WO2015131306A1

WO2015131306A1 - Dynamics of tie strength from social interaction

Info

Publication number: WO2015131306A1
Application number: PCT/CN2014/072785
Authority: WO
Inventors: Jing Yuan; Xing Xie; Yong Rui
Original assignee: Microsoft Technology Licensing, Llc
Priority date: 2014-03-03
Filing date: 2014-03-03
Publication date: 2015-09-11
Also published as: CN105308591A

Abstract

Systems and methods for observing friendships and social interactions between users of online communication platforms and ranking the users of the platforms to indicate likelihood of a user responding to a communication are described herein. At least one embodiment includes calculating a degree of intimacy using a degree of reciprocity, affinity calculation, sentiment value, self-disclosure value, and/or responsiveness value derived from communications initiated by a first user of a platform and directed communications from a second user of the platform in response to communications initiated by the first user. Degree of intimacy may be leveraged to rank two or more users of the platform and the ranking may be leveraged to predict which of the two or more users is more likely to respond to a communication from the first user. Various graphical user interfaces capable of producing visualizations of user intimacy and views of communication platforms are also described.

Description

DYNAMICS OF TIE STRENGTH FROM SOCIAL INTERACTION

BACKGROUND

[0001] Interpersonal ties represent connections between people (e.g., friendships, etc.). Generally, proximity and homophily provide a foundation on which interpersonal relationships are built. For example, people tend to befriend others who are geographically close by and/or similar to themselves in terms of age, ethnicity, education, genotypes, etc. However, not all friendships are equal and friendships change over time. That is, the strength of interpersonal ties may vary from person to person and from time to time. Often, these variations in the strength of interpersonal relationships result from variations in intimacy. Intimacy is the quality of close connection between people and the process of building this quality.

[0002] Social interactions, in both online and offline communication platforms, are diagnostic reflections of friendship relations, or social ties. The strength of social ties is described as tie strength, or a frequency of interaction between individuals. Intimacy is a key aspect of tie strength. Traditional approaches for calculating tie strength have been based on user retrospective data. Accordingly, tie strength has generally been represented as a static category (i.e., strong, weak, absent tie) or a fixed value. However, static categories and fixed values do not realistically represent tie strengths between users of social networks. SUMMARY

[0003] This disclosure describes systems and methods for observing friendships and online social interactions between users of communication platforms and ranking the users of the communication platforms to indicate a likelihood that a particular user will respond to a communication. In at least one embodiment, a degree of intimacy is calculated after determining a degree of reciprocity, frequency, affinity calculation, and at least one of sentiment values, self-disclosure values, or responsiveness values derived from communications between a first user of communication platforms and a second user of communication platforms.

[0004] In some embodiments, the degree of reciprocity between the first user and the second user is determined based at least in part on a frequency of directed communication from the second user to the first user in response to communication initiated by the first user. Further, in some embodiments, the frequency of communications may be measured in successive time periods. Additionally, in at least one embodiment, the systems and methods described herein may include an affinity calculation between the first user and the second user based at least in part on at least one of a profile affinity or a topic affinity. Moreover, several contextual features (e.g., sentiment, self-disclosure, and/or responsiveness) may be analyzed to identify sentiment values, self-disclosure values, and/or responsiveness values. In at least one embodiment, the degree of reciprocity, the frequency, the affinity calculation, the sentiment values, the self-disclosure values, and/or the responsiveness values may be leveraged to calculate the degree of intimacy between the first user and the second user.

[0005] In at least one embodiment, the degree of intimacy may be leveraged to rank two or more users of the communication platforms. Furthermore, the ranking may be leveraged to predict which user of the two or more users is likely to respond to a particular communication from the first user and/or an order in which the two or more users are likely to respond.

[0006] In at least one embodiment, the systems and methods may be configured to produce a graphical user interface capable of producing various visualizations of user intimacy and views of the communication platforms described herein.

[0007] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The Detailed Description is set forth with reference to the accompanying figures, in which the left-most digit of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in the same or different figures indicates similar or identical items or features. [0009] FIG. 1 is a diagram showing an example system for exploring dynamics of intimacy based on user communication in communication platforms.

[0010] FIG. 2 is a diagram showing an example system for exploring various dynamics of social interactions between users and leveraging the dynamics to rank the users based on degrees of intimacy

[0011] FIG. 3 illustrates a network view depicting topic discovery comparisons for users.

[0012] FIG. 4 illustrates classifications of emoticons.

[0013] FIG. 5 illustrates a network view depicting sentiments of users over time.

[0014] FIG. 6 illustrates an example responsiveness matrix.

[0015] FIG. 7 illustrates a network view depicting changes in the frequency of interaction between users over time.

[0016] FIG. 8 illustrates an overall network view user interface.

[0017] FIG. 9 illustrates an ego network view user interface.

[0018] FIG. 10 illustrates a community view user interface.

[0019] FIG. 11 illustrates a process for calculating a degree of intimacy between users of communication platforms.

[0020] FIG. 12 illustrates a process for ranking and predicting based on the degree of intimacy. [0021] FIG. 13 illustrates a process for calculating a degree of intimacy and leveraging the degree of intimacy to predict users who are likely to respond to particular communications.

DETAILED DESCRIPTION

[0022] Dynamics of tie strength are important in understanding friendship relationships. However, present methods of considering tie strength assign fixed values or descriptions (e.g., weak or strong) to classify tie strength. Fixed values, or static measurements, are unrealistic for truly understanding friendship relationships because static measurements fail to consider that the strength of a tie may not be reciprocal, that the strength of a tie may evolve over time, and/or that given different contexts such as location, emotion, and environment, people may perceive varying degrees of tie strength. Considering the dynamics of tie strength offers a more realistic approach to understanding friendship relationships.

[0023] This disclosure describes systems and methods for observing friendships and social interactions between users of communication platforms and ranking the users of the communication platforms to indicate a likelihood that a particular user will respond to a communication. In at least one embodiment, a degree of intimacy is calculated based on investigating dynamics of relationships between two people (e.g., dyadic relationships) through social interactions. Accordingly, a variety of features, such as a degree of reciprocity, frequency, affinity calculation, sentiment values, self-disclosure values, and/or responsiveness values are determined from communications between a first user of one or more communication platforms and a second user of the one or more communication platforms and used to calculate the degree of intimacy between the first user and the second user. In at least one embodiment, the degree of intimacy may be leveraged to rank two or more users of the communication platforms. Furthermore, the ranking may be leveraged to predict which user of the two or more users is likely to respond to a particular communication from the first user and/or an order in which the two or more users are likely to respond.

[0024] Such predictions can enable a variety of promising applications in communication platforms including online communication platforms. For example, users may be more likely to purchase or use a product that has been recommended by a close friend than a casual acquaintance. Accordingly, accurate predictions may guide advertisers and marketers to design better incentive mechanisms. Additionally, accurate predictions may be useful in assisting users to create content to engage reactions and increase social impact. For example, using the methods and systems described herein, users may have knowledge of the best time of a day to post a communication related to a certain topic that can trigger a discussion among contacts within a user's social circle. The methods and systems described herein may also be useful in analyzing friendship dynamics in communication platforms such as social-networking services, instant messaging services, social entertainment networks, microblogging services, text messaging services, etc. For instance, accurate predictions of tie strength may be leveraged to create and maintain contact groups across one or more of these types of platforms.

[0025] For the purposes of this disclosure, the dynamics of social interactions concern directed ego networks. Directed ego networks consist of a focal node representing a member of a communication platform (the "ego" user (u)) and nodes representing individuals to whom the ego user u directly connects ("friend" users (u)). In at least one embodiment, given a communication platform as a directed graph G = (U, F), where U are users, and F are the following relations ((u, v G F means a user u G U follows v G U on G), a "friend user" is a user u G U called a friend of user v G U (or u and v have a friendship relation), denoted by u ~ v, if v) G F and (v, u G F. In the same embodiment, an ego network is made of the neighbors of a user u G V that are defined as E_u ^° = {v\u~v, v G U}, and the ego network of u is E_u = E_u ^° U {ii}.

[0026] Relationships between the ego user u and the friend users u may be represented by ties indicative of a strength of a relationship between the ego user u and the friend users u. This disclosure describes using dynamics of social interactions between ego users u and friend users u to predict which friend users u of a plurality of friend users u are likely to respond to communications initiated by the ego user u. In some embodiments, the dynamics of social interactions may also be used to predict an order in which the friend users u will respond.

[0027] In at least one embodiment, interactions between ego users u and his or her friend users u via online communication platforms are observed to explore dynamics of intimacy from online social interactions. Examples of online communication platforms include, but are not limited to, social-networking services, which can include Facebook®, Google Plus®, Linkedln®, etc.; instant messaging services, which can include Skype®, WhatsApp, GoogleTalk, AOL Instant Messenger, Microsoft Instant Messenger, etc.; text messaging services, which can include MMS, SMS, etc., social entertainment networks, which can include Xbox Live®, hi5, etc.; and microblogging services, which can include Twitter®, Sina Weibo (a Chinese microblogging site), etc. As users of communication platforms, users may send communications to other users. Sending a communication may include posting a communication to a user's webpage, sharing a communication for other users to view, sending a communication to a personal account associated with a user, etc. Communications may include microblogs, text messages, phone calls, video calls, photos, publications, hyperlinks, and other online communication tools. For example, online social networks such as Twitter® or Facebook® permit users to send communications including microblogs, or small elements of content such as text, images, and/or links. In online social networks such as Twitter® or Facebook®, a user may send a microblog by posting the microblog on his or her Facebook® timeline or broadcasting the microblog on his or her Twitter® handle.

[0028] Once users send communications, other users of the communication platforms may receive the communications. For example, the other users may receive the communication by reading the communication, by viewing the communication, by converting the text of the communication to speech by known means of text-to- speech, by activating a hyperlink, etc. In online social networks such as Facebook® or Twitter®, when a user posts a microblog on a Facebook® timeline or a Twitter® handle, other users may receive the microblog when they read their Facebook® timeline or Twitter® feed.

[0029] In some embodiments, the users who receive the communications (e.g., microblogs, etc.) have an opportunity to reply to the communication or rebroadcast the communication to other users of the communication platforms. These responsive communications are called directed communications. Examples of directed communications include replying to a user's Tweet, retweeting a user's Tweet, tagging a friend in a microblog, sharing a fellow Facebook® friend's status, and other online communication tools that may be used to respond to a communication of a user of a communication platform.

[0030] In some embodiments, contents of the communications may be viewed by the public (e.g., public Tweets, public Facebook® status updates, etc.). However, in other embodiments, contents of the communications (e.g., text messages, video calls, etc.) may be private and may not be viewed by the public. In such embodiments (i.e., when the contents of the communication are private), the systems and methods described herein may not proceed without first obtaining permission from the authors of the communication to access the content of the communication.

[0031] In the embodiments where the contents of the communication are private or the contents of the information include personally identifiable information (PII) that identifies or can be used to identify, contact, or locate a person to whom such information pertains, a user may be provided with notice that the systems and methods herein are collecting PII. Additionally, prior to initiating PII data collection, users may have an opportunity to opt-in or opt-out of the PII data collection. For example, a user may opt-in to the PII data collection by taking affirmative action indicating that he or she consents to the PII data collection. Alternatively, a user may be presented with an option to opt-out of the PII data collection. An opt-out option may require an affirmative action to opt-out of the PII data collection, and in the absence of affirmative user action to opt-out, PII data collection may be impliedly permitted.

[0032] Embodiments described herein provide systems and methods for observing friendships and online social interactions between users of communication platforms and ranking the users of the communication platforms to indicate a likelihood that a particular user will respond to a communication. In various instances, a processing unit configured via programming from modules or APIs to perform techniques as described herein can include one or more of a central processing unit (CPU), a graphics processing unit (GPU), a field-programmable gate array (FPGA), another class of digital signal processor (DSP), or other hardware logic components that may, in some instances, be driven by the CPU. For example, and without limitation, illustrative types of hardware logic components that can be used include Application- Specific Integrated Circuits (ASICs), Application-Specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc. Illustrative Environment

[0033] The environment described below constitutes but one example and is not intended to limit application of the system described below to any one particular operating environment. Other environments may be used without departing from the spirit and scope of the claimed subject matter. The various types of processing described herein may be implemented in any number of environments including, but not limited to, stand along computing systems, network environments (e.g., local area networks or wide area networks), peer-to-peer network environments, distributed-computing (e.g., cloud- computing) environments, etc.

[0034] FIG. 1 illustrates an example operating environment 100 that includes a variety of devices and components that may be implemented in a variety of environments for exploring the dynamics of intimacy from social interactions of one or more users of communication platforms. More particularly, the example operating environment 100 may include a service provider 102, one or more network(s) 104, one or more users 106, and one or more user devices 108 associated with the one or more users 106. Alternatively, or in addition, the functionally described herein can be performed, at least in part, by one or more hardware logic components such as accelerators. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Application- specific Integrated Circuits (ASICs), Application -specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc. For example, an accelerator can represent a hybrid device, such as one from ZYLEX or ALTERA that includes a CPU course embedded in an FPGA fabric.

[0035] As shown, the service provider 102 may include one or more server(s)

110, which may include one or more processing unit(s) 112 and computer-readable media 114. In various embodiments, the service provider 102 may calculate a degree of intimacy between users 106 of one or more communication platforms in response to communications initiated by a first user 106 and directed communications from a second user 106 in response to the communications initiated by the first user 106. The service provider 102 may determine a degree of reciprocity, frequency, affinity calculation, sentiment values, self-disclosure values, and/or responsiveness values from the communications initiated by the first user 106 and the directed communications from the second user 106. The service provider 102 may leverage the degree of reciprocity, the frequency, the affinity calculation, the sentiment values, the self-disclosure values, and/or the responsiveness values to determine a degree of intimacy. In at least one embodiment, the service provider 102 may leverage the degree of intimacy to rank and predict users 106 of the communication platforms who are likely to respond to a communication initiated by the first user 106 and the order in which the users 106 will respond. In some embodiments, the service provider 102 may produce various graphical user interfaces for visualizing user intimacy.

[0036] In some embodiments, the network(s) 104 may be any type of network known in the art, such as the Internet. Moreover, the users 106 may communicatively couple to the network(s) 104 in any manner, such as by a global or local wired or wireless connection (e.g., local area network (LAN), intranet, etc.). The network(s) 104 may facilitate communication between the server(s) 110 and the user devices 108 associated with the users 106.

[0037] In some embodiments, the users 106 may operate corresponding user devices 108 to perform various functions associated with the user devices 108, which may include one or more processing unit(s), computer-readable storage media, and a display. Furthermore, the users 106 may utilize the user devices 108 to communicate with other users 106 via the one or more network(s) 104.

[0038] User device(s) 108 can represent a diverse variety of device types and are not limited to any particular type of device. Examples of device(s) 108 can include but are not limited to stationary computers, mobile computers, embedded computers, or combinations thereof. Example stationary computers can include desktop computers, work stations, personal computers, thin clients, terminals, game consoles, personal video recorders (PVRs), set-top boxes, or the like. Example mobile computers can include laptop computers, tablet computers, wearable computers, implanted computing devices, telecommunication devices, automotive computers, personal data assistants (PDAs), portable gaming devices, media players, cameras, or the like. Example embedded computers can include network enabled televisions, integrated components for inclusion in a computing device, appliances, microcontrollers, digital signal processors, or any other sort of processing device, or the like.

[0039] The service provider 102 may be any entity, server(s), platform, etc., that may leverage a collection of features from communication platforms, including online communication platforms, to measure the interaction dynamics between users of the communication platforms. Moreover, and as shown, the service provider 102 may include one or more server(s) 110, which may include one or more processing unit(s) 112 and computer-readable media 114 such as memory. The one or more server(s) 110 may include devices.

[0040] Embodiments support scenarios where device(s) that may be included in the one or more server(s) 110 can include one or more computing devices that operate in a cluster or other grouped configuration to share resources, balance load, increase performance, provide fail-over support or redundancy, or for other purposes. Device(s) included in the one or more server(s) 110 can belong to a variety of categories or classes of devices such as traditional server-type devices, desktop computer-type devices, mobile devices, special purpose-type devices, embedded-type devices, and/or wearable-type devices. Thus, although illustrated as desktop computers, device(s) can include a diverse variety of device types and are not limited to a particular type of device. Device(s) included in the one or more server(s) 110 can represent, but are not limited to, desktop computers, server computers, web-server computers, personal computers, mobile computers, laptop computers, tablet computers, wearable computers, implanted computing devices, telecommunication devices, automotive computers, network enabled televisions, thin clients, terminals, personal data assistants (PDAs), game consoles, gaming devices, work stations, media players, personal video recorders (PVRs), set-top boxes, cameras, integrated components for inclusion in a computing device, appliances, or any other sort of computing device.

[0041] Device(s) that may be included in the one or more server(s) 110 can include any type of computing device having one or more processing unit(s) 112 operably connected to computer-readable media 114 such as via a bus, which in some instances can include one or more of a system bus, a data bus, an address bus, a PCI bus, a Mini-PCI bus, and any variety of local, peripheral, and/or independent buses. Executable instructions stored on computer-readable media 114 can include, for example, the data platform module 116, the intimacy module 118, the ranking module 120, the output module 122, and other modules, programs, or applications that are loadable and executable by processing units(s) 112. Alternatively, or in addition, the functionally described herein can be performed, at least in part, by one or more hardware logic components such as accelerators. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Application-specific Integrated Circuits (ASICs), Application -specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc. For example, an accelerator can represent a hybrid device, such as one from ZYLEX or ALTERA that includes a CPU course embedded in an FPGA fabric. [0042] Device(s) that may be included in the one or more server(s) 110 can further include one or more input/output (I/O) interface(s) coupled to the bus to allow device(s) to communicate with other devices such as user input peripheral devices (e.g., a keyboard, a mouse, a pen, a game controller, a voice input device, a touch input device, gestural input device, and the like) and/or output peripheral devices (e.g., a display, a printer, audio speakers, a haptic output, and the like). Devices that may be included in the one or more server(s) 110 can also include one or more network interfaces coupled to the bus to enable communications between computing device and other networked devices such as user device(s) 108. Such network interface(s) 118 can include one or more network interface controllers (NICs) or other types of transceiver devices to send and receive communications over a network. For simplicity, some components are omitted from the illustrated device.

[0043] Processing unit(s) 112 and can represent, for example, a CPU-type processing unit, a GPU-type processing unit, a field-programmable gate array (FPGA), another class of digital signal processor (DSP), or other hardware logic components that may, in some instances, be driven by a CPU. For example, and without limitation, illustrative types of hardware logic components that can be used include Application- Specific Integrated Circuits (ASICs), Application- Specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc. In various embodiments, the processing unit(s) 112 may execute one or more modules and/or processes to cause the server(s) 110 to perform a variety of functions, as set forth above and explained in further detail in the following disclosure. Additionally, each of the processing unit(s) 112 may possess its own local memory, which also may store program modules, program data, and/or one or more operating systems.

[0044] In at least one configuration, the computer-readable media 114 of the server(s) 110 may include components that facilitate interaction between the service provider 102 and the users 106. For example, the computer-readable media 114 may include the data platform module 116, the intimacy module 118, the ranking module 120, and the output module 122. The modules (116, 118, 120, and 122) can be implemented as computer-readable instructions, various data structures, and so forth via at least one processing unit(s) 112 to configure a device to execute instructions and to perform operations implementing. Functionality to perform these operations may be included in multiple devices or a single device.

[0045] Depending on the exact configuration and type of the server(s) 110, the computer-readable media 114 may include computer storage media and/or communication media. Computer storage media can include volatile memory, nonvolatile memory, and/or other persistent and/or auxiliary computer storage media, removable and non-removable computer storage media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules, or other data. Computer memory is an example of computer storage media. Thus, computer storage media includes tangible and/or physical forms of media included in a device and/or hardware component that is part of a device or external to a device, including but not limited to random-access memory (RAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), phase change memory (PRAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, compact disc read-only memory (CD-ROM), digital versatile disks (DVDs), optical cards or other optical storage media, miniature hard drives, memory cards, magnetic cassettes, magnetic tape, magnetic disk storage, magnetic cards or other magnetic storage devices or media, solid-state memory devices, storage arrays, network attached storage, storage area networks, hosted computer storage or any other storage memory, storage device, and/or storage medium that can be used to store and maintain information for access by a computing device.

[0046] In contrast, communication media may embody computer readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave, or other transmission mechanism. The term "modulated data signal" means a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. Such signals or carrier waves, etc. can be propagated on wired media such as a wired network or direct- wired connection, and/or wireless media such as acoustic, RF, infrared and other wireless media. As defined herein, computer storage media does not include communication media. That is, computer storage media does not include communications media consisting solely of a modulated data signal, a carrier wave, or a propagated signal, per se. Exploring the Dynamics of Intimacy

[0047] FIG. 2 is a diagram showing an example system for exploring various dynamics of social interactions between users of communication platforms and leveraging the dynamics to rank the users based on degrees of intimacy. As shown in FIG. 2, the system 200 may include the data platform module 116, the intimacy module 118, the ranking module 120, and the output module 122.

[0048] The data platform module 116 may be configured to crawl user application programming interfaces (APIs) to identify friend users u and recursively explore a large number of users 106. The data platform module 116 may include a user scheduler 202, profiler crawler 204, relation crawler 206, communication crawler 208, and directed communication crawler 210. In some embodiments, crawlers are optional. For example, in embodiments where the user of the communication platform owns the user behavior data, crawlers may not be necessary. However, in embodiments where the user of the communication platform does not own the data, crawlers may be used to identify friend users u and to collect data.

[0049] The intimacy module 118 may include a variety of modules and components configured to store and/or process user behavior data collected in the data platform module 116. For example, the intimacy module 118 may include a user profiles storage component 212, a user relations storage component 214, a communications storage component 216, and a directed communications storage component 218. Additionally, intimacy module 118 may include an affinity calculation module 220, a topic extraction module 222, a segmentation module 224, a name-entity recognition module 226, a sentiment analysis module 228, a self-disclosure analysis module 230, and a responsiveness and reciprocity analysis module 232. The intimacy module 118 may include a variety of other components including profile affinity index 234 and topic affinity index 236 that may be stored together in the index compartment 238, communication-level features storage component 240, and responsiveness matrix 242.

[0050] The ranking module 120 may include a ranking model module 244 and an offline learning module 246. The output module 122 may be configured for producing graphical user interfaces to provide multiple aspects of user dynamic intimacy visualization.

[0051] In at least one embodiment, the data platform module 116 may be configured to crawl user application programming interfaces (APIs) to identify friend users u and recursively explore a large number of users 106. Additionally, the data platform module 116 may traverse and crawl the communications initiated by an ego user v and directed communications responsive the communications initiated by the ego user v. The user scheduler 202 controls the crawling process which collects user behavior data from communications between users 106 of communication platforms. The profiler crawler 204 crawls user profiles to collect user behavior data relating to the ego user u and friend users u. The relation crawler 206 crawls the friendship relationships (e.g., relationships between a user 106 and his or her Twitter® followers, relationships between a user 106 and his or her Skype® contacts, relationships between a user 106 and his or her friends on Xbox Live®, etc.) to collect user behavior data relating to relationships between the ego user u and friend users u. The communication crawler 208 crawls all communications between an ego user u and each friend user u of the ego user u. The directed communication crawler 210 may check whether the ego user u provided a directed communication in response to a communication from a friend user u. If the ego user v provided a directed communication in response to a communication, the directed communication crawler 210 crawls the directed communication.

[0052] In at least one embodiment, the intimacy module 118 may be configured to perform processing to the data collected in the data platform module 116. For example, the intimacy module 118 may apply one or more models including a reciprocity model, a temporality model, an affinity model, and/or a contextuality model to the data collected in the data platform module 116. The data processing performed by the intimacy module 118 may generate a series of features that may be user- level features or communication-level features. User-level features are features that are associated with user behavior (e.g., reciprocal rank, profile affinity, topic affinity, and/or responsiveness value). Communication- level features are features extracted from the communications or directed communications (e.g., sentiment value and/or self-disclosure value).

[0053] In some embodiments, the intimacy module 118 may include storage components configured for storing user profile information, user relation information, communications, directed communications, and communication-level features. For example, user behavior data collected from the profile crawler 204 may be stored in the user profile storage component 212 and user behavior data collected from the relation crawler 206 may be stored in the user relations storage component 214. Data collected from the communication crawler 208 may be stored in the communications storage component 216 and data collected from the directed communication crawler 210 may be stored in the directed communications storage component 218. The intimacy module 118 may process the data collected from the data platform module 116 by applying one or more models to the data using one or more modules in the intimacy module 118.

[0054] The intimacy module 118 may include a variety of modules configured to process data collected in the data platform module 116. For example, intimacy module 118 may include the affinity calculation module 220, the topic extraction module 222, the segmentation module 224, the name-entity recognition module 226, the sentiment analysis module 228, the self-disclosure analysis module 230, and the responsiveness and reciprocity analysis module 232. The intimacy module 118 may include a variety of other components including the profile affinity index 234 and topic affinity index 236, the communication-level features storage component 240, and the responsiveness matrix 242.

[0055] The affinity calculation module 220 may receive user behavior data stored in the user profiles storage component 212 and the user relations storage component 214. In at least one embodiment, the affinity calculation module 220 may be configured to calculate an affinity calculation between the ego user u and one or more of the friend users u. In some embodiments, the affinity calculation may be based on a profile affinity measurement and a topic affinity measurement. The two affinity measurements (e.g., profile and topic) may capture the homophily between the ego user u and friend users u.

[0056] The profile affinity measurement considers a similarity between features included in a user's profile. For example, features that are used to build indexes representing profile affinity may include gender, age, education, professions, geographic locations visited, city, common friends, etc. In some embodiments, the features may be binary. For instance, the gender feature and city feature simply indicate whether a friend user u of the ego user u has the same gender and/or lives in the same city as the ego user v. However, other features may require an application of an affinity function to calculate the profile affinity measurement. For example, a Jaccard Similarity may be used to calculate affinity measurements for features such as education affinity, common friends affinity, and age affinity. Equations (l)-(3) below illustrate nonlimiting examples of affinity measurements for education affinity, common friends affinity, and age affinity.

l¾n¾|

Affinity(Education) EQ. 1 where a friend user's u school list S_u and an ego user's u school list

Affinity(CommonFriends) EQ. 2

where E_u ^° represents an index of an ego user's u friends.

min(A_u,A_v)

Affinity(Age) EQ. 3

max(A_u,A_v)'

where A_u and Α_Ώ are the ages of a friend user u and the ego user u. [0057] If users 106 have not input the features described above and/or the information has not otherwise been provided, the affinity calculation module 220 may assign a value (e.g., -1) to represent the profile affinity measurement. The profile affinity measurements may be indexed in a sorted list and stored in a profile affinity index 234 of the intimacy module 118. In at least one embodiment, the profile affinity measurements may be sorted in descending order of affinity value.

[0058] In addition to performing calculations to determine a profile affinity measurement, the affinity calculation module 220 may also calculate a topic affinity measurement for summarizing and comparing users' 106 topics of interest. In at least one embodiment, the topic affinity measurement is based on keywords identified from a corpus of communications and directed communications from the ego user u and/or one or more friend users u. The keywords may be identified by a topic extraction process performed by the topic extraction module 222.

[0059] In at least one embodiment, to initiate the topic discovery and keyword extraction process, the topic extraction module 222 identifies named-entities from a corpus of communications and directed communications from the ego user u and/or one or more friend users u. In some embodiments, the communications that make up the corpus may be text or speech communication. In embodiments where the communication includes speech communication, a speech-to-text conversion may occur prior to starting the topic discovery and keyword extraction process. In some embodiments (i.e., in the case of Chinese words), individual words in the corpus are segmented via preprocessing in the segmentation module 224. Unlike formal communication (e.g., news reports, books, etc.), the communications described herein (e.g., Tweets, microblogs, text messages, etc.) are much shorter strings of text, having more diverse topics and numerous noisy or meaningless words. Accordingly, after the named- entities are identified, the corpus is filtered to remove any non-named-entities. Then, the topic extraction module 222 performs topic discovery and topical keyword extraction on the corpus. In at least one embodiment, a Twitter-LDA model is leveraged to discover topics, where each word in a Tweet (e.g., communication) is either assigned as a topical word or a background word according to a Bernoulli test. Then, a PageRank-based method is utilized to identify important words in a topic as topical key phrases. For example, the topic extraction module 222 obtains a topic distribution Θ_Μ = (Θ_Μ ⁽¹⁾, Θ_Μ ⁽²⁾...Θ_Μ ^{( )}) for each user u 106, where 9_M ^W, is a distribution over words, i= 1, 2, K. For each topic i, the topic extraction module 222 may identify a list of important words through context-aware topical PageRank (cTPR), as topical keywords of user u.

[0060] In some embodiments, communication platform users may self-identify several keywords as their interests. The affinity calculation module 220 may merge the identified keywords and the self-identified keywords into a single collection of final keywords. The collection of final keywords may be used to compute the topic affinity measurement between u and v £ E_u ^° using cosine similarity as shown in Equation 4 below. Affinity(Topic) =

The resulting topic affinity measurements may be indexed in a sorted list and stored in the topic affinity index 236 in the intimacy module 118. In at least one embodiment, the topic affinity measurements may be sorted in descending order of affinity value. Topic discovery is addressed further regarding FIG. 3, below.

[0061] FIG. 3 illustrates a network view 300 depicting topic discovery comparisons for an ego user u and friend users u. The topic affinity index 236 of FIG. 2 may use these comparisons to sort topic affinity measurements in descending order of affinity value. In FIG. 3, outer layer circles 302 denote the topic set and middle layer circles 304 identify specific topics identified during the topic discovery process. For the purposes of FIG. 3, there are six (6) middle layer circles 304 representing six (6) topics: children, service, jobs, pets, real estate, and celebrities. The inner layer circles 306 show users' distribution in each topic. In at least one embodiment, the length of the radius of the circles represents the probability distribution. For example, the topic of pets is a similar topic of communication between the ego user u and two friend users u (as shown by circles 306 A and 306B). However, the different radiuses represented by the inner layer circles 306A and 306B communicates to a user 106 that one friend 306A discusses pets more frequently than the other friend 306B. A user 106 may click on different levels of circles to zoom in and/or zoom out. Circle 308 represents a magnified (e.g., zoomed- in) view of the circle representing the identified topic of jobs. While FIG. 3 is discussed in terms of circles, other shapes and/or designs could be used to illustrate topic discovery and offer comparisons for an ego user u and friend users u.

[0062] The intimacy module 118 may include one or more indexes. In at least one embodiment, index 238 may receive data from the affinity calculation module 220 and may store the data in the profile affinity index 234 or the topic affinity index 236, depending on the type of data (e.g., profile affinity measurement or topic affinity measurement) received.

[0063] Depending on the context of a communication, users may perceive different degrees of intimacy. In at least one embodiment, the intimacy module 118 may extract rich contextual features from the communications and directed communication to diagnose a user's context (e.g., contextuality) when interacting with friends. Different contextual features (e.g., emotions, opinions, and level of self-disclosure) may be useful in observing the rich dynamics of tie strength. In at least one embodiment, the intimacy module 118 may receive data from the communications storage component 216 and the directed communications storage component 218 and may analyze contextual features of the data. For example, the intimacy module 118 may consider contextual features of the data including sentiment, self-disclosure, and responsiveness in the sentiment analysis module 228, the self-disclosure module 230, and the responsiveness and reciprocity module 232, respectively.

[0064] Sentiment describes feelings, attitudes, and/or opinions. In some embodiments, sentiment may be a contextual feature relevant to the dynamics of tie strength. In addition to words that may communicate emotions, many communications include emoticons to express emotions. The sentiment analysis module 228 may mine emotions (e.g., happy, sad, etc.) and polarity (e.g., positive, neutral, negative, etc.) from the corpus of communications and directed communications from the ego user u and/or one or more friend users u. In at least one embodiment, emoticons and individual words in a dictionary of words are classified as representing emotions. For instance, emoticons and individual words may be classified as happy or sad and/or positive, neutral, or negative. FIG. 4 illustrates classifications of emoticons 400 and Table 1 below describes classifications of individual words.

TABLE 1

[0065] The sentiment analysis module 228 of FIG. 2 mines the corpus and identifies words and/or emoticons. In some embodiments, the communications that make up the corpus may be text or speech communication. In embodiments where the communication includes speech communication, a speech-to-text conversion may occur prior to performing a sentiment analysis. Additionally, in some embodiments, (i.e., in the case of Chinese words), individual words in the corpus are segmented via preprocessing in the segmentation module 224 prior to the sentiment analysis module 228 mining the corpus. In at least one embodiment, after words and/or emoticons have been identified in the corpus, the words and emoticons may be classified by the emotions they represent (e.g., happy, sad, positive, neutral, and/or negative), based on the pre-identifled classes. In other embodiments, lexicon-based topic mining may be used to identify words associated with emotions and/or polarity. In some embodiments, a sentiment value may be determined by Equation 5 modeled below.

_ . _{Ύ Ύ} , #(happy word/emoticons) -#(sad words/emoticons) _π _ _

Sentiment Value = #( ^hap ^μp^μy^γ words/emoticons)— +#(sad words/emoticons) EQ. 5

In at least one embodiment, a sentiment value above a threshold may be indicative of a happy user 106 and a sentiment value below a threshold may be indicative of a sad user 106.

[0066] In other embodiments, the sentiment value may include a polarity value that is calculated by leveraging the emoticons as noisy labels to train a support vector machine (SVM) classifier using a linear kernel. Unigram, bigram, and other contextual features (e.g., punctuation, hashtags, and/or target-dependent features) may be used to train the SVM classifier. For the communications and directed communications that make up the corpus, the polarity value may be calculated using the trained model and classified based on the emoticons. In embodiments where no emoticons are used, a dummy class may be assigned as the emotion.

[0067] In some embodiments, the sentiment value may be considered a communication-level feature. Accordingly, in at least one embodiment, the sentiment value and related data may be stored in the communication-level features storage component 240 of FIG. 2.

[0068] FIG. 5 illustrates a network view 500 depicting sentiments of an ego user vj and eight friend users u over time. The vertical axis 502 denotes eight users 106 (A-H). The horizontal axis 504 denotes time. A variety of shapes may be used to represent a sentiment value. For the purpose of FIG. 5, circles 506 are used to represent the sentiment value between the ego user u and friend users u. The color, texture, and/or appearance of the shapes may vary to visually differentiate between sentiments. For example, in FIG. 5, circle 506A appears bold and circle 506B does not appear bold. Additionally, circle 506A appears to have a larger radius than circle 506B. The bold circle 506A represents a sentiment value associated with sadness. Conversely, circle 506B that is not bold represents a sentiment value associated with happiness.

[0069] In at least one embodiment, the length of the radius of the circles 506 represents a degree of sentiment value. For example, a relatively large, bold circle (e.g., 506A) may represent a very sad sentiment. Conversely, a relatively small circle (e.g., 506B) that is not bold may represent a mildly happy sentiment. However, a relatively large circle 506C that is not bold may represent a very happy sentiment. The sentiment network view illustrated in FIG. 5 may be used to show how sentiment varies between an ego user u and friend users u and how those sentiments may change over time.

[0070] Self-disclosure describes communications though which users 106 reveal information about himself or herself to other users 106. In some embodiments, self- disclosure may be a contextual feature relevant to the dynamics of tie strength. In some embodiments, self-disclosure may be a key characteristic in indicating the strength of social ties. The self-disclosure analysis module 230 of FIG. 2 may analyze a variety of linguistic features in a corpus of communications and directed communications to calculate a self-disclosure value indicative of the degree of self-disclosure in online interactions. In some embodiments, the communications that make up the corpus may be text or speech communication. In embodiments where the communication includes speech communication, a speech-to-text conversion may occur prior to performing a self- disclosure analysis. In some embodiments, (i.e., in the case of Chinese words), individual words in the corpus are segmented via preprocessing in the segmentation module 224 prior to the self-disclosure analysis module 230 analyzing the linguistic features of the corpus.

[0071] For example, the self-disclosure analysis module 230 may consider a length of communications, an occurrence of first person words, and/or geo- communications to calculate a self-disclosure value. In at least one embodiment, the self- disclosure analysis module 230 may consider a number of words included in a communication or directed communication (e.g., the length of a Tweet, microblog, text message, etc.). Generally, communications having a number of words over a threshold (i.e., longer communications) tend to include more self-disclosure due to users sharing personal information, feelings, or thoughts with others more often than communications comprised of a number of words under a threshold (i.e., shorter communications). In at least one embodiment, the self-disclosure analysis module 230 may consider the occurrence of first person words (e.g., I, me, my, mine, our, etc.) in the communications. A frequent occurrence of first person words is indicative of self-disclosure. The self- disclosure analysis module 230 may consider geo-communications, or communications wherein a geographic location is indicated in the communication or at the time of posting the communication. Geo-communications are indicative of self-disclosure.

[0072] Additionally, the self-disclosure analysis module 230 of FIG. 2 may consider the user device 108 from which a user 106 sends or posts a communication to determine a self-disclosure value. For example, if a user 106 posts a microblog (e.g., a Tweet, etc.) from his or her mobile user device 106, it is more likely that the user 106 is on-the-move and is less likely to self-disclose. Conversely, if a user 106 posts a microblog (e.g., a Tweet, etc.) from a web user device 108, it is more likely that the user 106 is in a more sedentary location (e.g., home or work) where he or she is more likely to self-disclose.

[0073] The various linguistic features discussed herein are leveraged to determine a degree of self-disclosure, measured by the self-disclosure value. In some embodiments, the self-disclosure value may be considered a communication-level feature. Accordingly, in at least one embodiment, the self-disclosure value and related data may be stored in the communication-level features storage component 240 of FIG. 2.

[0074] Responsiveness describes the empirical probability that a user 106 (e.g., ego user u) will respond to his or her friend users u. Reciprocity describes a user's 106 relative tendency to interact with friend users u in an ego network. Responsiveness may be a contextual feature relevant to the dynamics of tie strength. Reciprocity may be an additional feature relevant to the dynamics of tie strength. The responsiveness and reciprocity analysis module 232 of FIG. 2 may determine a responsiveness value and a degree of reciprocity.

[0075] In a responsiveness analysis, historical interaction data may be used to measure an empirical probability that a user 106 (e.g., ego user u) is likely to respond to a communication from a friend user u. In some embodiments, the responsiveness analysis may calculate a responsiveness value based on an availability of a user 106 to respond to a communication ("availability"), a capacity of a user 106 to read communications and interact with friend users u ("capacity"), and a tendency of a user 106 to interact with particular friend users u ("tendency"). In at least one embodiment, the availability, capacity, and tendency of a user 106 to interact with other users 106 are time- dependent variables.

[0076] In some embodiments, a responsiveness value with respect to each friend user u of an ego user u may be calculated by the equations modeled below in Equations 6-9. In Equations 6-9 below, Δ indicates a time period such that Δ = t_\, t₂]. Τ_υ ^Α may be used to denote the communications (T) posted by the ego user u within the time period Δ and R " y_U may be used to denote the directed communications (JV) (e.g., reply communications or rebroadcast communications, etc.) sent from friend users u to the ego user u within the time period Δ. A cost value associated with the ego user u reading communications and responding to the communications may be estimated by:

In some embodiments, the greater the cost value, the less available the ego user u may be to respond to a particular communication sent in a given time period.

[0077] In some embodiments, Equation 7 below may be used to approximate the capacity of the ego user u to read communications and interact (e.g., reply with a communication or rebroadcast a communication) with other friend users u. A capacity value may be estimated by:

[0078] Based on the cost value (that is inversely related to the ego user's u availability) and the capacity value, the responsiveness and reciprocity module 232 may calculate a responsiveness value relevant to reply communications using Equation 8 below.

'v→u \reply

In Equation 8, the numerator captures the tendency of an ego user v to respond to a friend user u and the capacity of the ego user u and the inverse of the denominator captures the availability of the ego user u. Additionally, the responsiveness and reciprocity module 232 may calculate a responsiveness value relevant to rebroadcasting (e.g., re-Tweeting or re -posting) communications as modeled by Equation 9.

In Equation 9, J? ,w ^was replaced with R _V ^A _W, where JIT represents rebroadcast (e Tweeted or re -posted) communications.

[0079] Using the following application of the responsiveness and reciprocity module, responsiveness data may be stored in one or more matrixes in the intimacy module 118 including a responsiveness matrix 242 as shown in FIG. 2. For example, the responsiveness matrix 242 may include two separate matrices (one for reply communications and one for rebroadcasting communications), where each row of the matrices may be associated with a user and each column indicates a time period.

[0080] FIG. 6 illustrates an example responsiveness matrix 600. Responsiveness matrix 600 may represent responsiveness matrix 242 described above. In FIG. 6, each column 602 of the responsiveness matrix 600 illustrates a time period. For the purpose of FIG. 6, the time periods are represented in one hour increments of a 24-hour day (e.g., 12a, la... l2p, lp... l lp). Each row 604 of the responsiveness matrix 600 represents individual friend users u, identified as users 106 A through user G. The various patterns in the squares represent different responsiveness values where the pattern in square 606 represents a low responsiveness value and the pattern in square 608 represents a high responsiveness value. The patterns in the squares between 606 and 608 represent increasing responsiveness values as a user 106 travels from square 606 to square 608. Other designs and/or configurations may be used to display the responsiveness matrix 600. Similarly, colors, alternative patterns, textures, etc., may be used to differentiate responsiveness values. The responsiveness matrix 600 allows an ego user u to observe his or her probability to respond (e.g., reply to a communication or rebroadcast a communication) to his or her friend users u during various time periods.

[0081] In addition to determining a responsiveness value, the responsiveness and reciprocity analysis module 232 of FIG. 2 may determine a degree of reciprocity. The degree of reciprocity is a value representing the frequency by which a friend user u sends a directed communication (e.g., reply communication, re -broadcast communication, etc.) in response to a communication initiated by an ego user u. In at least one embodiment, the responsiveness and reciprocity analysis module 232 quantifies reciprocity between an ego user u and a friend user u using a percentage rank (PR) that may be calculated by the equation modeled in Equation 10.

u

where f_u→w represents the outgoing interaction frequency from u to w (and w represents a friend user different from friend user u). In Equation 10, a PR below a threshold may be indicative of a friend user u having a higher percentage rank than a friend user u having a PR above a threshold.

[0082] The percentage rank measures the percentage of users 106 who have received a same number of or more interactions from ego user u when compared with friend user u. Using the percentage rank, the responsiveness and reciprocity analysis module 232 of FIG. 2 may calculate a reciprocal interaction rank (RIR) from ego user u to friend user u. In at least one embodiment, the reciprocal interaction rank may be calculated by the equation modeled in Equation 11.

RIR_U→V = (PR_u(v) - PR_v(uj) x J) EQ. 11

[0083] In Equation 11, the difference between PR_u v) and PR_v(ii) is scaled by

Dunbar's number (35), representing an approximate number of stable social ties in an ego network (150) correlated with a relative neocortex size. The reciprocal interaction rank represents a degree of inequality in terms of perceived tie strength. The reciprocal interaction rank may represent the degree of reciprocity between an ego user u and a friend user u and may be leveraged to calculate the degree of intimacy between the ego user vj and friend user u.

[0084] The tie strength between an ego user u and a friend user u may change over time. Thus, the intimacy module 118 may observe and analyze social interactions in successive time periods, including in some instances via a dedicated temporality module (not pictured). In at least one embodiment, the intimacy module 118 catches and interactively visualizes temporal dynamics related to the frequency of communications between an ego user u and a friend user u. In some embodiments, the intimacy module 118 catches and interactively visualizes the temporal dynamics of the frequency of directed communications from a friend user u to the ego user u.

[0085] FIG. 7 illustrates a network view 700 depicting changes in the frequency of interaction, or degree of reciprocity, between an ego user u and his or her friend users u over time. The vertical axis 702 denotes an interaction frequency, or a degree of reciprocity. The horizontal axis 704 indicates time. The lines 706 represent interaction frequencies between the ego user u and four friend users u over time. Each user 106 is represented by a different line pattern. The network view illustrated in FIG. 7 may be used to show how a degree of intimacy changes over time.

[0086] In some embodiments, the indexes (e.g., profile affinity index 234 and topic affinity index 236) and responsiveness matrix 242 are periodically updated to maintain user's temporal updates to their profile and topic affinity, as well as responsiveness. Communication-level features, however, may be streaming data that may be updated in real-time to reflect timely and subtle changes in a user's sentiment and degree of self-disclosure.

[0087] In some embodiments, the ranking module 120 as shown in FIG. 1 and

FIG. 2 may receive data from the intimacy module 118 to generate an intimacy score, or degree of intimacy, for one or more users 106 and may rank two or more users 106 according to the users' 106 intimacy scores. In at least one embodiment, the ranking module 120 may receive information from index 238 (e.g., profile affinity measurement and/or topic affinity measurement), communication-level features storage component 240 (e.g., sentiment values, self-disclosure values, etc.), and/or the responsiveness and reciprocity analysis module 232 (e.g., degree of reciprocity, responsiveness value, etc.). In some embodiments, the ranking module 120 may include a ranking model module 244 and an offline learning module 246 as shown in FIG. 2.

[0088] The offline learning module 246 may be configured to execute an offline learning phase to learn a ranking model combining all the contextual features described above (e.g. profile affinity measurement, topic affinity measurement, sentiment value, self-disclosure value, responsiveness value, etc.). In at least one embodiment, to learn the ranking model, sample sets of positive and negative instances may be used. For example, a feature vector \(q, v) may be generated for each pair of communication-user pairs (q, u) where q represents a communication sent by a user u and υ G E_u ^° , the ego user. In at least one embodiment, \(q, u) includes all of the extracted user-level features associated with friend user u and ego user u and communication-level features associated with the communication q. In at least one embodiment, Y = (1, 2, ··· , m) represents a label set for a user-communication pair (q, u), where the label represents a friend user's u intention to respond to an ego user's u communication. A training set may be represented by S ⁼ {(^xi_> where x_t G X represents the feature vector of the z^'th user- communication pair and y_t = Y. The ranking model F(5) = f(x) may be trained to assign a score (e.g., a degree of intimacy) for a feature vector x and rank friend users u according to the score. In at least one embodiment, the training may be a learning-to- rank framework for information retrieval that is well known in the art, for example, point- wise approaches (e.g., Logistic Regression), pair-wise approaches (e.g., RankingSVM), and list-wise approaches (e.g., LambdaMART) may be applied.

[0089] Equation 12 models the Logistic Regression function, where w represents the weight vector for the feature vector x, which may be estimated using maximum likelihood estimation. (x) = Pr(y = i|x) = _I^i<^ ^EQ- ¹²

[0090] Equation 13 models the RankingSMV function to learn a linear ranking function /(x) = (w, x), where the weight feature w is trained by minimizing the regularized hingle loss function, given by:

^ IMI² + C∑_fti)eP[<w, x_i - x_i)]₊, EQ. 13 where [x]₊ = max(0,l— x) and (i,y) G T means user i is preferred over user j.

[0091] Equation 14 models the LamdaMART function for learning a combination of regression tress as the ranking function, where each f_t (x) is a regression over the features. LambdaMART may be trained by empirically optimizing IR measures such as NDCG directly.

f (x) =∑₌₁ a_if_i (x) EQ. 14

[0092] The ranking module 120 may be trained to assign a score (e.g., degree of intimacy) for a feature vector, and can rank communications initiated by the ego user u.

Following the processing in the intimacy module 118, the ranking module 120 may receive data extracted from the intimacy module 1 18 and may sort friend users u according to the scores (e.g., degree of intimacy) generated by applying ranking function.

In at least one embodiment, friend users u may be ranked based on friend users u who are likely to respond to a communication from an ego user u based on the friend user's u intimacy scores (e.g., degree of intimacy). Alternatively, friend users u may be ranked based on an order in which friend users u 106 are likely to respond to a communication from an ego user u based on the friend users ' intimacy scores. Thus, in at least one embodiment, the ranking order may be used to predict which friend users u are most likely to respond to a particular communication from the ego user u and to predict an order in which the friend users u are likely to respond.

[0093] In addition to ranking users 106 to determine which users 106 are likely to respond and/or an order in which users 106 are likely to respond, the ranking module 120 may perform additional actions based on the degree of intimacy. For example, the ranking module 120 may leverage the degree of intimacy to suggest one or more users 106 who may be interested in some subject matter and/or media object being viewed by a user 106 and/or suggest one or more users 106 to share an experience (e.g., a game, a video, etc.). Additionally, the ranking module 120 may leverage the degree of intimacy to filter and/or order a displayed list of users 106 (e.g., contacts) for future connections or communications and/or may modify a displayed list of users 106 to add and/or remove users 106. Furthermore, the ranking module 120 may leverage the degree of intimacy to present icons on a displayed list of users 106 (e.g., contacts) reflective of the determined degree of intimacy. Similarly, the ranking module 120 may leverage the degree of intimacy to update and/or modify the icons based on changes in the degree of intimacy over time.

[0094] FIGS. 8-10 illustrate three example network views that may be used to visualize user intimacy: an overall network view, an ego network view, and a community network view, respectively.

[0095] FIG. 8 illustrates an overall network view user interface 800. In FIG. 8, users are represented by nodes 802 and edges 804 are shown to connect nodes. Various shapes may be used to represent the nodes 802. For the purposes of FIG. 8, the nodes appear as circles. The edges 804 represent ties between users 106 and may vary in strength. In at least one embodiment, variations on edge 804 thickness and direction may be shown in the overall network view to represent various strengths of ties between users 106. For example a strong tie between two users 106 may be represented by an edge 804 that appears thicker than an edge 804 that represents a weak tie between two users 106. In FIG. 8, a user may scroll a mouse over the user interface 800 causing individual nodes 802 and/or edges 804 to change color. In some embodiments, names of users 106 associated with the nodes 802 may appear when the user scrolls the mouse over the user interface 800. For the purposes of privacy, an ego user u interested in viewing his or her overall network may be required to sign-in or authenticate prior to exploring their overall network views as illustrated in FIG. 8.

[0096] FIG. 9 illustrates an alternative view that may be used to visualize user intimacy, an ego network view 900 user interface. In at least one embodiment, the ego network view includes a node 902 that represents the ego user u. The ego network view 900 may also include additional nodes 904 (e.g., 904A, 904B, 904C) that represent individual friend users u. Each friend user u may be associated with his or her own edge 906. The nodes 904 representing the friend users u may vary in color, appearance, size, etc. For the purposes of FIG. 9, the nodes are illustrated as circles. As can be seen in FIG. 9, nodes 904A, 904B, and 904C appear bold. The variations in color and/or appearance may identify an ego user's u most intimate friend users u. Accordingly, for the purposes of FIG. 9, friends represented by nodes 904A, 904B, and 904C are ego user's vj most intimate friend users u. Similarly, variations in size may represent the frequency of communication, or degree of reciprocity, between the ego user u and the particular friend user u identified by the node 904. For example, the ego user u communicates more frequently, or has a high degree of reciprocity, with the user represented by node 904A than the user represented by node 904C because the radius of node 904A measures more than the radius of node 904C. This ego network view 900 allows an ego user u to view his or her ego network. In some embodiments, an ego user u may click on his or her close friend user's u nodes and observe his or her close friends' ego networks.

[0097] FIG. 10 illustrates an alternative view that may be used by an ego user u to explore him or herself in a network via a community view 1000. FIG. 10 illustrates an interaction history of the ego user's u community with tension 1002 and time 1004 axes in an edge bundling graph. In the edge bundling graph, points 1008 on the perimeter of the circle represent individual users 106. Edges 1006 represent social ties and individual edges 1006 may vary in width. The variations in width represent variations in interaction frequency. In at least one embodiment, wider edges 1006 represent more frequent communication, or a high degree of reciprocity, than narrower edges 1006. In some embodiments, the ego user u may move his or her mouse over the user interface and the edges 1006 may change color. In at least one embodiment, statistics of reciprocal rank may be depicted on the user interface in a bar chart or some other means of communicating the reciprocal rank.

Example Processes

[0098] FIGS. 11-13 describe example processes for investigating dynamics of intimacy to calculate a degree of intimacy between an ego user u and friend users u. The example processes are described in the context of the environment of FIGS. 1 and 2 but are not limited to those environments. The processes are illustrated as logical flow graphs, each operation of which represents an operation in the illustrated or another sequence of operations that may be implemented in hardware, software, or a combination thereof. In the context of software, the operations represent computer-executable instructions stored on one or more computer-readable media 114 that, when executed by one or more processors 112, configure a computing device to perform the recited operations. Generally, computer-executable instructions include routines, programs, objects, components, data structures, and the like that configure a computing device to perform particular functions or implement particular abstract data types.

[0099] The computer-readable media 114 may include hard drives, floppy diskettes, optical disks, CD-ROMs, DVDs, read-only memories (ROMs), random access memories (RAMs), EPROMs, EEPROMs, flash memory, magnetic or optical cards, solid-state memory devices, or other types of storage media suitable for storing electronic instructions, as described above. Finally, the order in which the operations are described is not intended to be construed as a limitation, and any number of the described operations may be combined in any order and/or in parallel to implement the process.

[00100] FIG. 1 1 illustrates a process 1100 for calculating a degree of intimacy between an ego user u and friend users u of communication platforms.

[00101] Block 1 102 illustrates determining a degree of reciprocity between a first user (e.g. an ego user u) and a second user (e.g., a friend user u). As described above, the intimacy module 118 may determine the degree of reciprocity based at least in part on a frequency of directed communication from a friend user u to the ego user u in response to communication initiated by the ego user u.

[00102] Block 1104 illustrates determining a frequency of the directed communication in successive time periods. As described above, the indexes (e.g., profile affinity index 234 and topic affinity index 236) and responsiveness matrix 242 are periodically updated to maintain user's temporal updates to their profile and topic affinity, as well as responsiveness. Communication-level features, however, may be streaming data that may be updated in real-time to reflect timely and subtle changes in a users' sentiment and degree of self-disclosure.

[00103] Block 1 106 illustrates performing an affinity calculation between the first user and the second user. As described above, the intimacy module 118 may determine the affinity calculation based on a profile affinity and/or a topic affinity.

[00104] Block 1 108 illustrates analyzing contextual features. As described above, contextual features may include at least one of sentiment, self-disclosure, or responsiveness. The intimacy module 118 may analyze the contextual features to determine sentiment values, self-disclosure values, and/or responsiveness values.

[00105] Block 1110 illustrates calculating a degree of intimacy between the first user and the second user. As described above, the intimacy score, or degree of intimacy, may be based at least in part on the degree of reciprocity, the frequency, the affinity calculation, and at least one of the sentiment values, the self-disclosure values, or the responsiveness values derived from the communication initiated by the first user and directed communication.

[00106] FIG. 12 illustrates a process 1200 for leveraging the degree of intimacy to rank two or more users of a set of users of communication platforms and predicting which user of the two or more users is likely to respond to a particular communication from the first user.

[00107] Block 1202 illustrates ranking based on the degree of intimacy. As described above, the ranking module 120 may rank two or more users of the communication platforms based on the degree of intimacy between the first user and the two or more users.

[00108] Block 1204 illustrates predicting based on the ranking. As described above, the ranking module 120 may rank which user of the two or more users is likely to respond to a particular communication from the first user based at least in part on the ranking. Additionally, the ranking module 120 may determine an order in which the two or more users 106 who are likely to respond are likely to respond. The ranking module 120 may also determine an order in which the two or more users 106 who are likely to respond are likely to respond. Furthermore, the ranking module 120 may leverage the degree of intimacy to make other predictions as described above (e.g., suggest one or more users 106 who may be interested in some subject matter and/or media object being viewed by a user 106, filter and/or order a displayed list of users 106 (e.g., contacts) for future connections or communications, and/or present icons on a displayed list of users 106 (e.g., contacts) reflective of the determined degree of intimacy, etc.).

[00109] FIG. 13 illustrates a process 1300 for calculating a degree of intimacy between an ego user u and friend users u of communication platforms and leveraging the degree of intimacy to predict which user of the two or more users is likely to respond to a particular communication from the first user.

[00110] Block 1302 illustrates receiving communication and/or directed communication. As described above, the data platform module 116 collects communications initiated by an ego user u and directed communications from a friend user u in response the communication initiated by the ego user u. These communications may form a corpus for processing by the intimacy module 118.

[00111] Block 1304 illustrates determining profile affinity. As described above, the affinity calculation module 220 calculates a profile affinity measurement that considers a similarity between features included in a user's profile. For example, features that are used to build indexes representing profile affinity may include gender, age, education, professions, geographic locations visited, city, common friends, etc. [00112] Block 1306 illustrates performing pre-processing. As described above, prior to topic extraction, sentiment analysis, and/or self-disclosure analysis, in some embodiments (i.e., in the case of Chinese words), individual words in the corpus are segmented via preprocessing in the segmentation module 224. Prior to topic extraction, the corpus may be filtered to remove non-named-entities.

[00113] Block 1308 illustrates extracting topical keywords. As described above, the topic extraction module 222 may identify important words in a topic as topical key phrases.

[00114] Block 1310 illustrates performing the sentiment analysis. As described above, the sentiment analysis module 228 may mine emotions (e.g., happy, sad, etc.) and polarity (e.g., positive, neutral, negative, etc.) from a corpus of communications and directed communications from the ego user u and/or one or more friend users u to determine a sentiment value.

[00115] Block 1312 illustrates preforming the self-disclosure analysis. As described above, the self-disclosure analysis module 230 may analyze a variety of linguistic features to calculate a self-disclosure value indicative of the degree of self- disclosure in online interactions.

[00116] Block 1314 illustrates performing responsiveness analysis. As described above, the responsiveness and reciprocity analysis module 232 may perform a responsiveness analysis to calculate a responsiveness value that may be based on an availability of a user 106 to respond to a communication ("availability"), a capacity of a user 106 to read communications and interact with friend users u ("capacity"), and a tendency of a user 106 to interact with particular friend users u ("tendency").

[00117] Block 1316 illustrates determining the degree of reciprocity. As described above, the responsiveness and reciprocity analysis module 232 may determine a degree of reciprocity. The degree of reciprocity is a value representing the frequency by which a friend user u sends a directed communication (e.g., reply communication, re- broadcast communication, etc.) in response to a communication initiated by an ego user vj. In at least one embodiment, the reciprocity analysis module quantifies reciprocity between an ego user u and a friend user u using a percentage rank (PR) and reciprocal interaction rank (RIR).

[00118] Block 1318 illustrates providing the ranking module 120 with data. As described above, data extracted during process by the intimacy module 118 may be provided to the ranking module 120 for ranking two or more users.

[00119] Block 1320 illustrates retrieving a ranking list. As described above, the ranking module 120 may rank which user of the two or more users is likely to respond to a particular communication from the first user based at least in part on the ranking. Additionally, the ranking module 120 may determine an order in which the two or more users 106 who are likely to respond are likely to respond. In addition to determining which users 106 are likely to respond and/or an order in which users 106 are likely to respond, the ranking module 120 may leverage the degree of intimacy to perform other predictions as described above (e.g., suggest one or more users 106 who may be interested in some subject matter and/or media object being viewed by a user 106, filter and/or order a displayed list of users 106 (e.g., contacts) for future connections or communications, and/or present icons on a displayed list of users 106 (e.g., contacts) reflective of the determined degree of intimacy, etc.).

Conclusion

[00120] Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are described as illustrative forms of implementing the claims.

[00121] Conditional language such as, among others, "can," "could," "might" or

"may," unless specifically stated otherwise, are understood within the context to present that certain embodiments include, while other embodiments do not necessarily include, certain features, elements and/or steps. Thus, such conditional language is not generally intended to imply that certain features, elements and/or steps are in any way required for one or more embodiments or that one or more embodiments necessarily include logic for deciding, with or without user input or prompting, whether certain features, elements and/or steps are included or are to be performed in any particular embodiment. Conjunctive language such as the phrase "at least one of X, Y or Z," unless specifically stated otherwise, is to be understood to present that an item, term, etc. may be either X, Y, or Z, or a combination thereof.

Claims

1. A method comprising:

determining a degree of reciprocity between a first user of a communication platform and a second user of the communication platform based at least in part on a frequency of directed communication from the second user to the first user in response to communication initiated by the first user;

determining the frequency of the directed communication in successive time periods;

performing an affinity calculation between the first user and the second user based at least in part on at least one of a profile affinity or a topic affinity;

analyzing contextual features including at least one of sentiment, self-disclosure, or responsiveness to identify sentiment values, self-disclosure values, or responsiveness values; and

calculating a degree of intimacy between the first user and the second user based at least in part on the degree of reciprocity, the affinity calculation, and at least one of the sentiment values, the self-disclosure values, or the responsiveness values derived from the communication initiated by the first user and directed communication.

2. A method as claim 1 recites, wherein the directed communication includes at least one of:

the second user replying to the communication initiated by the first user; or the second user broadcasting the communication initiated by the first user to a plurality of users of the communication platform.

3. A method as either claim 1 or 2 recites, wherein at least one of:

the profile affinity is based at least in part on gender, age, education, geographic presence, and common friends associated with the first user and the second user; and the topic affinity is based at least in part on topic distributions associated with the first user and the second user, the topic distributions based at least in part on topics identified by topic discovery and topical keyword extraction from the communication initiated by the first user and the directed communication from the second user to the first user;

the self-disclosure value is based on at least one of:

a number of words comprising the communication and the directed communication;

an occurrence of first person words in the communication and the directed communication; or

a geographic location associated with the communication and directed communication; or the responsiveness value is based on at least one of:

availability of the second user to respond to the communication; capacity of the second user to receive the communication; or tendency of the second user to interact with the first user.

4. A method as any one of claims 1 to 3 recites, further comprising:

mining emotions from the communication initiated by the first user and the directed communication from the second user to the first user to generate a list of words; identifying words associated with a first emotion and words associated with a second emotion opposite the first emotion in the list of words;

counting the words associated with the first emotion and the words associated with the second emotion; and

determining the sentiment value based at least in part on a number of words associated with the first emotion and a number of words associated with the second emotion.

5. A method as any one of claims 1 to 4 recites further comprising:

ranking two or more users of a plurality of users of the communication platform based at least in part on the degree of intimacy between the first user and the two or more users; and predicting which user of the two or more users is likely to respond to a particular communication from the first user based at least in part on the ranking.

6. A computer-readable medium having encoded thereon computer executable instructions to configure a computer to perform a method as any one of claims 1 to 5 recites.

7. A system comprising:

a computer-readable media storing a plurality of modules;

a processing unit operably coupled to the computer-readable media, the processing unit adapted to execute modules of the plurality of modules comprising:

a data crawling module configured for collecting data from communication initiated by a first user of a communication platform and from directed communication from a second user of the communication platform in response to the communication initiated by the first user;

an intimacy module configured for processing the data by applying one or more of a reciprocity model, a temporality model, an affinity model, or a contextuality model to the data; and

an output module configured for producing a graphical user interface to visualize user intimacy based at least in part on the processing of the data.

8. A system as claim 7 recites, wherein the intimacy module is further configured for at least one of:

applying a reciprocity model to determine a degree of reciprocity between the first user of the communication platform and the second user of the communication platform based at least in part on a frequency of directed communication from the second user to the first user in response to communication initiated by the first user;

applying an affinity model to perform an affinity calculation between the first user and the second user based at least in part on at least one of a profile affinity or a topic affinity, wherein:

the profile affinity is based at least in part on gender, age, education, geographic presence, and common friends associated with the first user and the second user; and

the topic affinity is based at least in part on topic distributions associated with the first user and the second user;

applying a contextuality model to analyze contextual features including at least one of sentiment, self-disclosure, or responsiveness to identify at least one of:

a sentiment value based at least in part on mining emotions and polarity from the communication initiated by the first user and the directed communication;

a self-disclosure value based at least in part on a number of words comprising the communication initiated by the first user and directed communication, an occurrence of first person words in the communication initiated by the first user and directed communication, and a geographic location associated with the communication and directed communication; or

a responsiveness value based at least in part on an availability of the second user to respond to the one or more communications, a capacity of the second user to receive the one or more directed communications, and a tendency of the second user to interact with the first user; or

applying a temporality model to determine the frequency of the data in successive time periods.

9. A system as claim 8 recites, wherein the output module is further configured to produce a graphical user interface represented by at least one of:

an overall network view wherein a plurality of nodes represent a plurality of users of the communication platform and the plurality of nodes are connected by edges representing connections between individual nodes of the plurality of nodes; or

a community view wherein an edge bundling graph utilizes an axis of time and an axis of tension to represent an interaction history of the communication and directed communication.

10. A system as any one of claims 7 to 9 recites, wherein the output module is further configured to produce a graphical user interface to visualize user dynamic intimacy.

11. A system as any one of claims 7 to 10 recites further comprising a ranking module configured for:

calculating a degree of intimacy between the first user and the second user based at least in part on processing of the data; and

ranking two or more users of a plurality of users of the communication platform based at least in part on the degree of intimacy between the first user and the two or more users.