Family relationship identification method, apparatus and device, and computer readable storage medium

Abstract

The invention discloses a family relationship identification method, apparatus and device, and a computer readable storage medium. The method comprises the steps that first attribute information of a first user, second attribute information of a second user and a communication relation between the first user and the second user are acquired, the first attribute information comprises a first household registration address, and the second attribute information comprises a second household registration address; determining a household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address; and identifying the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship and the communication relationship by using a heterogeneous graph neural network model. According to the family relationship identification method provided by the embodiment of the invention, the accuracy of family relationship identification can be improved.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence technology, and in particular, relates to a method, device, equipment and computer-readable storage medium for identifying family relationships. Background Art

[0002] With the development of mobile communications, home users have become the key to the strategic layout of operators. At present, home users are no longer satisfied with the single voice and Internet needs. For mobile communication companies, users' needs have shifted to entertainment and life application services. Therefore, major telecom operators have begun to develop home services. In home services, the identification of family relationships is crucial.

[0003] At present, the method of identifying family relationships is usually to identify family relationships only based on the telephone communication behavior between members. However, with the development of the times, the frequency of telephone communication between family members is getting lower and lower. In addition, the identification of family relationships based on telephone communication behavior is easily interfered by intermediate points with large in-and-out degrees. For example, real estate agents, deliverymen, couriers, salesmen, etc. usually need to maintain customers by phone. In this way, identifying family relationships based only on telephone communication behavior can easily identify two non-family members as the same family, resulting in low accuracy in identifying family relationships. Summary of the invention

[0004] The embodiments of the present application provide a method, apparatus, device, computer-readable storage medium, and computer program product for identifying family relationships, which can improve the accuracy of identifying family relationships.

[0005] In a first aspect, an embodiment of the present application provides a method for identifying family relationships, the method comprising:

[0006] Acquire first attribute information of a first user, second attribute information of a second user, and a communication relationship between the first user and the second user, wherein the first attribute information includes a first registered permanent address, and the second attribute information includes a second registered permanent address;

[0007] Determining a household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address;

[0008] Using a heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship and the communication relationship.

[0009] In a possible implementation, before using the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household address relationship, and the communication relationship, the method further includes:

[0010] Acquire attribute information samples respectively corresponding to a plurality of users, household registration address relationship samples, communication relationship samples, and family relationship samples between the plurality of users;

[0011] Constructing a heterogeneous graph using attribute information samples respectively corresponding to the multiple users as nodes and household registration address relationship samples and communication relationship samples between the multiple users as edges;

[0012] Extracting features from nodes and edges in the heterogeneous graph to obtain feature vectors;

[0013] Using the initial heterogeneous graph neural network model, family relationship recognition is performed based on the feature vector to obtain predicted family relationships;

[0014] Determining a first loss function value according to the similarity between the predicted family relationship and the family relationship sample;

[0015] The model parameters are adjusted according to the first loss function value, and the heterogeneous graph neural network model is trained to obtain the heterogeneous graph neural network model.

[0016] In a possible implementation manner, before determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address, the method further includes:

[0017] Using an address classification model to identify the address types of the first registered address and the second registered address;

[0018] The determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address includes:

[0019] In a case where the address types of the first household registration address and the second household registration address are both private residences, a household registration address relationship between the first user and the second user is determined according to the first household registration address and the second household registration address.

[0020] In a possible implementation, determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address includes:

[0021] Using a word segmenter, decompose the first registered residence address into a plurality of first registered residence address elements, and decompose the second registered residence address into a plurality of second registered residence address elements;

[0022] Using a part-of-speech tagger to perform part-of-speech recognition on the plurality of first household registration address elements to obtain a first part-of-speech recognition result, and performing part-of-speech recognition on the plurality of second household registration address elements to obtain a second part-of-speech recognition result;

[0023] updating the plurality of first household registration address elements into a plurality of third household registration address elements according to the first part-of-speech recognition result, and updating the plurality of second household registration address elements into a plurality of fourth household registration address elements according to the second part-of-speech recognition result;

[0024] The household registration address relationship is determined according to the plurality of third household registration address elements and the plurality of fourth household registration address elements.

[0025] In a possible implementation manner, determining the household registration address relationship according to the plurality of third household registration address elements and the plurality of fourth household registration address elements includes:

[0026] Generating a first household registration address map according to the plurality of third household registration address elements and the associations therebetween, and generating a second household registration address map according to the plurality of fourth household registration address elements and the associations therebetween;

[0027] Based on a random walk strategy, randomly sampling the multiple first household registration address elements in the first household registration address graph to obtain a first node sequence, and randomly sampling the multiple second household registration address elements in the second household registration address graph to obtain a second node sequence;

[0028] generating a first node vector corresponding to the first node sequence and a second node vector corresponding to the second node sequence;

[0029] Determine whether the first node vector and the second node vector match using a multi-layer perceptron to obtain a matching result;

[0030] The household registration address relationship is determined according to the matching result.

[0031] In a possible implementation, before using the address classification model to identify the types of the first registered residence address and the second registered residence address, the method further includes:

[0032] Get household registration address samples and their corresponding address type labels;

[0033] Predicting the address type of the household registration address sample using the initial address classification model to obtain a predicted address type;

[0034] Determining a second loss function value according to the similarity between the predicted address type and the address type label;

[0035] The model parameters are adjusted according to the second loss function value, and the address classification model is obtained through training.

[0036] In a possible implementation, the first attribute information includes a first surname, the second attribute information includes a second surname, and before using the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship, and the communication relationship, the method further includes:

[0037] Determining a surname relationship between the first user and the second user according to the first surname and the second surname based on a surname graph, wherein the surname graph includes association relationships between a plurality of surnames;

[0038] The using of the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household address relationship, and the communication relationship includes:

[0039] Using the heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship, the communication relationship and the surname relationship.

[0040] In a possible implementation, before determining the surname relationship between the first user and the second user based on the surname graph according to the first surname and the second surname, the method further includes:

[0041] Get last name data;

[0042] Extracting surname elements from the surname data, wherein the surname elements include surname, family tree and clan tree;

[0043] Determine the association relationship between the surnames according to the family tree and the clan tree;

[0044] The surname graph is constructed with the surnames as nodes and the associations between the surnames as edges.

[0045] In a second aspect, an embodiment of the present application provides a device for identifying family relationships, the device comprising:

[0046] A first acquisition module, used to acquire first attribute information of a first user, second attribute information of a second user, and a communication relationship between the first user and the second user, wherein the first attribute information includes a first registered permanent address, and the second attribute information includes a second registered permanent address;

[0047] A first determining module, configured to determine a household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address;

[0048] The first identification module is used to use a heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship and the communication relationship.

[0049] In a third aspect, an embodiment of the present application provides an electronic device, the device comprising: a processor and a memory storing computer program instructions;

[0050] When the processor executes the computer program instructions, the processor implements any possible implementation method of the first aspect described above.

[0051] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implements a method in any possible implementation method of the first aspect described above.

[0052] In a fifth aspect, an embodiment of the present application provides a computer program product. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes a method as any possible implementation method in the first aspect above.

[0053] The family relationship identification method, device, equipment, computer-readable storage medium and computer program product of the embodiment of the present application can obtain multiple relationships between the first user and the second user in different dimensions and different types by obtaining the communication relationship between the first user and the second user, and determining the household address relationship between the first user and the second user based on the first household address and the second household address. By obtaining the first attribute information and the second attribute information, and using a heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship and the communication relationship. That is, when performing family relationship identification, the attribute information of the two users and the multiple relationships between the two users in different dimensions and different types are comprehensively considered, which can improve the accuracy of identifying family relationships compared to family relationship identification based only on telephone communication behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solution of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0055] Figure 1 It is a flowchart of a method for identifying family relationships provided in an embodiment of the present application;

[0056] Figure 2 is a schematic diagram of a first household registration address map provided in an embodiment of the present application;

[0057] Figure 3 is a schematic diagram of a heterogeneous graph provided in an embodiment of the present application;

[0058] Figure 4 It is a structural schematic diagram of a family relationship identification device provided in an embodiment of the present application;

[0059] Figure 5 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] The features and exemplary embodiments of various aspects of the present application will be described in detail below. In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without the need for some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by illustrating the examples of the present application.

[0061] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "include..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0062] As described in the background technology section, in order to solve the problems of the prior art, the embodiments of the present application provide a method, apparatus, device, computer-readable storage medium and computer program product for identifying family relationships.

[0063] The following first introduces the method for identifying family relationships provided in the embodiment of the present application.

[0064] Figure 1FIG. 1 is a flow chart of a method for identifying family relationships provided in an embodiment of the present application. Figure 1 As shown, the method for identifying family relationships provided in the embodiment of the present application includes the following steps:

[0065] S110, obtaining first attribute information of a first user, second attribute information of a second user, and a communication relationship between the first user and the second user, wherein the first attribute information includes a first registered permanent address, and the second attribute information includes a second registered permanent address;

[0066] S120, determining a household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address;

[0067] S130. Using a heterogeneous graph neural network model, identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship, and the communication relationship.

[0068] The family relationship identification method of the embodiment of the present application can obtain multiple relationships between the first user and the second user in different dimensions and different types by obtaining the communication relationship between the first user and the second user, and determining the household address relationship between the first user and the second user based on the first household address and the second household address. By obtaining the first attribute information and the second attribute information, and using a heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship and the communication relationship. That is, when performing family relationship identification, the attribute information of the two users and the multiple relationships between the two users in different dimensions and different types are comprehensively considered. Compared with performing family relationship identification based only on telephone communication behavior, the accuracy of identifying family relationships can be improved.

[0069] The specific implementation methods of the above steps are introduced below.

[0070] In some embodiments, in S110, the first user may be any one of the multiple users. The second user may be any one of the multiple users except the first user. Each user may correspond to an attribute information. The attribute information of the user may include, for example, the user's surname, gender, age, household registration address, user number and other information. Among them, each user may correspond to a household registration address. The first household registration address may be the household registration address of the first user. The second household registration address may be the household registration address of the second user. The household registration addresses of different users may be the same or different. The form of the household registration address may be, for example, "No. F, Community E, Street D, City B, County C, Province A".

[0071] In addition, user numbers may include fixed telephone numbers and mobile phone numbers. There may be a one-to-one correspondence between users and user numbers. Based on this, the communication relationship may include a telephone communication relationship and a base station signaling relationship between two users (i.e., a user number pair). Among them, the telephone communication relationship may include call duration and number of calls. The call duration may, for example, include the call duration during the daytime working hours on weekdays, the call duration during the night on weekdays, the call duration during the day on weekends, and the call duration during the night on weekends; the number of calls may, for example, include the number of calls during the daytime working hours on weekdays, the number of calls during the night on weekdays, the number of calls during the day on weekends, and the number of calls during the night on weekends; the base station signaling relationship may, for example, include the number of days that the user number pair overlaps with the base station at night on weekdays and the number of days that the base station overlaps at night on non-working days.

[0072] In some embodiments, in S120, the household registration address relationship may include the household registration address being consistent and the household registration address being inconsistent. If the first household registration address and the second household registration address are exactly the same, it can be determined that the first household registration address and the second household registration address are consistent. If the first household registration address and the second household registration address are substantially the same, it can also be determined that the first household registration address and the second household registration address are consistent. For example, if the first household registration address and the second household registration address are also exactly the same except for the "province", and the reason for the difference in the "province" is a formal error, it can be determined that the first household registration address and the second household registration address are consistent.

[0073] As an example, by performing a one-to-one comparison between the first registered residence address and the second registered residence address, the registered residence address relationship between the first user and the second user may be determined.

[0074] In actual situations, if the household registration address is not entered in a standardized manner, the same household registration address may be identified as a different household registration address. Therefore, in order to improve the accuracy of determining the household registration address relationship, in some embodiments, the above S120 may specifically include:

[0075] Using a word segmenter, the first registered permanent address is decomposed into a plurality of first registered permanent address elements, and the second registered permanent address is decomposed into a plurality of second registered permanent address elements;

[0076] Using a part-of-speech tagger to perform part-of-speech recognition on a plurality of first household registration address elements to obtain a first part-of-speech recognition result, and performing part-of-speech recognition on a plurality of second household registration address elements to obtain a second part-of-speech recognition result;

[0077] updating the plurality of first household registration address elements into a plurality of third household registration address elements according to the first part-of-speech recognition result, and updating the plurality of second household registration address elements into a plurality of fourth household registration address elements according to the second part-of-speech recognition result;

[0078] A household registration address relationship is determined based on multiple third household registration address elements and multiple fourth household registration address elements.

[0079] Here, the tokenizer may be a Transformer tokenizer obtained by fine-tuning a pre-trained Transformer tokenizer using a household registration address sample and a household registration address element sample. The Transformer tokenizer may process the household registration address (including the first household registration address element and the second household registration address element) and decompose it into multiple household registration address elements.

[0080] The part-of-speech tagger can be, for example, a POS tagger. The POS tagger can identify the part-of-speech (e.g., noun, adjective, verb, etc.) of the household registration address element, and can also identify symbols such as brackets, periods, double quotes, single quotes, etc. that appear in the household registration address element. Based on this, the part-of-speech tagging result can include the part-of-speech corresponding to each household registration address element. If there are symbols in the household registration address element, the part-of-speech tagging result can also include the symbols included in the household registration address element. For example, if a first household registration address element is "A% Province", the first part-of-speech tagging result can include that the part of speech of "A% Province" is a noun, and the symbol "%" is included therein.

[0081] In addition, updating the household registration address element according to the part-of-speech tagging result may be to delete the symbol in the original household registration address element. For example, based on the above example, the third household registration address element corresponding to the first household registration address element "A% Province" may be "A Province" obtained by deleting the "%" in the first household registration address element. If multiple first household registration address elements do not include symbols, then multiple third household registration address elements and multiple first household registration address elements may be the same. The second household registration address element and the fourth household registration address element are similar and will not be described in detail here.

[0082] In addition, determining the household registration address relationship according to the plurality of third household registration address elements and the plurality of fourth household registration address elements may be to match the plurality of third household registration address elements and the plurality of fourth household registration address elements one by one. Since the third household registration address element is updated based on the first part-of-speech recognition result, and the fourth household registration address element is updated based on the second part-of-speech recognition result, the accuracy of determining the household registration address relationship can be improved by determining the accuracy of the household registration address relationship according to the plurality of third household registration address elements and the plurality of fourth household registration address elements.

[0083] Based on this, in order to improve the efficiency of determining the household registration address relationship, in some embodiments, the above-mentioned determination of the household registration address relationship based on multiple third household registration address elements and multiple fourth household registration address elements may specifically include:

[0084] Generating a first household registration address map according to a plurality of third household registration address elements and associations therebetween, and generating a second household registration address map according to a plurality of fourth household registration address elements and associations therebetween;

[0085] Based on the random walk strategy, a plurality of first household registration address elements in the first household registration address graph are randomly sampled to obtain a first node sequence, and a plurality of second household registration address elements in the second household registration address graph are randomly sampled to obtain a second node sequence;

[0086] generating a first node vector corresponding to the first node sequence and a second node vector corresponding to the second node sequence;

[0087] Using a multi-layer perceptron to determine whether the first node vector and the second node vector match, and obtaining a matching result;

[0088] The household registration address relationship is determined based on the matching results.

[0089] Here, the association relationship between the household registration address elements can be determined according to the order in which the household registration address elements appear in the household registration address. For example, if the first household registration address is "Province A, City B, County C, Street D, Community E, No. F", then it can be determined that the third household registration address element (i.e., the first household registration address element) "Province A" is adjacent to "City B" and is in front of "City B", and similarly, it can be determined that the third household registration address element "City B" is adjacent to "County C" and is in front of "County C". The same is true for the remaining third household registration address elements, which will not be described in detail here. After determining the adjacent relationship between multiple third household registration address elements, the first household registration address graph can be established by using multiple third household registration address elements as nodes and the adjacent relationships between multiple third household registration address elements as edges. As shown in the above example, when the first household registration address is "Province A, City B, County C, Street D, Community E, No. F", the first household registration address graph can be as follows Figure 2 The second household registration address map is similar and will not be described in detail here.

[0090] In addition, the random walk strategy can be, for example, a random walk strategy based on node2vec. Specifically, by taking each node in the household address map (including the first household address map and the second household address map) as a starting point for random sampling, a node sequence corresponding to the household address map can be obtained. In addition, in order to reduce the sparsity of the data and improve the robustness of the vector representation, multiple random walks can be performed and more node sequences can be collected to obtain a more comprehensive and accurate node representation.

[0091] Specifically, the random walk strategy can be implemented by the following formulas (1)-(3).

[0092] The random walk formula can be expressed as follows:

[0093]

[0094] In formula (1), c i It can represent the i-th walk, π vxIt can represent the transition probability between nodes, and Z can be a parameter.

[0095] In addition, the non-standard transition probability can be expressed as follows:

[0096] π vx =a pq (t,x)·w vx (2)

[0097]

[0098] In formula (3), d tx It can represent the shortest path between node t and node x, and its value is {0,1,2}. The parameters p and q in formula (3) need to be properly tuned to ensure the balance and breadth of sampling. Adjusting these two parameters can control the way random walks explore in the graph, thereby avoiding the problem of sampling bias.

[0099] In addition, in the field of Natural Language Processing (NLP), the vector representation of a word can be obtained by predicting the context given an input word. Inspired by this, the node vectors in the graph can also be obtained by predicting adjacent nodes given a node sequence. In other words, by treating the first node sequence and the second node sequence as sentences, the word2vec method (for example, using the Skip-Gram model) can be used to generate a first node vector corresponding to the first node sequence and a second node vector corresponding to the second node sequence.

[0100] After generating the first node vector and the second node vector, a multilayer perceptron (MLP) can be used to determine whether the first node vector and the second node vector match to obtain a matching result. If the matching result is a match, the household registration address relationship can be consistent with the household registration address. If the matching result is not a match, the household registration address relationship can be inconsistent. In addition, in order to improve the accuracy of the matching, the tf-idf value between the first node vector and the second node vector can be calculated to enhance the difference of similar addresses, and then matching is performed based on the MLP method to obtain a matching result.

[0101] In this way, by converting the household registration address pair matching problem into a graph matching problem, the efficiency of determining the household registration address relationship can be improved.

[0102] In addition, if the first household registration address and the second household registration address are both company addresses, then the household registration address relationship is determined by the above method, and then the family relationship is identified based on the household registration address relationship, which may misjudge the relationship between colleagues as the same family.

[0103] Based on this, in order to further improve the accuracy of identifying family relationships, in some embodiments, before the above S120, the following steps may also be included:

[0104] Using an address classification model to identify the address type of the first registered address and the second registered address;

[0105] Based on this, the above S120 may specifically include:

[0106] In a case where the address types of the first registered residence and the second registered residence are both private residences, a registered residence address relationship between the first user and the second user is determined according to the first registered residence and the second registered residence.

[0107] Here, the address classification model can be a neural network model that can determine the household registration address type. The address classification model can be based on a supervised learning algorithm, such as a support vector machine (SVM) algorithm, a decision tree algorithm, a random forest algorithm, etc., or can be based on a semi-supervised learning method and a transfer learning method, which are not limited here. In addition, the address type can include private residences and non-private residences.

[0108] As an example, after the first registered residence address and the second registered residence address are input into the address classification model, the address classification model can identify the address types of the first registered residence address and the second registered residence address to obtain the address types corresponding to the first registered residence address and the second registered residence address. If the address type is a private residence, the registered residence address relationship can be further determined.

[0109] In addition, through a series of predefined rules and rule bases, the input household registration address can be matched and classified to identify whether the household registration address is a private residence.

[0110] In this way, by determining the household registration address relationship when the address types of the first household registration address and the second household registration address are both private residences, and then identifying the family relationship based on the household registration address relationship, the accuracy of identifying the family relationship can be improved in one step.

[0111] Based on this, in order to obtain the above-mentioned address classification model, in some embodiments, before the address classification model is used to identify the address type of the first household registration address and the second household registration address, the following may also be included:

[0112] Get household registration address samples and their corresponding address type labels;

[0113] The initial address classification model is used to predict the address type of the household registration address sample to obtain the predicted address type;

[0114] Determining a second loss function value according to the similarity between the predicted address type and the address type label;

[0115] The model parameters are adjusted according to the second loss function value, and the address classification model is obtained through training.

[0116] Here, the map software may generally include a large amount of address data. The address data may include detailed information of the address. Each address data may correspond to a POI (Point of Interes) type. The POI types may include, for example, "residential area", "community", "residential building", "shop", "company", "bus stop", etc. Among them, "residential area", "community", "residential building" may be marked as "private residence". "Shop", "company", "bus stop" may be marked as "non-private residence".

[0117] As an example, after obtaining address data from map software, the address data can be preprocessed. Preprocessing can include special symbols, conversion to lowercase, word segmentation and other operations. The preprocessed data can be converted into word vector representation, and the length of each word vector can be set by counting the vector lengths of all words. Generally, pre-trained word vector models such as Word2Vec, GloVe, etc. can be used, or a custom word vector model based on the current task can be used to standardize the feature vector so that its mean is 0 and its variance is 1, so as to better train the model. In this way, the household registration address sample can be the address data after preprocessing. In addition, the address type sample can be any one of "private residence" and "non-private residence". After obtaining the household registration address sample and its corresponding address type label, the household registration address sample and its corresponding address type label can be divided into a training set and a test set.

[0118] In addition, the initial address classification model can be a neural network model built based on the SVM algorithm. The goal of the SVM algorithm is to find an optimal segmentation hyperplane to separate data points of different address categories.

[0119] As an example, after constructing the initial address classification model, the training set can be used to train the initial address classification model to obtain the address classification model. During the training process, it is necessary to select a suitable kernel function (such as a linear kernel function, a polynomial kernel function, or a radial basis function, etc.) and adjust the model parameters (such as regularization parameters, penalty factors, etc.) to improve the accuracy of the model. Afterwards, the trained model can be evaluated using the test set, and the model's accuracy, precision, recall, and other indicators can be calculated to evaluate its performance and generalization ability.

[0120] In some embodiments, in S130, a heterogeneous graph neural network (HGN) model can be used to predict unknown link relationships in a heterogeneous graph in a link prediction task. Link prediction refers to inferring possible but unobserved edges in a graph based on known graph structures and node attributes, which can be used to determine whether different users are family relationships.

[0121] In an embodiment of the present application, a heterogeneous graph can be constructed with the attribute information corresponding to multiple users as nodes, and the household address relationship and communication relationship between multiple users as edges. A heterogeneous graph neural network model can be constructed based on the heterogeneous graph. Among them, the heterogeneous graph neural network model may include neural network structures such as an embedding layer of nodes and edges, a graph convolution layer, a pooling layer, and an attention layer. In this way, by constructing a heterogeneous graph neural network model based on a heterogeneous graph, the heterogeneous graph neural network model can identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household address relationship, and the communication relationship.

[0122] Based on this, in order to obtain a heterogeneous graph neural network model capable of identifying family relationships, in some embodiments, before the above S130, the following steps may also be included:

[0123] Obtaining attribute information samples corresponding to multiple users, household registration address relationship samples, communication relationship samples, and family relationship samples between multiple users;

[0124] A heterogeneous graph is constructed using attribute information samples corresponding to multiple users as nodes and household address relationship samples and communication relationship samples between multiple users as edges;

[0125] Extract features from nodes and edges in heterogeneous graphs to obtain feature vectors;

[0126] Using the initial heterogeneous graph neural network model, family relationships are identified based on feature vectors to obtain predicted family relationships;

[0127] Determining a first loss function value according to the similarity between the predicted family relationship and the family relationship sample;

[0128] The model parameters are adjusted according to the first loss function value, and the heterogeneous graph neural network model is trained.

[0129] Here, the attribute information sample may include information such as surname, gender, age, household registration address, user income, etc. The household registration address relationship sample may include information such as the household registration address is consistent and is a private residence, the household registration address is consistent and is a non-private residence, and the household registration address is inconsistent. The communication relationship sample may include information such as the number of calls, call duration, weekend night base station overlap, etc. for the user number pair. Among them, the sample data may come from network data in actual application scenarios, such as social networks, recommendation systems, etc. After obtaining sample data such as attribute information samples, household registration address relationship samples, communication relationship samples, and family relationship samples, the data may be divided into a training set and a test set.

[0130] In addition, after obtaining the sample data, the sample data can be constructed into a heterogeneous graph, where nodes can represent different types of entities and edges can represent different types of relationships. Graph data processing libraries (such as NetworkX, DGL, etc.) can be used to construct and process graph data.

[0131] By extracting features from nodes and edges in heterogeneous graphs, they can be converted into representation vectors that can be used in neural network models. Features of nodes and edges can be extracted through various graph neural network models (such as GCN, GAT, GraphSAGE, etc.).

[0132] An initial heterogeneous graph neural network model can be constructed based on the heterogeneous graph. The initial heterogeneous graph neural network model may include neural network structures such as node and edge embedding layers, graph convolution layers, pooling layers, and attention layers.

[0133] By training the initial heterogeneous graph neural network model with the training set, the model parameters can be optimized to obtain a heterogeneous graph neural network model. During the training process, the binary cross entropy loss function can be used to measure the gap between the predicted results and the actual link situation. That is, the first loss function value can be the binary cross entropy loss function value. In addition, the model parameters are updated by using optimization methods such as back propagation algorithm and gradient descent so that it can better fit the training data. After the model training is completed, the trained model can be used to predict the test data. The performance of the model can be evaluated by comparing the predicted results output by the model with the true labels. When evaluating the model prediction results, evaluation indicators (such as accuracy, precision, recall, etc.) can be used to evaluate the performance of the model for further tuning and improvement.

[0134] In addition, in order to further improve the accuracy of identifying family relationships, in some embodiments, before the above S130, the following steps may also be included:

[0135] Determining a surname relationship between the first user and the second user according to the first surname and the second surname based on a surname map, wherein the surname map includes association relationships between multiple surnames;

[0136] Using a heterogeneous graph neural network model, identifying a family relationship between a first user and a second user based on first attribute information, second attribute information, a household address relationship, and a communication relationship, including:

[0137] Using a heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household registration address relationship, the communication relationship and the surname relationship.

[0138] Here, the surname map includes associations between multiple surnames, which may include surnames that are the same or have the same pronunciation, the same origin, or the same clan.

[0139] As an example, if the relationship between users includes a surname relationship, and the communication relationship includes a telephone communication relationship and a base station signaling relationship, the heterogeneous graph can be as follows: Figure 3 As shown. Figure 3 Among them, the first type of relationship can be a telephone communication relationship, the second type of relationship can be a base station signaling relationship, the third type of relationship can be a household registration address relationship, and the fourth type of relationship can be a surname relationship.

[0140] In this way, by considering the surname relationship based on the first attribute information, the second attribute information, the household registration address relationship and the communication relationship, the accuracy of identifying family relationships can be further improved.

[0141] Based on this, in order to determine the surname relationship based on the surname map, in some embodiments, before determining the surname relationship between the first user and the second user based on the surname map according to the first surname and the second surname, the following steps may also be included:

[0142] Get last name data;

[0143] Extract surname elements from surname data, where the surname elements include surname, family tree and ancestry tree;

[0144] Determine the relationships between surnames based on genealogical and pedigree trees;

[0145] A surname graph is constructed with surnames as nodes and the relationships between surnames as edges.

[0146] Here, building a surname graph involves representing surnames and their associated information in the form of a graph. A surname graph can be a graph structure. Each node can represent a surname, and edges can represent the association between surnames, such as the same surname, homonym, common origin, and common clan. The surname graph can be used to study the origin, inheritance, and evolution of surnames, and can also help identify family relationships. The following are the general steps to build a surname graph:

[0147] First, surname data related to surnames are collected, including surnames and their origins, distribution, variants, and other information. Surname data can be obtained from censuses, surname databases, historical documents, and other channels. Then, the surname data is preprocessed, such as data cleaning and deduplication, to ensure the accuracy and completeness of the surname data.

[0148] Secondly, surname elements are extracted. For each surname, its surname elements need to be extracted, such as surname, given name, family tree, clan tree, etc. Technologies such as word segmentation and part-of-speech tagging can be used to achieve the extraction of surname elements.

[0149] Then, a surname graph is constructed based on the associations between surnames. A surname graph is a graph structure in which each node represents a surname and edges represent the associations between surnames. For example, if two surnames have the same pronunciation, the same clan, the same origin, or a historical connection, an edge can be established to represent the association between them.

[0150] Finally, the surname map can be expanded according to needs to add more information and associations. For example, the population distribution and historical evolution of surnames can be added to enrich the content of the surname map.

[0151] Based on the family relationship identification method provided in the above embodiment, the present application also provides a specific implementation of a family relationship identification device. Please refer to the following embodiment.

[0152] like Figure 4 As shown, the family relationship identification device 400 provided in the embodiment of the present application includes the following modules:

[0153] A first acquisition module 410 is used to acquire first attribute information of a first user, second attribute information of a second user, and a communication relationship between the first user and the second user, wherein the first attribute information includes a first registered permanent address, and the second attribute information includes a second registered permanent address;

[0154] A first determination module 420, configured to determine a household registration address relationship between a first user and a second user according to the first household registration address and the second household registration address;

[0155] The first identification module 430 is used to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household address relationship and the communication relationship by using a heterogeneous graph neural network model.

[0156] The above-mentioned family relationship identification device 400 is described in detail below, as follows:

[0157] In some embodiments, the family relationship identification device 400 may further include:

[0158] A second acquisition module is used to acquire attribute information samples corresponding to a plurality of users, household registration address relationship samples, communication relationship samples, and family relationship samples between a plurality of users before identifying the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship, and the communication relationship using the heterogeneous graph neural network model;

[0159] The first construction module is used to construct a heterogeneous graph using attribute information samples corresponding to multiple users as nodes and household registration address relationship samples and communication relationship samples between multiple users as edges;

[0160] The first extraction module is used to extract features from nodes and edges in the heterogeneous graph to obtain feature vectors;

[0161] The second recognition module is used to use the initial heterogeneous graph neural network model to identify family relationships based on feature vectors to obtain predicted family relationships;

[0162] A second determination module, configured to determine a first loss function value according to a similarity between the predicted family relationship and the family relationship sample;

[0163] The first training module is used to adjust the model parameters according to the first loss function value and train a heterogeneous graph neural network model.

[0164] In some embodiments, the family relationship identification device 400 may further include:

[0165] The third identification module is used to identify the address types of the first household registration address and the second household registration address by using an address classification model before determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address.

[0166] Based on this, the first determining module 420 may specifically include:

[0167] The first determination submodule is used to determine the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address when the address types of the first household registration address and the second household registration address are both private residences.

[0168] In some embodiments, the first determining module 420 may further include:

[0169] A disassembly submodule, used to disassemble the first registered residence address into a plurality of first registered residence address elements and to disassemble the second registered residence address into a plurality of second registered residence address elements by using a word segmenter;

[0170] A first identification submodule is used to perform part-of-speech identification on a plurality of first household registration address elements using a part-of-speech tagger to obtain a first part-of-speech identification result, and to perform part-of-speech identification on a plurality of second household registration address elements to obtain a second part-of-speech identification result;

[0171] An updating submodule, configured to update the plurality of first household registration address elements into a plurality of third household registration address elements according to the first part-of-speech recognition result, and to update the plurality of second household registration address elements into a plurality of fourth household registration address elements according to the second part-of-speech recognition result;

[0172] The second determination submodule is used to determine the household registration address relationship according to multiple third household registration address elements and multiple fourth household registration address elements.

[0173] In some embodiments, the second determining submodule may specifically include:

[0174] A first generating unit, configured to generate a first household registration address map according to a plurality of third household registration address elements and associations therebetween, and to generate a second household registration address map according to a plurality of fourth household registration address elements and associations therebetween;

[0175] A sampling unit, configured to randomly sample a plurality of first household registration address elements in the first household registration address graph based on a random walk strategy to obtain a first node sequence, and to randomly sample a plurality of second household registration address elements in the second household registration address graph to obtain a second node sequence;

[0176] A second generating unit, used to generate a first node vector corresponding to the first node sequence and a second node vector corresponding to the second node sequence;

[0177] A matching unit, used for determining whether the first node vector and the second node vector match by using a multi-layer perceptron, and obtaining a matching result;

[0178] The determination unit is used to determine the household registration address relationship according to the matching result.

[0179] In some embodiments, the family relationship identification device 400 may further include:

[0180] The third acquisition module is used to obtain household registration address samples and their corresponding address type labels;

[0181] A prediction module, used to predict the address type of the household registration address sample using the initial address classification model to obtain a predicted address type;

[0182] A third determination module, used to determine a second loss function value according to the similarity between the predicted address type and the address type label;

[0183] The second training module is used to adjust the model parameters according to the second loss function value and train the address classification model.

[0184] In some embodiments, the first attribute information includes a first surname, and the second attribute information includes a second surname. Based on this, the family relationship identification device 400 may also include:

[0185] The fourth determination module is used to determine the surname relationship between the first user and the second user based on the surname map according to the first surname and the second surname before using the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship and the communication relationship, wherein the surname map includes the association relationship between multiple surnames.

[0186] Based on this, the first identification module 430 may specifically include:

[0187] The second identification submodule is used to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship, the communication relationship and the surname relationship by using a heterogeneous graph neural network model.

[0188] In some embodiments, the family relationship identification device 400 may further include:

[0189] A fourth acquisition module, configured to acquire surname data before determining a surname relationship between the first user and the second user according to the first surname and the second surname;

[0190] The second extraction module is used to extract surname elements from the surname data, and the surname elements include surnames, family trees and pedigrees;

[0191] A fifth determination module is used to determine the association relationship between surnames based on the family tree and the clan tree;

[0192] The second construction module is used to construct a surname graph using surnames as nodes and associations between surnames as edges.

[0193] The family relationship identification device of the embodiment of the present application can obtain multiple relationships between the first user and the second user in different dimensions and different types by obtaining the communication relationship between the first user and the second user, and determining the household address relationship between the first user and the second user based on the first household address and the second household address. By obtaining the first attribute information and the second attribute information, and using a heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship and the communication relationship. That is, when performing family relationship identification, the attribute information of the two users and the multiple relationships between the two users in different dimensions and different types are comprehensively considered. Compared with performing family relationship identification based only on telephone communication behavior, the accuracy of identifying family relationships can be improved.

[0194] Based on the family relationship identification method provided in the above embodiment, the embodiment of the present application also provides a specific implementation of the electronic device. Figure 5 A schematic diagram of an electronic device 500 provided in an embodiment of the present application is shown.

[0195] The electronic device 500 may include a processor 510 and a memory 520 storing computer program instructions.

[0196] Specifically, the processor 510 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0197] The memory 520 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 520 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. Where appropriate, the memory 520 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 520 may be inside or outside the electronic device 500. In a particular embodiment, the memory 520 is a non-volatile solid-state memory.

[0198] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Thus, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of the present application.

[0199] The processor 510 implements any one of the family relationship identification methods in the above embodiments by reading and executing computer program instructions stored in the memory 520.

[0200] In one example, the electronic device 500 may further include a communication interface 530 and a bus 540. Figure 5 As shown, the processor 510, the memory 520, and the communication interface 530 are connected via a bus 540 and communicate with each other.

[0201] The communication interface 530 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0202] Bus 540 includes hardware, software or both, and the parts of electronic equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front side bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 540 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the application considers any suitable bus or interconnection.

[0203] Exemplarily, the electronic device 500 may be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA).

[0204] The electronic device can execute the method for identifying family relationships in the embodiment of the present application, thereby realizing the combination Figures 1 to 4 A method and apparatus for identifying family relationships is described.

[0205] In addition, in combination with the family relationship identification method in the above embodiment, the embodiment of the present application can provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any family relationship identification method in the above embodiment is implemented.

[0206] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0207] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0208] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0209] The above reference is according to the method of the embodiment of the present application, the flow chart of the device (system) and the computer program product and / or the block diagram described various aspects of the present application.It should be understood that each square box in the flow chart and / or the block diagram and the combination of each square box in the flow chart and / or the block diagram can be realized by computer program instructions.These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the realization of the function / action specified in one or more square boxes of the flow chart and / or the block diagram.Such a processor can be but is not limited to a general-purpose processor, a special-purpose processor, a special application processor or a field programmable logic circuit.It can also be understood that each square box in the block diagram and / or the flow chart and the combination of the square boxes in the block diagram and / or the flow chart can also be realized by the dedicated hardware that performs the specified function or action, or can be realized by the combination of dedicated hardware and computer instructions.

[0210] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A method for identifying family relationships, characterized in that: include: Acquire first attribute information of a first user, second attribute information of a second user, and a communication relationship between the first user and the second user, wherein the first attribute information includes a first registered permanent address, and the second attribute information includes a second registered permanent address; Determining a household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address; Using a heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship and the communication relationship.

2. The method according to claim 1, characterized in that Before using the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household address relationship, and the communication relationship, the method further includes: Acquire attribute information samples respectively corresponding to a plurality of users, household registration address relationship samples, communication relationship samples, and family relationship samples between the plurality of users; Constructing a heterogeneous graph using attribute information samples respectively corresponding to the multiple users as nodes and household registration address relationship samples and communication relationship samples between the multiple users as edges; Extracting features from nodes and edges in the heterogeneous graph to obtain feature vectors; Using the initial heterogeneous graph neural network model, family relationship recognition is performed based on the feature vector to obtain predicted family relationships; Determining a first loss function value according to the similarity between the predicted family relationship and the family relationship sample; The model parameters are adjusted according to the first loss function value, and the heterogeneous graph neural network model is trained to obtain the heterogeneous graph neural network model.

3. The method according to claim 1, characterized in that: Before determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address, the method further includes: Using an address classification model to identify the address types of the first registered address and the second registered address; The determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address includes: In a case where the address types of the first household registration address and the second household registration address are both private residences, a household registration address relationship between the first user and the second user is determined according to the first household registration address and the second household registration address.

4. The method according to any one of claims 1 to 3, characterized in that: The determining the household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address includes: Using a word segmenter, decompose the first registered residence address into a plurality of first registered residence address elements, and decompose the second registered residence address into a plurality of second registered residence address elements; Using a part-of-speech tagger to perform part-of-speech recognition on the plurality of first household registration address elements to obtain a first part-of-speech recognition result, and performing part-of-speech recognition on the plurality of second household registration address elements to obtain a second part-of-speech recognition result; updating the plurality of first household registration address elements into a plurality of third household registration address elements according to the first part-of-speech recognition result, and updating the plurality of second household registration address elements into a plurality of fourth household registration address elements according to the second part-of-speech recognition result; The household registration address relationship is determined according to the plurality of third household registration address elements and the plurality of fourth household registration address elements.

5. The method according to claim 4, characterized in that The determining the household registration address relationship according to the plurality of third household registration address elements and the plurality of fourth household registration address elements comprises: Generating a first household registration address map according to the plurality of third household registration address elements and the associations therebetween, and generating a second household registration address map according to the plurality of fourth household registration address elements and the associations therebetween; Based on a random walk strategy, randomly sampling the multiple first household registration address elements in the first household registration address graph to obtain a first node sequence, and randomly sampling the multiple second household registration address elements in the second household registration address graph to obtain a second node sequence; generating a first node vector corresponding to the first node sequence and a second node vector corresponding to the second node sequence; Determine whether the first node vector and the second node vector match using a multi-layer perceptron to obtain a matching result; The household registration address relationship is determined according to the matching result.

6. The method according to claim 3, characterized in that Before using the address classification model to identify the types of the first registered residence address and the second registered residence address, the method further includes: Obtain household registration address samples and their corresponding address type labels; Using the initial address classification model to predict the address type of the household registration address sample, to obtain a predicted address type; Determining a second loss function value according to the similarity between the predicted address type and the address type label; The model parameters are adjusted according to the second loss function value, and the address classification model is obtained through training.

7. The method according to claim 1, characterized in that The first attribute information includes a first surname, the second attribute information includes a second surname, and before using the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship, and the communication relationship, the method further includes: Determining a surname relationship between the first user and the second user according to the first surname and the second surname based on a surname graph, wherein the surname graph includes association relationships between a plurality of surnames; The using of the heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household address relationship, and the communication relationship includes: Using the heterogeneous graph neural network model, the family relationship between the first user and the second user is identified based on the first attribute information, the second attribute information, the household address relationship, the communication relationship and the surname relationship.

8. The method according to claim 7, characterized in that Before determining the surname relationship between the first user and the second user according to the first surname and the second surname based on the surname graph, the method further includes: Get last name data; Extracting surname elements from the surname data, wherein the surname elements include surname, family tree and clan tree; Determine the association relationship between the surnames according to the family tree and the clan tree; The surname graph is constructed with the surnames as nodes and the associations between the surnames as edges.

9. A device for identifying family relationships, characterized in that: The device comprises: A first acquisition module, used to acquire first attribute information of a first user, second attribute information of a second user, and a communication relationship between the first user and the second user, wherein the first attribute information includes a first registered permanent address, and the second attribute information includes a second registered permanent address; A first determining module, configured to determine a household registration address relationship between the first user and the second user according to the first household registration address and the second household registration address; The first identification module is used to use a heterogeneous graph neural network model to identify the family relationship between the first user and the second user based on the first attribute information, the second attribute information, the household registration address relationship and the communication relationship.

10. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the method for identifying family relationships as described in any one of claims 1-8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the method for identifying family relationships as described in any one of claims 1-8.

12. A computer program product, characterized in that When the instructions in the computer program product are executed by a processor of an electronic device, the electronic device executes the method for identifying family relationships as described in any one of claims 1-8.

Images