WO2011148222A1

WO2011148222A1 - Spreading viral messages in social networks

Info

Publication number: WO2011148222A1
Application number: PCT/IB2010/001646
Authority: WO
Inventors: Samik Datta; Anirban Majumder; Nisheeth Shrivastava
Original assignee: Alcatel Lucent
Priority date: 2010-05-25
Filing date: 2010-05-25
Publication date: 2011-12-01

Abstract

Spreading Viral Messages in Social Networks. The present invention relates to social networks and, more particularly, to spreading messages through social networks. At least one node is selected from at least one social network wherein the node maximizes the influence of a seed set on other nodes in the social network on adding the node to the seed set. The node is added to the seed set and at least one viral message is sent to the nodes in the seed set. A maximum number of the nodes can be present in each seed set and there is a constraint on the maximum number of the viral messages received by each of the nodes.

Description

Spreading Viral Messages in Social Networks

TECHNICAL FIELD

[001] The present invention relates to social networks and, more particularly, to spreading messages through social networks.

BACKGROUND

[002] Word of mouth information exchange, also called as Viral Propagation, has been a successful vehicle for spreading messages. With the advent of digital communication and social networking sites, more and more of our social interactions has become through these social networking sites. Also, social interactions as digital traces of usage of the social networks have become available. Digital traces give us access to very detailed information about user interactions and social behavior that can be used in an online viral marketing campaign. The earliest works in viral marketing studied the cascading characteristics of word of mouth propagation, where the information cascades using the interactions of individuals. A user who knows the information tells it to his friends, who then tell it to their friends and so on. These cascades typically start from a few individuals who are the initial carriers (seeds) and over time cover portions of the social network far beyond the friend network of the initial carriers. Such cascades are commonplace phenomena, common in many different kinds of information spreads such as news, advertisements, product recommendations and health warnings. There are instances of viral marketing campaigns, such as the Nike commercial, that use social networking sites, such as orkut.com and facebook.com, to market products.

[003] Several researchers have explored social networks for designing algorithms for spreading messages by finding influential users and communities. One popular practice employed by many brands is to broadcast the same message to multiple users. However, the broadcast method would not scale with increasing number of messages, as the users will start considering the broadcast messages as spam and ignore the messages.

[004] Other methods use various measures of centrality in social networks to measure the influence of a node on other nodes. Once the centrality scores are calculated for every node in the graph, top scoring nodes are selected as seeds. Degree centrality of a node 'v' is proportional to the degree of the node. The highest degree nodes are chosen with the assumption that since the nodes have many friends the nodes can influence many as well. But, since the node is in active contact with only a small fraction of the neighboring nodes, often the top scoring nodes are much less influential than what the scores indicate. Betweenness centrality of a node 'v' is the fraction of all shortest paths from node V to node 't', for all possible choices of s and t, that pass through the node 'v' . If the score is high, then many nodes communicate through the node. However, studies indicate that communication does not necessarily occur via the shortest path, so often the betweenness centrality leads to erroneous choices. Closeness centrality of a node 'v' is inversely proportional to the average of distances of vertices reachable from node 'v'. For a connected graph, higher the score, faster is the possibility of message spread. However, communication may not always occur via the shortest path therefore closeness centrality is not a good metric.

[005] Another approach uses an Eigen vector centrality score of a node 'v', which is proportional to the sum of the scores of neighbors of node 'v'. The score captures the authority of the node V. However most of the links in social networks are rarely used for exchanging information and therefore, the Eigen vector centrality measure may not capture the essence of influences in social networks. Also, current algorithms used to determine seeds do not generalize to situations wherein there are multiple products to be marketed simultaneously.

SUMMARY

[006] In view of the foregoing, an embodiment herein provides a method for sending viral messages in at least one social network. A set of seed nodes of size at least one is selected from a social network wherein the set maximizes the influence on other nodes in the social network. The seed sets are determined for multiple products and for each of them viral messages are sent to all the nodes in the corresponding seed set. A maximum number of the nodes can be present in each seed set and there is a constraint on the maximum number of the viral messages received by each of the nodes. The influence of a node on other nodes decreases with the increase in the number of the viral messages received by that node. The seed set is determined using at least one of an Independent Cascade Model or a Linear Threshold Model. The seed set is first determined for one of said viral messages having a maximum influence, wherein said influence is determined based on said nodes in said seed set of said one viral message.

[007] Embodiments further disclose an analysis module for tracking the spread of viral messages in at least one social network. The analysis module analyzes statistics involved when the viral messages spread across the nodes in the social network, wherein the viral message propagations are initialized through the nodes in a seed set; further the seed set maximizes influence on other nodes in the social network. The statistics include at least one of how each of the node received a viral message, number of the viral messages received by the node, number of the viral messages required to influence the node, probability of one of the nodes influencing a neighboring node and popularity of the viral messages. The analysis module sends the statistics to a Seed Optimizer.

[008] Embodiments herein also disclose a Seed Optimizer module for optimizing seed selection in a social network. The Seed Optimizer module selects a set of seed nodes of size at least one from the social network wherein the set maximizes the influence on other nodes in the social network. The Seed Optimizer module performs the selection after receiving statistical information from an Analysis module. The statistical information includes information about the spread of the viral messages across the nodes in the social network.

[009] These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings.

BRIEF DESCRIPTION OF THE FIGURES

[0010] The embodiments herein will be better understood from the following detailed description with reference to the drawings, in which:

[0011] FIG. 1 illustrates a block diagram of a seed optimizer connected to social networks, according to an embodiment herein;

[0012] FIGS. 2a and 2b are flowcharts illustrating selection of seeds using the Greedy algorithm with hard thresholds, according to an embodiment herein; and

[0013] FIGS. 3a and 3b are flowcharts illustrating selection of seeds using the Fair-Greedy algorithm with hard thresholds, according to an embodiment herein. DETAILED DESCRIPTION OF EMBODIMENTS

[0014] The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein may be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.

[0015] The embodiments herein disclose a system and method for optimizing seed selection for spreading viral messages in a social network. Referring now to the drawings, and more particularly to FIGS. 1 through 7, where similar reference characters denote corresponding features consistently throughout the figures, there are shown embo diments .

[0016] In a social network, the interactions between users of the social network play an important role in spreading information. When the user of the social network receives a message, the user has the option of relaying the message to his friends. For example, a user may receive an advertisement about a product and the user may decide to send the advertisement to his friends in the social network. If any of his friends decide to use the product, then the original user can be said to have an influence on his friends in the social network. His friends may decide to pass on the advertisement to their friends and so on in a cascading manner. The initial carriers of information in the social network are called as seeds and it is imperative to select the seeds that can influence maximum number of users in the social network. For example, the adoption of a new cell phone model by college students. At the onset, students would be reluctant to switch to a new product. However, if a few students switch to the new product initially, then with time other students would also be influenced by those few students and would eventually switch to the new product. Thus, after some time a significant number of students would have switched to the new product. In this example, the first few students who switch to the new product initially are the seeds. The example shows the influence of seeds on other users in a social network. Any message posted in a social networking site may be viewed by some users and circulated to other users through the social network using e-mail, blog-post, by word of mouth or by any other information transfer means. Thus, it is prudent to select seeds having optimum influence, in circulating messages in the social network, when looking to spread viral messages in the social network.

[0017] FIG. 1 illustrates a block diagram of a seed optimizer connected to social networks. A social network is a network made of users 107 who can interact with each other to send information/messages or simply to communicate with each other. For example, Facebook is a social networking site using which users 107 can create a profile, edit the profile, add/delete friends or send messages to his friends. There are different social networking sites using which users 107 can interact with each other and there may be different applications running on the different social networking sites. For example, Orkut and Twitter are two different social networking sites. A social networks 106 module is a set of all social networking sites. A user 107 of one social networking site can interact with a second user 107 of another social networking site. An application components module 105 provides a library of features in the social network that can be used by the user 107. For example, the feature may be a poke feature wherein user 1 107 flies a virtual paper plane to user 3 107. The feature may also be such that, user 4 107 can send virtual gifts to user 5 107. The features provided by the application components module 105 help provide fun and entertainment to users 107 who can choose to use the features at their leisure. The application components module 105 is a library that can be used for writing applications. Users 107 benefit from the application components module 105 but users 107 may not use the application components module 105 directly.

[0018] In a social network, the interactions between the users 107 of the social network play an important role in the spreading of information. When a seed in the social network receives a message, the seed has the option of relaying the message to his friends in the social network. It is important to select the seed that can influence maximum number of users 107 in the social network. A seed optimizer 103 module selects the seeds for sending the messages. The seed optimizer 103 module has a global knowledge of the graph topology of the social network site. The seed optimizer 103 module runs an algorithm and determines the optimum seeds. For example, the seed optimizer 103 module may run Greedy algorithm or Fair-Greedy algorithm to determine the optimum seeds and the algorithm may be run by considering the propagation model as an Independent Cascade Model or as a Linear Threshold Model. Considering the network topology as a graph G(V, E) where the nodes, V, in the graph represent identities of users 107 and the edges, E, represent the interactions between the users 107. In the Independent Cascade Model probabilities, p_u,_v <= 1 , are given for each edge (u,v) in E and it is assumed that any user 107, u, has a probability p_u,v of activating a neighbor v. A node is said to be activated when the node is influenced by the message or by another node. For example, the message may be received from a friend of the user 107 or from anyone who initiates sending of messages in the network. The influence among users 107 is propagated in a random manner. After becoming active, each node, u, gets one chance of activating a neighbor, v, with a probability p_u when v is not already activated by any node. In the Linear Threshold Model a node, v, gets an influence of W_u by its neighbor u such that (∑w_{u v}≤ \) . Each node, v, specifies a pre-determined threshold Q_v chosen from u the interval [0, 1]. A node gets activated when the total influence from the node's active neighbors is at least Q_v . That is, _{u :} ¾_ve ^{Λ u,v} .

[0019] An analysis module 102 keeps track of how the message is spreading in the social network. For example, the analysis module 102 may keep track of how a user 107 received the message, to how many users did the user 107 send the message or how many messages a user 107 viewed before sending the message to his friends. The analysis module 102 calculates p_u,v and sends the calculated values of /¾_,v to the seed optimizer module 103. The analysis module 102 also keeps track of the popularity of the features offered by the message. For example, a marketing company specializing in designing and running viral marketing campaigns designs a viral advertisement, selects seeds and sends the viral advertisement to the selected seeds. Then analytics of progress and popularity of the viral advertisement are collected and analyzed. The success of the viral advertisement depends on selecting the right set of seeds since the seeds help spread the viral advertisement through the social network. In the example, the seed optimizer 103 module selects the seeds and the analysis module 102 collects and analyzes the progress and popularity of the viral advertisement. A user channel 104 enables users 107 to send messages to his friends in the network. The user channel 104 provides an interface for users 107 to read and write messages. For example, the user channel 104 may help users 107 read e-mail messages. If a user 107 sends it to seeds in order to virally propagate the message, then the user 107 can monitor the effectiveness of the created message using a Social Marketing Interface 101. The message sent by the user may be a message created by him. The message sent by the user may also be a message created by the viral marketing campaigner. The Social Marketing Interface 101 shows demographics of users 107 currently viewing or sending the message. The Social Marketing Interface 101 also shows other statistics and metrics related to the created message. For example, the Social Marketing Interface 101 may show the number of users 107 influenced by the message.

[0020] FIGS. 2a and 2b are flowcharts illustrating selection of seeds using the Greedy algorithm with hard thresholds. A social network is a network made of users 107 who can interact with each other to send information/messages or simply to communicate with each other. In a social network the interactions with the users 107 of the social network play an important role in the spreading of information. When a seed in the social network receives a message, the seed has the option of relaying the message to his friends in the social network. It is important to select the seed that can influence maximum number of users in the social network. A seed optimizer 103 module selects the seeds for sending the messages using the Greedy algorithm. To run the algorithm, the social network topology may be considered as a graph G(V, E) where the nodes, V, in the graph represent identities of users and the edges, E, represent the interactions between the users. When multiple messages pertaining to multiple products have to be sent to seeds, too many messages should not be sent since the seeds might start considering the messages as spam messages. The willingness of a seed, u, to relay the message to his friends in the social network can be expressed as a response function f_(UiV)(m_u ) , where (u, v) is the edge on which the influence propagates and f_{( v)}(m_u ) is the probability with which 'u' influences V and 'ntu is the number of products for which 'u' has been selected as a seed and hence received messages. Thus, if 'u' is picked as a seed for more and more products, then the influence of 'u' on V will gradually decrease.

[0021] The Greedy algorithm may be run using a Hard threshold model. In the Hard threshold model 'u' may be selected as a seed for S_u number of products.

Therefore, f_(u m_u ) = p_{u v} ifm_u < S_u , otherwise, f_(u m_u ) = 0 . For example, if there are two nodes V and V, /¾_jV = 0.7 and S_u = 3, then when 'u' receives the first viral message f_(uv)(l ) = 0.7. On receiving the second and third viral messages, f_(U:V 2) =

0.7 and f_(uv)(3 ) = 0.7. However, if 'u' receives a fourth viral message /_(U;V) (4) = 0 since 'u' may not be selected for more than for 5_U number of products. Given the social network and the model of influence propagation through the social network to run the Greedy algorithm and find the seeds, first the information about the different products, P_y , is obtained (201) where 't' is the number of products. Then the vector of seed requirements, k_t ... _t , is obtained (202), where ' ¾' is the target number of seeds for product The aim of the Greedy algorithm is to select the seed sets S = S_t ... _t , where 'Si' is the seed set for and the product constraint is obtained

(203) as I S_i |≤ k_; , Vz e {1,2,...,t) . Further, <5_v is denoted as the number of products for which V can be chosen as a seed. The node constraints can then be obtained (203) using δ_ν as I { : v e S_t ) |≤ δ_ν,ν e V . I(S) is defined as the expected number of nodes being influenced by choosing 'S' as the seed sets. The influence propagation may be independent for each product. The total number of nodes influenced t

is I(S) = ^/(5^* _; ) . The Greedy algorithm chooses 'S' such that the overall influence is i=l maximized. [0022] If there are more products (204), then for all the products to be considered, the node constraints and product constraints for the new products are obtained. At each step, the Greedy algorithm picks (205) a node V without violating the node constraint and the seed set 'Si'. The Greedy algorithm further picks (206) a product and checks (207) if the addition of node V to the seed set 'Si' gives the maximum increase, I(v,S_; ) , to the overall influence, I(S). If the addition of V to 'Si' gives the maximum increase to I(S), then node V is added (208) to the seed set if node 'v' does not violate the node constraint. /(v,5^*.) is the increase in influence after adding node V to the seed set. The Greedy algorithm checks (209) to see if all the products have been assigned the required number of seeds. The Greedy algorithm is run until all the products have been assigned the required number of seeds or there are no nodes left to be assigned. The various actions in method 200 may be performed in the order presented, in a different order or simultaneously. Further, in some embodiments, some actions listed in FIG. 2 may be omitted.

[0023] FIGS. 3a and 3b are flowcharts illustrating selection of seeds using the

Fair-Greedy algorithm with hard thresholds. A social network is a network made of users 107 who can interact with each other to send information/messages or simply to communicate with each other. In a social network the interactions with the users 107 of the social network play an important role in the spreading of information. When a seed in the social network receives a message, the seed has the option of relaying the message to his friends in the social network. A seed optimizer 103 module selects the seeds for sending the messages using the Fair-Greedy algorithm. To run the algorithm the social network topology may be considered as a graph G(V, E) where the nodes, V, in the graph represent identities of users and the edges, E, represent the interactions between the users. When multiple messages pertaining to multiple products have to be sent to seeds, too many messages should not be sent to the same seed since the seeds might start considering the messages as spam messages. The willingness of a seed, u, to relay the message to his friends in the social network can be expressed as a response function f_{(u v)} (m_u ) , where (u, v) is the edge on which the influence propagates and /_(M;V) («_M ) is the probability with which 'u' influences V and 'm_u' is the number of products for which 'u' has been selected as a seed and hence received messages. Thus, if 'u' is picked as a seed for more and more products, then the influence of 'u' on V will gradually decrease.

[0024] The Fair-Greedy algorithm can be run using a Hard threshold model. In the Hard threshold model 'u' can be selected as a seed for S_u number of products.

Therefore, _(M;V) (mJ = p_u>v if m_u < S_u , otherwise, f_{(u v)} (m_u ) = 0 . Given the social network and the model of influence propagation through the social network to run the Fair-Greedy algorithm and find the seeds, first the information about the different products, P_1;...;t , is obtained (301) where 't' is the number of products. Then the vector of seed requirements, k_{t ... t} , is obtained (302), where is the target number of seeds for product The aim of the Fair-Greedy algorithm is to select the seed sets S = S_{t ... t} , where 'Si' is the seed set for and the product constraint is obtained

(303) as I S_i |≤ k_; , Vz e {1,2,...,t) . Further, <5_v is denoted as the number of products for which V can be chosen as a seed. The node constraints can then be obtained (303) using δ_ν as | {i : v e £. } |< δ_ν , ν e V . I(S) is defined as the expected number of nodes being influenced by choosing 'S ' as the seed set. The influence propagation may be independent for each product. The total number of nodes influenced t

is I(S) = ^ /(5^* _; ) . The Fair-Greedy algorithm seeks to allocate seeds fairly across all i=l products having the same seed requirement. If products and 'p require the same number of seeds, then the Fair-Greedy algorithm seeks to allocate seeds to the

I(S ) products so as to make the ratio of the influences of the two products, — , close to

1.0.

[0025] For all the products to be considered (304), the node constraints and product constraints for the products are obtained. At each step, the Fair-Greedy algorithm picks (305, 306) the product, p_i: with the minimum influence and a seed and checks (307) if that seed, v, increases the influence of pi by the maximum value, then v is added (308) to the seed set if it does not violate the node constraint. The Fair- Greedy algorithm checks (309) to see if all the products have been assigned the required number of seeds. The Fair-Greedy algorithm is run until all the products have been assigned the required number of seeds or there are no nodes left to be assigned. The various actions in method 300 may be performed in the order presented, in a different order or simultaneously. Further, in some embodiments, some actions listed in FIG. 3 may be omitted.

[0026] The embodiments disclosed herein allow messages to be targeted to users satisfying specific requirements and demographics. For example, the messages may be targeted towards students or towards software professionals. In a second example, viral message regarding sports wear may be targeted towards sportsmen. In other embodiments, products may have influence propagation probabilities, on each edge, different from other products. Then, the seed selection would be performed by using an influence vector, p defining the probability of the influence of product

P. propagated on edge (u, v). The influence computation for different products is independent and thus the influence computation may be performed with each product having different influence probabilities.

[0027] The embodiments disclosed herein can be implemented through at least one software program running on at least one hardware device and performing network management functions to control the network elements. The network elements shown in Fig. 1 include blocks which can be at least one of a hardware device, or a combination of hardware device and software module.

[0028] The embodiment disclosed herein specifies a system for optimizing seed selection and sending viral messages to the selected seeds. Therefore, it is understood that the scope of the protection is extended to such a program and in addition to a computer readable means having a message therein, such computer readable storage means contain program code means for implementation of one or more steps of the method, when the program runs on a server or mobile device or any suitable programmable device. The method is implemented in a preferred embodiment through or together with a software program written in e.g. Very high speed integrated circuit Hardware Description Language (VHDL) another coding language, or implemented by one or more VHDL or several software modules being executed on at least one hardware device. The hardware device can be any kind of device which can be programmed including e.g. any kind of computer like a server or a personal computer, or the like, or any combination thereof, e.g. one processor and two FPGAs. The device may also include means which could be e.g. hardware means like e.g. an ASIC, or a combination of hardware and software means, e.g. an ASIC and an FPGA, or at least one microprocessor and at least one memory with software modules located therein. The embodiments described herein could be implemented in pure hardware or partly in hardware and partly in software. Alternatively, the embodiments may be implemented on different hardware devices, e.g. using a plurality of CPUs.

[0029] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the claims as described herein.

Claims

CLAIMS What is claimed is:

1. A method for sending viral messages in at least one social network, the method comprising steps of: selecting at least one node from said at least one social network wherein said node maximizes influence of a seed set on other nodes in said social network on adding said node to said seed set;

adding said node to said seed set; and

sending at least one viral message to said nodes in said seed set.

2. The method, as claimed in claim 1, wherein said seed set is determined for multiple said viral messages.

3. The method, as claimed in claim 1, wherein a maximum number of said nodes can be present in said seed set.

4. The method, as claimed in claim 1, wherein said node has a constraint on a maximum number of said viral messages received by each of said node.

5. The method, as claimed in claim 1, wherein influence of said node decreases with the increase in number of said viral messages received by said node.

6. The method, as claimed in claim 1, wherein said seed set is determined using at least one of: an Independent Cascade Model; or

a Linear Threshold Model.

7. The method, as claimed in claim 1, wherein said seed set is first determined for one of said viral messages having a minimum influence, wherein said influence is determined based on said nodes in said seed set of said one viral message.

8. A analysis module (102) for tracking the spread of viral messages in at least one social network, said module (102) having at least one means adapted for: analyzing statistics involved when said viral messages spread across said nodes in said social network, wherein said viral messages are spread through nodes in a seed set, further said seed set maximizes influence on other nodes in said social network.

9. The analysis module (102), as claimed in claim 8, wherein said statistics include at least one of: how each of said node received a said viral message;

number of said viral messages received by said node;

number of said viral messages required to influence said node;

probability of one of said nodes influencing a neighboring said node; and popularity of said viral messages.

10. The analysis module (102), as claimed in claim 8, wherein said analysis module (102) sends said statistics to a Seed Optimizer (103) Module.

11. A Seed Optimizer module (103) for optimizing seed selection in a social network, said Seed Optimizer module (103) having at least one means adapted for: selecting at least one node from said at least one social network wherein said node maximizes influence of a seed set on other nodes in said social network on adding said node to said seed set.

12. The Seed Optimizer module (103), as claimed in claim 11, wherein said Seed Optimizer module (103) is configured to perform said selection after receiving statistical information from an Analysis module (102).

13. The Seed Optimizer module (103), as claimed in claim 12, wherein said statistical information includes information about spread of said viral messages across said nodes in said social network.