A method and device for searching adversarial communities in symbolic social networks

Abstract

The present invention discloses an adversarial community search method and device for symbolic social networks. The method comprises: obtaining a query vertex and a symbolic social network that satisfies a degree constraint, and for any vertex, calculating its follower set and the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; obtaining an independent set in the symbolic social network using a strategy that reduces the minimum eigenvalue the fastest; deleting the follower set of each vertex in the independent set, and if the connected component where the query vertex is located in the symbolic social network after deletion satisfies the balance condition, then initializing the adversarial community accordingly, otherwise repeating the above calculation and deletion process; adding the deleted vertices that do not affect the balance constraint and the degree constraint after being added back to the adversarial community back to the adversarial community to obtain the final adversarial community.

Description

Technical Field

[0001] The present invention relates to the field of social network analysis and mining, and in particular to an adversarial community search method and device for symbolic social networks. Background Art

[0002] With the development of social media, World Wide Web search and knowledge base, graph data management has received more and more attention, and one of its core tasks is community search. Unlike traditional social networks, symbolic social networks contain both positive and negative social relationships, which can describe synonymous and antonymous relationships between word networks. Given a query point, in order to find two adversarial communities containing the query point, a balance is introduced. The core requires that there is no positive social relationship between the two communities, no negative social relationship in a single community, and for each vertex in the two adversarial communities, there is at least vertices have active social relationships with them, at least vertices with whom they have negative social relations. Based on balance Kernel model,designing an adversarial community search method for symbolic social networks.

[0003] In the process of implementing the present invention, the inventors found that the prior art has at least the following problems:

[0004] First, the current community models in symbolic social networks generally only describe one community, rather than two adversarial communities. As a result, only the community where the query point is located can be found, but the community adversarial to the community cannot be found at the same time.

[0005] In addition, the adversarial community models in current symbolic social networks generally cannot guarantee both the closeness and adversarial nature of the community, and are therefore difficult to apply to adversarial community searches in actual social networks.

[0006] In addition, the constraints of community models in some current symbolic social networks are too strong, resulting in the fact that in real symbolic social networks, there are basically no subgraphs that meet the constraints of the community model. Summary of the invention

[0007] In view of the above technical problems and the shortcomings in the art, the present invention provides an adversarial community search method and device for symbolic social networks, which uses the relationship between the minimum eigenvalue of the Laplace matrix of the symbolic social network and its balance constraint to heuristically search for a subgraph that satisfies the balance constraint; uses the estimation formula of the minimum eigenvalue of the Laplace matrix of the symbolic social network to perform approximate calculations efficiently and accurately; and verifies whether the currently deleted vertices can be added back to the community to ensure the integrity of the community as much as possible.

[0008] A method for searching adversarial communities in symbolic social networks, comprising the following steps:

[0009] S11: Get symbolic social network and query vertex;

[0010] S12: Iteratively delete the vertices that meet any of the conditions A) and B) in the obtained symbolic social network to complete the degree constraint: A) the positive degree is less than the positive threshold; B) the negative degree is less than the negative threshold;

[0011] S13: for any vertex in the current symbolic social network, calculate its follower set, where the follower set is the set of the vertex and the points that are cascaded deleted due to the deletion of the vertex, and further calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; traverse all vertices in the current symbolic social network to obtain the minimum eigenvalues ​​corresponding to these vertices one by one;

[0012] S14: Acquire an independent set in the current symbolic social network by adopting a strategy that reduces the minimum eigenvalue at the fastest speed;

[0013] S15: deleting the follower set of each vertex in the independent set in the current symbolic social network, and taking the connected component where the query vertex is located in the deleted symbolic social network as the obtained symbolic social network;

[0014] S16: If the obtained symbolic social network satisfies the equilibrium condition, the adversarial community is initialized as the obtained symbolic social network; otherwise, the process jumps to step S13; the equilibrium condition is satisfied, which means that the vertex set of the obtained symbolic social network can be divided into two sets, so that the two endpoints of any positive edge are not in the two sets respectively, and the two endpoints of any negative edge cannot be in the same set;

[0015] S17: Check whether all the vertices deleted in step S15 can be added back to the confrontation community. If adding them back to the confrontation community does not affect the balance constraint and the degree constraint, then add them back to the confrontation community. After processing all the vertices deleted in step S15, the final confrontation community is obtained.

[0016] In some embodiments, in step S12, the positive degree of any vertex in the symbolic social network that completes the degree constraint is not less than a given positive threshold , the negative degree is not less than the given negative threshold .

[0017] In some embodiments, in step S13, for any vertex in the current symbolic social network, the positive degree list and the negative degree list in the current symbolic social network are copied to maintain the follower set of the vertex.

[0018] In some embodiments, in step S13, the step of calculating the follower set of any vertex in the current symbolic social network specifically includes:

[0019] (1) For any vertex Follow Collection Initialize the vertex Add to follow collection In the vertex The positive neighbor node of the vertex is decremented by 1 in the positive degree list. The negative neighbor node of , reduces its negative degree by 1 in the negative degree list;

[0020] (2) The vertices The number of positive neighbor nodes with positive degree less than The number of vertices and negative neighbor nodes with negative degrees less than The set of vertices Add to follow collection middle;

[0021] (3) Traversing a collection , for any vertex , if the vertex First time being added to the following set , then repeat with the vertex The same operation is repeated until the vertex currently visited With Vertex The number of hops between them reaches the threshold, and the final follow-up set is obtained .

[0022] In some embodiments, in step S13, the minimum eigenvalue upper bound is used to calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the following set. , the calculation formula is:

[0023] ,

[0024] in, represents the vertex index, and Deleting any vertex Follow Collection The smallest eigenvalue and corresponding unit eigenvector of the Laplacian matrix of the pre-symbolic social network, , , Respectively represent the unit eigenvector with index , , Elements of and Respectively represent the index The neighbor set and number of neighbors of the vertex, represents the Laplacian matrix Row, No. Elements of a column.

[0025] In some embodiments, step S14 specifically includes the steps of:

[0026] S41: Combine all the minimum eigenvalues ​​obtained in step S13 into one Dimensional vector , is the number of vertices in the current symbolic social network, and selects the vertices corresponding to The vertex corresponding to the one-dimensional with the smallest median value , the independent set Initially, it contains only one element, the vertex ;

[0027] S42: from the current symbol social network according to the corresponding vector The elements in the set are selected in order from small to large and are independent sets The vertices without edges are added to the independent set , forming the final independent set.

[0028] In some embodiments, the number of vertices in the independent set is 100.

[0029] The present invention also provides a computer device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, and when the computer program is run, the processor executes the adversarial community search method for symbolic social networks.

[0030] The present invention also provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a computer device, the computer device executes the adversarial community search method for symbolic social networks.

[0031] The present invention also provides a computer program product, including a computer program, which, when executed by a computer device, enables the computer device to execute the adversarial community search method for symbolic social networks.

[0032] The present invention also provides an adversarial community search device for symbolic social networks, comprising:

[0033] The acquisition module is used to obtain the symbolic social network and query vertices;

[0034] The preprocessing module is used to iteratively delete the vertices that meet any of the conditions A) and B) in the symbolic social network obtained by the acquisition module, and complete the degree constraint: A) the positive degree is less than the positive threshold; B) the negative degree is less than the negative threshold;

[0035] The calculation module is used to: I) calculate the follower set of any vertex in the current symbolic social network, where the follower set is the set of the vertex and the points that are cascaded deleted due to the deletion of the vertex, and further calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; traverse all vertices in the current symbolic social network to obtain the minimum eigenvalue corresponding to these vertices one by one; II) adopt the strategy that makes the minimum eigenvalue decrease the fastest to obtain an independent set in the current symbolic social network; III) delete the follower set of each vertex in the independent set in the current symbolic social network, and use the connected component where the query vertex is located in the deleted symbolic social network as the obtained symbolic social network; IV) if the obtained symbolic social network meets the balance condition, initialize the adversarial community as the obtained symbolic social network, otherwise, jump to I); the satisfying balance condition means: the vertex set of the obtained symbolic social network can be divided into two sets, so that the two endpoints of any positive edge are not in the two sets respectively, and the two endpoints of any negative edge cannot be in the same set;

[0036] The verification module is used to check whether all the vertices deleted in the calculation module III) can be added back to the adversarial community. If adding them back to the adversarial community does not affect the balance constraint and the degree constraint, they will be added back to the adversarial community. After processing all the vertices deleted in the calculation module III), the final adversarial community is obtained.

[0037] The adversarial community search device for symbolic social networks can execute the adversarial community search method for symbolic social networks when running.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention constructs an efficient and highly approximate adversarial community search method for symbolic social networks. The present invention uses the relationship between the minimum eigenvalue of the Laplace matrix of the symbolic social network and its equilibrium constraint to heuristically search for a subgraph that satisfies the equilibrium constraint; uses the estimation formula of the minimum eigenvalue of the Laplace matrix of the symbolic social network to perform approximate calculations efficiently and accurately; verifies whether the currently deleted vertices can be added back to the community to ensure the integrity of the community as much as possible. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a flowchart of an adversarial community search method for symbolic social networks introduced in a specific implementation plan.

[0041] Figure 2 for Figure 1 The adversarial community search method for symbolic social networks shown in step S13 follows the set calculation flow chart.

[0042] Figure 3 for Figure 1 The specific flow chart of step S14 of the adversarial community search method for symbolic social networks is shown.

[0043] Figure 4 It is a schematic diagram of an adversarial community search device for symbolic social networks introduced in a specific implementation scheme. DETAILED DESCRIPTION

[0044] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0045] Combination Figure 1 , an adversarial community search method for symbolic social networks, comprising the steps of:

[0046] S11: Get symbolic social network and query vertex;

[0047] S12: Iteratively delete the vertices that meet any of the conditions A) and B) in the obtained symbolic social network to complete the degree constraint: A) the positive degree is less than the positive threshold; B) the negative degree is less than the negative threshold;

[0048] S13: for any vertex in the current symbolic social network, calculate its follower set, where the follower set is the set of the vertex and the points that are cascaded deleted due to the deletion of the vertex, and further calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; traverse all vertices in the current symbolic social network to obtain the minimum eigenvalues ​​corresponding to these vertices one by one;

[0049] S14: Acquire an independent set in the current symbolic social network by adopting a strategy that reduces the minimum eigenvalue at the fastest speed;

[0050] S15: deleting the follower set of each vertex in the independent set in the current symbolic social network, and taking the connected component where the query vertex is located in the deleted symbolic social network as the obtained symbolic social network;

[0051] S16: If the obtained symbolic social network satisfies the equilibrium condition, the adversarial community is initialized as the obtained symbolic social network; otherwise, the process jumps to step S13; the equilibrium condition is satisfied, which means that the vertex set of the obtained symbolic social network can be divided into two sets, so that the two endpoints of any positive edge are not in the two sets respectively, and the two endpoints of any negative edge cannot be in the same set;

[0052] S17: Check whether all the vertices deleted in step S15 can be added back to the confrontation community. If adding them back to the confrontation community does not affect the balance constraint and the degree constraint, then add them back to the confrontation community. After processing all the vertices deleted in step S15, the final confrontation community is obtained.

[0053] In step S11, each of the symbolic social networks is a tuple including a positive social relationship and a negative social relationship, and each of the query vertices is a vertex in the symbolic social network.

[0054] For example, the present invention can be applied to multiple practical fields such as word retrieval. The present invention can represent the symbolic social network as ,in, is a set of nodes, representing the set of users in the symbolic social network. is a set of edge sets, representing the social relationships between users. The edge set can be divided into ,in Contains positive edges, i.e., positive social relationships between users, Contains negative edges, that is, negative social relationships between users. Specifically, in symbolic social networks, there are two types of social relationships between users: support and opposition. If a user and users Mutual support can be and users The social relationship is represented by a positive edge , if the user and users If they oppose each other, users and users The social relationship is represented as a negative edge In addition, adversarial communities cover multiple aspects in social networks, including synonyms and antonyms of words. For a word in the dictionary, its synonyms and antonyms constitute a community respectively, and there is an adversarial relationship between the two communities.

[0055] In a specific implementation of step S12, vertices whose positive degrees are less than a positive threshold or whose negative degrees are less than a negative threshold in the symbolic social network may be iteratively deleted according to the positive degrees and negative degrees of the vertices in the current symbolic social network.

[0056] In step S12, the positive degree of any vertex in the symbolic social network that completes the degree constraint is not less than a given positive threshold , the negative degree is not less than the given negative threshold .

[0057] For example, in order to quickly delete points that do not meet the degree constraint, maintain the positive degree list and negative degree list of all vertices in the current symbolic social network, and then calculate the list of vertices whose positive and negative degrees meet the degree constraint. If all vertices are in the list, exit the loop, otherwise update the current symbolic social network to the derived subgraph of the vertices in the list in the original network, and then update the positive degree list and negative degree list, delete the degrees corresponding to the deleted vertices, and adjust the degrees of the remaining vertices, and iterate.

[0058] In step S13, for any vertex in the current symbolic social network, the positive degree list and the negative degree list in the current symbolic social network may be copied to maintain the follower set of the vertex.

[0059] Combination Figure 2 In step S13, the calculation of the follower set of any vertex in the current symbolic social network specifically includes:

[0060] (1) For any vertex Follow Collection Initialize the vertex Add to follow collection In the vertex The positive neighbor node of the vertex is decremented by 1 in the positive degree list. The negative neighbor node of , reduces its negative degree by 1 in the negative degree list;

[0061] (2) The vertices The number of positive neighbor nodes with positive degree less than The number of vertices and negative neighbor nodes with negative degrees less than The set of vertices Add to follow collection In the operation ;

[0062] (3) Traversing a collection , for any vertex , if the vertex First time being added to the following set , then repeat with the vertex The same operation is repeated until the vertex currently visited With Vertex The number of hops between them reaches a threshold (for example, it can be set to 8), and the final follow-up set is obtained. .

[0063] For example, vertices can be constructed in a loop Follow Collection , first check the vertex All positive neighbor vertices of and all negative neighbor vertices Can After being deleted, the degree constraint is satisfied, and then the current follower set is Any vertex in Perform the same operation on the final follow-up set Any vertex of , which is related to the vertex The previous hop count should not be greater than the set threshold ,Right now .

[0064] In step S13, the minimum eigenvalue upper bound can be used to calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the following set , the calculation formula is:

[0065] ,

[0066] in, represents the vertex index, and Deleting any vertex Follow Collection The smallest eigenvalue and corresponding unit eigenvector of the Laplacian matrix of the pre-symbolic social network, , , Respectively represent the unit eigenvector with index , , Elements of and Respectively represent the index The neighbor set and number of neighbors of the vertex, represents the Laplacian matrix Row, No. Elements of a column.

[0067] In the present invention, for any vertex in the symbolic social network , inaccurate calculation deletes its follow-up set After that, the minimum eigenvalue of the Laplace matrix of the symbolic social network is not obtained, but the estimation formula is used to estimate the value, which can improve the operation efficiency of the method.

[0068] Combination Figure 3, step S14 specifically includes the steps of:

[0069] S41: Combine all the minimum eigenvalues ​​obtained in step S13 into one Dimensional vector , is the number of vertices in the current symbolic social network, and selects the vertices corresponding to The vertex corresponding to the one-dimensional with the smallest median value , the independent set Initially, it contains only one element, the vertex , ,vertex The number meets ;

[0070] S42: from the current symbol social network according to the corresponding vector The elements in the set are selected in order from small to large and are independent sets The vertices without edges are added to the independent set , forming the final independent set.

[0071] In the present invention, for any vertex in each symbolic social network , whose number is , then its corresponding vector element Assign to delete vertex The smallest eigenvalue of the Laplacian matrix of the symbolic social network after the set of follows.

[0072] Furthermore, a heuristic strategy can be adopted to select an independent set in the symbolic social network with the strategy that reduces the eigenvalue the fastest, that is, ( To set the quantity, you can usually set , that is, the number of vertices in the independent set is 100) rounds of selection, so that the vertices selected in each round have no edges with the existing vertices in the independent set, and are the vertices that meet this condition in the vector The one with the smallest corresponding value in .

[0073] For example, in During the round update operation, first correspond the positive neighbors and negative neighbors of all vertices in the current independent set to the vector Set the elements in to infinity, and then select the vector The smallest component in , adds the vertex corresponding to the number to the independent set.

[0074] In step S15, in order to overcome the challenge that after deleting a vertex in the symbolic social network, the follower sets of other vertices change accordingly, before deleting, the list of vertices to be deleted is pre-calculated. First, the list of vertices is initialized to the independent set obtained in S14, and then for any vertex in it, a queue is established containing only the vertex, and the following operations are repeated until the queue is empty: pop a vertex from the queue , check the vertex Can the positive and negative neighbors of After the degree constraint is satisfied, if not, the vertex is added to the deletion list and queue. After processing all vertices in the independent set, select the connected component where the query vertex is located in the symbolic social network. The method is to iteratively find the vertex set in the connected component, first initialize the set to the query vertex and its positive neighbors and negative neighbors, and then perform a deletion on any vertex in the set. , and do the same thing until the cardinality of the set no longer changes.

[0075] In step S16, it is checked whether the symbolic social network satisfies the equilibrium condition, that is, the vertex set of the symbolic social network is Divide into two sets and , so that the edge set The two endpoints of any positive edge in the set are not in two sets respectively, and the two endpoints of any negative edge cannot be in the same set. Any positive edge ,satisfy ,right Any negative edge ,satisfy ;

[0076] For example, the adjacency matrix of the symbolic social network can be used, and the matrix and vector operation methods can be combined to judge the balance of the symbolic social network to improve the calculation efficiency.

[0077] In step S16, if the equilibrium condition is not met, it is necessary to continue to delete the vertices in the symbolic social network under the degree constraint according to the equilibrium condition, so jump to step S13; otherwise, the current symbolic social network has met the equilibrium condition and the degree constraint, and the output adversarial community is initialized as the current symbolic social network.

[0078] In step S17, for example, the degree constraint may refer to the vertex having at least positive neighbors and Negative neighbors, , is the input threshold.

[0079] In the community search field of symbolic social networks, two opposing communities on symbolic social networks can be searched by the method of the present invention. For example, in an electronic dictionary, there are a large number of words, and the relationship between these words may be synonyms or antonyms. When a user searches for a word, he may want to get its synonyms and antonyms. The present invention can efficiently and accurately find the synonyms and antonyms of a word.

[0080] A computer device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, and when the computer program is run, the processor executes the adversarial community search method for symbolic social networks.

[0081] A computer-readable storage medium stores a program or instruction, and when the program or instruction is executed by a computer device, the computer device executes the adversarial community search method for symbolic social networks.

[0082] A computer program product includes a computer program. When the computer program is executed by a computer device, the computer device is enabled to execute the adversarial community search method for symbolic social networks.

[0083] Combination Figure 4 , an adversarial community search device for symbolic social networks, wherein adversarial communities are two opposing communities in the symbolic social network and contain query vertices, and the adversarial community search device for symbolic social networks comprises:

[0084] An acquisition module 21, used to acquire a symbolic social network and query vertices;

[0085] The preprocessing module 22 is used to iteratively delete the vertices satisfying any of the conditions A) and B) in the symbolic social network obtained by the obtaining module 21, and complete the degree constraint: A) the positive degree is less than the positive threshold; B) the negative degree is less than the negative threshold;

[0086] The calculation module 23 is used to: I) calculate the follower set of any vertex in the current symbolic social network, where the follower set is the set of the vertex and the points that are cascaded deleted due to the deletion of the vertex, and further calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; traverse all vertices in the current symbolic social network to obtain the minimum eigenvalue corresponding to these vertices one by one; II) adopt the strategy that makes the minimum eigenvalue decrease the fastest to obtain an independent set in the current symbolic social network; III) delete the follower set of each vertex in the independent set in the current symbolic social network, and use the connected component where the query vertex is located in the deleted symbolic social network as the obtained symbolic social network; IV) if the obtained symbolic social network satisfies the balance condition, initialize the adversarial community as the obtained symbolic social network, otherwise, jump to I); satisfying the balance condition means: being able to divide the vertex set of the obtained symbolic social network into two sets, so that the two endpoints of any positive edge are not in the two sets respectively, and the two endpoints of any negative edge cannot be in the same set;

[0087] The verification module 24 is used to check whether all the vertices deleted in the calculation module 23 III) can be added back to the confrontation community. If adding them back to the confrontation community does not affect the balance constraint and the degree constraint, they are added back to the confrontation community. After all the vertices deleted in the calculation module 23 III) are processed, the final confrontation community is obtained.

[0088] The above-mentioned adversarial community search device for symbolic social networks can execute the above-mentioned adversarial community search method for symbolic social networks when running.

[0089] The adversarial community search device for symbolic social networks provided in the above specific implementation manner and the method flow embodiment executed by it belong to the same concept. The specific implementation process is detailed in the above method and flow introduction, which will not be repeated here.

[0090] In addition, it should be understood that after reading the above description of the present invention, those skilled in the art may make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the claims attached to this application.

Claims

1. A method for searching adversarial communities in symbolic social networks, characterized in that: The symbolic social network includes both positive social relations and negative social relations, which describe synonymy and antonymy relations between word networks; the adversarial community includes synonyms and antonyms of words; for a word in the dictionary, its synonyms and antonyms respectively constitute a community, and there is an adversarial relationship between the two communities; The adversarial community search method for symbolic social networks is used for word retrieval, and comprises the steps of: S11: Get symbolic social network and query vertex; S12: Iteratively delete the vertices that meet any of the conditions A) and B) in the obtained symbolic social network to complete the degree constraint: A) the positive degree is less than the positive threshold; B) the negative degree is less than the negative threshold; S13: for any vertex in the current symbolic social network, calculating its follower set, where the follower set is the set of the vertex and the points that are cascade deleted due to the deletion of the vertex, and further calculating the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; Traversing all vertices of the current symbolic social network to obtain minimum eigenvalues ​​corresponding to these vertices one by one; S14: Acquire an independent set in the current symbolic social network by adopting a strategy that reduces the minimum eigenvalue at the fastest speed; S15: deleting the follower set of each vertex in the independent set in the current symbolic social network, and taking the connected component where the query vertex is located in the deleted symbolic social network as the obtained symbolic social network; S16: If the obtained symbolic social network satisfies the equilibrium condition, the adversarial community is initialized as the obtained symbolic social network; otherwise, the process jumps to step S13; the equilibrium condition is satisfied, which means that the vertex set of the obtained symbolic social network can be divided into two sets, so that the two endpoints of any positive edge are not in the two sets respectively, and the two endpoints of any negative edge cannot be in the same set; S17: Check whether all the vertices deleted in step S15 can be added back to the confrontation community. If adding them back to the confrontation community does not affect the balance constraint and the degree constraint, then add them back to the confrontation community. After processing all the vertices deleted in step S15, the final confrontation community is obtained.

2. The adversarial community search method for symbolic social networks according to claim 1, characterized in that: In step S12, the positive degree of any vertex in the symbolic social network that completes the degree constraint is not less than a given positive threshold , the negative degree is not less than the given negative threshold .

3. The adversarial community search method for symbolic social networks according to claim 2, characterized in that: In step S13, for any vertex in the current symbolic social network, a positive degree list and a negative degree list in the current symbolic social network are copied to maintain a follower set of the vertex; In step S13, the calculation of the follower set of any vertex in the current symbolic social network specifically includes: (1) For any vertex Follow Collection Initialize the vertex Add to follow collection In the vertex The positive neighbor node of the vertex is decremented by 1 in the positive degree list. The negative neighbor node of , reduces its negative degree by 1 in the negative degree list; (2) The vertices The number of positive neighbor nodes with positive degree less than The number of vertices and negative neighbor nodes with negative degrees less than The set of vertices Add to follow collection middle; (3) Traversing a collection , for any vertex , if the vertex First time being added to the following set , then repeat with the vertex The same operation is repeated until the vertex currently visited With Vertex The number of hops between them reaches the threshold, and the final follow-up set is obtained .

4. The adversarial community search method for symbolic social networks according to claim 1, characterized in that: In step S13, the minimum eigenvalue upper bound is used to calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the following set. , the calculation formula is: ， in, represents the vertex index, and Deleting any vertex Follow Collection The smallest eigenvalue and corresponding unit eigenvector of the Laplacian matrix of the pre-symbolic social network, , , They represent the unit eigenvectors with indices , , Elements of and Respectively represent the index The neighbor set and number of neighbors of the vertex, represents the Laplacian matrix Row, No. Elements of a column.

5. The adversarial community search method for symbolic social networks according to claim 1, characterized in that: Step S14 specifically includes the following steps: S41: Combine all the minimum eigenvalues ​​obtained in step S13 into one Dimensional vector , is the number of vertices in the current symbolic social network, and selects the vertices corresponding to The vertex corresponding to the one-dimensional with the smallest median value , the independent set Initially, it contains only one element, the vertex ; S42: from the current symbol social network according to the corresponding vector The elements in the set are selected in order from small to large and are independent sets The vertices without edges are added to the independent set , forming the final independent set.

6. The adversarial community search method for symbolic social networks according to claim 1 or 5, characterized in that: The number of vertices in the independent set is 100.

7. A computer device comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, wherein: When the computer program is executed, the processor executes the adversarial community search method for symbolic social networks as described in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program or instruction, and when the program or instruction is executed by a computer device, the computer device executes the adversarial community search method for symbolic social networks as described in any one of claims 1 to 6.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a computer device, the computer device executes the adversarial community search method for symbolic social networks as described in any one of claims 1 to 6.

10. An adversarial community search device for symbolic social networks, characterized in that: The symbolic social network includes both positive social relations and negative social relations, which describe synonymy and antonymy relations between word networks; the adversarial community includes synonyms and antonyms of words; for a word in the dictionary, its synonyms and antonyms respectively constitute a community, and there is an adversarial relationship between the two communities; The adversarial community search device for symbolic social networks includes: The acquisition module is used to obtain the symbolic social network and query vertices; The preprocessing module is used to iteratively delete the vertices that meet any of the conditions A) and B) in the symbolic social network obtained by the acquisition module, and complete the degree constraint: A) the positive degree is less than the positive threshold; B) the negative degree is less than the negative threshold; The calculation module is used to: I) calculate the follower set of any vertex in the current symbolic social network, where the follower set is the set of the vertex and the points that are cascaded deleted due to the deletion of the vertex, and further calculate the minimum eigenvalue of the Laplacian matrix of the symbolic social network after deleting the follower set; traverse all vertices in the current symbolic social network to obtain the minimum eigenvalue corresponding to these vertices one by one; II) adopt the strategy that makes the minimum eigenvalue decrease the fastest to obtain an independent set in the current symbolic social network; III) delete the follower set of each vertex in the independent set in the current symbolic social network, and use the connected component where the query vertex is located in the deleted symbolic social network as the obtained symbolic social network; IV) if the obtained symbolic social network meets the balance condition, initialize the adversarial community as the obtained symbolic social network, otherwise, jump to I); the satisfying balance condition means: the vertex set of the obtained symbolic social network can be divided into two sets, so that the two endpoints of any positive edge are not in the two sets respectively, and the two endpoints of any negative edge cannot be in the same set; The verification module is used to check whether all the vertices deleted in the calculation module III) can be added back to the adversarial community. If adding them back to the adversarial community does not affect the balance constraint and the degree constraint, they will be added back to the adversarial community. After processing all the vertices deleted in the calculation module III), the final adversarial community is obtained.

Images