Account management method, electronic device, storage medium and computer program product

Abstract

The embodiment of the invention relates to the technical field of computers, and discloses an account management method, electronic equipment, a storage medium and a computer program product. The account management method comprises the following steps: clustering content data of a plurality of accounts, and constructing an association relationship between the accounts and the content data; determining at least one target group in the association relationship based on a group discovery algorithm; performing semantic analysis on the content data to identify a group category of the target group; and managing the account of the target group according to a processing strategy corresponding to the group category. The beneficial effects of the invention are that the method can achieve the precise recognition of the group behavior of the account, and achieves the targeted management of the account.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of computer, in particular to an account management method, an electronic device, a storage medium and a computer program product. BACKGROUND

[0002] At present, an Internet platform usually manages accounts based on analysis of the accounts to improve the rationality of management, such as personalized content recommendation, permission adjustment and other management for normal accounts.

[0003] In the process of Internet user activity, homogenization behavior may occur, and then an account group is formed. In related technologies, accounts are identified by operation links, account relationship chains or behavior log similarity, and subsequent unified management is performed according to the identified groups. However, group identification based on operation links, account relationship chains or behavior log similarity is insufficient in group profiling, and the subsequent account management is often not targeted, which affects user experience. SUMMARY

[0004] Embodiments of the present application aim to provide an account management method, an electronic device, a storage medium and a computer program product, which can accurately identify group behavior of accounts and then manage the accounts in a targeted manner.

[0005] To solve the above technical problems, embodiments of the present application provide the following technical solutions: In a first aspect, an account management method is provided, including: clustering content data of a plurality of accounts and constructing an association relationship between the accounts and the content data; determining at least one target group in the association relationship based on a group discovery algorithm; performing semantic analysis on the content data to identify a group category of the target group; and managing the accounts of the target group according to a processing strategy corresponding to the group category.

[0006] In a second aspect, an electronic device is provided, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described above.

[0007] In a second aspect, a computer readable storage medium is provided, and the computer readable storage medium stores computer instructions, which are executed by a processor to implement the method described above.

[0008] In a fourth aspect, a computer program product is provided, including a computer program, which is executed by a processor to implement the steps of the method described above.

[0009] The account management method in the embodiments of the present application clusters content data of multiple accounts, constructs an association relationship between the accounts and the content data, determines at least one target group in the association relationship based on a group discovery algorithm, performs semantic analysis on the content data to identify a group category of the target group, and manages the accounts of the target group according to a processing strategy corresponding to the group category. The account management method in the embodiments of the present application clusters content data of multiple accounts to construct an association relationship between accounts and content data, and then determines an account group that may have a specific association as a target group based on a group discovery algorithm. Subsequently, the target group is more accurately determined in terms of a group category based on content data sent by the accounts, for example, the accounts of different categories can be accurately divided based on the content published by the accounts. Then, different processing strategies are configured for the target groups of different categories for management, so as to improve the accuracy and rationality of account management, and thus ensure user experience in the platform. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 Flowchart of the account management method provided by an embodiment of the present application Figure 1 ; Figure 2 Flowchart of the account management method provided by an embodiment of the present application Figure 2 ; Figure 3 Flowchart of the account management method provided by an embodiment of the present application Figure 3 ; Figure 4 Flowchart of the account management method provided by an embodiment of the present application Figure 4 ; Figure 5 Flowchart of the account management method provided by an embodiment of the present application Figure 5 ; Figure 6 Structural block diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0011] In the process of active Internet users, homogenization behavior may occur, and then an account group is formed. In related technologies, accounts are identified by operation links, account relationship chains, or behavior log similarity, and then the identified groups are uniformly managed. However, group identification based on operation links, account relationship chains, or behavior log similarity is insufficient in terms of group portrait, and the account management is often not targeted, for example: In the process of personalized content recommendation for normal accounts, the content of interest of the above-identified group may not be uniform, and therefore the personalized content pushed may not meet the actual needs of the accounts in the group, thereby reducing the user experience.

[0012] In the platform, in addition to normal accounts, there may also be abnormal accounts, or risk accounts that affect the operation of the platform or the user experience, for example, accounts related to groups, based on platform accounts, through content submission, private chat information, bullet screen, and comment reply, etc. In the process of content management and account management of accounts that may have risks, management errors are prone to occur, for example, normal group accounts mastered by studios or individuals of the same operation link are mistakenly injured, and due to the lack of analysis of the content published by the account, the one-size-fits-all management method causes the average allocation of management resources, resulting in insufficient impact on high-risk behavior, while too many restrictions on low-risk or normal users, so as to adapt to complex scenarios, seriously affecting the user experience.

[0013] To solve the above problems, the embodiments of the present application provide an account management method, an electronic device, a storage medium and a computer program product. The account management method comprises: clustering content data of a plurality of accounts, and constructing an association relationship between the accounts and the content data, determining at least one target group in the association relationship based on a group discovery algorithm; performing semantic analysis on the content data to identify a group category of the target group; and managing the accounts of the target group according to a processing strategy corresponding to the group category. The account management method in the embodiments of the present application clusters the content data of a plurality of accounts to construct the association relationship between the accounts and the content data, and then determines the account group that may have a specific association as the target group based on the group discovery algorithm. Subsequently, the target group is more accurately determined based on the content data sent by the account, such as the content published by the account to accurately divide the account group of different categories, and then different processing strategies are configured based on the target group of different categories for management, so as to improve the accuracy and rationality of account management, and thereby guarantee the user experience in the platform.

[0014] To make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the drawings. However, those skilled in the art can understand that in the embodiments of the present application, many technical details are proposed in order to make the readers better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed by the present application can be implemented. The division of the following embodiments is for the convenience of description, and should not constitute any limitation on the specific implementation mode of the present application. The embodiments can be combined and referenced with each other without contradiction.

[0015] It can be understood that in the specific embodiments of the present application, the collection, use or processing of data requires the permission or consent of the data subject when the embodiments of the present application are applied to specific products or technical implementations, and the collection, use or processing of relevant data must comply with relevant laws, regulations and standards of the data source and implementation location, and through desensitization technology to ensure that the final use is desensitized data that has been safely processed, and the rights and interests of the data subject and data security are protected.

[0016] Referring to Figure 1 An embodiment of the present application relates to an account management method, comprising: Step 100, clustering content data of multiple accounts, and constructing an association relationship between the accounts and the content data.

[0017] Among them, the content data of the account can be the content sent by the account, for example, the content sent by the account can be video content, audio content, article content, reply content and barrage content, etc.

[0018] These data can be preprocessed after being obtained, such as text coding, invalid character filtering, fuzzy word normalization, etc., and after data cleaning, the data is used as standardized data for subsequent processing. In an optional embodiment, only Chinese and English characters and necessary information can be retained for subsequent processing. For video content or audio content, etc. Content data, text recognition can be used for text extraction for subsequent processing.

[0019] Similar or identical content data sent by different accounts is divided into the same class, which can be achieved by clustering methods such as divisive clustering, hierarchical clustering and network structure clustering, and then the association relationship is constructed, that is, the corresponding relationship between each account and the clustered content data. Among them, each account can correspond to multiple clustered content data, and each clustered content data can correspond to multiple accounts sending these content data.

[0020] In an optional embodiment, the association relationship between the account and the content data is constructed, comprising: A bipartite graph or projection graph constructed by the content data and the account as graph nodes.

[0021] For example, for an account, the account ID can be taken as a type node, and the clustered content data can be taken as another type node, thereby establishing a bipartite graph structure, and then based on the association between the nodes in the bipartite graph structure, the above-mentioned association relationship is obtained. Further, the frequency of the account node connecting the content data node can also be used to construct an account projection graph to determine the structural association between accounts as the above-mentioned association relationship.

[0022] In step 200, at least one target group is determined in the association relationship based on a group discovery algorithm.

[0023] The association relationship between the clustered content data of the multiple accounts and the accounts can determine that the accounts with specific associations between the accounts can be regarded as accounts in a target group, that is, the determined target group is an account group with associated content data or similar content data, thereby mining the organized behavior between the accounts with content as the association. For example, if the content sent by multiple accounts is similar or the same, these accounts can be determined as the same target group. In other embodiments, the content sent by the accounts and the attributes of the accounts can be combined to determine which accounts in these accounts belong to a target group.

[0024] For the association relationship between the accounts and the clustered content data, such as the bipartite graph or the projection graph, a group discovery algorithm can be used according to actual needs, such as Louvain method, label propagation algorithm, Infomap, hierarchical clustering, or graph neural network, to identify the group behavior of the accounts, and further determine at least one target group. The group identified by the group discovery algorithm can be directly used as the target group, or can be further evaluated as the target group in the embodiments of the present application.

[0025] The association relationship between the accounts and the clustered content data is identified to determine a group structure composed of multiple accounts.

[0026] In an optional embodiment of the present application, the identification of the association relationship between the accounts and the clustered content data to determine a group structure composed of multiple accounts includes: The association relationship between the accounts and the clustered content data is identified based on a group discovery algorithm to determine a group structure composed of multiple accounts.

[0027] In step 300, the content data is subjected to semantic analysis to identify a group category of the target group.

[0028] The content data subjected to semantic analysis can be the content data sent by the accounts in each determined target group. The content data can be the data text formed by the video content, audio content, article content, reply content, and bullet screen content. The content data can be subjected to a deep semantic mining process based on natural language processing technology, which can capture the semantic association and theme tendency implied in the text.

[0029] In an optional embodiment, the content data can be semantically analyzed based on an artificial intelligence model such as a large language model, and a category label can be generated as a group category. The determined group category can include a plurality of different normal user categories (for example, categories of content of interest), and a plurality of categories that can have adverse effects on other users.

[0030] Step 400: managing the accounts of the target group according to the processing strategy corresponding to the group category.

[0031] The determined group category is applied to the management of the target group, wherein a corresponding processing strategy can be preset for different group categories. For example, for a target group composed of different types of normal accounts, a content recommendation strategy for recommending different types of content to the accounts in the group can be preset to improve user experience. For example, for a target group composed of different types of risk accounts, a content filtering strategy for intercepting or limiting the content published by the accounts in the target group, a risk control strategy for marking the target group or the accounts in the target group for close observation, an account management strategy for managing the permissions of the target group or the accounts in the target group or banning them, and a reminding strategy for generating prompt information for the accounts to remind other users, and the like. Based on the determined group category, different processing strategies are assigned to the management of the target group, thereby realizing standardized hierarchical management, which is different from the traditional "one-size-fits-all" unified management, and has category adaptability and precision.

[0032] Thus, the account management method in the embodiments of the present application clusters the content data of a plurality of accounts, constructs the association relationship between the accounts and the content data, determines at least one target group in the association relationship based on a group discovery algorithm, performs semantic analysis on the content data to identify the group category of the target group, and manages the accounts of the target group according to the processing strategy corresponding to the group category. The account management method in the embodiments of the present application clusters the content data of a plurality of accounts to construct the association relationship between the accounts and the content data, and then determines a group of accounts that can have a specific association as a target group based on a group discovery algorithm. Subsequently, the target group is more accurately determined in terms of group category based on the content data sent by the accounts, such as the content published by the accounts, so as to accurately divide a group of accounts of different categories, and then different processing strategies are configured for the management of target groups of different categories to improve the accuracy and rationality of account management, thereby ensuring the user experience in the platform.

[0033] In an optional embodiment of the present application, the account management method further includes the steps of: Filtering the repeatedly published content of each account based on a preset frequency to use the repeatedly published content as the content data.

[0034] Generally, for the account existing in the form of a group, it usually repeatedly publishes similar content, for example, for the account of bad information, it usually repeatedly sends text in different article comment areas, etc. In this embodiment, the content published by the account is screened, specifically, the content repeatedly published by the account is screened based on a preset frequency, only the content repeatedly published by the account is taken as the content data of the account, so as to perform data dimension reduction processing at the content level, to avoid the interference of irrelevant content, accurately identify the account group, especially the account and the group formed by the account that may exist risks, and improve the processing efficiency and reduce the equipment cost in subsequent analysis.

[0035] In the above embodiment, the content data sent by the account is preprocessed such as normalization and data cleaning, to improve the consistency of the data and avoid interference items. In this embodiment, the screening of the content repeatedly published by the account can be based on the preprocessed data and the preset frequency.

[0036] The preset frequency can be set according to actual needs, for example, the preset frequency is 2, that is, the user publishes the same content twice to determine that it is repeatedly published content, which is taken as content data. In other embodiments, the preset frequency can be set to a larger value, so that the user publishes the same content multiple times to be considered as repeatedly published content.

[0037] Referring to Figure 2 In an optional embodiment of the present application, the group discovery algorithm determines at least one target group in the association relationship, including: Step 201, identifying the association relationship based on a group discovery algorithm to determine a group structure composed of a plurality of accounts; Step 202, determining whether the group structure is the target group according to the attribute index of the account.

[0038] In this embodiment, the association relationship determined based on the content data and the attribute index of the account jointly determine the target group. The content data includes data extracted from the content sent by the account as described above. The attribute index can be other data other than the content sent by the account, such as account attention, account level, registration time, registration address, login address, login time, login method and identity information of the login person, etc.

[0039] Compared with identifying the group only by the operation link, the account relationship chain or the behavior log similarity, the group of accounts is identified by the content data of the account, which can better identify the frequently occurring repeated content or variant content sent by the text template, and further more accurately determine the target group, to realize the management of the account that may exist risks.

[0040] Wherein, the preliminary identification of the group structure is performed through the association relationship constructed by the content data and the group discovery algorithm, and then whether the group structure is the target group expected to be managed in the application is judged through the attribute index, so as to improve the accuracy of management and prevent false management. For example, when the account group with possible risks is managed, different accounts in the comment area of a certain article post similar replies, and based on the association relationship, it is identified that these accounts may be located in the same group structure, and then based on the attribute index, it is judged, for example, the registration time of these accounts is consistent, so the risk degree of the group is further improved, and then the group structure is determined as the target group. The specific judgment method can be set according to actual needs, which is not absolutely limited here.

[0041] Referring to Figure 3 In an optional embodiment of the application, the clustering of the content data of the plurality of accounts comprises: Step 111, encoding the content data.

[0042] In the embodiment of the application, the high-dimensional content data is mapped to a short binary fingerprint (i.e. SimHash value) through a local sensitive hashing algorithm such as SimHash. Specifically, first, the content data is segmented, and each word is assigned a weight. Each word is mapped to a fixed-length binary hash value (e.g. 64 bits) through a standard hash function (such as MD5). The above weight is applied to the corresponding hash value, which can be specifically: for the bit of the hash value of 1, add the weight of the word; for the bit of the hash value of 0, subtract the weight of the word. Add all the weighted vectors bit by bit to obtain a comprehensive feature vector representing the entire document. Then, the final feature vector is dimensionally reduced. For each bit of the vector, if the bit value is greater than 0, generate 1; if less than or equal to 0, generate 0. Finally, this binary string (such as "1101...") is obtained, which is the SimHash encoding of the content data.

[0043] It can be understood that the data obtained by encoding similar content data is also similar.

[0044] In other embodiments, other encoding methods can also be used to encode the content data, for example, the MinHash algorithm can be used to encode the content data.

[0045] Step 121, the encoded content data is divided into buckets, wherein each bucket includes a data segment of the plurality of encoded content data.

[0046] Specifically, the complete SimHash code (such as 64 bits) is evenly divided into several pieces for bucketing, and finally each bucket includes multiple data pieces, each of which belongs to one content data.

[0047] In step 131, the difference distance between the data pieces in the bucket is determined to determine whether the content data is in the same category according to the difference distance.

[0048] In this embodiment, by comparing the difference distance between the data pieces in the bucket, if the difference distance is less than a set threshold, it indicates that the two data pieces are similar, and thus it can be determined that the two content data to which the data pieces belong are in the same category, thereby realizing clustering of the content data. For content data encoded by SimHash and subjected to bucketing, the difference distance can be a hash distance such as Hamming distance to make the above determination.

[0049] In other embodiments, for content data encoded by MinHash and subjected to bucketing, the difference distance can be a hash distance such as Jaccard distance or Hamming distance to make the above determination.

[0050] In more embodiments, other encoding methods can also be used for encoding of content data, which will not be described here.

[0051] Thus, by encoding and bucketing the content data for content classification, it can be realized without global data comparison, but only based on local data comparison in the bucket to realize content classification, thereby reducing memory overhead and computational cost, and requiring lower processing performance. Compared with the traditional content similarity algorithm and the cosine similarity method, which face the problems of memory explosion and high complexity when processing millions of texts, the account management method in the present application can reduce the real-time detection cost and processing delay, thereby better meeting the management needs of high-concurrency platforms.

[0052] Subsequently, the clustered content data can be used as nodes to obtain the association between the accounts and the clustered content data.

[0053] Referring to Figure 4 In an optional embodiment of the present application, determining whether the group structure is the target group according to the attribute index of the account includes: In step 112, the aggregation degree of the attribute index in the group structure is determined, where the aggregation degree of the attribute index includes the ratio of the number of accounts with similar attribute indexes in the group structure to the total number of accounts in the group structure. In step 122, it is determined whether the group structure is the target group according to the aggregation degree of the attribute index.

[0054] For a group structure preliminarily identified based on the association relationship between the account and the content data, further identification is performed based on the attribute indicators. For example, a large number of accounts in the group structure have similar or identical fingerprint indicators, which indicates that the group structure has a higher group behavior. Similarly, a large number of accounts have similar or identical account level indicators (such as all being newly opened low-level accounts), attention number indicators, and operation link indicators, which can reflect that the group structure has more group behaviors to some extent. Based on this, the embodiment of the present application constructs the aggregation degree of the attribute indicators based on these indicators to reflect the aggregation of the group structure. The aggregation degree can include the device fingerprint similarity aggregation degree, the operation link aggregation degree, the account level distribution aggregation degree, the attention number aggregation degree, and the behavior time window aggregation degree.

[0055] The aggregation degree includes the ratio of the number of accounts with similar attribute indicators in the group structure to the total number of accounts in the group structure. It can be understood that the higher the ratio, the higher the aggregation of the group structure with respect to the attribute indicators.

[0056] The aggregation degree can be used alone to determine whether the group structure is a target group. For example, if the fingerprint similarity aggregation degree is high, it can be determined that the same user controls multiple accounts to send the same content, and the group structure can be regarded as a target group. The aggregation degree can also be used in combination to determine whether the group structure is a target group. When identifying the account group that may have risks, the account level distribution aggregation degree or other attribute indicator aggregation degrees can be combined to more accurately represent the risk degree of the group structure, and thus determine the target group. All the aggregation degrees described above can be used in combination to determine the target group more accurately.

[0057] Referring to Figure 5 In an optional embodiment of the present application, each account includes a plurality of attribute indicators. The method for determining whether the group structure is the target group based on the aggregation degree of the attribute indicators includes: Step 1221, determining a comprehensive score of the group structure based on the aggregation degree of each attribute indicator and a preset weight corresponding to each attribute indicator; Step 1222, determining whether the group structure is the target group based on the comprehensive score.

[0058] Regarding the clustering degree of multiple attribute indicators, it is understandable that different attribute indicators may have different impacts on the risks that may exist in the group structure. As mentioned above, attribute indicators such as fingerprints and operational links may be better able to characterize the risks of the group structure than indicators such as attention volume. Therefore, preset weights can be set for different attribute indicators, and then a comprehensive risk assessment formula can be constructed based on the clustering degree and preset weights to score the group structure. Based on the scoring results, it can be determined whether the group structure is the target group. The scoring can also provide managers with more intuitive feedback and subsequent management.

[0059] In an optional embodiment, the formula for determining the comprehensive score includes:

[0060] in, M This represents the overall score; Indicates the first i The degree of clustering of the aforementioned attribute indicators. This represents a mapping function that maps the clustering degree to scores. Indicates the first i The preset weights corresponding to the aforementioned attribute indicators.

[0061] In this embodiment, the clustering degree can be mapped to a score using a mapping function. This mapping function can be set according to actual conditions. For example, different mapping functions can be set for different attribute indicators to strengthen or weaken the influence of each attribute indicator. For instance, for high-risk influencing factors such as operational links, the clustering degree can be mapped to a higher score to participate in the overall score calculation. For example, in a group structure, if 90% of accounts use the same operational link, the operational link clustering degree is 0.9. This indicator is a high-risk factor and can be mapped to 1 point, with a corresponding higher weight, thus participating in the overall score calculation. Conversely, in a group structure, if 70% of accounts have a low account level, the account level clustering degree is 0.7. Account level is a low-risk factor, mapped to 0.6 points, with a corresponding lower weight, thus participating in the overall score calculation.

[0062] In an optional embodiment of this application, the step of performing semantic analysis on the content data to identify the group category of the target group includes: Semantic analysis of the content data is performed based on a large language model to generate the group category of the target group.

[0063] The large language model can be a language model trained by historical content data as training corpus, has the ability of understanding context semantics, and the ability of analyzing and extracting corresponding group categories. In one specific embodiment, after inputting the content data into the large language model, a high-dimensional semantic vector is generated, which is compared with the group categories in the category library, and the group category matched by the content data is determined based on the matching degree, and the group category of the target group is determined.

[0064] For example, when managing the account group that may have risks, the target group based on the comprehensive score and reaching the threshold can be classified as a group that has certain risks and needs to be managed. Further, a large language model is introduced to perform semantic analysis on the content data of the account in the target group, so as to determine the group category.

[0065] In one optional embodiment of the present application, the semantic analysis of the content data by the large language model to generate the group category of the target group comprises: extracting keywords of the content data and aggregated features of attribute indicators of the account; inputting the keywords of the content data and the aggregated features of the attribute indicators into the large language model to generate the group category.

[0066] In this embodiment, for the classification of the target group, the semantic analysis is mainly based on the keywords of the content data as representative content. Multiple group categories, such as 3-10 representative contents, can be extracted, and the content keywords are analyzed. In addition, the aggregated features of the attribute indicators of the target group can be combined as important context clues for the judgment of the large language model to improve the accuracy of the analysis. For example, the average account level, average attention, and content sending platform (such as private message, bullet screen, and comment area) of the target group are determined as the aggregated features of the attribute indicators, and then the keywords of the content data are input into the large language model to generate the group category.

[0067] In one optional embodiment, the keywords of the content data and the aggregated features of the attribute indicators of the account can be extracted by constructing a prompt word template, for example, a Prompt-based template as input of the large language model. Specifically, the content format part of the preset Prompt template is filled according to the content data, so as to extract the keywords of the content data, and the behavior attribute part of the preset Prompt template is filled according to the attribute indicators of the account, so as to extract the aggregated features of the attribute indicators, thereby generating a target Prompt template.

[0068] The prompt template can further include a label space constraint part, for example, a label space constraint part with labels such as [first type of exception, second type of exception, third type of exception, …, "normal"] is set. After the target prompt template is input into the large language model, a label is generated as the category of the target group.

[0069] Subsequently, different processing strategies are provided for different categories of target groups to achieve management of the target groups and accounts in the target groups, so as to achieve precise management. In addition, a feedback mechanism can be introduced for rule callback or training iteration of the large language model, so that the recognition and processing are more accurate.

[0070] Figure 6 A structural schematic diagram of an electronic device is provided for an embodiment of the present application. As shown in Figure 6 The instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the above method embodiment.

[0071] The memory and the processor can be connected in a bus manner, the bus can include any number or kind of interconnected buses and bridges, and the bus connects various circuits of the one or more processors and the memory together. The bus can also connect various other circuits such as peripheral devices together, which are well known in the art, and therefore, further description thereof will not be given herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements such as multiple receivers and transmitters, which provide units for communicating with various other devices on a transmission medium. The data processed by the processor is transmitted on a wireless medium through the antenna, and further, the antenna also receives data and transmits the data to the processor.

[0072] The processor is responsible for managing the bus and general processing, and can also provide various functions including timing, peripheral interface, voltage regulation, power management, and other control functions. The memory can also be used to store data used by the processor during execution of operations.

[0073] Another embodiment of the present application relates to a computer readable storage medium, the computer readable storage medium stores computer instructions, the computer instructions are executed by the processor to implement the method described in the above method embodiment.

[0074] Another embodiment of the present application relates to a computer program product, including a computer program, the computer program is executed by the processor to implement the steps of the above method.

[0075] That is, a person skilled in the art can understand that all or part of the steps in the above-mentioned embodiment methods can be completed by a program instructing relevant hardware, the program being stored in a storage medium and including a plurality of instructions for causing a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0076] A person of ordinary skill in the art can understand that the above-mentioned embodiments are specific embodiments for implementing the present application, and in actual applications, various changes can be made in form and details without departing from the spirit and scope of the present application.

Claims

1. An account management method, characterized by, The method comprises the following steps: Clustering content data of multiple accounts and constructing an association relationship between the accounts and the content data; Determining at least one target group in the association relationship based on a group discovery algorithm; Performing semantic analysis on the content data to identify a group category of the target group; Managing the accounts of the target group according to a processing strategy corresponding to the group category.

2. The account management method of claim 1, wherein, The method of determining at least one target group in the association relationship based on a group discovery algorithm comprises the following steps: Identifying the association relationship based on a group discovery algorithm to determine a group structure composed of multiple accounts; Determining whether the group structure is the target group according to an attribute index of the accounts.

3. The account management method according to claim 1 or 2, characterized by, The method of constructing the association relationship between the accounts and the content data comprises the following steps: Constructing a bipartite graph or a projection graph with the content data and the accounts as graph nodes.

4. The account management method of claim 2, wherein, The method of determining whether the group structure is the target group according to the attribute index of the accounts comprises the following steps: Determining a concentration degree of the attribute index in the group structure, wherein the concentration degree of the attribute index comprises a ratio of a number of accounts with an approximate attribute index in the group structure to a total number of accounts in the group structure; Determining whether the group structure is the target group according to the concentration degree of the attribute index.

5. The account management method of claim 4, wherein, Each of the accounts comprises multiple attribute indexes; the method of determining whether the group structure is the target group according to the concentration degree of the attribute index comprises the following steps: Determining a comprehensive score of the group structure according to the concentration degree of each attribute index and a preset weight corresponding to each attribute index; Determining whether the group structure is the target group according to the comprehensive score.

6. The account management method of any one of claims 1-5, wherein, The method of performing semantic analysis on the content data to identify a group category of the target group comprises the following steps: Extracting keywords of the content data and aggregated features of the attribute indexes of the accounts; Inputting the keywords of the content data and the aggregated features of the attribute indexes into a large language model to generate the group category.

7. The account management method according to any one of claims 1 to 6, wherein The method of clustering content data of multiple accounts comprises the following steps: Encoding the content data; Dividing the encoded content data into buckets, wherein each bucket comprises multiple data segments of the encoded content data; Determining a difference distance between the data segments in the bucket to determine whether the content data is in the same category according to the difference distance.

8. The account management method of any of claims 1-7, wherein, The method further comprises the following steps: Filtering repeatedly published content of each account based on a preset frequency to take the repeatedly published content as the content data.

9. An electronic device, comprising: The method comprises the following steps: At least one processor; And A memory connected in communication with the at least one processor; wherein The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are executed by the processor to implement the method of any one of claims 1 to 8.

11. A computer program product comprising a computer program, characterized in that, The computer program, which when executed by the processor, implements the steps of the method as claimed in any one of claims 1 to 8.

Images