Social network similar account identification device and method
By obtaining and analyzing the speech data of social network accounts, calculating the similarity between content, location, time and basic information, the problem of identifying similar accounts in social networks is solved, and a more accurate recognition effect is achieved.
Patent Information
- Application Number
- CN202411760482.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-03-28
AI Technical Summary
There are a large number of similar accounts in social networks, which are used for repeated promotion or screen swiping, affecting user experience and destroying the platform's ecological environment.
By obtaining the speech data of the account to be identified, semantic features, location information, release time information and basic information are extracted, the similarity is calculated, and the comprehensive similarity is judged whether the account is a similar account.
It realizes more accurate identification of similar accounts on social networks, and combines multi-dimensional data analysis such as content, behavior, and time to make up for the shortcomings of single-dimensional recognition.
Smart Images
Figure CN119719797A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of computer information, and more specifically, to a device and method for identifying similar accounts in a social network. Background Art
[0002] In social networks, some users create a large number of similar accounts for repeated promotion or screen swiping for various purposes. These behaviors not only interfere with ordinary users and affect the user experience, but also may damage the ecological environment of social network platforms. Therefore, it is necessary to design a technical solution that can overcome the above defects. Summary of the invention
[0003] An object of the present invention is to provide a device and method for identifying similar accounts on a social network, which can more accurately identify similar accounts on a social network.
[0004] In order to achieve these purposes and other advantages of the present invention, according to one aspect of the present invention, the present invention provides a method for identifying similar accounts on a social network, including: S1: acquiring speech data of at least two accounts to be identified; S2: extracting semantic features of the speech data and calculating content similarity; S3: extracting location information of the speech data and calculating location similarity; S4: extracting release time information of the speech data and calculating time similarity; S5: calculating basic information similarity of at least two accounts to be identified; S6: judging whether at least two accounts to be identified are similar accounts based on the content similarity, the location similarity, the time similarity and the basic information similarity.
[0005] Further, a comprehensive similarity is calculated. If the comprehensive similarity is greater than a first predetermined threshold, it is determined whether at least two of the accounts to be identified are similar accounts; the comprehensive similarity is a weighted average of the content similarity, the location similarity, the time similarity and the basic information similarity.
[0006] Furthermore, it also includes: acquiring known similar accounts and known dissimilar accounts in real time, and adjusting the weights of the comprehensive similarity, which is the content similarity, the location similarity, the time similarity and the basic information similarity.
[0007] Furthermore, in S2, NLP technology is used to extract semantic features and calculate the content similarity; if the speech data is a picture, OCR image recognition technology is used to extract the content in the picture, and then NLP technology is used to extract semantic features and calculate the content similarity.
[0008] Furthermore, in S5, the basic information similarity is calculated based on the account name, account profile, registration time, and IP location of at least two of the accounts to be identified.
[0009] Furthermore, it also includes: S7: collecting the accounts to be identified whose comprehensive similarity is greater than a second predetermined threshold, and establishing an identification group, wherein the second predetermined threshold is less than the first predetermined threshold; S8: extracting the number of interactions and the interaction time between the accounts to be identified in the identification group; S9: calculating the association index of each account to be identified in the identification group based on the number of interactions and the interaction time, and determining the accounts to be identified whose association index is greater than a third predetermined threshold as similar accounts.
[0010] According to another aspect of the present invention, a device for identifying similar accounts on a social network is also provided, including: an acquisition module for acquiring speech data of at least two accounts to be identified; a first calculation module for extracting semantic features of the speech data and calculating content similarity; a second calculation module for extracting location information of the speech data and calculating location similarity; a third calculation module for extracting release time information of the speech data and calculating time similarity; a fourth calculation module for calculating basic information similarity of at least two accounts to be identified; and a judgment module for judging whether at least two accounts to be identified are similar accounts based on the content similarity, the location similarity, the time similarity and the basic information similarity.
[0011] Furthermore, the judgment module calculates a comprehensive similarity, and if the comprehensive similarity is greater than a first predetermined threshold, determines whether at least two of the accounts to be identified are similar accounts; the comprehensive similarity is a weighted average of the content similarity, the location similarity, the time similarity and the basic information similarity.
[0012] Furthermore, it also includes: an adjustment module, which is used to obtain known similar accounts and known dissimilar accounts in real time, and adjust the weights of the comprehensive similarity, which is the content similarity, the location similarity, the time similarity and the basic information similarity.
[0013] Furthermore, it also includes: a collection module, used to collect the accounts to be identified whose comprehensive similarity is greater than a second predetermined threshold, and establish an identification group, wherein the second predetermined threshold is less than the first predetermined threshold; an extraction module, used to extract the number of interactions and the interaction time between the accounts to be identified in the identification group; a fifth calculation module: calculates the association index of each account to be identified in the identification group according to the number of interactions and the interaction time, and determines the accounts to be identified whose association index is greater than a third predetermined threshold as similar accounts.
[0014] The present invention has at least the following beneficial effects: The present invention obtains speech data of at least two accounts to be identified, calculates content similarity, location similarity, time similarity, and basic information similarity, and then determines whether at least two accounts to be identified are similar accounts; the present invention combines content, behavior, time and other multi-dimensional data for analysis, which makes up for the shortcomings of single-dimensional identification and can more accurately identify similar accounts.
[0015] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 This is a framework diagram of an embodiment of the present application; Figure 2 This is a flow chart of an embodiment of the present application. DETAILED DESCRIPTION
[0017] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.
[0018] It should be understood that the terms such as "having", "including" and "comprising" used in the embodiments of the present application do not exclude the existence or addition of one or more other elements or their combinations. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain the relative position relationship, movement, etc. between the components in a certain specific posture. If the specific posture changes, the directional indication will also change accordingly. When an element is referred to as "fixed on" or "set on" another element, it can be directly on the other element or there may be a centering element at the same time. When an element is referred to as "connecting" another element, it can be directly connected to another element or it can be indirectly connected to another element through a centering element. The description of "first", "second", etc. in the embodiments of the present application is only for descriptive purposes, and cannot be understood as indicating or implying its relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" can explicitly or implicitly include at least one of the features.
[0019] It should be noted that the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.
[0020] The embodiment of the present application provides a method for identifying similar accounts on a social network, including: S1: Obtain speech data of at least two accounts to be identified; Exemplarily, content posted by a designated account is captured from a social media platform or a pre-established database; S2: extracting semantic features of the speech data and calculating content similarity; For example, natural language processing (NLP) technology is used to pre-process the text (word segmentation, stop word removal, etc.), TF-IDF, Word2Vec or other vectorization models are applied to convert the text into a numerical feature vector, and indicators such as cosine similarity and Jaccard index are used to measure the similarity of the content; For example, if the speech data is a picture, the OCR picture recognition technology is used to extract the content in the picture, and then the NLP technology is used to extract semantic features and calculate the content similarity; S3: extracting the location information of the speech data and calculating the location similarity; Exemplarily, the geographic location tags are parsed from the information posted by the account, and the proportion of shared locations between accounts is calculated or the geographic distance formula is used to quantify the location similarity; S4: extracting the release time information of the speech data and calculating the time similarity; For example, the distribution time (e.g., the distribution from one hour to another hour every day) is analyzed to compare the consistency of the peak activity hours between different accounts, or statistical methods such as the Pearson correlation coefficient can be used to quantify the similarity in the time series; S5: Calculate the similarity of basic information of at least two accounts to be identified; Exemplarily, the basic information similarity is calculated based on the account name, account profile, registration time, and IP location of at least two of the accounts to be identified; For example, the account name, account profile, registration time, IP location, profile picture and other information are compared, and the profile picture is compared using image processing technology to quantify the matching degree between these attributes and give a comprehensive score, which is the basic information similarity; S6: judging whether at least two of the accounts to be identified are similar accounts according to the content similarity, the location similarity, the time similarity and the basic information similarity; Exemplarily, a comprehensive similarity is calculated, and if the comprehensive similarity is greater than a first predetermined threshold, it is determined whether at least two of the accounts to be identified are similar accounts; the comprehensive similarity is a weighted average of the content similarity, the location similarity, the time similarity, and the basic information similarity; Exemplarily, a reasonable first predetermined threshold is determined according to actual needs, and if the threshold is exceeded, it is considered that two or more accounts may be similar accounts; Exemplarily, the weights of content similarity, location similarity, time similarity and basic information similarity are determined based on expert experience, or are set equal; It can be seen that the present embodiment conducts a multi-dimensional comprehensive analysis of the account to be identified, and conducts analysis in combination with multi-dimensional data such as content, behavior, and time, which makes up for the deficiency of single-dimensional identification and can more accurately identify similar accounts (such as advertising accounts); the present embodiment can conduct a unified analysis of similar accounts on multiple platforms and has a wide range of uses; the present embodiment can automate data analysis methods, reduce the need for manual intervention, and can continuously track similar accounts and persistently track and analyze incremental accounts.
[0021] In another embodiment, the method further includes: obtaining known similar accounts and known dissimilar accounts in real time, and adjusting the weights of the comprehensive similarity, which is the content similarity, the location similarity, the time similarity, and the basic information similarity; This embodiment obtains confirmed similar accounts and dissimilar accounts according to different platforms or needs, and adjusts the weights to facilitate analysis of different types of similar accounts and improve the scope of application.
[0022] In another embodiment, it further includes: S7: collecting the accounts to be identified whose comprehensive similarity is greater than a second predetermined threshold, and establishing an identification group, wherein the second predetermined threshold is less than the first predetermined threshold; S8: extracting the number of interactions and the interaction time between the accounts to be identified in the identification group; S9: calculating the association index of each account to be identified in the identification group according to the number of interactions and the interaction time, and determining the accounts to be identified whose association index is greater than a third predetermined threshold as similar accounts; In this embodiment, a lower second predetermined threshold (e.g., 0.7) is set based on the comprehensive similarity calculated previously, and this threshold is lower than the first predetermined threshold (e.g., 0.8) used to directly determine the similarity; accounts whose comprehensive similarity with at least one other account exceeds the second predetermined threshold are selected from all accounts to be identified, and these accounts are grouped into an identification group for more in-depth analysis; interaction data between accounts in the identification group are captured from the social media platform or database, including but not limited to likes, comments, and reposts, and the total number of times each account interacts with other accounts is recorded, and the specific time point of each interaction is recorded; it is analyzed whether there is overlap in the time period of the interaction, such as whether they interact in the same daily time period, and a comprehensive correlation index formula is constructed, such as the correlation index I= QUOTE , where i represents the interaction type (like, comment, and forward), v i Indicates the temporal overlap between this type of interaction and the interaction object (speech data), l iIndicates the number of interactions of this type; illustratively, the time overlap can be determined by constructing a time distribution curve of this type of interaction and calculating the curve overlap as the time overlap; a third predetermined threshold is set, and the third set threshold is determined based on historical data statistics, and accounts in the identification group that are greater than the third predetermined threshold are all highly suspicious and are identified as similar accounts; this embodiment is suitable for use when the comprehensive similarity does not reach the first predetermined threshold but is close to the first predetermined threshold, and similar accounts are quickly determined through correlation, reducing missed detections caused by improper human setting of the first predetermined threshold, and further improving recognition accuracy.
[0023] Based on the same inventive concept, an embodiment of the present application also provides a device for identifying similar accounts on a social network, comprising: an acquisition module, used to acquire speech data of at least two accounts to be identified; a first calculation module, used to extract semantic features of the speech data and calculate content similarity; a second calculation module, used to extract location information of the speech data and calculate location similarity; a third calculation module, used to extract release time information of the speech data and calculate time similarity; a fourth calculation module, used to calculate basic information similarity of at least two accounts to be identified; a judgment module, used to judge whether at least two accounts to be identified are similar accounts based on the content similarity, the location similarity, the time similarity and the basic information similarity; This embodiment uses a computer program to construct an acquisition module, a first calculation module, a second calculation module, a third calculation module, a fourth calculation module and a judgment module, and conducts a comprehensive analysis of multiple dimensions to be identified, combining multi-dimensional data such as content, behavior, and time for analysis, thereby making up for the deficiency of single-dimensional identification and being able to more accurately identify similar accounts (such as advertising accounts); this embodiment can also perform a unified analysis of similar accounts on multiple platforms, has a wide range of uses, and can automate data analysis methods, reduce the need for manual intervention, and can continuously track similar accounts and persistently track and analyze incremental accounts.
[0024] In another embodiment, the judgment module calculates the comprehensive similarity, and if the comprehensive similarity is greater than a first predetermined threshold, it determines whether at least two of the accounts to be identified are similar accounts; the comprehensive similarity is a weighted average of the content similarity, the location similarity, the time similarity and the basic information similarity; the judgment module of this embodiment provides a specific method for multi-dimensional comprehensive analysis.
[0025] In another embodiment, it further includes: an adjustment module, which is used to obtain known similar accounts and known dissimilar accounts in real time, and adjust the weights of the comprehensive similarity, which is the content similarity, the location similarity, the time similarity and the basic information similarity; The adjustment module of this embodiment obtains confirmed similar accounts and dissimilar accounts according to different platforms or needs, and adjusts the weights to facilitate analysis of different types of similar accounts and improve the scope of application.
[0026] In another embodiment, it further includes: a collection module, used to collect the accounts to be identified whose comprehensive similarity is greater than a second predetermined threshold, and establish an identification group, wherein the second predetermined threshold is less than the first predetermined threshold; an extraction module, used to extract the number of interactions and the interaction time between the accounts to be identified in the identification group; a fifth calculation module: calculating the association index of each account to be identified in the identification group according to the number of interactions and the interaction time, and determining the accounts to be identified whose association index is greater than a third predetermined threshold as similar accounts; In this embodiment, based on the comprehensive similarity calculated previously, a lower second predetermined threshold is set, an identification group is constructed, the interaction data between the accounts in the identification group is captured from the social media platform or database, a comprehensive correlation index formula is constructed, and the correlation between the accounts in the identification group is analyzed. Accounts in the identification group that are greater than the third predetermined threshold are all highly suspicious and are identified as similar accounts.
[0027] like Figure 1 and Figure 2 As shown, the embodiments of the present application can be further implemented using the following devices and methods: basic data acquisition device 101; data cleaning device 102; speech analysis device 103; basic information analysis device 104; comprehensive analysis device 105; continuous updating device 106; Device 101 collects and processes the data required by each platform. Device 102 performs keyword acquisition processing on the collected data, such as filtering out account data that has not posted advertisements, and extracting advertisement accounts that need to be identified. Device 103 is used to traverse the accounts obtained in device 102 to obtain the original speeches posted by them for analysis and processing. Device 104 is used to determine the similarity of the accounts based on the basic information of the accounts and the identification basis of the basic content of the accounts posted in step 3. Device 105 performs comprehensive evaluation and processing on the analysis results in 103 and 104. The device comprehensively evaluates the results of speech analysis 103 and basic information analysis 104. Device 106 is used to perform long-term calculations on identified similar accounts; by continuously analyzing the behavior and content publishing of these accounts, the recognition model is updated and adjusted in real time to ensure that the platform can make different adjustments in a timely manner according to needs.
[0028] Step s21 (data input): filter the text data in the library, and perform keyword acquisition processing on the collected data, such as filtering out account data that has not sent advertisements, extracting advertisement accounts that need to be identified, and then inputting them into the program for analysis.
[0029] Step s22 (speech data analysis): For similar accounts to be identified, search for posts within a recent time frame, and then conduct a comprehensive analysis to find similar posts. There are mainly four points: Text data: Use natural language processing (NLP) technology to extract semantic features such as keywords and themes of the published content, and find similar texts. Image data: Use OCR image recognition technology to extract text content in the image, convert it into text data, and then follow the process in step 1. Extract the publishing location information and find accounts with high text similarity in steps 1 and 2 and consistent publishing locations. Extract the publishing time information and determine whether the account active time periods are the same by the consistency of the speaking time periods. Through the above four points, if two accounts have the same speaking time period, close or consistent locations, and similar published content, it can be basically determined that the two are similar accounts, and it may even be that the same content is published by two accounts operated by the same person at the same time.
[0030] Step s23 (basic information analysis): Through the similar posting content obtained in s23, the main account of the posting is analyzed and judged, mainly based on the account name, account profile, account registration time, IP location and other information for comprehensive judgment.
[0031] Step s24 (comprehensive similar account score judgment): Integrate all the information obtained in s22 and s23, perform weighted scoring, and use the scores to determine whether the two are highly similar, slightly similar, etc.
[0032] By integrating data from multiple dimensions such as text similarity, behavior patterns, location and time, a similarity score or result is generated. If the score exceeds the set threshold, these accounts can be judged as similar accounts and corresponding treatment measures can be taken. The formula is as follows: Stotal=w1*S content +w2*S location +w3*S time +w4*S basic-info in: S total is the final similarity score, with a value range of 0 to 1, where 1 means the two accounts are completely similar and 0 means they are completely dissimilar.
[0033] S content Indicates the similarity of text or image content, usually calculated through natural language processing (NLP) or image similarity.
[0034] S location Indicates the similarity of the account's posting location, calculated based on the distance of the geographic location coordinates.
[0035] S time Indicates the temporal similarity of the content published by the accounts, based on the active time periods of the two accounts.
[0036] S basic_info Indicates the similarity of basic account information, such as registration information, IP address, etc.
[0037] w1, w2, w3, w4, and w5 are weight coefficients of each dimension, representing the contribution of each dimension to the similarity score. The sum of the weights should be 1 and can be adjusted according to the actual application scenario. For example, if content similarity and behavior similarity are more important in advertising account identification, they can be given higher weights. The calculation results based on this formula are shown in the following table. In the table below, the account score is 0.64, which meets the criteria for similar accounts and is a common similar account. Because it is less than 0.7, it is not included in the identification group.
[0038] Step s25 (output similar accounts): similar accounts and similarities obtained in step s24 are output as non-similar accounts (filtered out directly), similar accounts, and highly similar accounts, which are put into storage for use in subsequent data analysis and data processing. If in advertising account identification, similar advertising accounts can be directly filtered out during subsequent data analysis to improve analysis accuracy.
[0039] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.
Claims
1. A method for identifying similar accounts on a social network, characterized in that: include: S1: Obtain speech data of at least two accounts to be identified; S2: extracting semantic features of the speech data and calculating content similarity; S3: extracting the location information of the speech data and calculating the location similarity; S4: extracting the release time information of the speech data and calculating the time similarity; S5: Calculate the similarity of basic information of at least two accounts to be identified; S6: Determine whether at least two of the accounts to be identified are similar accounts according to the content similarity, the location similarity, the time similarity and the basic information similarity.
2. The method for identifying similar accounts in a social network according to claim 1, characterized in that: Calculating a comprehensive similarity, and if the comprehensive similarity is greater than a first predetermined threshold, determining whether at least two of the accounts to be identified are similar accounts; The comprehensive similarity is a weighted average of the content similarity, the location similarity, the time similarity and the basic information similarity.
3. The method for identifying similar accounts in a social network according to claim 2, characterized in that: Also includes: Known similar accounts and known dissimilar accounts are obtained in real time, and the weights of the comprehensive similarity, which is the content similarity, the location similarity, the time similarity, and the basic information similarity, are adjusted.
4. The method for identifying similar accounts in a social network according to claim 2, characterized in that: In S2, semantic features are extracted using NLP technology, and the content similarity is calculated; If the speech data is a picture, the OCR picture recognition technology is used to extract the content in the picture, and then the NLP technology is used to extract semantic features and calculate the similarity of the content.
5. The method for identifying similar accounts in a social network according to claim 2, characterized in that: In S5, the similarity of the basic information is calculated based on the account name, account profile, registration time, and IP location of at least two of the accounts to be identified.
6. The method for identifying similar accounts in a social network according to claim 2, wherein: Also includes: S7: collecting the accounts to be identified whose comprehensive similarity is greater than a second predetermined threshold and establishing an identification group, wherein the second predetermined threshold is less than the first predetermined threshold; S8: extracting the number of interactions and the interaction time between the accounts to be identified in the identification group; S9: Calculate the association index of each of the accounts to be identified in the identification group according to the number of interactions and the interaction time, and determine the accounts to be identified whose association index is greater than a third predetermined threshold as similar accounts.
7. A device for identifying similar accounts on a social network, characterized in that: include: An acquisition module, used to acquire speech data of at least two accounts to be identified; A first calculation module, used to extract semantic features of the speech data and calculate content similarity; A second calculation module is used to extract the position information of the speech data and calculate the position similarity; A third calculation module is used to extract the release time information of the speech data and calculate the time similarity; A fourth calculation module, used to calculate the similarity of basic information of at least two accounts to be identified; A judgment module is used to judge whether at least two accounts to be identified are similar accounts according to the content similarity, the location similarity, the time similarity and the basic information similarity.
8. The device for identifying similar accounts in a social network as claimed in claim 7, characterized in that: The judgment module calculates the comprehensive similarity, and if the comprehensive similarity is greater than a first predetermined threshold, determines whether at least two of the accounts to be identified are similar accounts; The comprehensive similarity is a weighted average of the content similarity, the location similarity, the time similarity and the basic information similarity.
9. The device for identifying similar accounts in a social network as claimed in claim 8, characterized in that: Also includes: The adjustment module is used to obtain known similar accounts and known dissimilar accounts in real time, and adjust the weights of the comprehensive similarity, which is the content similarity, the location similarity, the time similarity and the basic information similarity.
10. The device for identifying similar accounts in a social network as claimed in claim 8, characterized in that: Also includes: A collection module, used for collecting the accounts to be identified whose comprehensive similarity is greater than a second predetermined threshold, and establishing an identification group, wherein the second predetermined threshold is less than the first predetermined threshold; An extraction module, used for extracting the number of interactions and the interaction time between the accounts to be identified in the identification group; The fifth calculation module is used to calculate the association index of each of the accounts to be identified in the identification group according to the number of interactions and the interaction time, and determine the accounts to be identified whose association index is greater than a third predetermined threshold as similar accounts.