluetooth connection to another computing device is a second example of a short-range connection, or a near-field communication connection. A long-range connection may include a connection using, by way of example and not limitation, one or more of CDMA, GPRS, GSM, TDMA, and 802.16 protocols.

Many different arrangements of the various components depicted, as well as components not shown, are possible without departing from the scope of the claims below. Embodiments of the present invention have been described with the intent to be illustrative rather than restrictive. Alternative embodiments will become apparent to readers of this disclosure after and because of reading it. Alternative means of implementing the aforementioned can be completed without departing from the scope of the claims below. Certain features and sub-combinations are of utility and may be employed without reference to other features and sub-combinations and are contemplated within the scope of the claims.

Claims

CLAIMS What is claimed is:
1. A computer- implemented method for performing digital assistant operations, comprising: receiving, by a digital assistant device, a command that corresponds to a requested set of operations to be performed by the digital assistant device; obtaining, by the digital assistant device, a plurality of action datasets that are each selected based on a determined relevance of the action dataset to the received command, each action dataset in the selected plurality of action datasets having a corresponding set of instructions that is interpretable by the digital assistant device to reproduce a corresponding set of operations; and interpreting, by the digital assistant device, a set of instructions included in one of the selected plurality of action datasets that is determined most relevant to the received command based on contextual data collected by the digital assistant device, the included set of instructions being interpreted to perform the requested set of operations.
2. The computer- implemented method of claim 1, wherein the relevance of the selected action datasets is determined by one of the digital assistant device or the digital assistant server device.
3. The computer-implemented method of claim 2, wherein the selected action datasets are determined relevant by the digital assistant device based on the action datasets being stored locally in a memory of on the digital assistant device.
4. The computer-implemented method of claim 2, wherein the selected action datasets are determined relevant by the digital assistant server device based on the action datasets being stored by the digital assistant server device in one of a memory or a database.
5. The computer- implemented method of claim 1, wherein an action dataset is determined relevant to the received command based at least in part on another determination that a command template of the action dataset corresponds at least in part to the received command.
6. The computer-implemented method of claim 1, wherein the collected contextual data includes at least one of detected location data, message data, or application data, and wherein the determined most relevant action dataset is interpreted based further on a search of the collected contextual data for at least a portion of the received command or one or more keywords defined as associated with the portion of the received command.
7. The computer-implemented method of claim 6, wherein a result of the search is determined to correspond to an application installed on the digital assistant device.
8. The computer-implemented method of claim 6, wherein the one or more keywords defined associated with the portion of the received command is stored at least in part in one of the digital assistant server device, the digital assistant device, or an action dataset.
9. A non-transitory computer storage media storing computer-usable instructions that, when used by the one or more processors, cause the one or more processors to: receive a command that corresponds to a requested set of operations to be performed by the one or more processors; obtain a plurality of action datasets that are each selected based on a determined relevance of the action dataset to the received command, each action dataset in the selected plurality of action datasets having been generated by another digital assistant device and having a corresponding set of instructions that is interpretable by the one or more processors to reproduce a corresponding set of operations; and interpret a set of instructions included in one of the selected plurality of action datasets that is determined most relevant to the received command based on collected contextual data, the included set of instructions being interpreted to perform the requested set of operations.
10. The non-transitory media of claim 9, wherein the relevance of the selected action datasets is determined by the one or more processors or a digital assistant server device.
11. The non-transitory media of claim 10, wherein the selected action datasets are determined relevant by the one or more processors based on the action datasets being stored in a local memory.
12. The non-transitory media of claim 10, wherein the selected action datasets are determined relevant by the digital assistant server device based on the action datasets being stored by the digital assistant server device in one of a memory or a database.
13. The non- transitory media of claim 9, wherein an action dataset is determined relevant to the received command based at least in part on another determination that a command template of the action dataset corresponds at least in part to the received command.
14. The non-transitory media of claim 9, wherein the collected contextual data includes at least one of detected location data, message data, or application data, and wherein the determined most relevant action dataset is interpreted based further on a search of the collected contextual data for at least a portion of the received command or one or more keywords defined as associated with the portion of the received command.
15. The non-transitory media of claim 14, wherein a result of the search is determined to correspond to a locally-installed application.
16. The non-transitory media of claim 14, wherein the one or more keywords defined associated with the portion of the received command is stored at least in part in one of the digital assistant server device, in a local memory, or an action dataset.
17. A computerized system for performing digital assistant operations, comprising: one or more processors; and one or more non-transitory computer storage media storing computer-usable instructions that, when used by the one or more processors, cause the one or more processors to: receive a command that corresponds to a requested set of operations to be performed by the one or more processors; obtain a plurality of action datasets that are each selected based on a determined relevance of the action dataset to the received command, each action dataset in the selected plurality of action datasets having a corresponding set of instructions that is interpretable by the one or more processors to reproduce a corresponding set of operations; and interpret a set of instructions included in one of the selected plurality of action datasets that is determined most relevant to the received command based on collected contextual data, the included set of instructions being interpreted to perform the requested set of operations.
18. The computerized system of claim 17, wherein the relevance of the selected action datasets is determined by the one or more processors or a digital assistant server device.
19. The computerized system of claim 18, wherein the selected action datasets are determined relevant by the one or more processors based in part on the action datasets being stored in a local memory.
20. The computerized system of claim 18, wherein the selected action datasets are determined relevant by the digital assistant server device based in part on the action datasets being stored by the digital assistant server device in one of a memory or a database.
PCT/US2018/048064 2017-05-18 2018-08-27 Crowdsourced training for commands matching WO2019083602A1 (en)

Priority Applications (6)

Application Number Priority Date Filing Date Title
US201762576766P true 2017-10-25 2017-10-25
US62/576,766 2017-10-25
US15/984,122 2018-05-18
US15/984,122 US20180336049A1 (en) 2017-05-18 2018-05-18 Crowdsourced on-boarding of digital assistant operations
US16/110,103 2018-08-23
US16/110,103 US20180366108A1 (en) 2017-05-18 2018-08-23 Crowdsourced training for commands matching

Publications (1)

Publication Number Publication Date
WO2019083602A1 true WO2019083602A1 (en) 2019-05-02

Family

ID=66246986

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2018/048064 WO2019083602A1 (en) 2017-05-18 2018-08-27 Crowdsourced training for commands matching

Country Status (1)

Country Link
WO (1) WO2019083602A1 (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100088100A1 (en) * 2008-10-02 2010-04-08 Lindahl Aram M Electronic devices with voice command and contextual data processing capabilities
US20150348551A1 (en) * 2014-05-30 2015-12-03 Apple Inc. Multi-command single utterance input method
US20160259623A1 (en) * 2015-03-06 2016-09-08 Apple Inc. Reducing response latency of intelligent automated assistants

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20100088100A1 (en) * 2008-10-02 2010-04-08 Lindahl Aram M Electronic devices with voice command and contextual data processing capabilities
US20150348551A1 (en) * 2014-05-30 2015-12-03 Apple Inc. Multi-command single utterance input method
US20160259623A1 (en) * 2015-03-06 2016-09-08 Apple Inc. Reducing response latency of intelligent automated assistants

Similar Documents

Publication Publication Date Title
US7643985B2 (en) Context-sensitive communication and translation methods for enhanced interactions and understanding among speakers of different languages
US9633004B2 (en) Better resolution when referencing to concepts
KR101418163B1 (en) Speech recognition repair using contextual information
EP3241213B1 (en) Discovering capabilities of third-party voice-enabled resources
TWI566107B (en) A method for processing of multi-part voice commands, a non-transitory computer-readable storage medium and an electronic device
US9430463B2 (en) Exemplar-based natural language processing
US8938394B1 (en) Audio triggers based on context
Chu et al. Balancing energy, latency and accuracy for mobile sensor data classification
US9299342B2 (en) User query history expansion for improving language model adaptation
US10387410B2 (en) Method and system of classification in a natural language user interface
TWI585744B (en) A method of operating a virtual assistant, system, and computer-readable storage medium
CN102301358B (en) Textual disambiguation using social connections
US20160378747A1 (en) Virtual assistant for media playback
US9865264B2 (en) Selective speech recognition for chat and digital personal assistant systems
CN102884569B (en) Integration of embedded and network speech recognizers
US20160163312A1 (en) Disambiguating heteronyms in speech synthesis
DE102016214955A1 (en) Latency-free digital assistant
US9805733B2 (en) Method and apparatus for connecting service between user devices using voice
US20160098393A1 (en) Natural language understanding (nlu) processing based on user-specified interests
US20160179787A1 (en) Extensible context-aware natural language interactions for virtual personal assistants
US9286910B1 (en) System for resolving ambiguous queries based on user context
US20140222422A1 (en) Scaling statistical language understanding systems across domains and intents
US9009025B1 (en) Context-based utterance recognition
JP2016524190A (en) Environment-aware interaction policy and response generation
EP3243199B1 (en) Headless task completion within digital personal assistants

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18869869

Country of ref document: EP

Kind code of ref document: A1