Method, device and system for identifying key users in a derived topic network

Abstract

The application belongs to the field of social network information processing, and particularly relates to a key user identification method, device and system in a derivative topic network; the method comprises the following steps: obtaining data information in a topic network; extracting a native topic, a derivative topic user relationship network, a native topic and derivative topic propagation network, user basic attributes and user intimacy from the data information in the topic network; extracting a user feature vector representation through a community label propagation algorithm according to the user basic attributes and user intimacy; extracting a topic vector representation of the native topic and the derivative topic through a joint distribution adaptive algorithm according to the native topic and the derivative topic propagation network; inputting the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and outputting an opinion leader user or a long tail user. The application can accurately identify key users in a derivative topic network.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of Internet application, and relates to data analysis in a social network, in particular to a key user identification method, device and system in a derivative topic network. BACKGROUND

[0002] The rise and popularity of social media have enabled a large number of users to freely express their opinions, share information, and participate in extensive discussions and interactions. This interactive behavior has built a vast user network in social networks, in which the influence of users plays a crucial role in information dissemination. However, the rapid development of social networks has also brought a series of problems. In particular, in some special events, a series of new topics derived from them have aroused intense discussions among netizens, with fast information dissemination speed and wide range, posing challenges to information management in social networks, and also potentially triggering widespread online public opinion with a very large influence range and transmission speed.

[0003] Derivative topic user influence analysis has important influence in multiple fields. On social network platforms, derivative topic user influence analysis can reveal the degree of user participation in topics, interaction frequency, and information dissemination paths, helping platforms better understand user needs and preferences, identify potential target customer groups, understand their needs and preferences, optimize recommendation algorithms, and thus develop more accurate recommendation strategies to improve user experience.

[0004] Currently, researchers' studies on derivative topic node influence analysis mainly fall into three categories. The first is based on social network structure, focusing on the relationships and connection structures between nodes in social networks, as well as community structures and information dissemination paths in social networks. Key nodes in social networks are identified through the above indicators and methods. The second is based on user behavior, which mainly analyzes a series of dissemination behaviors and activities such as likes, comments, and reposts of users on topics in social networks. By analyzing these behavior data, user influence scores are calculated to identify users who have made significant contributions to information dissemination and social influence in social networks. The third is based on topic information, which mainly analyzes text sentiment and distinguishes user emotional tendencies to analyze user influence. The above technologies fundamentally analyze node influence from different angles based on existing core content of social networks.

[0005] With the continuous expansion of social network data and the gradual maturity of deep learning technology, neural networks, deep learning, and other nonlinear models have gradually become the mainstream choice for user influence analysis models. However, these existing technologies cannot accurately depict the dissemination scale of original topics and derivative topics, as well as the cognitive differences of users on original topics and derivative topics, making it difficult to discover key users in the early stage of derivative topic dissemination. SUMMARY

[0006] In view of the problems of the prior art, the present application provides a key user identification method, device and system in a derivative topic network, which realizes early user influence analysis of a derivative topic based on propagation scale and cognitive difference.

[0007] In a first aspect, the present application provides a key user identification method in a derivative topic network, comprising:

[0008] Obtaining data information in a topic network; the data information comprises user basic information, topic basic information and user relationship network;

[0009] Extracting a primary topic, a derivative topic user relationship network, a primary topic and derivative topic propagation network, user basic attributes and user intimacy from the data information in the topic network;

[0010] According to the user basic attributes and the user intimacy, a user feature vector representation is extracted by a community label propagation algorithm;

[0011] According to the primary topic and derivative topic propagation network, a topic vector representation of the primary topic and the derivative topic is extracted by a joint distribution adaptive algorithm;

[0012] The user feature vector representation and the topic vector representation are input into a user influence dynamic space-time model based on an attention mechanism, and an opinion leader user or a long tail user is output.

[0013] In a second aspect, the present application further provides a key user identification device in a derivative topic network, comprising:

[0014] A data acquisition module is configured to obtain data information in a topic network; the data information comprises user basic information, topic basic information and user relationship network;

[0015] A data extraction module is configured to extract a primary topic, a derivative topic user relationship network, a primary topic and derivative topic propagation network, user basic attributes and user intimacy from the data information in the topic network;

[0016] A user feature extraction module is configured to extract a user feature vector representation according to the user basic attributes and the user intimacy by a community label propagation algorithm;

[0017] The topic feature extraction module is configured to extract topic vector representations of the original topic and the derivative topic according to the original topic and the derivative topic propagation network through a joint distribution adaptive algorithm.

[0018] The key user identification module is configured to input the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and output opinion leader users or long tail users.

[0019] In a third aspect of the present application, the present application further provides a key user identification system in a derivative topic network, comprising:

[0020] The transceiving module is configured to obtain data information in a topic network, wherein the data information comprises user basic information, topic basic information and a user relationship network.

[0021] The processing module is configured to extract an original topic, a derivative topic user relationship network, an original topic and a derivative topic propagation network, user basic attributes and user intimacy from the data information in the topic network, extract a user feature vector representation from the user basic attributes and the user intimacy through a community label propagation algorithm, extract topic vector representations of the original topic and the derivative topic from the original topic and the derivative topic propagation network through a joint distribution adaptive algorithm, and input the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and output opinion leader users or long tail users.

[0022] The display module is configured to display the opinion leader users or the long tail users output by the processing module.

[0023] The present application has the following beneficial effects:

[0024] The key user identification method, device and system in the derivative topic network described above are configured to obtain data information in a topic network, extract an original topic, a derivative topic user relationship network, an original topic and a derivative topic propagation network, user basic attributes and user intimacy, extract a user feature vector representation from the user basic attributes and the user intimacy through a community label propagation algorithm, extract topic vector representations of the original topic and the derivative topic from the original topic and the derivative topic propagation network through a joint distribution adaptive algorithm, and input the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and output opinion leader users or long tail users, thereby improving the identification accuracy of key users in a derivative topic network in a social network platform. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 FIG. 1 is a key user identification scene schematic diagram of an embodiment of the present application.

[0026] Figure 2 is a key user identification method flowchart of an embodiment of the present application;

[0027] Figure 3 is a user implicit influence diagram based on user intimacy of an embodiment of the present application;

[0028] Figure 4 is a topic vector representation method based on emotion of an embodiment of the present application;

[0029] Figure 5 is a user influence dynamic space-time model diagram of an embodiment of the present application;

[0030] Figure 6 is a key user identification device structure diagram of an embodiment of the present application;

[0031] Figure 7 is a key user identification system structure diagram of an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to better illustrate the technical solutions of the present application and make the advantages more concise and clear, the problems to be solved by the present application will be specifically explained first, and then the specific embodiments of the present application will be further described in detail with reference to the accompanying drawings.

[0033] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0034] The terms "first", "second", etc. in the description, claims, and drawings of the application, and the above, are used to distinguish like objects, and are not necessarily used to describe a particular sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances such that the embodiments described herein are capable of operating in other sequences than illustrated or described herein. Moreover, the terms "comprise", "have" and any variations thereof are intended to cover a non-exclusive inclusion, for example, a process, method, article, or apparatus that comprises a list of steps or modules is not necessarily limited to those specifically listed, but can include additional steps or modules not expressly listed or inherent to such process, method, article, or apparatus. The division of modules appearing herein is merely a logical division, and in actual application, other division manners can be implemented, for example, a plurality of modules can be combined or integrated into another system, or some features can be omitted or not executed, in addition, the coupling or direct coupling or communication connection between the modules shown or discussed can be through some interfaces, and the indirect coupling or communication connection between the modules can be electrical or other similar forms, which are not limited herein. In addition, the modules or sub-modules described as separate components can or can not be physically separated, and can or can not be physical modules, or can be distributed into a plurality of circuit modules, and part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments of the application.

[0035] In view of the current status of user influence research, there are still several challenges in identifying key users in the derivative topic network using original topic related information:

[0036] 1. It is difficult to discover the hidden influence of key users in the early stage of derivative topic outbreak. At present, it is difficult to discover key users with influence due to the reasons such as new theme, small user participation and influence, and unclear views and attitudes of participating users.

[0037] 2. Cognitive differences caused by user domain migration. User domain migration leads to cognitive differences in the knowledge domain of the original topic domain and the derivative topic domain, which further affects the user's attitude and behavior towards the topic and also has a certain impact on the user's influence.

[0038] 3. The influence of derivative topic users is in dynamic change. The key users of the original topic may lack relevant professional knowledge in the derivative topic domain, thus changing from key users to long-tail users; at the same time, new key users with greater influence may emerge.

[0039] Based on the above technical problems, Figure 1 The application environment diagram for the key user identification method in one embodiment. Please refer to Figure 1The key user identification method is applied in the terminal 101 and the server 102. The terminal 101 and the server 102 are connected through a network. The terminal 101 sends data information in a topic network to the server 102; and receives a result that a user is an opinion leader user or a long tail user from the server 102; wherein the server 102 obtains the data information in the topic network from the terminal 101; the data information includes user basic information, topic basic information and a user relationship network; the original topic, the derivative topic user relationship network, the original topic, the derivative topic propagation network, the user basic attribute and the user intimacy between users are extracted from the data information in the topic network; the user feature vector representation is extracted by a community label propagation algorithm according to the user basic attribute and the user intimacy between users; the topic vector representation of the original topic and the derivative topic is extracted by a joint distribution adaptive algorithm according to the original topic and the derivative topic propagation network; the terminal 110 inputs the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and outputs an opinion leader user or a long tail user. It can be understood that the user identification method in the derivative topic network provided by the embodiment of the application can also be not limited to be applied to a system including a terminal and a server, and is realized through the interaction of the terminal and the server. The terminal 101 can include but is not limited to various personal computers, notebook computers, tablet computers and the like. The server 102 can be realized by an independent server or a server cluster composed of multiple servers.

[0040] As shown in Figure 2 In some embodiments, the application provides a user identification method in a derivative topic network, the method comprising:

[0041] 201, obtaining data information in a topic network; the data information includes user basic information, topic basic information and a user relationship network;

[0042] In the embodiment of the application, the data is obtained by directly querying real-time social network data in a database provided by an enterprise. The database stores active data, social data and the like generated by all users in various social network platforms. The user can be a user who has logged in a social network platform. Here, the basic information (such as age, region, number of fans, number of friends and the like) of the user, the interaction of the original topic and the derivative topic (such as likes, collections and forwarding) and the historical behavior data of the user who has the interaction are needed to be obtained. The basic information of the interactive user and the basic information of the topic are needed to be obtained for the participation of the original topic and the derivative topic. The historical behavior of the original topic and the derivative topic includes the original topic and the derivative topic related information interacted by the user in the past. And the data is preprocessed.

[0043] 202、extracting the original topic, the derived topic, the user relationship network, the original topic propagation network, the derived topic propagation network, the user basic attribute and the user intimacy from the data information in the topic network;

[0044] In the social network, the influence of a user is affected by multiple factors, including social relationship, interest, historical behavior, etc. Based on this, the user-related attributes are extracted:

[0045] In some embodiments, the single message propagation network is composed of all user sets participating in the topic propagation and all propagation path sets of the users in the topic propagation process; which can be expressed as:

[0046]

[0047] Formula (1) is the single message propagation network. represents all user sets participating in T i topic propagation, represents all propagation path sets of the users in T i topic propagation process, both of which constitute T i topic propagation network

[0048] In some embodiments, the original topic propagation network N OT and the derived topic propagation network N DT are respectively composed of user sets participating in the original topic propagation and the derived topic propagation and the total number of messages under the corresponding topic propagation network; which can be expressed as:

[0049]

[0050] Formula (2) represents the original topic propagation network. Formula (3) represents the derived topic propagation network. is the i-th user in the original topic propagation network, is the single message propagation network, |S| is the total number of messages under the original topic propagation network, and |M| is the total number of messages under the derived topic propagation network.

[0051] In some embodiments, the original topic user relationship network N OR and the derived topic user relationship network N DR are respectively composed of user sets participating in the original topic propagation and the derived topic propagation and user relationship sets, which can be expressed as:

[0052]

[0053] Formula (4) represents the original topic user relationship network. Common user relationship networks include like, comment, forward and mention four networks, and the network type is represented by R, represents the i-th user participating in the original topic, represents that the i-th user and the j-th user in the original topic network have R types of relationships. Formula (5) represents the derived topic user relationship network.

[0054] In some embodiments, the user basic attribute can be constituted by the number of fans, the number of friends, the authentication field such as the entertainment field, the authentication community such as the star super community, and the emotional characteristics of the user. The user basic attribute Att(u i ) is represented as:

[0055] Att(u i ) = [fansNum(u i ), friNum(u i ), field(u i ), cerCom(u i ), emo(u i )] (6)

[0056] Formula (6) represents the user basic attribute. fansNum(u i ) represents the number of fans of the user u i , friNum(u i ) represents the number of friends of the user u i , field(u i ) represents the authentication field of the user u i , which can be empty, cerCom(u i ) represents the authentication community of the user u i , which can have multiple values, and emo(u i ) represents the emotional characteristics of the user u i , which can be the result of emotional analysis based on the social behavior, speech content, etc. of the user.

[0057] In some embodiments, the user intimacy is determined by the similarity score, the difference score of the same attribute between users, and the weighted score of the same interaction mode between users. The user intimacy I(u i ,u j ) is represented as:

[0058]

[0059] wherein I(u i ,u j ) represents the intimacy between the user u i and the user u j , w m represents the weight of the attribute m, and S(u i ,u j ,m) is represented as the useri and user u j The similarity score on attribute m, P(u i ,m)-P(u j (m) represents user u i and user u j Regarding the difference in attribute m, w n B(u) represents the weight of the nth interaction method. i ,u j (n) represents user u i and user u j The score for the nth interaction method.

[0060] 203. Based on users' basic attributes and the intimacy between users, extract user feature vector representations using a community tag propagation algorithm;

[0061] In this embodiment of the invention, the strong correlation between original topics and derived topics is utilized. Starting from the original topic, the characteristics of the community where the node is located are obtained by designing user intimacy and introducing a community tag propagation algorithm.

[0062] In some embodiments, the extraction of the user feature vector representation includes processing the intimacy between users and the frequency of tag occurrences using a speaker listener-based propagation algorithm, calculating the acceptance probability of each tag; calculating the probability that a user belongs to a tag based on the acceptance probability of each tag; obtaining the node tag set, the number of connected communities, and the community size based on the probability that a user belongs to a tag, and generating the user feature vector representation through an embedding model.

[0063] like Figure 3 As shown, assuming the derivative topic user relationship network contains 14 users, this embodiment uses a community tag propagation algorithm to divide the network into communities based on the user relationship network, obtaining the community to which each user belongs and its characteristics. It can be seen that users 2, 5, and 11 form one user community; users 1, 4, 6, 7, and 12 form another; and users 3, 8, 9, 10, 12, 13, and 14 form yet another. These community characteristics are used as the basic attributes of the users to generate preliminary user feature vector representations. Then, a two-layer graph convolutional network is used to further update and optimize these features, thereby obtaining more accurate user feature vector representations. Specifically, this embodiment uses the intimacy between users as the basis for selecting labels for neighboring nodes, calculating the acceptance probability of each label based on the intimacy between users and the frequency of label occurrence. Assuming u i As the initial node, u j As the speaker, I(u) i ,u jrepresents the closeness between user u i and user u j . Then the probability of each label received is:

[0064]

[0065] where l is a label in the label list, counts(l) is the frequency of the label, and sum(l) is the total number of labels.

[0066] In the propagation algorithm based on the speaker listener, each user node obtains labels from its neighbor nodes, and based on the probability of each label received, the label with the highest probability is selected for acceptance and propagated to the neighbor nodes. After multiple rounds of label propagation and update, each user node has a final label set; the number of connected communities and the community size are obtained; and the user feature vector representation is generated through the embedding model according to the user label set, the number of connected communities, and the community size. Each label set represents one or more communities to which the node belongs. The number of communities connected by the node is the number of different labels to which the node finally belongs. Assuming that the label set of node u i is L(u i ), the number of communities connected by the node is the number of labels in the label set. The number of communities connected by user node u i is:

[0067] Num cerCom (u i )=|L(u i )| (9)

[0068] The size of the community in which the node is located refers to the number of all other nodes in the community to which the node belongs. The label set L(u i ) output by SLPA can determine the community to which a node belongs. If node u i belongs to label l, then the community corresponding to this label is the community composed of all nodes marked as l. The community size is the number of nodes in this community. Let N l be the node set of label l, then:

[0069] Num size (u i ,l)=|N l | (10)

[0070] where N l is the node set of all labels l, and |N l | represents the size of the community, i.e., the number of nodes in the community corresponding to label l.

[0071] The number of connections between nodes is referred to as the degree of node u i , that is, the degree of node u i . The degree of node u i , deg(u i ), represents the number of connections between node u i and its neighbor nodes. That is:

[0072] Num degree (u i ) = deg(u i ) = |N(u i )| (11)

[0073] where N(u i ) represents the neighbor node set of node u i , and |N(u i )| represents the number of neighbor nodes connected to node u i .

[0074] The embodiment of the present application adopts modularity as a community division evaluation index, and the greater the modularity value is, the better the division quality is. Modularity is a quantitative index and is not affected by subjective judgment, and can objectively reflect the pros and cons of community division. The number of connected communities and the size of the community are optimized to form the basic attributes of the user, so as to further optimize the user feature vector representation; and the user feature vector representation obtained finally is more accurate.

[0075] Furthermore, the embodiment of the present application adopts the number of communities connected by the node, the size of the community, and the number of connections between nodes as the user's own attributes, according to formulas (9)-(11), the single user feature vector M(u i ) = {Num cerCom (u i ), Num size (u i , l), Num degree (u i )}, and the community user feature vector representation is X = N x M(u i ), where N represents the number of nodes in a single community network.

[0076] In some embodiments, the embodiment extracts the user feature vector representation through an embedding model composed of a two-layer graph convolution network. The graph convolution network updates the embedding vector of the node by propagating the node feature information layer by layer. When processing graph data, the graph convolution network can capture the local neighbor information of the node, thereby generating useful node embedding. By processing the initial user feature vector representation X, the adjacency matrix A, and the degree matrix D through the embedding model, the final user feature vector representation can be more accurately extracted.

[0077] 204. Based on the propagation network of the original topic and the derived topic, extract the topic vector representations of the original topic and the derived topic through the joint distribution adaptive algorithm;

[0078] In social networks, the influence of users on the dissemination of original and derivative topics changes due to domain migration, and the resulting cognitive differences lead to inconsistencies in content data distribution. To improve the generalization ability of the recognition model on derivative topics, this embodiment of the invention employs a joint distribution adaptive method to optimize the model's performance in both topic content spaces. Specifically, the extraction methods for the topic vector representations of the original and derivative topics include:

[0079] Extract message content and user content through native topic propagation networks and derivative topic propagation networks;

[0080] Keywords are extracted from message content and user content to obtain feature vectors for topic content and feature vectors for user sentiment.

[0081] By combining the feature vectors of topic content and the feature vectors of user sentiment with the number of users participating in the original and derivative topics, a joint distribution adaptive algorithm is used to obtain the topic vector representations of the original and derivative topics.

[0082] like Figure 4 As shown, this embodiment of the invention can extract the features of topics and users from the propagation network of original topics and derived topics through user feature vector representation and topic semantic information matrix, thereby obtaining user sentiment difference features; combined with joint distribution adaptive algorithm, by optimizing the model's performance between the source domain and the target domain, a more accurate topic vector representation can be obtained.

[0083] Specifically, this embodiment of the invention divides the data into two parts: topic content and user content, and uses the propagation network N from original topics and derived topics. OT N DT The process involves extracting message content and user sets, embedding a topic semantic information matrix, and mining the similarities and differences in user sentiment. Next, the topic content and user content are preprocessed, and keywords are extracted using Term Frequency-Inverse Document Frequency (TF-IDF) and TextRank algorithms to obtain features of topic and user sentiment. Finally, the TE2Vec method is used to generate feature vectors for the original topic and derived topics as follows:

[0084]

[0085] Among them, T O and T Dthe feature vector representing the topic content, and the feature vector representing the user emotion, C O and C D respectively represent the number of users participating in the original topic and the derivative topic.

[0086] It should be noted that the emo(u i ) in formula (6) reflects the overall emotional tendency of the user, and are the specific manifestations of the emotional features corresponding to these overall emotional tendencies in the original topic and the derivative topic. and are the emotional vectors generated according to the emotional performance of the user in different topics. The emo(u i ) is a relatively basic emotional feature representation, and are high-dimensional emotional feature vectors generated according to the application and changes of the emotional feature in specific topic scenarios.

[0087] The source domain wherein, The target domain wherein, Assume that V O = V D , Y O = Y D , By using pseudo-labels, combining edge MMD metrics and class conditional probability distribution MMD metrics, an exchange variable W is found, in the self-space corresponding to W, The differences of all the above are significantly reduced, and the content feature W T E.

[0088] In this embodiment, the topic vector representation of the original topic is regarded as the source domain, and the topic vector representation of the derivative topic is regarded as the target domain; by minimizing the square of the marginal distribution difference of the source domain and the target domain, and the square of the difference of each class of samples in the source domain and the target domain, the optimal topic vector representation of the original topic and the derivative topic is obtained. The total optimization target is:

[0089]

[0090] wherein, H is a center matrix, U is all participating users of the original topic and the derivative topic, tr(W T UM0U T W) represents the square of the marginal distribution difference of the source domain and the target domain, tr(W T UM V U TW) represents the squared difference between the V-th class samples in the source and target domains. It is calculated as follows:

[0091]

[0092] The source and target domains contain multiple categories, each corresponding to a certain number of samples. The Vth category refers to samples belonging to that category, and this classification can be based on different features, such as sentiment categories or topic categories.

[0093] 205. Input the user feature vector representation and the topic vector representation into the user influence dynamic spatiotemporal model based on the attention mechanism, and output the opinion leader user or the long-tail user.

[0094] In this embodiment of the invention, it is considered that a user's influence may increase or decrease over time because their social relationships, activities, and content may change. Therefore, in order to analyze user influence more accurately, this embodiment considers the dynamic changes of user influence in time and space.

[0095] The attention-based dynamic spatiotemporal model of user influence treats user influence as a dynamically changing spatiotemporal process, optimizing the analysis of user behavior patterns and time-series data by introducing an attention mechanism. It not only captures the interaction relationships between users but also determines how these relationships evolve over time and space.

[0096] like Figure 5 As shown, by combining user feature vector representations and topic vector representations, the dynamic spatiotemporal model utilizes GRU to capture time-series features, leverages an attention mechanism to improve the judgment of the importance of different time steps and inputs, and finally uses GCN to capture graph structure information for global modeling. Ultimately, based on these dynamic spatiotemporal features, the model can predict a user's influence in a social network and classify them as opinion leaders or long-tail users.

[0097] Since the GRU can capture the features before and after the time sequence, it has a significant advantage in predicting the time sequence, so the embodiment introduces the GRU to capture the features before and after the topic time sequence. For each node in the topic network, the GRU performs feature mining on the input topic vector representation and user feature vector representation, and under the joint action of the activation function and the weight matrix, the vector representation information is updated and reset, the time feature-based gated recurrent neural network module effectively obtains the presequence and periodic features of the vector representation information in the topic network, and has excellent memory function. After fusion by the attention mechanism, the user feature vector representation and the topic vector representation weighted by attention can be obtained; by passing the fused representation through the softmax layer, the probabilities that the user is a long tail user and an opinion field can be determined. In fact, the final binary classification result is obtained by using the softmax function; the output corresponding to each user node is a one-dimensional vector, and the one-dimensional vector has two values, 0 representing a long tail user and 1 representing an opinion leader. This way can accurately identify the key users in the derivative topic network.

[0098] In some preferred embodiments, the user influence dynamic spatio-temporal model based on the attention mechanism can be optimized and trained by the classification loss to achieve better recognition effect.

[0099] In one embodiment, please refer to Figure 6 A key user identification device in a derivative topic network, comprising:

[0100] A data acquisition module 301 is configured to acquire data information in a topic network, wherein the data information comprises user basic information, topic basic information and user relationship network.

[0101] A data extraction module 302 is configured to extract original topics, derivative topic user relationship networks, original topic and derivative topic propagation networks, user basic attributes and user intimacy from the data information in the topic network.

[0102] A user feature extraction module 303 is configured to extract a user feature vector representation by a community label propagation algorithm according to the user basic attributes and the user intimacy.

[0103] A topic feature extraction module 304 is configured to extract topic vector representations of original topics and derivative topics by a joint distribution adaptive algorithm according to the original topics and the derivative topic propagation networks.

[0104] A key user identification module 305 is configured to input the user feature vector representation and the topic vector representation into a user influence dynamic spatio-temporal model based on an attention mechanism, and output an opinion leader user or a long tail user.

[0105] In one embodiment, please refer to Figure 7 A key user identification system in a derivative topic network, comprising:

[0106] A transceiver module 401 is configured to acquire data information in a topic network; the data information comprises user basic information, topic basic information and user relationship network;

[0107] A processing module 402 is configured to extract original topics, derivative topic user relationship networks, original topic and derivative topic propagation networks, user basic attributes and user intimacy from the data information in the topic network; to extract user feature vector representations according to the user basic attributes and user intimacy through a community label propagation algorithm; to extract topic vector representations of the original topics and the derivative topics according to the original topic and derivative topic propagation networks through a joint distribution adaptive algorithm; and to input the user feature vector representations and the topic vector representations into a user influence dynamic space-time model based on an attention mechanism, and output opinion leader users or long tail users;

[0108] A display module 403 is configured to display the opinion leader users or long tail users output by the processing module.

[0109] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiments can be completed by programs instructing related hardware, and the programs can be stored in a computer readable storage medium, which can include ROM, RAM, magnetic disk or optical disk, etc.

[0110] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for identifying key users in a derived topic network, characterized by, The method comprises: acquiring data information in a topic network; the data information comprises user basic information, topic basic information, and a user relationship network; extracting a primary topic, a derivative topic, a user relationship network, a primary topic propagation network, a derivative topic propagation network, user basic attributes, and user intimacy from the data information in the topic network; extracting a user feature vector representation by a community label propagation algorithm according to the user basic attributes and the user intimacy; the extraction method of the user feature vector representation comprises processing the user intimacy and the frequency of label occurrence by a propagation algorithm based on a speaker monitor, calculating the acceptance probability of each label, calculating the probability of a user belonging to a label according to the acceptance probability of each label, obtaining a user label set, the number of connected communities, and community size, and generating the user feature vector representation by an embedding model according to the user label set, the number of connected communities, and community size; extracting a topic vector representation of the primary topic and the derivative topic by a joint distribution adaptive algorithm according to the primary topic propagation network and the derivative topic propagation network; inputting the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and outputting an opinion leader user or a long tail user. 2.The method of identifying key users in a derived topic network according to claim 1, characterized in that, The user intimacy is determined by a similarity score, a difference score, and a weighted score of the same interaction mode between users. 3.The method of claim 1, wherein, The acceptance probability of each label is determined by the product of the user intimacy and the frequency of label occurrence and the ratio of the total number of labels. 4.The method of claim 1, wherein, The extraction method of the topic vector representation of the primary topic and the derivative topic comprises: extracting message content and user content through the primary topic propagation network and the derivative topic propagation network; extracting keywords from the message content and the user content to obtain a feature vector of topic content and a feature vector of user emotion; obtaining the topic vector representation of the primary topic and the derivative topic by a joint distribution adaptive algorithm through the feature vector of topic content and the feature vector of user emotion, and the number of users participating in the primary topic and the derivative topic.

5. The method of identifying key users in a derived topic network according to claim 4, wherein, The extraction method of the topic vector representation of the primary topic and the derivative topic comprises: regarding the topic vector representation of the primary topic as a source domain and regarding the topic vector representation of the derivative topic as a target domain; obtaining the optimal topic vector representation of the primary topic and the derivative topic by minimizing the square of the marginal distribution difference between the source domain and the target domain and the square of the difference between the samples in the source domain and the target domain.

6. A key user identification device in a derivative topic network, characterized in that, comprises: a data acquisition module configured to acquire data information in a topic network; the data information comprises user basic information, topic basic information, and a user relationship network; a data extraction module configured to extract a primary topic, a derivative topic, a user relationship network, a primary topic propagation network, a derivative topic propagation network, user basic attributes, and user intimacy from the data information in the topic network; a user feature extraction module configured to extract a user feature vector representation by a community label propagation algorithm according to the user basic attributes and the user intimacy; The topic feature extraction module is configured to extract topic vector representations of the original topic and the derivative topic according to the original topic and the derivative topic propagation network by using a joint distribution adaptive algorithm. The key user identification module is configured to input the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and output opinion leader users or long tail users.

7. A system for identifying key users in a derived topic network, characterized by, The method comprises the following steps: The transceiving module is configured to acquire data information in a topic network. The data information comprises user basic information, topic basic information, and a user relationship network. The processing module is configured to extract an original topic, a derivative topic user relationship network, an original topic and a derivative topic propagation network, user basic attributes, and user intimacy from the data information in the topic network; extract a user feature vector representation from the user basic attributes and the user intimacy by using a community label propagation algorithm; extract topic vector representations of the original topic and the derivative topic from the original topic and the derivative topic propagation network by using a joint distribution adaptive algorithm; input the user feature vector representation and the topic vector representation into a user influence dynamic space-time model based on an attention mechanism, and output opinion leader users or long tail users; The display module is configured to display the opinion leader users or the long tail users output by the processing module.

Images