High-fidelity data management for cross-domain analytics
A controller for managing data exchange between business and search engines addresses integration challenges, enhancing predictive accuracy and enabling monetization by automating data sharing and model improvements.
Patent Information
- Application Number
- JP2023537586
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2021-12-14
- Filing Date
- 2021-12-15
- Publication Date
- 2024-12-05
- Estimated Expiration
- 2041-12-15
AI Technical Summary
Current analytics systems face challenges in integrating business-owned data with search engine data, leading to inaccurate predictions and missed opportunities for mutual benefit due to lack of automated data transfer and integration.
A controller, potentially implemented as a rules engine, manages data exchange between business and search engine APIs, ensuring privacy and security while improving predictive models by integrating relevant data attributes, and enabling billing and auditing for monetization.
Enhances predictive accuracy of machine learning models by ensuring data convergence, automates data sharing based on privacy policies, and facilitates monetization through improved data utilization.
Smart Images

Figure 0007595175000001 
Figure 0007595175000002 
Figure 0007595175000003
Abstract
Description
[Background technology]
[0001] Analytics in the form of machine learning and cognitive networks are now commonly used for predictive calculations. In the case of search engines, machine learning models are used to deliver prioritized search results (called "surfacing results" in computer science terms) to users and other requesters. In the case of businesses, they collect data and make predictions for operations. For example, in inventory management, machine learning models are used to predict when to replenish stock and forecast consumer demand.
[0002] Businesses have an interest in being surfaced by search engines, and search engines, in turn, have an interest in providing accurate results to search result consumers. However, businesses store a lot of metadata and contextual data that search engines cannot access, and search engine platforms provide a lot of data and functionality that business stores do not necessarily utilize. Thus, there is a need to mediate the transfer of information, and in particular the transfer of data, for the mutual benefit of businesses and search engine platforms. [Brief description of the drawings]
[0003] The detailed description is described with reference to the accompanying drawings, in which the leftmost digit(s) of a reference number identifies the drawing in which the reference number first appears. Use of the same reference number in different drawings indicates similar or identical items.
[0004] [Figure 1] 1 is an exemplary configuration for implementing an exemplary embodiment of high fidelity data management.
[0005] [Diagram 2] FIG. 1 is a block diagram of an example environment, including hardware, software, and a communications environment, for implementing high fidelity data management.
[0006] [Diagram 3] FIG. 2 is a block diagram of an example controller for controlling data exchange between a business API and a search engine API.
[0007] [Figure 4a] 1 illustrates a flow diagram of an example process used by a controller to manage data access to data stored in a data store for a business domain. [Figure 4b] 1 illustrates a flow diagram of an example process used by a controller to manage data access to data stored in a data store for a business domain.
[0008] [Figure 5a] 1 illustrates a flow diagram of an exemplary process for controlling data exchange between a business API and a search engine API. [Figure 5b] 1 illustrates a flow diagram of an exemplary process for controlling data exchange between a business API and a search engine API. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] The present disclosure is directed to techniques for performing high fidelity management of data analyzed by machine learning models in business and search engine contexts. Machine learning models, including cognitive networks, for purposes of this disclosure are summaries of a dataset, called the training dataset. Specifically, the training dataset consists of data that represents a statistically significant number of observations of interest. The training dataset is then placed into a set of neural networks that capture the statistical behavior of selected attributes of those observations, called the machine learning model. The machine learning model is much smaller than the training dataset and can be used to predict new behaviors based on new observations, called input data. Specifically, input data is received and attribute values corresponding to those in the machine learning model are extracted. The attributes are then input into the machine learning model, and the predicted attribute values are then output. Depending on the statistics used, a confidence level may be calculated and output as well.
[0010] Consider an example in which a statistically large number of photos of dogs and cats are collected in a training dataset for the purpose of creating a machine learning model for identifying whether an image constituting new input data contains a dog or a cat. The training dataset is input to a machine learning model generator, and a machine learning model is created. The generator may be programmed to look for certain attributes, such as the shape of the nose, the distance between the eyes, etc. When the generated machine learning model receives and scans the new input data photo, it extracts those attributes, and the machine learning model then generates a probability that there is a dog or a cat in the picture. For example, the machine learning model may say that there is an 85% confidence that the input data image is of a dog.
[0011] Businesses use analytical applications, also known as analytics, for similar reasons: to make predictions about inventory, customer numbers, revenue, etc., based on historical data. Similarly, search engines use analytics to improve the accuracy and relevance of search results. However, search engines suffer from not having the most accurate data available about businesses, i.e., data that is owned by the businesses themselves.
[0012] As an example, a search engine can tell whether a user clicked on a link, but it cannot tell whether a customer actually went to a store and made a purchase, information that is held by the business itself. Similarly, a business cannot know the amount of interest in its offerings, but data from search engines provides this information.
[0013] Currently, search engines expose application programming interfaces ("APIs") to provide a means to provide business information and perform search engine optimization ("SEO"). However, currently, while some businesses expose business data via APIs, there is no automation governing the potential combination of business-owned data with search engine-owned data.
[0014] The controller can be used to advantageously manage a combination of data from both the business and the search engine. The controller may include, be implemented as, or be replaced by a rules engine. In various embodiments, the data may include metadata, operational data, and / or contextual data, and the data may be related to the business and / or the search engine. A first advantage is privacy and security management. Privacy and security laws are constantly in flux and becoming increasingly restrictive. As a result, not all business data is made public.
[0015] A second advantage of the controller is that it ensures that a machine learning model is developed that improves statistical confidence rather than invalidating the prediction. The introduction of additional attributes may improve the accuracy of the machine learning model, but only if there is a true correlation between the additional attributes and the prediction. In the previous dog and cat example, the machine learning model can be programmed to look at the background color of the image. Except for anomalies, there should be no correlation between the background color of the image and whether the image contains a dog or a cat. Thus, introducing background color as an attribute would either provide a neutral prediction, or worse, degrade the prediction of the machine learning model.
[0016] Thus, the controller may implement and / or provide the necessary logic to ensure that data received from search engine APIs and data shared by businesses via those APIs improves the predictive value of the machine learning model. If data introduced into a machine learning model improves the predictive value, the data model is said to "converge positively" or simply "converge."
[0017] A third advantage is that the controller has the ability to monitor what data is shared via the business API and quantify the benefits to the business of improved machine learning models and data models in general. Thus, if the controller exists as a third-party service, the controller may act as a billing and auditing model to monetize the services provided by the controller. In this way, such a controller can be the basis of a platform for managing data between businesses and search engines. Exemplary Configurations for High-Fidelity Data Management
[0018] FIG. 1 is an exemplary configuration 100 for implementing an exemplary embodiment of high fidelity data management. In various embodiments, such high fidelity data management may be performed in the context of a domain, which includes computing assets for an entity. These assets may reside in the cloud or outside the cloud. A business domain 102 is a set of computing assets of an enterprise 104. The enterprise 104 may have some computing assets dedicated to analytics 106 and a data store 108 that includes its operational data, metadata, etc. In general, the data in the data store 108 provides training data for the data models used by the analytics 106. Search engine platforms are also businesses and therefore have their own domains, such as a search engine domain 110. Similarly, search engines may have their own analytics 112 and data stores 114.
[0019] Due to the volume of data and computational requirements to generate the data models, the analytics 106 and 112 and data stores 108 and 114 typically reside on a cloud 116. The cloud 116 is generally a set of servers separated by a hypervisor that provides computing and storage on demand in the form of virtual machines. The cloud 116 is described in more detail with respect to FIG.
[0020] Businesses such as those encapsulated in the business domain 102 and the search engine domain 110 may expose their data and receive data via APIs. For business domains 102 that wish to collaborate with the search engine domain 110, the business domain may use a controller 118. In various embodiments, the controller 118 may include, be implemented as, or replace a rules engine. The controller 118 controls what data and functions are exposed by controlling the behavior of the business APIs 120. The controller 118 also controls which search engine APIs 122 are invoked. The controller 118 is described in more detail with respect to FIG. 3. Thus, the controller 118 provides a fully functional two-way data exchange 124, not just data exchange or function invocation via the search engine API 122 alone, where the data exchange 124 complies with the business logic of the enterprise 104. An Exemplary Environment for High-Fidelity Data Management
[0021] 2 is a block diagram of an exemplary environment 200 including hardware, software, and communications environments for implementing high fidelity data management. The functionality for performing high fidelity data management is typically hosted on a computing device. Exemplary computing devices include, but are not limited to, on the client side, mobile devices (including smartphones), tablet computers, laptops, desktop personal computers, in-vehicle navigation devices, and kiosks. Exemplary computing devices on the server side include, but are not limited to, mainframes, physical servers, and virtual machines. Generally, the computing devices are networked.
[0022] The client-side computing device or client 202 may have a processor 204, a memory 206. The processor may be a central processing unit, a repurposed graphical processing unit, and / or a dedicated controller such as a microcontroller. The computing device may further include an input / output (I / O) interface 208 and / or a network interface 210. The I / O interface 208 may be any controller card such as a Universal Asynchronous Receiver Transmitter (UART) used in conjunction with a standard I / O interface protocol such as RS-232 and / or Universal Serial Bus (USB). The network interface 210 may potentially work in conjunction with the I / O interface 208 and may be a network interface card supporting Ethernet and / or Wi-Fi and / or any number of other physical and / or data link protocols. Alternatively, the network interface 210 may be an interface to an interface system.
[0023] The memory 206 is any computer-readable medium capable of storing several software components, including an operating system 212 and software components 214, and / or other applications 216, including an Internet browser or application that integrates Internet browsing capabilities. Generally, a software component is a set of computer-executable instructions stored together as a separate whole. The operating system 212 and applications 216 are themselves software components or integrated collections of software components. Examples of software components 214 include static libraries, dynamically linked libraries, and binary executable files, such as executable programs. Other examples of software components 214 include interpreted executable files that run on a runtime, such as servlets, applets, p-code binaries, and Java binaries. The software components 214 may run in kernel mode and / or user mode.
[0024] Computer readable media includes at least two types of computer readable media: computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing 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 (DVD), Blu-Ray, or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that may be used to store information for access by a computing device. In contrast, communication media may embody computer readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transmission mechanism. As defined herein, computer storage media does not include communication media.
[0025] The server-side computing device or server 218 is any computing device capable of participating in a network. The network may be, but is not limited to, a local area network ("LAN"), a virtual private network ("VPN"), a cellular network, or the Internet. The server 218 has hardware components similar to the client-side computing device, i.e., the client 202. Specifically, it includes a processor 220, a memory 222, an input / output interface 224, and / or a network interface 226. The memory 222 may implement an operating system 228, server-side software components 230, and server-side applications 232. The server 218 differs from the client 202 in that it generally has more powerful processing power to handle concurrently running processes and a larger network capacity to communicate with multiple clients 202. The server-side software components 230 may include libraries and runtimes (e.g., for executing interpreted code). The server-side applications 232 may include web servers (also called "application servers") and database servers, as well as server software that provides functionality not typically found in legacy system applications. Exemplary server software may include a transaction monitor, a single sign-on server, an identity server, a security server (including access control list (ACL) functionality), and a network session manager (allowing multiple concurrent sessions on legacy system applications).
[0026] In general, high fidelity data management may be implemented as a software service on a physical server 218. However, such software services may also be hosted on the cloud 116 via a cloud service 236. Specifically, the cloud service 236 is comprised of multiple physical computer servers separated via a hypervisor. Each physical computer server may have one or more processors, memory, at least an input / output interface, and / or a network interface. Features and variations of the processors, memory, input / output interfaces, and network interfaces are substantially similar to those described for the physical server 218 above.
[0027] Cloud services 236 include a hypervisor that can delegate calls to any piece of hardware within the underlying physical server, and upon request, create virtual machines from separate pieces of hardware, regardless of the physical server (a process called "disaggregation"). Similar to physical servers 218, the virtual machines may host components including software applications, services, as well as virtual web server 238 functionality and virtual storage / data store 240 functionality.
[0028] The virtual machines themselves may be further divided into containers, which allow programs to run within isolated subsets of the virtual machine. Software such as Kubernetes, Mesos, and Docker are examples of container management software. Unlike virtual machines, which have startup delays due to the need to provision an entire OS, containers can be created on-demand more quickly because the underlying virtual machine is already provisioned.
[0029] Cloud services 236 may embody a service abstraction. Common examples include service abstractions such as Platform as a Service ("PAAS"), Infrastructure as a Service ("IAAS"), and Software as a Service ("SAAS"). Controller for data management and exchange
[0030] 3 is a block diagram 300 of an example controller 118 for controlling data exchange 124 between business API 120 and search engine API 122. Controller 118 manages several functions and subfunctions (herein referred to as "manager", "data store", "ACL", "rules engine" and the like) that are comprised of software components implemented via computer executable instructions and / or computer readable data. Specifically, it has an access management function 304, a data management function 306, an external API manager function 308, and an administration panel 310 for administrator 312. Functions 304-308 and administration panel 310 may be implemented via routines, program instructions, objects, and / or data structures that execute in environment 200 to perform particular tasks or implement particular abstract data types.
[0031] The access management function 304 is responsible for determining access to specific APIs and access to data within the business domain 102. It consists of a user account data store 314 for storing accounts that have access to data and / or APIs within the business domain 102. In an alternative example, the access management function 304 may not directly perform key and account management. Instead, the access management function 304 can delegate such functions to other security infrastructure within the business domain 102. The user account data store 314 includes at least identifiers of unique accounts for accessing APIs and / or data.
[0032] The access control list indicates which privileges are associated with each particular account. Thus, the access management function 304 includes a data access control list 316 and an API access control list. The data access control list 316 specifies database and data access privileges for the data stores 108 in the business domain 102. These may be implemented as SQL authorization privileges in a relational database management system or equivalent for the data stores 108 in the business domain 102. The API access control list 318 stores privileges for user accounts in the user account data store 314 to access the business API 120.
[0033] The data management function 306 determines the behavior of the business API 120 when accessed by a user account in the user account data store 314. Specifically, generally when a request to access data via the business API 120 is received, the data management function 306 identifies the accessing user and associated privileges via the access management function 304. Based on the received privileges, the view and filter manager 320 identifies the particular version of the data view that may be accessed and any additional data filters to apply. The view and filter manager 320 maps and applies the version of the view and filter to the user account.
[0034] When managing views for a relational database, a set of data can expose a limited set of fields by defining an SQL view. Some views may have many fields and other views may limit the fields. Depending on the account accessing the business API 120, the view and filter manager 320 selects which version of the view to access.
[0035] Additionally, in a relational database, an SQL view can have additional filters and conditions set on the data, thereby limiting the number of data rows returned. Thus, depending on the account accessing business API 120, view and filter manager 320 selects which filters to apply to the view.
[0036] In effect, the view and filter manager 320, upon receiving an API request and identifying the access account, applies a combination of both view management and filter management before allowing the result set to be returned via the API. In this manner, the view and filter manager 320 can manage which fields and database records are shared from the same API 120.
[0037] The view and filter manager 320 works in cooperation with the privacy manager 322. The privacy manager 322 may interface with an external automated privacy policy or store an automated privacy policy itself. In particular, access to data and APIs via the data access control list 316 and the API access control list 318, as well as the application / selection of views and filters in the view and filter manager 320, are modified according to the recommendations of the privacy manager 322. The recommendations are displayed to the administrator 312 via the administration panel 310. The administrator 312 may select which changes to implement based on the recommendations of the privacy manager 322. Alternatively, the administration panel 310 may be configured to automatically accept all privacy manager 322 recommendations. In this way, privacy and cybersecurity improvements may be fully automated.
[0038] The controller 118 is also present to ensure that the machine learning models (e.g., data models also referred to as "DM") that depend on the external APIs have data that converges to improve their predictive capabilities. The DM-API dependency manager 324 of the external API manager function 308 may store a set of identifiers of internal machine learning models from analytics 106 in the business domain 102 that are cross-correlated with the external search engine API 122. Similarly, the DM-API dependency manager 324 of the external API manager function 308 may similarly store identifiers machine learning models from analytics 112 of the search engine domain 110 that have configured sharing of machine learning model information cross-correlated with the business API 120. This allows the external API manager function 308 to track which machine learning models have dependencies on which APIs.
[0039] The data model history data store 326 stores the historical performance of the data model. In particular, it stores the predicted statistical confidence in the data model and in some cases, it stores the actual predictive performance of the data model. The rules engine 328 reviews the history and makes recommendations on what changes should be made to the API access based on predefined thresholds set by the administrator 312 via the administrator panel 310 or based on the machine learning model itself. In particular, recommendations from the rules engine 328 may specify access to additional APIs, removal of APIs that do not contribute to the model's predictive value, or may recommend changes to applied views and filters. These results are presented to the administrator 312 via the administration panel 310. Upon approval of the administrator 312, the access management function 304 and / or the data management function 306 may implement all or part of the recommendations of the rules engine 328. In some cases, the administration panel 310 may be configured to automatically accept recommendations from the rules engine 328. In this manner, data and API modifications may be fully automated. Process Example
[0040] 4 and 5 present exemplary processes 400 and 500 for implementing a remote trigger of a personal device to perform an action in response to a detected event. Each of the processes 400 and 500 is illustrated as a collection of blocks in a logical flow chart that represents a sequence of operations that may be implemented in hardware, software, or a combination thereof. In the context of software, the blocks represent computer-executable instructions that, when executed by one or more processors, perform the recited operations. Generally, computer-executable instructions may include routines, programs, objects, components, data structures, etc. that perform particular functions or implement particular abstract data types. The order in which the operations are described is not intended to be construed as a limitation, and any number of the described blocks can be combined in any order and / or in parallel to implement the process. For purposes of illustration, the processes 400 and 500 are described with reference to the configuration 100 of FIG. 1 and the environment 200 of FIG. 2.
[0041] 4a and 4b show a flow diagram of an example process 400 used by the controller to manage data access to data stored in a data store of a business domain. At block 402, the controller 118 may receive an access request including a user account identifier from an application on a computing device, the access request requesting access to data stored in the data store of the business domain. At block 404, the controller 118 may determine, based on a data access control list of the business domain, whether a user account having the user account identifier has access to the data stored in the data store of the business domain.
[0042] If, at decision block 406, the controller 118 determines that the user account having the user account identifier has access to the data (“Yes” at decision block 406), the process 400 may proceed to block 408. At block 408, the controller 118 may determine, based on an API access control list for the business domain, whether the user account having the user account identifier has access to one or more business APIs of the business domain that provide access to data stored in the data store.
[0043] If, at decision block 410, the controller 118 determines that the user account having the user account identifier has access to one or more business APIs based on the API access control list for the business domain ("Yes" at decision block 410), the process 400 may proceed to block 412. At block 412, the controller 118 may select a data view from a plurality of views for viewing the data and one or more data filters to apply to the data based on one or more data access privileges for the user account having the user account access identifier. In some embodiments, the data access privileges for the user account having the user account access identifier may be stored in the data access control list.
[0044] At block 414, the controller 118 may determine whether to modify the data view selected for viewing based on the one or more privacy policies. In various embodiments, the one or more privacy policies may be stored in an internal database of the controller 118 or an external database. If, at decision block 416, the controller 118 determines that the data view selected for viewing should be modified ("Yes" at decision block 416), the process 400 may proceed to block 418. At block 418, the controller 118 may select an alternative view of the multiple data views specified by the one or more privacy policies for viewing the data. In various embodiments, the selection may be made automatically by the controller 118 or in response to an input made by an administrator via an administrative interface of the controller 118.
[0045] At block 420, the controller 118 may determine whether to modify one or more data filters to apply to the data based on the one or more privacy policies. If at decision block 422 the controller 118 determines that one or more data filters should be modified ("Yes" at decision block 422), the process 400 may proceed to block 424. At block 424, the controller 118 may select one or more alternative data filters specified by the one or more privacy policies to apply to the data. In various embodiments, the selection may be made automatically by the controller 118 or in response to input made by an administrator via an administrative interface of the controller 118.
[0046] At block 426, the controller 118 may apply the one or more selected data filters to the data stored in the data store of the business domain to generate filtered data. At block 428, the controller 118 may provide the selected data view of the filtered data to an application on the computing device via one or more business APIs. For example, the controller 118 may allow an application on the computing device to access the selected view of the filtered data via the one or more business APIs.
[0047] Returning to decision block 406, if the controller 118 determines that the user account having the user account identifier does not have access to the data ("no" at decision block 406), the process 400 may proceed to block 430. At block 430, the controller 118 may deny the application on the computing device access to the data stored in the data store of the business domain. Returning to decision block 410, if the controller 118 determines that the user account having the user account identifier does not have access to one or more business APIs based on the API access control list of the business domain ("no" at decision block 410), the process 400 may proceed to block 430.
[0048] 5a and 5b show a flow diagram of an example process 500 for controlling data exchange between a business API and a search engine API. In block 502, the controller 118 may store a first plurality of machine learning model identifiers for a first machine learning model of a business domain, such as the business domain 102, in a data store of the business domain. In block 504, the controller 118 may determine one or more corresponding search engine APIs of the plurality of search engine APIs used by each of the first machine learning models of the business domain to receive data from the search engine domain. For example, the controller 118 may make a decision based on information the controller 118 obtains from the analytics 106.
[0049] At block 506, the controller 118 may store, based on the determination, cross-correlation information in a data store that cross-correlates each of the first plurality of machine learning model identifiers of the first machine learning model with one or more corresponding search engine APIs.
[0050] At block 508, the controller 118 may receive a second plurality of machine learning model identifiers for a second machine learning model of a search engine domain, such as the search engine domain 110. In various embodiments, the controller 118 may receive such identifiers from the search domain. At block 510, the controller 118 may receive additional cross-correlation information correlating the second plurality of machine learning model identifiers with one or more corresponding business APIs of the business domain, each of the second machine learning models being configured to receive data from the business domain via one or more corresponding business APIs. In various embodiments, the controller 118 may receive such information from the search domain. At block 512, the controller 118 may store the second plurality of machine learning model identifiers and the additional cross-correlation information in a data store.
[0051] At block 514, the controller 118 may determine whether to configure the machine learning model for the business domain or the search engine domain to receive data from one or more newly cross-correlated APIs or to terminate receiving data from at least one currently cross-correlated API. In various embodiments, the controller 118 may make such a determination based on performance data of the machine learning model, such as one or more of the historical performance of the machine learning model, the predicted statistical confidence in the machine learning model, or the actual predictive performance of the machine learning model.
[0052] At decision block 516, if the machine learning model in the business domain or the search engine domain is to receive data from one or more newly cross-correlated APIs ("YES" at decision block 516), the process 500 may proceed to block 518. At block 518, the controller 118 may store information about a new cross-correlation between a machine learning model identifier of the machine learning model in the business domain or the search engine domain and at least one API of a different domain of the business domain or the search engine domain when the machine learning model is newly configured to receive data from at least one API. For example, the new cross-correlation may be of a machine learning identifier of a machine learning model in the business domain to at least one API of the search engine domain, or vice versa.
[0053] However, if the machine learning model for the business domain or the search engine domain is not to receive data from one or more newly cross-correlated APIs ("no" at decision block 516), process 500 may proceed to decision block 520. If, at decision block 520, the machine learning model for the business domain or the search engine domain should cease receiving data from one or more currently cross-correlated APIs ("yes" at decision block 520), process 500 may proceed to block 522.
[0054] At block 522, the controller 118 may delete information about an existing cross-correlation between a machine learning model identifier of the machine learning model in the business domain or the search engine domain and one or more APIs of a different domain of the business domain or the search engine domain when the machine learning model is configured to terminate receiving data from the one or more APIs. For example, the termination may include terminating an existing cross-correlation between a machine learning model identifier of the machine learning model in the business domain and one or more APIs of the search engine domain, or vice versa.
[0055] Process 500 may then loop back to block 514. However, if the machine learning model for the business domain or the search engine domain should not finish receiving data from the one or more currently cross-correlated APIs ("No" at decision block 520), process 500 may also loop back to block 514. Exemplary Use Cases for High-Fidelity Data Management
[0056] There are many use cases for high-fidelity data management as a platform. Depending on the use case, different monetization models can be applied. Improved surfacing of stores
[0057] In one use case, current search engines are known to not provide accurate recommendations. For example, a person searching for a coffee shop may be directed to a coffee shop that is farther away than other sources of coffee. This may occur because the search engines only identify coffee-specific outlets as candidates for surfacing. This ignores alternative general sources of coffee, such as convenience stores.
[0058] By providing search engines with data directly from businesses with a full listing of services, search engines can surface not only specialty stores, but general stores as well. Examples of data surfacing include store locations and availability hours, as well as listings of services such as car washes, fuel / diesel, electric vehicle charging, alcohol, food, etc. Machine learning model evolution matching data evolution
[0059] Businesses are constantly collecting data and modifying their collections. Many businesses, such as consumer stores, have separate mobile applications for users to crowdsource information. Thus, as different sources of data are collected as training data, the rules engine 328 allows the data model to improve over time. In this way, the machine learning model evolves as the data quality evolves.
[0060] The administrator 312 is aware of changes in data sources. For example, the initial data source may only be that received from the search engine via the search engine API 122. Local business data stored in the data store 108 may then be integrated with the training data. Finally, crowdsourced information collected via users' mobile applications may be further integrated. The rules engine 328 enables the administrator 312 to determine which data to integrate with which access privileges while ensuring positive data convergence. White-labeling API
[0061] Since the controller 118 resides in the business domain 102 and has full control over the enterprise 104, it may provide an additional business API 120 that provides an amalgam of search engine data that is exposed via a search engine API 122. An example may be a business API 120 that exposes recommendations based on a machine learning model consisting of data from both the search engine domain 110 and the business domain 102.
[0062] In another example, simply sharing data from both the search engine domain 110 and the business domain 102 in a merged set via the business API 120 may be performed.
[0063] If permitted to do so, the business domain 102 may expose its own APIs that pass through search engine domain data. This is called a white-labeling API. In this way, partners of the enterprise 104 may have a single location and source of APIs. Monetization
[0064] In some embodiments, the functionality of the controller 118 may be resold. The controller 118 may house a monitor that communicates with an administration panel 310. The administration panel 310 may include an application interface that provides access to the business API 120, access to the data in the data store 108, and accepted improvements to the machine learning models via a rules engine 328. This data may be further stored in a separate data store called an access tracking data store (not shown).
[0065] Using the data in the access tracking data store, a business may create a monetization model to charge for the functionality of the controller 118. The monetization model may include a per-access model or a subscription model. One of the main challenges in monetization is demonstrating value. Since the controller 118 stores the improved performance of the machine learning model via the data model history data store 326, it can not only demonstrate value but also charge for the monetization model based on the degree of improvement, if desired.
[0066] Although embodiments of high-fidelity data management are described above with respect to the business domain and the search engine domain, alternative embodiments of such high-fidelity data management may be applied to other types of data management between two domains. conclusion
[0067] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing any claims based on this disclosure.
Claims
1. One or more non-transitory computer-readable media storing computer-executable instructions that, when executed, cause one or more processors to: receiving an access request, including a user account identifier, from an application on a computing device requesting access to data stored in a data store of a business domain; determining whether a user account having the user account identifier has access to the data stored in the data store of the business domain based on a data access control list of the business domain, the user account providing access to the data stored in the data store of the business domain; determining whether the user account having the user account identifier has access to one or more business APIs of the business domain that provide access to the data stored in the data store of the business domain based on an application program interface (API) access control list of the business domain; selecting, in response to the user account having the user account identifier having access to the data stored in the data store of the business domain and the one or more business APIs of the business domain, based on one or more data access privileges for the user account having the user account identifier, one data view of a plurality of data views for viewing the data stored in the data store of the business domain and one or more data filters to apply to the data, wherein the data stored in the data store of the business domain is provided to the application on the computing device via the one or more business APIs of the business domain; The one or more non-transitory computer-readable media configured to cause a computer to perform operations including:
2. 10. The one or more non-transitory computer-readable media of claim 1, further comprising providing a data view of the data filtered by the one or more data filters to the application on the computing device via the one or more business APIs of the business domain.
3. 10. The one or more non-transitory computer-readable media of claim 1, further comprising: denying the application on the computing device from accessing the data stored in the data store of the business domain in response to at least one of the user account having the user account identifier not having access to the data stored in the data store of the business domain or not having access to the one or more business APIs of the business domain.
4. The operation includes: determining whether the data view for viewing the data stored in the data store of the business domain and the one or more data filters for filtering the data stored in the data store of the business domain have been modified based on one or more privacy policies; in response to selecting an alternative data view for viewing the data and applying one or more alternative data filters to the data based on the one or more privacy policies, providing the alternative data view of the data filtered by the one or more alternative data filters to the application on the computing device via the one or more business APIs of the business domain; in response to selecting the alternative data view for viewing the data without applying the one or more alternative data filters to the data based on the one or more privacy policies, providing the alternative data view of the data filtered by the one or more data filters to the application on the computing device via the one or more business APIs of the business domain; in response to applying the one or more alternative data filters without selecting the alternative data view for viewing the data, providing the data view of the data filtered by the one or more alternative data filters to the application on the computing device via the one or more business APIs of the business domain; in response to not selecting the alternative data view for viewing the data and not applying the one or more alternative data filters to the data based on the one or more privacy policies, providing the data view of the data filtered by the one or more data filters to the application on the computing device via the one or more business APIs of the business domain; 13. The one or more non-transitory computer-readable media of claim 1, further comprising:
5. 10. The one or more non-transitory computer-readable media of claim 1, wherein the data view is configured to provide a limited set of fields of the data for viewing by the application, and the one or more data filters are configured to at least limit a number of data rows of the data provided to the application.
6. The one or more non-transitory computer-readable media of claim 1 , wherein the data store of the business domain stores operational data and metadata associated with one or more businesses.
7. The one or more non-transitory computer-readable media of claim 6 , wherein the one or more businesses include at least one convenience store.
8. The one or more non-transitory computer-readable media of claim 1 , wherein the user account identifier identifies a user account associated with a search engine domain.
9. 10. The one or more non-transitory computer-readable media of claim 8, wherein the computing device is included in the search engine domain, the search engine domain including a search engine that provides search results for a search based at least on the data provided to the computing device via the one or more business APIs.
10. storing in the data store for the business domain a plurality of machine learning model identifiers for machine learning models for the business domain; determining one or more corresponding search engine APIs of a plurality of search engine APIs used by each of the machine learning models of the business domain to receive data from the search engine domain; and storing, based on the determining, cross-correlation information in the data store that cross-correlates each of the plurality of machine learning model identifiers of the machine learning model with one or more search engine APIs; Attribute values in the data from the search engine domain that correspond to attribute values in the machine learning model of the business domain are used by the machine learning model of the business domain to generate predicted attributes associated with the search engine domain.
9. One or more non-transitory computer-readable media as recited in claim 8.
11. receiving additional machine learning model identifiers for additional machine learning models for the search engine domain; and receiving additional cross-correlation information correlating the additional plurality of machine learning model identifiers with one or more corresponding business APIs of the business domain, each of the additional machine learning models configured to receive data from the business domain via the one or more corresponding business APIs; storing the additional plurality of machine learning model identifiers and the additional cross-correlation information in the data store for the business domain; additional attribute values in the data from the business domain that correspond to additional attribute values in the machine learning model of the search engine domain are used by the machine learning model of the search engine domain to generate additional predicted attributes related to the business domain.
11. One or more non-transitory computer-readable media as recited in claim 10.
12. storing information about a new cross-correlation between a machine learning model identifier of a machine learning model in the business domain or the search engine domain and at least one API of a different domain of the business domain or the search engine domain when the machine learning model is newly configured to receive data from the at least one API; or When the machine learning model is configured to terminate receiving data from the one or more APIs, deleting information regarding an existing cross-correlation between the machine learning model identifier of the machine learning model in the business domain or the search engine domain and one or more APIs of a different domain of the business domain or the search engine domain; 12. The one or more non-transitory computer-readable media of claim 11, further comprising:
13. 13. The one or more non-transitory computer-readable media of claim 12, wherein configuring the machine learning model to receive data from the at least one API or configuring the machine learning model to terminate receiving data from the one or more APIs is initiated based at least on one or more of a past performance of the machine learning model, a predicted statistical confidence in the machine learning model, or an actual predictive performance of the machine learning model.
14. 13. The one or more non-transitory computer-readable media of claim 12, wherein the configuring of the machine learning model to receive data from the at least one API or the configuring of the machine learning model to terminate receiving data from the one or more APIs is initiated by user input entered in response to a recommendation generated based on one or more of a historical performance of the machine learning model, a predicted statistical confidence in the machine learning model, or an actual predictive performance of the machine learning model.
15. A computer-implemented method comprising: storing, via one or more computing devices, a plurality of machine learning model identifiers for machine learning models of a first domain in a data store of the first domain; determining a corresponding search engine API of one or more of a plurality of second domain APIs used by each of the machine learning models of the first domain to receive data from a second domain; storing, based on the determining, cross-correlation information in the data store that cross-correlates each of the plurality of machine learning model identifiers of the machine learning model with one or more second domain APIs; 4. A computer-implemented method comprising:
16. receiving, via the one or more computing devices, an additional plurality of machine learning model identifiers for additional machine learning models of a second domain; and receiving, via the one or more computing devices, additional cross-correlation information correlating the additional plurality of machine learning model identifiers with one or more corresponding first domain APIs of the first domain, wherein each of the additional machine learning models is configured to receive data from the first domain via the one or more corresponding first domain APIs; storing, via the one or more computing devices, the additional plurality of machine learning model identifiers and the additional cross-correlation information in the data store of the first domain; 20. The computer-implemented method of claim 15, further comprising:
17. storing, via the one or more computing devices, information regarding a new cross-correlation between a machine learning model identifier of a machine learning model in the first domain or the second domain and at least one API of a different domain of the first domain or the second domain when the machine learning model is newly configured to receive data from the at least one API; or When the machine learning model is configured to terminate receiving data from the one or more APIs, deleting, via the one or more computing devices, information regarding an existing cross-correlation between the machine learning model identifier of the machine learning model in the first domain or the second domain and one or more APIs of a different domain of the first domain or the second domain; 20. The computer-implemented method of claim 15, further comprising:
18. 20. The computer-implemented method of claim 17, wherein configuring the machine learning model to receive data from the at least one API or configuring the machine learning model to terminate receiving data from the one or more APIs is initiated based on at least one or more of a past performance of the machine learning model, a predicted statistical confidence in the machine learning model, or an actual predictive performance of the machine learning model.
19. 18. The computer-implemented method of claim 17, wherein attribute values in the data from the second domain that correspond to attribute values in the machine learning model of the first domain are used by the machine learning model of the first domain to generate predicted attributes related to the second domain, and additional attribute values in the data from the first domain that correspond to additional attribute values in the machine learning model of the second domain are used by the machine learning model of the second domain to generate additional predicted attributes related to the first domain.
20. One or more processors and a memory including a plurality of computer-executable components executable by the one or more processors, the plurality of computer-executable components comprising: an access management functionality component that controls access by user accounts associated with a search engine domain to data stored in a data store of the business domain based on a user account access control list and an application program interface (API) access control list that controls access to a business API of the business domain; a data management functional component that selects one of a plurality of data views for viewing the data stored in the data store of the business domain and one or more data filters to apply to the data based on at least one of one or more data access privileges for the user account associated with the search engine domain or one or more privacy policies; an external API manager component that stores, within the data store of the business domain, at least cross-correlation information correlating a plurality of machine learning model identifiers of machine learning models of the search engine domain with one or more corresponding business APIs of the business domain, each of the machine learning models being configured to receive data from the business domain via the one or more corresponding business APIs; A system including:
Citation Information
Patent Citations
Archival data identification
JP2017228302A
Enabling chatbots by detecting and supporting affective argumentation
US20200410166A1
