US20160188733A1

US20160188733A1 - Generating Recommendations Based on Clustered Application States

Info

Publication number: US20160188733A1
Application number: US14/981,714
Authority: US
Inventors: Eric Glover; Leigh L. Klotz, Jr.; Dinesh MISHRA
Original assignee: Quixey Inc
Current assignee: Samsung Electronics Co Ltd
Priority date: 2014-12-29
Filing date: 2015-12-28
Publication date: 2016-06-30
Also published as: US10157230B2; US10387505B2; US20160189225A1; US20160188719A1

Abstract

A deep linking system includes a storage system and a recommendation engine. The storage system stores a plurality of application records. The storage system further stores a plurality of cluster records, each cluster record defining a respective cluster of a plurality of clusters, each cluster including a respective plurality of clustered state identifiers, whereby the state identifiers are clustered according to one or more features. The recommendation engine includes one or more processors configured to receive a recommendation request and to identify cluster records from the plurality of cluster records based on the recommendation request. The identified cluster records indicate one or more clusters to which the state identifier of the recommendation request is related. The one or more processors are further configured to select one or more state identifiers from the identified cluster records, generate recommendation results based on the selected state identifiers, and transmit the recommendation results to the remote device.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of U.S. Provisional Application No. 62/097,508, filed on Dec. 29, 2014. The entire disclosure of the application referenced above is incorporated herein by reference.

TECHNICAL FIELD

This disclosure relates to clustering states of software applications and using the clusters to generate recommendations to application states.

BACKGROUND

Many software applications offer multiple functions to users. Further, most users do not use all the functions of the software application. For example, a restaurant related software application may offer functions for finding restaurants by cuisine, viewing photographs of dishes, search menus of restaurants, and read reviews of specific restaurants. Most users of this software application, however, may only use the software application to view photographs of dishes. Another software application may offer similar functions; however, users of the other application may overwhelmingly prefer the search by cuisine and read review functions of this application.

SUMMARY

The details of one or more implementations of the disclosure are set forth in the accompanying drawings and the description below. Other aspects, features, and advantages will be apparent from the description and drawings, and from the claims.
In one example, the present disclosure is directed to a deep linking system comprising a storage system and a recommendation engine. The storage system includes one or more storage devices storing a plurality of application records, each application record including i) a state identifier that indicates a state of a respective software application, and ii) application state information corresponding to the state of the software application. The one or more storage devices further store a plurality of cluster records, each cluster record defining a respective cluster of a plurality of clusters identified by the deep linking system, each cluster including a respective plurality of clustered state identifiers, each clustered state identifier identifying a state of a respective software application, wherein the plurality of clusters are clustered according to one or more features. The recommendation engine includes a processing system, the processing system comprising one or more processors that execute computer-readable instructions. The one or more processors may be configured to receive a recommendation request containing a received state identifier from a remote device. The one or more processors may further be configured to identify one or more cluster records from the plurality of cluster records using the received state identifier of the recommendation request. The one or more identified cluster records respectively indicate one or more clusters to which the state of the software application defined in the recommendation request is related. The one or more processors may further be configured to select one or more state identifiers from the identified cluster records, the selected state identifiers respectively corresponding to one or more application states to recommend in response to the recommendation request. The one or more processors may further be configured to generate recommendation results and transmit the recommendation results to the remote device.
In another example, the present disclosure is directed to a method comprising maintaining, by a processing system including one or more processors, a plurality of application records on a storage device, each application record including i) a state identifier that indicates a state of a respective software application, and ii) application state information corresponding to the state of the application. The method further comprises maintaining a plurality of cluster records on a storage device, each cluster record defining a respective cluster of a plurality of clusters, each cluster including a respective plurality of clustered state identifiers, each clustered state identifier identifying a state of a respective software application, wherein the plurality of clusters are clustered according to one or more features. The method further comprises receiving a recommendation request containing a function identifier from a remote device. The method further comprises identifying one or more cluster records from the plurality of cluster records using the state identifier of the recommendation request, wherein the identified one or more cluster records respectively indicate one or more clusters to which the state of the software application defined by the recommendation request belongs. The method further comprises selecting one or more state identifiers from the one or more cluster records, the selected state identifiers respectively corresponding to one or more application states to recommend to the user. The method further comprises generating recommendation results based on the states indicated by the one or more selected state identifiers, the recommendation results including one or more result objects, each result object capable of being rendered into a user-selectable link. The method further comprises transmitting the recommendation results to the remote device.

DESCRIPTION OF DRAWINGS

FIG. 1A is a schematic view of an example environment of a deep-linking system.

FIG. 1B is a schematic views of a user device displaying example recommended links.

FIG. 2 is a schematic view of example components of a user device

FIG. 3A is a schematic view of example components of an analytics engine.

FIG. 3B is a schematic view of an example application state record.

FIG. 3C is a schematic view of an example cluster record.

FIG. 3D is a schematic view of an example entity ontology.

FIG. 3E is a schematic view of an example functional ontology.

FIG. 4 is a schematic view of example components of a recommendation engine.

FIG. 5 is a flow chart illustrating an example set of operations of a method for processing a recommendation request.

Like reference symbols in the various drawings indicate like elements.

DETAILED DESCRIPTION

The figures and following description relate to example implementations by way of illustration only. It should be noted that from the following discussion, alternative implementations of the structures and methods disclosed herein will be readily recognized as viable alternatives that may be employed without departing from the scope of the disclosure. This disclosure describes techniques for generating recommendations to application states using functional clusters.
The present disclosure relates to recommending states of software applications. A deep-linking system of the present of the present disclosure can include an analytics engine and a recommendation engine. The analytics engine of the present disclosure can be configured to collect data from consenting user devices. The analytics engine can further generate and store information regarding applications and their application states in application records, whereby each application record corresponds to a particular application state. The analytics engine can further identify and group application records into clusters based on the function the corresponding application state performs and the various features associated with that particular application state (e.g., geographic location, time of day, popularity).
The recommendation engine can receive a recommendation request (e.g., from a search system and/or user device), identify one or more application states that satisfy the recommendation request, and generate a set of recommendation results. The recommendation engine can transmit the recommendation results to a requesting user device, server device, or any other suitable recipient. The recommendation engine can leverage the clusters of application states generated by the analytics engine to provide recommendations. The recommendation engine can communicate with the analytics engine to identify clusters of application states that are clustered according to features indicated by the recommendation request. In some implementations, the recommendation engine can select application states from clusters clustered according to features not explicitly indicated by the recommendation request, such as the time of day a recommendation request was sent. In this way, the recommendation engine can accurately select application states that optimally satisfy the recommendation request. In other words, the recommendation engine can leverage the analytics engine to narrow the universe of application states from which to make recommendations to a cluster of application states that have already been determined to be relevant, by the analytics engine, to a particular set of features.
FIG. 1A illustrates an example environment 10 of a deep-linking system 100. A deep-linking system 100 is a collection of computing devices that generates user-selectable links that link to states of software applications and provides the user-selectable links to one or more user devices 200. A user selectable link (or link) is an object that is displayed by a user device 200 that includes one or more underlying access mechanisms (described in greater detail below). When a user selects a user selectable link, the user device 200 can access a state of a software application using an access mechanism included in the selected link. While the user device 200 in FIG. 1A is depicted as a smartphone, a user device 200 can be any suitable user computing device including, but not limited to, a tablet computing device, a personal computing device, a laptop computing device, a gaming device, a vehicle infotainment device, and/or a smart appliance (e.g., smart refrigerator or smart television).
The deep-linking system 100 can include a recommendation engine 400. In the illustrated example, the deep-linking system 100 further includes an analytics engine 300. The analytics engine 300 collects usage data 102 from a plurality of user devices 200. Usage data 102 can include any data that indicates the states of software applications that a user device 200 is accessing.
A software application can refer to a software product that causes a computing device to perform a function. In some examples, a software application may also be referred to as an “application,” “an app,” or a “program.” Example software applications include, but are not limited to, productivity applications, social media applications, messaging applications, media streaming applications, social networking applications, and games. Software applications can perform a variety of different functions for a user. For example, a restaurant reservation application can make reservations for restaurants. As another example, an internet media player application can stream media (e.g., a song or movie) via the Internet. In some examples, a single software application can provide more than one function. For example, a restaurant reservation application may also allow a user to retrieve information about a restaurant and read user reviews for the restaurant in addition to making reservations. As another example, an internet media player application may also allow a user to perform searches for digital media, purchase digital media, generate media playlists, and share media playlists. The functions of an application can be accessed using native application editions of the software application and/or web application editions of the software application.
A native application edition (or “native application”) is, at least in part, installed on a user device 200. In some scenarios, a native application is installed on a user device 200, but accesses an external resource (e.g., an application server) to obtain data and/or instruction from the external resource. For example, social media applications, weather applications, news applications, and search applications may respectively be accessed by one or more native application editions that execute on various user devices 200. In such examples, a native application can provide data to and/or receive data from the external resource while accessing one or more functions of the software application. In other scenarios, a native application is installed on the user device 200 and does not access any external resources. For example, some gaming applications, calendar applications, media player applications, and document viewing applications may not require a connection to a network to perform a particular function. In these examples, the functionality of the software application is encoded in the native application editions itself. The native application edition is able to access the functions of the software application without communicating with any other external devices.
Web application editions (also referred to as “web applications”) of a software application may be partially executed by a user device 200 (e.g., by a web browser executed by the user device 200) and partially executed by a remote computing device (e.g., a web server or application server). For example, a web application may be an application that is executed, at least in part, by a web server and accessed by a web browser (e.g., a native application) of the user device 200. Example web applications may include, but are not limited to, web-based email, online auctions websites, social-networking websites, travel booking websites, and online retail websites. A web application accesses functions of a software product via a network. Examples implementations of web applications include websites and/or HTML-5 application editions.
In some scenarios, a software application may be accessed by one or more native application editions of the software application and/or one or more web application editions of the software application. In these scenarios, there may be overlap between the states or functions that the native application edition(s) can access and the states or functions that the web application edition can access. For example, a restaurant review application may have reviews of thousands of restaurants and may also provide an on-line ordering function from some of the restaurants. The restaurant review application may be accessed by a first native application edition configured for a first operating system (e.g., the ANDROID operating system maintained by Google, Inc.), a second native application edition configured for a second operating system (e.g., the IOS operating system developed by Apple, Inc.), and a web application edition (e.g., via a web browser) of the restaurant review application. The restaurant review application may allow all the editions (native and web) to access the various reviews of restaurants but may only allow on-line orders to be placed using the native application editions. In this way, some states or functions of the restaurant review application cannot be accessed by the web application edition but there is overlap between the states or functions that can be accessed by the native application editions and the web application edition.
A state of a software application can refer to a parameterized function of the software application. A software application can perform one or more functions. A function is a service of the software application that can be accessed by a user device 200 via an edition of the software application. Non-limiting examples of functions can include “making a restaurant reservation” (which may parameterized with a restaurant identifier, a date, and a time), “searching for a cuisine” (which may be parameterized with a cuisine type and a location), “view flight prices” (which may be parameterized with departure and arrival airport codes, arrival and departure dates, and round trip flags), “request a driver” (which may be parameterized with a pick-up location), and “view a file” (which may be parameterized with a file identifier). A state of a software application can be accessed from a user device using an edition of the software application. An operating system of a user device 200 can instruct an edition of a software application to access a state of the software application using an access mechanism (e.g., a resource identifier 106). In some implementations, the state of a software application may be represented by a function ID 104.
A user device 200 can access a state of a software application via an edition of the software application using an access mechanism. When rendering a user selectable link (e.g., in the set of search results 460) a user device 200 displays the user selectable link such that the link can be selected by a user of the user device 200. A user selectable link may include one or more underlying access mechanisms. A user selectable link, when selected by a user, causes the user device 200 to access a state of the software application using an edition of the software application identified by the access mechanism.
Access mechanisms may include at least one of a native application access mechanism (hereinafter “application access mechanism”), a web access mechanism, and an application download mechanism. The user device 200 may use the access mechanisms to access functionality of applications. For example, the user may select a user selectable link including an access mechanism in order to access functionality of an application indicated in the user selectable link. As described herein, the deep-linking system 100 may transmit one or more application access mechanisms, one or more web access mechanisms, and one or more application download mechanisms to the user device 200 in the recommendation results 460.
An application access mechanism may be a string that includes a reference to a native application e.g., one of native applications 212 installed on the user device 200) and indicates one or more operations for the user device 200 to perform. If a user selects a user selectable link including an application access mechanism, the user device 200 may launch the native application referenced in the application access mechanism and perform the one or more operations indicated in the application access mechanism.
A web access mechanism may include a resource identifier that includes a reference to a web resource (e.g., a page of a web application/website). For example, a web access mechanism may include a uniform resource locator (URL) (i.e., a web address) used with hypertext transfer protocol (HTTP). If a user selects a user selectable link including a web access mechanism, the user device 200 may launch the web browser application 216 and retrieve the web resource indicated in the resource identifier. Put another way, if a user selects a user selectable link including a web access mechanism, the user device 200 may launch the web browser application 216 and access a state (e.g., a page) of a web application/website. In some examples, web access mechanisms may include URLs for mobile-optimized sites and/or full sites.
An application download mechanism may indicate a site (e.g., a digital distribution platform where a native application can be downloaded in the scenario where the native application is not installed on the user device 200. If a user selects a user selectable link including an application download address, the user device 200 may access a digital distribution platform from which the referenced native application may be downloaded. The user device 200 may access a digital distribution platform using at least one of the web browser application 216 and one of the native applications 212.
As previously described, a function ID 104 is a string of alphabetic, numeric, and/or symbolic characters (e.g., punctuation marks) that uniquely identifies a state of an application. Put another way, a function ID 104 is a unique reference to a state of an application. In some implementations, a function ID 104 can be in the format of a resource identifier. For example, the function ID 104 may be a uniform recourse locator (URL) or an application resource identifier. In these implementations, the function ID 104 may be used by a user device to access a software application via a web application edition or one or more native application editions of the software application, respectively.
In some implementations, a function ID 104 can map to one or more access mechanisms. In these implementations, a function ID 104 may map to a web resource identifier (e.g., a URL), one or more application resource identifiers, and/or one or more scripts. For instance, a state of an example software application, “exampleapp,” may be accessed via a web application edition and two native application editions (e.g., an edition configured for the ANDROID operating system and an edition configured for the WINDOWS PHONE operating system). In this example, the web resource identifier may be www.exampleapp.com/param1=abc&param2=xyz, the first application resource identifier may be android.exampleapp::param1=abc&param2=xyz, and the second application resource identifier may be windows.exampleapp::param1=abc&param2=xyz. In this example, a function ID 104 may map to the web resource identifier and the two application resource identifiers.
In some implementations, a function ID 104 may have a URL-like structure that utilizes a namespace other than http://, such as “func://” which indicates that the string is a function ID 104. In the example of “exampleapp” above, the function ID 104 corresponding to the example state may be func://exampleapp::param1=abc&param2=xyz, which may map to the access mechanisms described above. In this example, the function ID 104 can be said to be parameterized, whereby the value of “param1” is set to “abc” and the value of “param2” is set equal to “xyz.”
In some implementations, a function ID 104 may take the form of a parameterizable function. For instance, a function ID 104 may be in the form of “app_id[action(param_1, param_2, . . . , parameter_n)]”, where app_id is an identifier (e.g., name) of a software application, action is an action that is performed by the application (e.g., “view menu”), and parameter_1 . . . parameter_n are n parameters that the software application receives in order to access the state corresponding to the action and the parameters. Drawing from the example above, a function ID 104 may be “exampleapp[example action(abc, xyz)]”. In this example, the function ID 104 can be said to be parameterized, whereby the value of “param1” is set to “abc” and the value of “param2” is set equal to “xyz.” Given this function ID 104 and the referencing schema of the example application, the foregoing function ID 104 may be used to generate or look up the access mechanisms defined above. Furthermore, while function IDs 104 have been described with respect to resource identifiers, a function ID 104 may be used to generate or look up one or more scripts that access a state of a software application. Further, a function ID 104 may take any other suitable format. For example, the function ID 104 may be a human-readable string that describes the state of the application to which the function ID 104 corresponds.
As previously mentioned, the analytics engine 300 collects usage data 102 from a plurality of user devices 200, whereby the usage data 102 can indicate a resource identifier 106, a function ID 104, and/or search activity log 108. In some implementations, the analytics engine 300 identifies the states of software applications accessed by users of user devices 200 that provided the usage data 102. The usage data may be anonymized, thereby minimizing risks to users' privacy. A user device 200 can be configured to collect and transmit usage data 102 only when a user explicitly agrees to share such usage data 102.
In some implementations, the analytics engine 300 generates or updates application state records 332 (FIG. 3B) corresponding to the received usage data 102. As will be discussed, an application state record corresponds to a particular state of a software application. In some implementations, the application state record includes a function ID 104 corresponding to the state as well as a set of features. The features can include ontological features corresponding to the state and/or statistical features regarding the state. For example, ontological features of the state may include entity information as well as function information (e.g., view menu or make flight reservation).
Entity information corresponding to a state of a software application can define one or more entities corresponding to the state and the entity types of the respective entities. An entity can refer to a value (e.g., a noun or number) that is known to the deep-linking system 100 and able to be categorized by an ontology of the deep-linking system 100. An entity type can refer to a categorization of an entity. An entity can have more than one entity type. For example, the entity “New York Yankees” may have the entity types “professional sports team,” “New York professional sports team,” and “Major League Baseball team” associated therewith. Furthermore, entity types may be subtypes of other entity types. For example, the entity type “New York professional sports teams” may be a subtype of the entity type “professional sports teams.” As will be discussed in greater detail below, the relationship between entities and entity types may be defined in accordance with an entity ontology (FIG. 3D).
Functional information corresponding to a state of an application can define a function performed by the software application when in the given state. In some implementations, a given state can perform more than one function (e.g., make reservations for a restaurant and see reviews of the restaurant). The functional information corresponding to a known collection of software applications may be defined according to a functional ontology, which is discussed below (FIG. 3E).
Statistical features can include any statistics regarding a state of a software application. The statistics may be based, at least in part, on the usage data 102 received by the analytics engine 300. Examples of statistical features may include a value indicating how many times the state of the software application is accessed by user devices and a rate at which the state is accessed. The record may further store geographical features. Geographical features may indicate a location from which the state was accessed and a language corresponding to the state.
The analytics engine 300 can identify clusters of application state records 332 (or function IDs 104 thereof) based on the various features. The process by which a cluster is identified may be referred to as “clustering.” In some implementations, the analytics engine 300 clusters application state records based on a specific set of features. For example, the analytics engine 300 may cluster records based on entity types and location. The resulting clusters 110 may, for example, result in a first cluster that corresponds to popular states of software applications relating to restaurants in Mountain View, Calif., a second cluster that corresponds to popular states of software applications relating to bars in New York City, and a third cluster that corresponds to unpopular states of software applications relating to bakeries in Des Moines, Iowa. In the foregoing example, there are likely more clusters not explicitly discussed. The states of the software applications may be represented by the function IDs 104.
Thus, the function IDs 104 in the first cluster may include a function ID 104 identifying a state of a first software application where users can read reviews of a particular restaurant in Mountain View, a second function ID 104 identifying a state of a second software application where users can view a photograph of a particular dish served at a different restaurant in Mountain View, and a third function ID 104 identifying a state of a third software application where users can make a reservation at yet another restaurant in Mountain View.
The analytics engine 300 may identify hundreds, thousands, or millions of clusters, as the analytics engine 300 clusters the application state records 332 with respect to certain set of features. Furthermore, the analytics engine 300 may cluster the application state records with respect to different sets of features, thereby identifying clusters 110 that identify different information. For example, clusters 110 that are identified based on function, geographic location, and time of the day may indicate the types of actions certain users perform on their user devices 200 at certain times of the day. For instance, such clusters may reveal that users in Los Angeles are more likely to view movie reviews on a Friday night than users in San Diego. Implementations of the analytics engine 300 are described in greater detail below.
The deep-linking system 100 may utilize the clusters 110 in performing various tasks. In some implementations, a recommendation engine 400 may utilize the clusters 110 to improve recommendations to users. A recommendation engine 400 receives a recommendation request 402 from a user device indicating a state of a software application that the user device 200 is currently set to and identifies other states to recommend to the user of the user device 200 based on the data contained in the request. In these implementations, the recommendations may indicate other states of software applications that are indicated in the same cluster 110 as the state identified in the recommendation request 402. For example, if a user is viewing a video about a famous reggae artist using a first software application, the recommendation engine 400 may identify an article about the famous reggae artist that is accessible via a second software application because the two states (i.e., the state corresponding to the video and the state corresponding to the article) are indicated in the same cluster 110. In this example, the application state records 332 corresponding to the states may be clustered according to entity type and the type of action being performed. Implementations of the recommendation engine 400 are described in greater detail below.
The deep-linking system 100 can provide data and instructions to a user device that causes the user device to render and display user selectable links. In some implementations, individual recommendation results 460 can be communicated in result objects. A result object can contain data and instructions that, when rendered by a user device 200, result in a recommendation result that includes one or more user selectable links. A result object can include a function ID 104 and/or one or more access mechanisms that correspond to a state of a software application. In the former scenario, a user device 200 utilizes the function ID 104 to determine the one or more access mechanisms (e.g., a URL, one or more application resource identifiers, and/or a script). A user selectable link, when selected (e.g., pressed on or clicked on) by a user, instructs the user device 200 to access the resource identified by the underlying access mechanism(s).
FIG. 1B illustrates an example of a user device 200 displaying a recommended link 408. A recommended link 408 may refer to a user selectable link that is provided by a recommendation engine 400. In the illustrated example, the user device 200 is executing a native application. The native application may be configured to request recommendations from the deep-linking system 100. The deep-linking system 100 provides result objects containing recommendations, which the user device 200 renders and displays in a GUI of the native application.
The user selectable links of FIG. 1B are provided for example only and are not intended to limit the scope of the disclosure. Any suitable types of user selectable links may be implemented by the deep-linking system 100.
FIG. 2 illustrates an example user device 200 and example components thereof In the illustrated example, the user device 200 includes a processing device 210, a storage device 220, a network interface 230, and a user interface 240. The user device 200 may include additional components not shown in FIG. 2. The components of the user device 200 may be interconnected by, for example, a bus or other communication circuitry.
The processing device 210 can include one or more processors that execute computer-executable instructions and associated memory (e.g., RAM and/or ROM) that stores the computer-executable instructions. In implementations where the processing device 210 includes more than one processor, the processors can execute in a distributed or individual manner. The processing device 210 can execute an operating system 214, one or more native applications 212 (which may include a search application), a web browser 216, and/or a behavior monitor 218, all of which can be implemented as computer-readable instructions. One or more of the native applications may include a native application module 215 that communicates with the deep-linking s system 100.
The storage device 220 can include one or more computer-readable mediums (e.g., hard disk drives, solid state memory drives, and/or flash memory drives). The storage device 220 can store any suitable data that is utilized by the operating system of the user device 200. The storage device 220 can be in communication with the processing device 210, such that the processing device 210 can retrieve any needed data therefrom.
The network interface 230 includes one or more devices that are configured to communicate with the network 150. The network interface 230 can include one or more transceivers for performing wired or wireless communication. Examples of the network interface 230 can include, but are not limited to, a transceiver configured to perform communications using the IEEE 802.11 wireless standard, an Ethernet port, a wireless transmitter, and a universal serial bus (USB) port.
The user interface 240 includes one or more devices that receive input from and/or provide output to a user. The user interface 240 can include, but is not limited to, a touchscreen, a display, a QWERTY keyboard, a numeric keypad, a touchpad, a microphone, and/or speakers.
The processing device 210 executes one or more native applications 212. In some implementations, a native application 212 includes a native application module (hereafter referred to as “NAM”) 215. The NAM 215 is a set of computer readable instructions embedded (e.g., by the developer) in the native application 212. In some implementations, the developers may utilize a software developer kit (hereafter referred to as “SDK”) provided by the provider of the recommendation engine 400 to implement the NAM 215. When the user accesses a state of the native application 212 that calls the NAM 215, the NAM 215 generates a recommendation request 402 based on the state of the native application that is currently being accessed (referred to as the “accessed state”). An example recommendation request 402 can define a state of the native application (e.g., a function identifier 104 corresponding to the accessed state or a resource identifier 106 corresponding to the accessed state) and can further define zero or more context parameters 404 that may be used by the recommendation engine 400 to make recommendations. In some implementations, the NAM 215 determines an application resource identifier of the accessed state and includes the application resource identifier in the recommendation request 402. Additionally or alternatively, the NAM 215 can determine a function ID 104 corresponding to the accessed state and can include the function ID 104 in the recommendation request 402. The NAM 215 may further request a geolocation of the user device 200 from the operating system 214 of the user device 200, which the NAM 215 includes in the recommendation request 402. Further, in some implementations, the NAM 215 can include user information (information which the user has agreed to share, such as age, gender, and/or preferences) in the recommendation request 402. Moreover, in some implementations, the NAM 215 may seek consent from the user to include usage data acquired by the behavior monitor 218 (discussed in detail below). Additionally or alternatively, the NAM 215 can include request parameters in the recommendation request. Request parameters can include any type of data that relates to the contents of the request 402. Examples of request parameters include, but are not limited to, a number of recommendation objects to include in the recommendation results 460 and the types of links to include (e.g., only to the requesting native application, to other native applications, to advertisements, etc.). The NAM 215 transmits the recommendation request 402 to the recommendation engine 400.
Upon receiving the recommendation results 460, the NAM 215 can render and display the recommendation results 460 in a graphical user interface (hereafter referred to as “GUI”) of the native application 212. The recommendation results 460 can include one or more recommended links 408 (FIG. 1B). Each recommended link 408 may be included in a recommendation result object (discussed in greater detail below). A recommendation result object can include one or more access mechanisms, a layout file, and result data. The result data can include the content that is displayed in the recommended link 408. Examples of content may be text descriptions of a state of a software application, icons, screenshots of the state, and data used to populate the individual result. For instance, if the recommendation result object corresponds to a state of a restaurant review application (e.g., a review of a specific restaurant), the result data may include an icon of the application, a number of star ratings of the restaurant, a price rating (e.g., is the restaurant expensive or cheap), and any information that may be displayed in the user selectable link. The NAM 215 renders the recommended links 408 from the recommendation result objects and outputs recommended links 408 to the GUI. If the user selects a recommended link 408, the NAM 215 can initiate the accessing of the state of the software application indicated by the recommended link 408. If the software application indicated by the selected link is the same as the software application of the native application, then the NAM 215 can access the state of the native application to the state indicated by the access mechanism contained in the recommended link 408. If, however, the software application indicated by the selected link 408 is different than the software application of the native application, the NAM 215 may instruct the operating system 214 to access the state of the software application. For example, the NAM 215 may provide the access mechanism of the selected link 408 to the operating system 214. The operating system 214 can then perform any necessary operations to access the state indicated by the access mechanism (e.g., launch a native application indicated by an access mechanism and set the state of the native application to the state indicated by the access mechanism). In some examples, the operating system 214 or another component of the user device 200 may recognize the software application indicated by the selected link 408 and proceed to access the indicated state without instruction from the NAM 215.
The processing device 210 further executes a behavior monitor 218. A behavior monitor 218 is a set of computer-readable instructions that may be a standalone application or may be incorporated into the operating system 214, the web browser 216, and/or any other native applications 212. The behavior monitor 218 monitors the use of the user device 200 and generates usage data 102 based thereon. The behavior monitor 218 may execute as a background process that monitors the current state of the user device and/or monitors a user's response to a set of search results 460.
In some implementations, the behavior monitor 218 monitors the current state of the user device 200. In these implementations, the behavior monitor 218 can monitor the operation of the user device 200 to determine when the user device 200 switches to a state of a software application that is different from the current state. For example, a user may be using a movie database application to view a bio of a famous actor and may switch the state of the movie database application to a state that lists start times of newly released movies. In another example, a user may open a media streaming application and begin playing a song. In yet another example, a user may open a search application, enter a search query, and click on a search result that takes the user to a different application. In all of these examples the user device 200 switches to a different state of a software application from a current state. The different state may be within the same software application (e.g., going from a bio of an actor of a movie to time listings of the movie) or may be across different software applications (e.g., switching from a state within a search application to a state of an unrelated software application listed in a set of search results).
Each time the user device 200 switches to a different state, the behavior monitor 218 records the transition and generates usage data 102 corresponding to the different state. In some implementations, the behavior monitor 218 determines a function ID 104 or a resource identifier of the different state. For example, if the behavior monitor 218 is integrated into the operating system 214 or an application edition corresponding to the different state, the behavior monitor can determine a resource identifier 106 indicating the different state. In another example, when the different state is accessed using a web application, the behavior monitor 218 can record the web resource identifier corresponding to the different state.
In some of the implementations where the behavior monitor 218 identifies resource identifiers 106 instead of function IDs 104, the behavior monitor 218 translates the resource identifier 106 into a function ID 104. In some examples, the behavior monitor 218 obtains custom URL schemes of various software applications (e.g., those that are installed on the user device 200). Examples of how to implement a custom URL scheme may be found at, for example, appurl.org, (maintained by Quixey, Inc.) or schma.org (maintained by Google, Inc., Yahoo, Inc., Microsoft Corporation, and Yandex). The behavior monitor 218 utilizes the custom URL scheme to translate the resource identifier into a function ID 104.
In implementations where the user device executes a search application, the behavior monitor 218 can monitor the search application to determine search activity log data 108. In these implementations, the behavior monitor 218 must first receive consent from the user to begin gathering activity log data 108. For example, once the behavior monitor 218 obtains consent, the behavior monitor 218 can record search queries that the user device 200 transmits and a user's responses to the search results returned in response to search queries. For example, the behavior monitor 218 can identify the user selectable links (i.e., search results) that were selected by the user. In some implementations, a result object of a selected link, which links to a state of a software application, can include the function ID 104 of the linked-to state. Thus, when the user selects the link, the behavior monitor 218 can obtain the function ID 104 of the linked-to state. The behavior monitor 218 can record the function ID 104 and the search query in the usage data 102.
The behavior monitor 218 identifies additional information. For example, the behavior monitor 218 can determine a date or day of the week, a time, and/or geolocation of the user device 200. The behavior monitor 218 can include the additional data (e.g., an identifier of the different state, a time, a day of the week, and a geolocation) in the usage data 102 of the user device 200.
The behavior monitor 218 transmits the usage data to the analytics engine 300. Prior to transmission, the behavior monitor 218 can scrub the usage data 102 to remove any data that may be used to identify the user of the user device 200. The behavior monitor 218 can transmit the usage data 102 at each state transition or at predetermined intervals (e.g., every five minutes, every six hours, every day). In order to further protect the privacy of the user, the behavior monitor 218 may purge the collected usage data 102 from memory (e.g., from the storage device 220) each time it transmits the usage data 102 to the analytics engine 300.
FIG. 3A illustrates example components of an analytics engine 300. In the illustrated example the analytics engine 300 includes a processing system 310, a storage system 320, and a network interface 370. The analytics engine may include additional components not explicitly shown in FIG. 3A. The components of the analytics engine 300 may be interconnected, for example, by a bus and/or any other form or medium of digital data communication, e.g., a communication network 150.
The processing system 310 is a collection of one or more processors that execute computer readable instructions. In implementations having two or more processors, the two or more processors can operate in an individual or distributed manner. In these implementations, the processors may be connected via a bus and/or a network. The processors may be located in the same physical device or may be located in different physical devices. The processing system executes a data collection module 312, an entity matching module 314, and a clustering module 316.
The network interface device 370 includes one or more devices that perform wired or wireless (e.g., Wi-Fi or cellular) communication. Examples of the network interface devices include, but are not limited to, a transceiver configured to perform communications using the IEEE 802.11 wireless standard, an Ethernet port, a wireless transmitter, and a universal serial bus (USB) port.
The storage system 320 includes one or more storage devices. The storage devices may be any suitable type of computer readable mediums, including but not limited to read-only memory, solid state memory devices, hard disk memory devices, and optical disk drives. The storage devices may be connected via a bus and/or a network. Storage devices may be located at the same physical location (e.g., in the same device and/or the same data center) or may be distributed across multiple physical locations (e.g., across multiple data centers). The storage system 320 stores an application state record data store 330, a knowledge data store 350, and a cluster record data store 360. Example contents of the respective data stores 330, 350, 360 are discussed in detail below.
The application state record data store 330 includes a plurality of different application state records 332. FIG. 3B illustrates an example state record 332. Each state record 332 may include data related to a state of the software application resulting from performance of the function. In some implementations, a state record 332 includes a function identifier (ID) 104, application state information 334, entity information 336, one or more access mechanisms 338 used to access the state of the software application, and state statistics 340. The application state data store 330 may include one or more databases, indices (e.g., inverted indices), tables, files, or other data structures which may be used to implement the techniques of the present disclosure.
The function ID 104 may be used to identify the application state record 332 among the other state records 332 included in the application state data store 330. The function ID 104 may be represented in any suitable format, as described above. An application state record 332 further includes one or more access mechanisms 338. The access mechanism(s) 338 may include one or more application access mechanisms, one or more web access mechanisms, one or more application download addresses, and/or one or more scripts. A user device 200 may use the one or more application access mechanisms and the one or more web access mechanisms to access the same, state of the software application using a corresponding edition of the software application. For example, the user device 200 may use the different access mechanism(s) 338 to retrieve similar information, play the same song, or play the same movie. The application download addresses may indicate locations where the native application editions referenced in the application access mechanisms can be downloaded.
The application state information 334 may include data that describes the application state to which the record corresponds. The state of the application may be accessed by an edition of the software application using one of the access mechanisms 338 in the application state record 332. Additionally, or alternatively, the application state information 334 may include data that describes the function performed according to the access mechanism(s) 338 included in the application state record 332. The application state information 334 may include a variety of different types of data. For example, the application state information 334 may include structured, semi-structured, and/or unstructured data.
In some implementations, the application state information 334 may include the content that is provided in a result object. The content corresponds to the data provided by a software application when the software application is set in the application state defined by the access mechanism(s) 338 defined in the application state record 332. The types of data included in the application state information 334 may depend on the type of information associated with the application state and the functionality defined by the access mechanism(s) 338. For example, if the application state record 332 is for an application that provides reviews of restaurants, the application state information 334 may include information (e.g., text and numbers) related to a restaurant, such as a category of the restaurant, reviews of the restaurant (e.g., textual reviews and/or a star rating), and a price rating of the restaurant. In another example, if the application state record 332 is associated with a shopping application, the application state information may include data that describes products (e.g., names, ratings, and prices) that are shown when the shopping application is set to a state defined by an access mechanism(s) 338 stored in the record 332.
The application state information 334 may further define the function performed by the software application when set to the state. The application state information 334 may include a name of an action that is performed by the software application. The application state information 334 may also define the types of parameters that the software application receives in order to access the state. For example, if a state of a software application allows a user to view a particular photograph or file, the action may be “view document” and the software application requires a filename parameter in order to access the state.
The entity information 336 defines the entities and entity types corresponding to the state of the software application defined by the application state record 332. The entities may be entities that are identified in the content corresponding to the state of the software application or entities that can be accepted by the software application when set to the state of the software application. For instance, if a state record 332 defines a state of a shopping application where a user can purchase a particular product, the entities may be a name of the product, a maker of the product, and a price of the product. In another example, the application state record 332 may correspond to a state of an airline application where a user can make flight reservations. In this example, the entities may include all known airport codes that the airline flies in and out of. The entity types may be the entity types of the entities defined in the entity information. As previously discussed an entity may have multiple entity types.
The entity information 336 may further include geographic entities relating to the state of the software application. For example, if the state of a software application refers to a restaurant in a particular city, the geographic entities may include the city, the state, and/or the country of the restaurant. In another example, if the state of a software application allows a user to purchase a particular product, the geographic entity may identify countries where the product may be shipped or countries where the state of the software application may accessed. To the extent the state of the software application is not limited by geography (e.g., a level of a game), the entity information 336 may identify the state as having global relevance. The entity information 336 may further include time related information. For example, the time related entities may indicate a date of publication.
The statistics 340 can define statistics relating to the state of the software application. Examples of the statistics are the number of times the state of the software application is accessed, how often (on average) the state is accessed over a period of time (e.g., per hour or per day), the times of the day when the state is accessed, how many times the state is accessed for each day of the week, and the geolocations of the user devices when the state is accessed. The statistics can include any other additional statistics relating to the state of the software application.
The knowledge data store 350 stores a knowledge base 352. In some implementations, the knowledge base 352 includes one or more entity tables. In these implementations, an entity table is a lookup table that relates a term or combination of terms to the possible entity types of the term or combination of terms. Each relation can also have an associated entity score that is a probability value that indicates a likelihood that the term is of that entity type. The entity scores can be determined, for example, heuristically by analyzing large sets of text and documents.
The knowledge base 352 can include any other additional or alternative data structures. For example, in some implementations at least a portion the knowledge base 352 is structured in accordance with an entity ontology 354 (FIG. 3D) and/or a functional ontology (FIG. 3E). FIG. 3D illustrates an example of an entity ontology 354. The entity ontology 354 may define a formal framework for expressing relationships between different items in the knowledge base 352. The entity ontology 354 may define relationships between general entity types to app-specific entity types. For example, the “name” general entity type may relate to a “Biz ID” app-specific entity type for a first software application and “Rest name” app-specific entity type for a second software application. In this way, the first software application's schema refers to a restaurant name as “Biz ID” and the second software application's schema refers to a restaurant name as “Rest_name.” Furthermore, entity types may relate to other entity types. For example, the general entity type “Thai cuisine” may reference an “Asian cuisine” entity type as Thai cuisine may be thought of as a subclass of “Asian Food.” Further, the entity type “restaurant” entity type may relate to an “address” entity type, a “cuisine” entity type, and any other relevant classifications. An “address” entity type may include a “street address” entity type, a “state” entity type, a “city” entity type, and a “zip code” entity type. The knowledge base 352 includes data points that populate the ontology. For example, the string “Thai” may be related to the “Thai cuisine,” while the string “Tom's Thai” may relate to “Thai cuisine” entity type and “restaurants” entity type. As the analytics engine 300 learns about new entities, the analytics engine 300 can connect the new entity to its corresponding entity types. In this way, the knowledge base 352 indicates how an entity relates to other entities and the entity type(s) of the entity given the entity ontology 354. For instance, the entity “Tom's Thai” may be linked to a state entity “California,” a city entity “Mountain View,” and a zip code entity “94040.”
Furthermore, as the ontology also includes app-specific entities, the analytics engine 300 is able to represent the restaurant name “Tom's Thai” in a manner that is understood by third party applications (e.g., “1234” for a first application and “Toms_Thai” for a second application). In some implementations, the ontology and its corresponding data points (i.e., the specific entities) may be indexed and stored in the knowledge base 352. For example, the analytics engine may index the ontology and corresponding data points into one or more entity tables. In these implementations, components of the analytics engine 300 can query the entity tables with a query term, and if the query term (or combination of query terms) is listed in the entity table as an entity, the entity table returns to potential entity type(s) of the query term (or query terms).
FIG. 3E illustrates an example of a portion of a knowledge base 352 that is structured according to a functional ontology 356. A functional ontology can define the relationships between known verticals, functions that support the vertical, the software applications that perform those functions, and the entity types that the software applications require in order to perform the function. The example of FIG. 3E relates to a vertical focused on movies. Typical functions relating to movies are purchasing movie tickets, finding listings of times a movie is playing, finding information about the title, streaming a movie, finding reviews of a movie, and finding information on actors and actresses. In the illustrated example, the knowledge base identifies purchasing movie tickets, finding listings of times a movie is playing, finding information of the title as possible functions associated with the movie vertical. A first software application (App_1) and a second software application (App_2) can be used to find movie times.
The first software application can also be used to purchase tickets and the second software application can be used to get information about a movie. According to the knowledge base 352 the first software application requires a movie name entity, a theatre name entity, and an address entity in order to perform the purchase movie tickets function. Further, the knowledge base 352 indicates that the second software application requires a movie name entity and an address to perform the purchase movie tickets function. The example of FIG. 3E is a simplified example of the movie vertical. The functional ontology 356 may include any number of verticals and any number of functions defined therein. Furthermore, the functional ontology 356 may be structured in other suitable manners.
The cluster data store 360 stores cluster records 362. Each cluster record 362 relates to a unique cluster 110 identified by the analytics engine 300. FIG. 3C illustrates an example of a cluster record 362. A cluster record may include a cluster ID 364, function ID data 366, and feature information 368.
A cluster ID 364 may be a string made up of letters, numbers, and/or characters that identify a cluster 110 from other clusters 110. The cluster ID 364 may be a random value (a random arrangement of letters and numbers) or a human understandable value (e.g., a name of the cluster). As will be discussed, when the clustering module 316 identifies a new cluster 110, the clustering module 316 can create a new cluster record 362 and can assign a new cluster ID 364 to the new cluster record 362.
The function ID data 366 identifies the states of the software application that belong to the cluster. The states of the software applications may be represented by function IDs 104. Thus, the function ID data 366 can define the function IDs that were included in the cluster 110 corresponding to the cluster record 362.
The feature information 368 identifies the features that were used to identify the cluster 110 corresponding to the cluster record 362. As will be discussed, the clustering module 316 can cluster states of applications based on different features. For example, when recommending a state of an application, functional features and entities defined by the state may be relevant, but statistics that tend to show popularity may not be relevant. In this example, some clusters 110 may be clustered based on whether an application contains a relevant function or associated entity. Thus, the feature information 368 can define the features that were used to identify the cluster 110 defined by the cluster record 362.
The data collection module 312 receives usage data 102 from a plurality of user devices 200. In some implementations, the data collection module 312 generates and/or updates state records 332 based on the usage data 102. When the data collection module 312 receives usage data 102 from a user device 200, the data collection module 312 identifies one or more states of one or more software applications accessed by a user device 200. As previously discussed, the state of the software application may be represented in the usage data 102 by a function ID 104 or a resource identifier 106. In the case that the state is represented by a resource identifier 106, the data collection module 312 can translate the resource identifier 106 to a function ID 104 using the custom URL scheme of the software application to which the resource identifier 106 corresponds.
For each state of a software application identified in the usage data 102, the data collection module 312 determines whether the application state data store 330 includes a state record 332 corresponding to the state. For instance, the data collection module 312 can search the application state record data store 330 using the function ID 104 representing a particular state. If the data collection module 312 finds a state record 332 corresponding to the function ID 104, the data collection module 312 can update the statistics 340 of the application state record 332 with information contained in the usage data 102. For instance, the data collection module 312 can increment a total number of times that the state was accessed, can indicate a time that the state was accessed, a geolocation from which the state was accessed, and a day of the week that the state was accessed.
In the event that the data collection module 312 does not find a state record 332 corresponding to a function ID 104, the data collection module 312 can create a state record 332 corresponding to the state of the software application represented by the function ID 104. The data collection module 312 can create a new state record 332. The data collection module 312 can further instruct a crawler (not shown) to crawl the software application at the given state, so as to identify the application state information 334 and the access mechanisms used to access the state. The data collection module 312 can also update the statistics of the new application state record 332 in the manner described above.
In the event the data collection module 312 creates a new state record 332, the entity matching module 314 can identify the ontological features of the new state record 332. In some implementations, the entity matching module 314 obtains ontological features corresponding to the function ID 104 provided by data collection module 312. Ontological features can include entity features and functional features. An entity feature identifies entities that are relevant to the state and the entity types thereof. For example, if a function ID corresponds to a software application that, amongst other functions, allows users to view photos of food, the entity features may include a name of the restaurant, a name of the dish, or a type of cuisine. The entity matching module 314 can utilize the portion of the knowledge base 352 organized according to the entity ontology 354 to identify the entity features and the portion of the knowledge base 352 organized according to the function ontology 356 to identify the functional features. The entity matching module 314 can also fetch other features pertaining to the function ID. The entity matching module can store the identified ontological features in the application state record 332.
The clustering module 316 identifies clusters 110 of states of software applications and generates/updates cluster records 362 based thereon. Identifying clusters 110 of states of software applications can refer to the clustering of application state records 332 and/or of function IDs 104 defined in the application state records 332 based on one or more features defined in the application state records 332. A developer (e.g., a developer of the recommendation engine 400) can define a set of features on which to cluster the states (e.g., function IDs 104). For instance, a developer wishing to identify the most popular restaurants in an area can define the following set of features on which to cluster the states: business type, number of accesses over a given period of time, and geographic location. In this example, the clustering module 316 can cluster the states based on these features to identify a plurality of clusters 110. One of the clusters 110 may include function IDs 104 corresponding to states of software applications that relate to restaurants that are popular and within a specific geographic region. Furthermore, states of software applications that do not relate to restaurants may be clustered together in a non-relevant cluster 110. In another example, a developer may want to identify states of applications that correspond to similar actions. In this example, the developer can define the following set of features: action type and entity types. In this example, function IDs 104 corresponding to similar actions and/or states that receive or define similar entity types may reside in the same clusters 110. A developer can select any set of features on which to cluster the states of the software applications. The clustering module 316 can be configured to perform any suitable clustering algorithm. Examples of clustering algorithms include, but are not limited to, k-means clustering, bi-clustering, tri-clustering, and k-nearest neighbors.
In operation the clustering module 316 receives a set of feature types. The clustering module 316 then clusters the function IDs 104 based on the feature types. The result of the clustering is one or more clusters 110. For each cluster 110 identified by the clustering module 316, the clustering module 316 generates a cluster record 362 corresponding to the cluster 110. The clustering module 316 may assign a cluster ID 364 to the cluster record 362. The cluster ID 364 may be any suitable value. In some implementations, the clustering module 316 increments a counter each time it identifies a new cluster 110 and uses the new counter value as the cluster ID 364.
The function ID data 366 defines the membership of the cluster 110. The clustering module 316 includes the function IDs 104 of the application state records 332 that were grouped in the identified cluster 110 in the function ID data 366 of the cluster record 362. In this way, the function ID data 366 identifies which application state records 332 belong to a particular cluster 110.
The clustering module 316 further includes the feature types that were used to identify the cluster 110 in the feature information 368 of the cluster record 362. In this way, the feature information 368 indicates the features that were used to identify the cluster 110.
Each time the clustering module 316 is called, the clustering module 316 can identify new clusters 110 of function records 332 and/or update preexisting clusters 110. For example, when the clustering module 316 receives a set of feature types that it has not yet clustered, the clusters 110 resulting from the clustering on the new set of feature types are new clusters. In such a situation, the clustering module 316 identifies new clusters 110 and, therefore, generates new cluster records 362. In the event that the clustering module 316 is operating with a previously used set of feature types, the clustering module 316 may identify new members of preexisting clusters 110 (e.g., newly discovered states of software applications and/or states of software applications whose features have changed over time) or may identify clusters 110 with new members altogether. In the former scenario, the clustering module 316 may update clusters 110 that are previously identified clusters 110 with new members (e.g., newly discovered states of software applications and/or states of software applications whose features have changed over time). In the latter scenario, the clustering module 316 may identify one or more new clusters 110 all while updating one or more preexisting clusters 110.
The cluster records 362 identified by the analytics engine 300 can be used for many suitable applications. In some implementations, the cluster records 362 can be used by the recommendation engine 400 to improve recommendation results 460.
FIG. 4 illustrates an example recommendation engine 400 according to some implementations of the present disclosure. In the illustrated example, a recommendation engine 400 includes a processing system 410, a storage system 420, and a network interface 450. The processing system 410, storage system 420, and network interface 450 of the recommendation engine 400 may be substantially similar to those of the analytics engine 400. In some instances, one or more of the components may be shared amongst the recommendation engine 400, the analytics engine 300, and/or any other engines of the deep linking system 100. The processing system 410 executes a recommendation module 412. The recommendation module 412 accesses the cluster data store 360 and the application state record data store 330.
The recommendation engine 400 receives a recommendation request 402 and, in some implementations, context parameters 404 from a user device 200 and provides the recommendation results 460 to the user device 200 in response to the recommendation request 402. The recommendation results 460 indicate one or more recommended states of software applications. A recommended state of a software application can refer to a state of a software application that the user of the user device 200 may want to view given the user's current use of the user device 200 (e.g., the state of the software application currently being accessed) or given the context parameters 404 (e.g., the time of day and location). The recommendation engine 400 references the cluster data store 360 to determine the recommended states.
In some implementations, the recommendation engine 400 determines the recommended states based on the recommendation request 402 received from a user device 200. As previously discussed, the recommendation request 402 may identify a state of a software application currently being accessed by the user device 200 (referred to as the “current state”). In some implementations, the user device 200 provides the function ID 104 or resource identifier 106 corresponding to the current state in the recommendation request 402. In the case that the recommendation request 402 includes a resource identifier 106 identifying the current state, the recommendation module 412 translates the resource identifier to a function ID 104 corresponding to the state. The recommendation module 412 can query the cluster data store 360 using the function ID 104 of the current state. The cluster data store 360 returns the cluster IDs 364 of the clusters 110 that contain the function ID 104 of the current state. In some implementations, the cluster data store 360 may return the cluster IDs 364 of the clusters 110 clustered according to features related to the function ID 104 of the current state. The returned clusters IDs 364 indicate the clusters 110 that contain function IDs 104 of states of software applications that are similar or related to the current state. The recommendation module 412 identifies the recommended states from the clusters 110 of the returned cluster IDs 364.
In an example, the current state being accessed by a user device 200 may be an article about a professional baseball team (e.g., the Detroit Tigers). The function ID 104 of the current state may result in a number of clusters 110 being identified by the recommendation module 412. For example, a first cluster 110 may identify other current articles relating to professional baseball teams (e.g., articles about Major League Baseball teams), a second cluster 110 about professional sports teams in the same city or state as the professional baseball team (e.g., articles about the Detroit Tigers, Detroit Lions, Detroit Pistons, and Detroit Redwings), and a third cluster 110 identifying different states of software applications where baseball tickets or baseball memorabilia of professional baseball teams may be purchased (e.g., tickets to sporting events or team jerseys). In this example, the cluster record 362 of the first cluster 110 may indicate that the first cluster 110 was clustered using entity types, time, and action type. Thus, the states defined in the first cluster 110 may all relate to professional baseball teams, have the action “view article,” and were recently published. The cluster record 362 of the second cluster 110 may indicate that the second cluster 110 was clustered according to entity type, geography, time, and action type. Thus, the second cluster 110 may indicate states that all relate to professional sports teams from the city of Detroit, that have the action “view article,” and that were recently published. The cluster record 362 of the second cluster 110 may indicate that the third cluster 110 was clustered according to entity type. Thus, the states defined in the third cluster 110 may all indicate states that relate to professional sports. Accordingly, this cluster 110 may define states that relate to professional sports teams in any capacity. The foregoing example assumes an entity ontology that includes a professional sports team entity and a professional baseball team entity and a functional ontology that defines a “view article” action.
The recommendation module 412 then determines which states from the clusters to recommend to a user. The recommendation module 412 can be configured to make this determination in a number of different manners. In some implementations, the recommendation module 412 identifies the function IDs 104 other than the function ID 104 of the current state appearing in the greatest number of the identified clusters 110. In these implementations, the recommendation module 412 can retrieve the cluster records 362 of the identified clusters 110. For each cluster record 362, the recommendation module 412 can identify the function IDs appearing in the respective cluster 110 from the function ID data 366 identified therein. The recommendation module 412 can then identify the N function IDs 104 appearing in the most clusters 110, wherein N is an integer greater than or equal to one.
Additionally or alternatively, the recommendation module 412 can identify one or more states most similar to the current state. In these implementations, the recommendation module 412 may utilize a distance formula to calculate distances between a state appearing in one or more of clusters 110 of the current state. Examples of distance formulas include Euclidean distance formulas and cosine similarity formulas. The recommendation module 412 can then select the N “nearest” states of the software applications to recommend to a user.
In some implementations, the distance between two states can be calculated using the application state information 334 defined in the respective application state records 332 of the two states. In some implementations, the recommendation module 412 retrieves the application state record 332 corresponding to the current state and, for each function ID 104 appearing in an identified cluster 110, retrieves the application state record 332 corresponding to the function ID 104. The recommendation module 412 then calculates the distance (e.g., Euclidean distance or cosine similarity) between the application state record 332 of the current state and the application state record 332 of the function ID 104. The recommendation module 412 can calculate the distance using a set of predetermined features. For instance, the recommendation module 412 can calculate the distance using the entity types defined in the records 332, the actions defined in the records 332, geographic entities defined in the records 332, and time related information defined in the records 332. In some implementations, the recommendation request 402 may identify the relevant features on which to calculate the distance. For example, the NAM 215 that issues the recommendation request 402 may define one or more features on which to base the distance calculation. It is noted that many features may be binary in nature. For example, with respect to entity types and actions, two application state records 332 can either include the same entity types or actions or they do not. In the case that they do contain the same entity type or action, the distance between the two records 332 with respect to the entity type or action may be set to zero and another value (e.g., one) if they do not contain the same entity type. Further, with respect to these types of features, the distance between two application state records 332 may be weighted. For example, more specific entity types may be assigned a greater weight than less specific entity types.
In some implementations, the recommendation module 412 determines recommended states based on context parameters 404 that are provided with a recommendation request 402. In these implementations, the recommendation request 402 can indicate that the recommended states are to be identified based on the context parameters 404. A NAM 215 of a native application 212 executing on a user device 200 may issue the recommendation request 402 and may specify that the recommendations are to be based on the context parameters 404 provided by the user device 200.
In these implementations, the recommendation module 412 receives a set of context parameters 404 from the user device 200. The context parameters 404 may indicate a time that the recommendation request 402 is sent and a geographic location from which the recommendation request 402 is sent. The recommendation request 402 may indicate the current state of the user device 200. In these implementations, the recommendation module 412 can identify clusters corresponding to the current state of the user device 200 and the context parameters 404 surrounding the recommendation request. For example, if the native application that issues the recommendation request is a restaurant review application and the user is accessing a state having a review of a restaurant in New York City, and the context parameters 404 indicate that the time of the request is 6:00 PM and the geographic location of the request is in New York City, the recommendation module 412 may reference a cluster 110 that identifies states for making dinner reservations for sushi restaurants because the most popular restaurant related states of software applications during the evenings in New York City include states that allow users to make reservations for sushi restaurants. Conversely, if the native application that issues the recommendation request is a restaurant review application and the user is accessing a state having a review of a restaurant in Dubuque Iowa, and the context parameters 404 indicate that the time of the request is 6:00 PM and the geographic location of the request is in Dubuque Iowa, the recommendation module 412 may reference a cluster 110 that identifies states of software applications that allows users to read menus for all-you-can eat buffets in Dubuque Iowa, as the most popular states of software applications during evenings in northeast Iowa include states that identify all-you-can-eat buffets in Dubuque, Iowa and the menus of said buffets. In this way, the recommendation module takes into account the geographic and/or temporal context of a user when making recommendations.
In operation, the recommendation module 412 receives a recommendation request 402 and context parameters 404. The context parameters 404 can include a time and a geographic location corresponding to the user. The context parameters 404 may include additional information, such as a user profile ID or demographic information (e.g., male or female, age group).
In some implementations, the recommendation module 412 queries the cluster record data store using the geographic location and any other relevant parameters. For instance, if the recommendation module 412 wants to make geographically and temporally relevant recommendations, the recommendation module 412 queries the cluster record data store 360 using the time and geographic location of the user device 200 (received in the context parameters). In these implementations, the cluster data store 360 returns cluster IDs 364 of cluster records 362 that pertain to the geographic region and time range surrounding the recommendation request 402. Drawing from the examples above, if the recommendation request 402 was transmitted from a location in or around New York City at 6:00 PM, the clusters IDs 364 returned may correspond to clusters 110 that contain function IDs 104 of states of software applications that were accessed by user devices located in New York City in evenings. Assuming that one of the features on which the clusters 110 were clustered included statistics that indicate popularity (e.g., accesses per hour), the clusters 110 represent states that are accessed from New York City in the evening with varying degrees of popularity. It is noted that other clusters 110 that are clustered on the same features would correspond to other locations and/or other time ranges and identify states of varying popularity (e.g., western Ohio in the mornings, Beijing China in the evenings, or New York City late night). In other words, the recommendation module can utilize the context parameters 404 provided by a user device 200 to select clusters that are relevant to the time and geographic location indicated by the context parameters 404.
It is further noted that while requesting clusters 110 of states of software applications clustered according to geographic location, time ranges, and popularity, the recommendation module 412 can request clusters 110 clustered according to other features as well. For instance, a particular application developer that includes the NAM in its native applications may request that the identified states be clustered on age, sex, and popularity of the user. In this scenario, the returned clusters 110 may include a first cluster 110 that identifies very popular states with males between the ages of 26 and 35, a second cluster 110 that identifies moderately popular state with females between the ages 18 and 25, and a third cluster 110 that identifies very popular states with females between the ages 36 and 50. The recommendation module 412 can request clusters 110 clustered according to any suitable features, so long as the analytics engine 300 has previously clustered states of software applications according to those features or is able to cluster states of software applications according to those features.
The recommendation module 412 selects a recommended state from the returned clusters 110. In some implementations, the recommendation module 412 can select N recommended states from one or more of the returned clusters 110 based on the relative distance to the currently accessed state. In these implementations, the recommendation module 412 can calculate similarity between the currently accessed state and states represented by one or more of the returned clusters 110 (e.g., the cluster 110 identifying the most popular states) in the manner defined above. Additionally or alternatively, the user device 200 providing the recommendation request 402 may provide a set of conditions that must be adhered to. For example, the NAM issuing the recommendation request 402 may require that the recommended states must correspond to a particular software application or set of software applications. In this way, the NAM 215 can ensure that the user of the user device 200 is directed to a software application that is affiliated with the native application 212 executing the NAM 215.
Once the recommendation module 412 selects the recommended states, the recommendation module 412 can generate the recommendation results 460 based on the recommended states. As previously discussed, the recommended states are represented by function IDs 104. The recommendation module 412 can retrieve the application state records 332 corresponding to the recommended states using the function IDs thereof. The recommendation module 412 can, for each recommended state, generate a recommendation result object based on the contents contained in the application state record 332 corresponding to the recommended state. In some implementations, the recommendation module 412 can utilize a recommendation result object template that defines the look and feel of a recommendation link 408. In these implementations, the recommendation module 412 can instantiate a new recommendation result object by populating the fields of the result object template with information contained in the application state information 334 of the application state record 332. A recommendation result object can contain data and instructions that, when rendered by a user device 200, provide a recommendation that includes one or more user selectable links. In this way, the recommendation module 412 generates a recommendation result object that may be rendered by the user device 200 when delivered in the recommendation results 460.
The recommendation result object template may include an access mechanism field that receives an access mechanism to access the recommended state. The recommendation result object template may also include a field that receives a textual description of the recommended state. The recommendation module 412 may obtain the textual description from the application state information 334 of the application state record 332. The recommendation result object template may also include a field that receives a file containing an image of an icon of the software application. The recommendation module 412 may obtain the address of the file from the application state information 334 of the application state record 332. The recommendation module 412 may include any other suitable information in the result object. The recommendation module 412 can encode the recommendation results (e.g., the recommendation result objects) in a container, such as a hyper-text markup language (HTML) document that can be embedded in another HTML document (e.g., via an iFrame) or a Java script objection notation (JSON) object. The recommendation module 412 may transmit the recommendation results 460 to the user device 200 that issued the recommendation request 402.
FIG. 5 illustrates an example set of operations of a method 500 for processing a recommendation request 402. For purposes of explanation, the method 500 is explained with respect to the recommendation engine 400 of FIG. 4 and is executed by the processing device 410 thereof. The method 500 may, however, be executed in any suitable computing device.
At operation 510, the recommendation engine 400 receives a recommendation request 402. The recommendation request 402 may include, but is not limited to, a current state of the requesting user device, a function ID 104 or resource identifier associated with the current state of the requesting user device, or any other suitable information associated with the user device 200 or the current state of the application that is currently being accessed by the user device 200. For example, if a user device 200 is accessing a video game review application and currently viewing a review for the video game STARCRAFT® (by Blizzard Entertainment, Inc.), the recommendation request 402 may include the function ID 104 corresponding to the particular state of the video game review application for the review of STARCRAFT®. The recommendation request 402 may include, but is not limited to, other suitable information such as the function ID 104 of a previous application state accessed by the user device 200, the names of other applications on the user device 200 capable of accessing video game reviews, and/or the function ID of an application state frequently accessed by the user device 200. In examples where the recommendation request 402 is transmitted by a server device, such as a search engine, the recommendation request 402 may contain one or more function IDs 104 corresponding to application states in a set of search results in addition to any other suitable information. In some implementations, the recommendation engine 400 may also receive context parameters 404 (see, e.g., FIG. 1A). The context parameters 404 may indicate a time that the recommendation request 402 is sent and a geographic location from which the recommendation request 402 is sent.
At operation 512, the recommendation module 412 identifies one or more clusters by querying the cluster data store 360. The recommendation module 412 uses the information provided by the recommendation request 402 to query the cluster data store 360 for relevant clusters 110. For example, the recommendation module 412 can search the cluster data store 360 using the function ID 104 of the current state of the user device 200. In some implementations, the cluster data store 360 is indexed using an inverted index that maps state identifiers (e.g., function identifiers or resource identifiers) to clusters containing the state identifiers. In these implementations, the recommendation module 412 searches the cluster data store 360 by querying the inverted index using the state identifier received in the recommendation request 402. The inverted index outputs the cluster IDs 364 of the clusters 110 that contain the function ID 104 of the current state. Because the clusters 110 were clustered according to the attributes of the underlying states of applications, the returned clusters 110 contain function IDs 104 of states of software applications that are similar or related to the current state of the requesting user device 200 (e.g., the state of the software application being accessed by the user device 200). For example, if the current state being accessed by the user device 200 is an article about a video game (e.g., an article about the game STARCRAFT® by Blizzard Entertainment, Inc.), an example returned cluster 110 may identify other articles relating to the developer of the video game (e.g., an article about the sequel, STARCRAFT® II, also by Blizzard Entertainment, Inc.) while yet another cluster 110 may identify message board forums regarding the game itself
At operation 514, the recommendation module 412 selects the states from the identified clusters 110 to recommend to a user. The recommendation module 412 can be configured to make this determination using various techniques. In one example, the recommendation module may recommend a state associated with a function ID 104 appearing the greatest number of identified clusters 110. In this example, the recommendation module 412 retrieves the cluster records 362 of the identified clusters 110. For each cluster record 362, the recommendation module 412 can identify the function IDs appearing in the respective cluster 110 from the function ID data 366 identified therein. The recommendation module 412 identifies the N function IDs 104 appearing in the most clusters 110, where N is an integer greater than or equal to one. In another example, the recommendation module 412 may recommend a state that is most similar to the current state. In this example, the recommendation module 412 utilizes a distance formula to calculate distances between a state appearing in one or more clusters 110 and the current state. Examples of distance formulas include Euclidean distance formulas and cosine similarity formulas. The recommendation module 412 selects the N “nearest” states of the software applications to recommend. More specifically, the recommendation module 412, for each function ID 104 (i.e., the function ID of the recommendation request 402 and the function IDs 104 of the identified clusters 110), populates a feature vector (using binary or numerical values) based on a predefined set of features (e.g., time, location, category, cuisine, etc.). In this way, the recommendation module 412 may implement a distance formula using the feature vectors associated with each respective function ID 104 to determine the relative distance between the function IDs 104 appearing in the one or more identified clusters 110 and the function ID 104 indicated by the recommendation request 402. The recommendation module 412 may then select the one or more states corresponding to the function IDs 104 of the identified clusters 110 that have the shortest distance to the function ID 104 of the recommendation request 402. In the STARCRAFT® example above, the feature vectors utilized by the recommendation request 402 may be populated based on features commonly associated with review applications such as rating and category. In this way, the recommendation module 402 selects the function IDs of application states corresponding to reviews of highly rated video games in the same category as STARCRAFT®. Feature vectors may be based on, but not limited to, features such as a time frame, a geo-location, a category, and a cuisine. Additionally or alternatively, the recommendation module may recommend a state from a cluster 110 identified using the context parameters 404 that are provided with a recommendation request 402. In this example, the context parameters 404 may indicate a time, geographic location, and/or any other suitable information. The context parameters 404 may further indicate a set of conditions from the NAM 215, such as a condition indicating that the recommended states should correspond to a particular software application or set of software applications. In this situation, the recommendation module 412 can provide recommendations that comply with the conditions set forth by the NAM 215 and are also relevant to other information provided by the context parameters 404. In a specific example, the current state may correspond to a news application and the user may be in San Jose, CA. The NAM issuing the recommendation request may indicate that any recommended states must be from the same application. In this example, the recommendation module 412 may identify a cluster that has articles concerning Northern California. From this cluster, the recommendation module 412 limits the selected states to states pertaining to the news application of the current state. The recommendation module 412 may utilize any other suitable techniques to determine which states from the identified clusters 110 to recommend to a user.
At operation 516, the recommendation module 412 generates recommendation results 460 based on the recommended states. The recommendation module 412 retrieves the application state records 332 corresponding to the recommended states using their function IDs. For each retrieved application state record 332, the recommendation module 412 can generate a recommendation result object based on the contents of the retrieved application state record 332. The recommendation module 412 can utilize a result object template to generate the recommendation result object. The recommendation module 412 can encode content from the retrieved application state record 332 in the result object template, such as a title of the application and/or state of the application and an access mechanism used to access the state. The foregoing may be used by the user device 200 to generate a link to the recommended state. The recommendation module 412 encodes the recommendation result objects in a container, such as a HTML document or JSON object. At operation 518, the recommendation module transmits the recommendation results 460 to the user device 200 that issued the recommendation request 402.
The method 500 of FIG. 5 is provided for example only and not intended to limit the scope of the disclosure. A recommendation engine 400 may provide recommendations that utilize the cluster records 362 in other suitable manners as well without departing from the scope of the disclosure.
Various implementations of the systems and techniques described here can be realized in digital electronic and/or optical circuitry, integrated circuitry, specially designed ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.
These computer programs (also known as programs, software, software applications or code) include machine instructions for a programmable processor, and can be implemented in a high-level procedural and/or object-oriented programming language, and/or in assembly/machine language. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, non-transitory computer readable medium, apparatus and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal used to provide machine instructions and/or data to a programmable processor.
Implementations of the subject matter and the functional operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them. Moreover, subject matter described in this specification can be implemented as one or more computer program products, i.e., one or more modules of computer program instructions encoded on a computer readable medium for execution by, or to control the operation of, data processing apparatus. The computer readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, a composition of matter effecting a machine-readable propagated signal, or a combination of one or more of them. The terms “data processing apparatus,” “computing device” and “computing processor” encompass all apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, or multiple processors or computers. The apparatus can include, in addition to hardware, code that creates an execution environment for the computer program in question, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them. A propagated signal is an artificially generated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal that is generated to encode information for transmission to suitable receiver apparatus.
A computer program (also known as an application, program, software, software application, script, or code) can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file in a file system. A program can be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.
The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application specific integrated circuit).
Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. Generally, a processor will receive instructions and data from a read only memory or a random access memory or both. The essential elements of a computer are a processor for performing instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices for storing data, e.g., magnetic, magneto optical disks, or optical disks. However, a computer need not have such devices. Moreover, a computer can be embedded in another device, e.g., a mobile telephone, a personal digital assistant (PDA), a mobile audio player, a Global Positioning System (GPS) receiver, to name just a few. Computer readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto optical disks; and CD-ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.
To provide for interaction with a user, one or more aspects of the disclosure can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube), LCD (liquid crystal display) monitor, or touch screen for displaying information to the user and optionally a keyboard and a pointing device, e.g., a mouse or a trackball, by which the user can provide input to the computer. Other kinds of devices can be used to provide interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input. In addition, a computer can interact with a user by sending documents to and receiving documents from a device that is used by the user; for example, by sending web pages to a web browser on a user's client device in response to requests received from the web browser.
One or more aspects of the disclosure can be implemented in a computing system that includes a backend component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a frontend component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the subject matter described in this specification, or any combination of one or more such backend, middleware, or frontend components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (“LAN”) and a wide area network (“WAN”), an inter-network (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks).
The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. In some implementations, a server transmits data (e.g., an HTML page) to a client device (e.g., for purposes of displaying data to and receiving user input from a user interacting with the client device). Data generated at the client device (e.g., a result of the user interaction) can be received from the client device at the server.
While this specification contains many specifics, these should not be construed as limitations on the scope of the disclosure or of what may be claimed, but rather as descriptions of features specific to particular implementations of the disclosure. Certain features that are described in this specification in the context of separate implementations can also be implemented in combination in a single implementation. Conversely, various features that are described in the context of a single implementation can also be implemented in multiple implementations separately or in any suitable sub-combination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a sub-combination or variation of a sub-combination.
Similarly, while operations are depicted in the drawings in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multi-tasking and parallel processing may be advantageous. Moreover, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.
A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. Accordingly, other implementations are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results.

Claims

What is claimed is:

1. A deep linking system comprising:

a storage system that stores:

a plurality of application records, each application record including i) a state identifier that indicates a state of a respective software application, and ii) application state information corresponding to the state of the software application; and

a plurality of cluster records, each cluster record defining a respective cluster of a plurality of clusters identified by the deep linking system, each cluster including a respective plurality of clustered state identifiers, each clustered state identifier identifying a state of a respective software application, wherein the plurality of clusters are clustered according to one or more features; and

a recommendation engine including a processing system, the processing system including one or more processors that execute computer-readable instructions, the computer-readable instructions, when executed by the processing system, causing the processing system to:

receive a recommendation request containing a received state identifier from a remote device;

identify one or more cluster records from the plurality of cluster records using the received state identifier of the recommendation request, the one or more identified cluster records respectively indicating one or more clusters to which the state of the software application defined in the recommendation request is related;

select one or more state identifiers from the identified cluster records, the selected state identifiers respectively corresponding to one or more application states to recommend in response to the recommendation request;

generate recommendation results based on the states indicated by the one or more selected state identifiers, the recommendation results including one or more result objects, each result object capable of being rendered into a user-selectable link; and

transmit the recommendation results to the remote device.

2. The deep linking system of claim 1, wherein a state identifier comprises at least one of a function identifier, a resource identifier, and/or an application access mechanism.

3. The deep linking system of claim 2, wherein the plurality of clusters records are determined based on usage data collected from a plurality of user devices, the usage data indicating states of software applications accessed by users of the user devices.

4. The deep linking system of claim 3, wherein at least a subset of the plurality of clusters are clustered according to a popularity feature, a geography feature, and one or more entity features.

5. The deep linking system of claim 3, wherein at least a subset of the plurality of clusters are clustered according to a popularity feature and a function feature.

6. The deep linking system of claim 3, wherein each cluster record indicates a set of feature types on which the cluster represented by the cluster record was clustered.

7. The deep linking system of claim 6, further comprising:

an analytics engine comprising:

a second processing system that executes computer-readable instructions, the computer-readable instructions, when executed by the second processing system, causes the second processing system to:

receive an instruction to generate clusters of application records, the instruction including a set of feature types;

cluster the application records into two or more different clusters based on the application state information and the set of feature types;

for each of the two or more different clusters, generate a new cluster record based on the cluster, the new cluster record including the state identifiers of the application records in the cluster; and

store the two or more new cluster records in the plurality of clusters.

8. The deep linking system of claim 3, wherein the usage data received from each of the user devices includes a search activity log, the search activity log indicating actions performed by a user in response to being presented with search results.

9. The deep linking system of claim 1, wherein the recommendation request further includes one or more context parameters and/or one or more request parameters, wherein each context parameter respectively indicates contextual information corresponding to the recommendation request and each request parameter respectively indicates condition for the recommendation results.

10. The deep linking system of claim 9, wherein selecting one or more state identifiers comprises selecting the one or more state identifiers that appear in the greatest number of identified cluster records.

11. The deep linking system of claim 9, wherein selecting one or more state identifiers comprises:

generating a feature vector based on the application state record corresponding to the received state identifier;

for each state identifier in the one or more identified clusters:

generate a feature vector based on the corresponding application state record of each state identifier;

calculate a distance value between the feature vector corresponding to each state identifier and the feature vector corresponding to the received state identifier;

selecting one or more state identifiers from the identified clusters to include in the recommendation results based on their calculated distance value.

12. The deep linking system of claim 9, wherein selecting state identifiers comprises selecting one or more state identifiers from the state identifiers indicated by the identified cluster records that comply with the one or more request parameters indicated by the recommendation request.

13. A method comprising:

maintaining, by a processing system including one or more processors, a plurality of application records on a storage device, each application record including i) a state identifier that indicates a state of a respective software application, and ii) application state information corresponding to the state of the application;

maintaining, by the processing system, a plurality of cluster records on a storage device, each cluster record defining a respective cluster of a plurality of clusters, each cluster including a respective plurality of clustered state identifiers, each clustered state identifier identifying a state of a respective software application, wherein the plurality of clusters are clustered according to one or more features;

receiving, by the processing system, a recommendation request containing a function identifier from a remote device;

identifying, by the processing system, one or more cluster records from the plurality of cluster records using the state identifier of the recommendation request, wherein the identified one or more cluster records respectively indicate one or more clusters to which the state of the software application defined by the recommendation request belongs;

selecting, by the processing system, one or more state identifiers from the identified cluster records, the selected state identifiers respectively corresponding to one or more application states to recommend to the user;

generating, by the processing system, recommendation results based on the states indicated by the one or more selected state identifiers, the recommendation results including one or more result objects, each result object capable of being rendered into a user-selectable link; and

transmitting, by the processing system, the recommendation results to the remote device.

14. The method of claim 13, wherein a state identifier comprises at least one of a function identifier, a resource identifier, and/or an application access mechanism.

15. The method of claim 14, wherein the plurality of clusters are determined based on usage data collected from a plurality of user devices, the usage data indicating states of software applications accessed by users of the user devices.

16. The method of claim 14, wherein at least a subset of the plurality of clusters are clustered according to a popularity feature, a geography feature, and one or more entity features.

17. The method of claim 14, wherein at least a subset of the plurality of clusters are clustered according to a popularity feature and a function feature.

18. The method of claim 14, wherein each cluster record indicates a seat of feature types on which the cluster represented by the cluster record was clustered.

19. The method of claim 18, further comprising:

receiving, by a second processing system including one or more processors, an instruction to generate clusters, the instruction including a set of feature types;

clustering, by the second processing system, the application records into two or more different clusters based on the application state information and the set of feature types;

generating, by the second processing system, for each of the two or more different clusters, a new cluster record based on the cluster, the new cluster record including the state identifiers of the application records in the cluster; and

storing, by the second processing system, the two or more new cluster records with the plurality of clusters.

20. The method of claim 14, wherein the usage data received from each of the user devices includes a search activity log, the search activity log indicating actions performed by a user in response to being presented with search results.

21. The method of claim 13, wherein the recommendation request further contains one or more context parameters and/or one or more request parameters, wherein each context parameter respectively indicates contextual information corresponding to the recommendation request and each request parameter respectively indicates conditions that the recommendation results should satisfy.

22. The method of claim 21, wherein selecting state identifiers comprises selecting the one or more states that appear in the greatest number of clusters indicated by the one or more cluster records.

23. The method of claim 21, wherein selecting state identifiers comprises:

for each state identifier in the one or more identified clusters:

calculate a distance value between the feature vector corresponding to each state identifier and the feature vector corresponding to received state identifier;

24. The method of claim 21, wherein selecting state identifiers comprises selecting one or more state identifiers from the state identifiers indicated by the identified cluster records that comply with the one or more request parameters indicated by the recommendation request.