Social robot detection method and device based on heterogeneous graph

Abstract

The invention provides a social robot detection method and device based on a heterogeneous graph, and the method comprises the steps: obtaining user texts of a specified event in a social network, and analyzing the implicit association between the texts through a large language model; taking the account, the content and the topic of the user text as nodes, and constructing a heterogeneous graph under the specified event; extracting the feature representation of each account in the heterogeneous graph according to a preset meta path; modeling the feature representation by using a graph neural network to obtain a user node representation under the meta-path; fusing all the user node representations of the account to obtain a final representation of the account; and inputting the final representation into a classifier to obtain a detection result that whether the account belongs to the social robot or not.

Description

Technical Field

[0001] The present invention relates to the fields of network security technology and social media analysis technology, and in particular to a social robot detection method, device, electronic device, computer-readable storage medium and computer program product based on heterogeneous graphs. Background Art

[0002] With the rapid development of social media, social bots have become an indispensable presence in cyberspace. These automated programs can imitate the behavior of human users to post, comment, like, and other interactions. Their purpose may be to influence public sentiment, spread false information, manipulate public opinion, etc. The existence of social robots poses a serious threat to network security, information authenticity, and social order. Therefore, how to effectively identify and detect social robots has become an important issue in maintaining network security and information ecology. The development of social robot detection technology is not only related to personal privacy and data security, but also to network security and social stability.

[0003] At present, social robot detection methods are mainly divided into several categories: feature matching-based methods, which match the user account's registration information, published content and other features with the known social robot feature library; behavior analysis-based methods, which identify abnormal behaviors by analyzing user behavior patterns, such as posting frequency, interaction time, etc.; machine learning-based methods, which use machine learning algorithms to analyze user data, automatically learn the characteristics of social robots and classify them; and deep learning-based methods, especially graph neural network technology, which captures complex social relationships by constructing interaction graphs between users to improve detection accuracy. Although these methods have improved the efficiency and accuracy of detection to a certain extent, the continuous evolution and diversification of social robots make detection work still challenging. For example, social robots may imitate the behavior patterns of real users, making it difficult for behavior analysis-based methods to distinguish; and feature matching methods are limited by the update speed and coverage of feature libraries.

[0004] Although existing social robot detection methods have been applied in many practical scenarios, social robot training technology is also constantly evolving, and existing social robot detection methods still have some shortcomings and deficiencies.

[0005] First, when processing text content, existing social robot detection methods mainly treat it as part of the account features, and obtain the representation of content features through custom artificial features or deep learning. However, in actual application scenarios, the content information of social robot accounts also contains implicit relationship structures. For example, when multiple social robots collaborate to complete a specific manipulation task, the content they publish often has a strong correlation, and this correlation also reflects the implicit collaborative relationship between social robot accounts. Mining and utilizing these relationships will help detect social robot accounts.

[0006] Secondly, the existing social robot detection methods based on graph neural networks are mainly based on homogeneous graphs, and all nodes are account nodes, lacking content nodes. At the same time, the existing graph neural network methods often ignore the impact of different types of edges when performing calculations. The social robot relationship network contains many different types of edges, such as attention, fans, forwarding, and replies. Different types of edge structures reflect different behavior patterns between accounts. Current social robot detection methods ignore this behavior pattern, or are unable to use this behavior pattern to detect and classify social robots in social networks.

[0007] In summary, existing text sentiment analysis methods have two main problems: (1) content information is only used as part of the feature, and there is a lack of discovery of implicit correlation structures based on content information; (2) existing graph neural networks are all based on homogeneous graphs and lack the use of multiple types of nodes and edges. Summary of the invention

[0008] The purpose of the present invention is to solve the problem that the implicit relationship modeling on the content side of the existing social robot detection methods is insufficient and it is difficult to use multi-source heterogeneous information, and a social robot detection method and system based on heterogeneous graphs are proposed. The method and system use the powerful semantic understanding and logical reasoning capabilities of the large language model to mine the implicit relationship between different texts under the same topic, thereby enriching the node and edge information of the heterogeneous graph; at the same time, the method also adopts a meta-path-based heterogeneous graph analysis method to support business personnel to set up efficient social robot detection meta-paths based on their own business experience, so as to better collect multi-source information to enrich the representation of accounts and achieve more accurate social robot detection effects.

[0009] In view of the shortcomings of existing technologies, such as Figure 3 As shown, the present invention proposes a social robot detection method based on heterogeneous graphs, which includes:

[0010] In the initial step, we obtain user texts of specified events in social networks and use large language models to analyze implicit associations between texts;

[0011] The analysis step is to construct a heterogeneous graph under the specified event with the account, content and topic of the user text as nodes; extract the feature representation of each account in the heterogeneous graph according to the preset meta-path; and use the graph neural network to model the feature representation to obtain the user node representation under the meta-path; and fuse all the user node representations of the account to obtain the final representation of the account;

[0012] In the classification step, the final representation is input into the classifier to obtain the detection result of whether the account belongs to a social robot.

[0013] The social robot detection method based on heterogeneous graph, wherein the initial step includes: inputting the user text into the large language model through a prompt template; the prompt template includes a system instruction for informing the large language model of the specific task requirements and a pair of user texts of a specified event in the social network.

[0014] The social robot detection method based on heterogeneous graph, wherein the edge types between account-account nodes in the heterogeneous graph include attention, forwarding, and commenting; the edge type between account-content is a publishing relationship; the edge relationship between account-topic is a participation relationship, the edge relationship between content-content includes a support relationship, an opposition relationship, and an irrelevant relationship, the relationship between content-topic is a subordinate relationship, and the relationship between topic-content is an inclusion relationship;

[0015] The social robot detection method based on heterogeneous graph, wherein the meta-path includes: a meta-path for describing the connection between pairs of users who publish content on the same topic; and / or a meta-path for describing the connection between pairs of users holding the same position; and / or a meta-path for describing the connection between pairs of users holding opposite positions.

[0016] The social robot detection method based on heterogeneous graph, wherein the feature representation includes numerical information, classification information and text information; wherein the numerical information includes the number of account follows, the number of tweets posted, the number of followers and the length of the user name; the classification information is identity authentication information; and the text information includes: the account's self-introduction and tweets.

[0017] like Figure 4 As shown, the present invention also proposes a social robot detection device based on a heterogeneous graph, which includes:

[0018] The initial module obtains user texts of specified events in social networks and uses a large language model to analyze implicit associations between texts;

[0019] The analysis module uses the account, content and topic of the user text as nodes to construct a heterogeneous graph under the specified event; extracts the feature representation of each account in the heterogeneous graph according to the preset meta-path; and uses the graph neural network to model the feature representation to obtain the user node representation under the meta-path; and fuses all the user node representations of the account to obtain the final representation of the account;

[0020] The classification module inputs the final representation into the classifier to obtain the detection result of whether the account belongs to a social robot.

[0021] The social robot detection device based on heterogeneous graph, wherein the initial module includes: inputting the user text into the large language model through a prompt template; the prompt template includes a system instruction for informing the large language model of specific task requirements and a pair of user texts of a specified event in the social network;

[0022] The edge types between account-account nodes in this heterogeneous graph include following, forwarding, and commenting; the edge type between account-content is publishing relationship; the edge relationship between account-topic is participation relationship; the edge relationship between content-content includes support relationship, opposition relationship, and irrelevant relationship; the relationship between content-topic is affiliation relationship; the relationship between topic-content is inclusion relationship;

[0023] The social robot detection device based on heterogeneous graph, wherein the meta-path includes: a meta-path for describing the connection between pairs of users who publish content on the same topic; and / or a meta-path for describing the connection between pairs of users holding the same position; and / or a meta-path for describing the connection between pairs of users holding opposite positions; the social robot detection method based on heterogeneous graph, wherein the feature representation includes numerical information, classification information and text information; wherein the numerical information includes the number of account followers, the number of tweets posted, the number of followers and the length of the user name; the classification information is identity authentication information; the text information includes: the account's self-introduction and tweets.

[0024] The present invention also proposes an electronic device, which includes the social robot detection device described above, and the electronic device may be connected to an information display device, and the information display device is used to display the detection result according to display parameters and attributes set by the user or through an artificial intelligence model.

[0025] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which implements the steps of the social robot detection method when executed by a processor.

[0026] The present invention also proposes a computer program product, comprising a computer program, wherein the computer program implements the steps of the social robot detection method when executed by a processor.

[0027] It can be seen from the above scheme that the advantages of the present invention are:

[0028] 1. In response to the problem that existing social robot detection does not sufficiently mine text associations, the present invention proposes a text association analysis method based on a large language model. This method infers implicit associations between texts through carefully designed prompt templates, thereby enriching the structural features used for social robot detection.

[0029] 2. Aiming at the deficiency that the existing social robot detection methods based on graph neural networks only consider homogeneous graphs, the present invention proposes a social robot identification framework based on meta-paths. The framework supports business personnel to customize the association paths between social robot accounts, and aggregates multi-source heterogeneous feature information based on meta-paths to achieve a more comprehensive representation of accounts, thereby achieving more accurate social robot identification.

[0030] 3. The present invention constructs a social robot identification system based on heterogeneous graphs, covering functions such as content association structure mining, heterogeneous graph construction, meta-path design, account feature extraction, and social robot detection, which can integrate multi-source information to accurately identify social robot accounts. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of a social robot detection method based on heterogeneous graph;

[0032] Figure 2 It is a business logic diagram of a social robot detection system based on heterogeneous graph;

[0033] Figure 3 is a flow chart of the method of the present invention;

[0034] Figure 4 It is a module diagram of the device of the present invention;

[0035] Figure 5 This is a schematic diagram of the structure of a first electronic device of the present invention;

[0036] Figure 6 This is a schematic diagram of the application environment structure of the first electronic device of the present invention;

[0037] Figure 7 It is a schematic diagram of the structure of a second electronic device of the present invention.

[0038] Reference numerals:

[0039] A-First electronic device;

[0040] B-Social robot detection device;

[0041] C-data acquisition equipment;

[0042] D-information display device;

[0043] 1000 - second electronic device;

[0044] Ⅰ-computational unit;

[0045] II-ROM;

[0046] III-RAM;

[0047] IV-bus;

[0048] V-interface;

[0049] VI - input unit;

[0050] VII-output unit;

[0051] VIII- Storage medium;

[0052] Ⅸ-Communication unit. DETAILED DESCRIPTION

[0053] It should be noted that, in this application, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0054] Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0055] The processor described in the present invention is the control center of the electronic device, which can be a processor or a general term for multiple processing elements. For example, it can be one or more central processing units (CPUs), or application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement the embodiments of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (FPGAs).

[0056] Optionally, the processor can perform various functions of the electronic device by running or executing a software program stored in the memory, and calling data stored in the memory.

[0057] In a specific implementation, as an embodiment, the processor may include one or more CPUs. Each of these processors may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions). Electronic devices may include: servers, desktop computers, laptops, smart phones, tablet computers, embedded computers, etc., wherein the embedded computers include vehicles and robots, etc.

[0058] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment, which will not be repeated here.

[0059] It should be noted that the structure of the electronic device shown in the drawings of the present invention does not constitute a limitation thereto, and the actual knowledge structure recognition device may include more or fewer components than shown in the drawings, or a combination of certain components, or a different arrangement of components.

[0060] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0061] It should also be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0062] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0063] It should also be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0064] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0065] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0066] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0067] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0068] In view of the problem that existing social robot detection methods lack implicit association structure modeling on the content side, the present invention proposes a text association recognition method based on a large language model. The method implements unsupervised text association analysis by designing a suitable prompt template. The association analysis results between texts include categories such as "support, opposition, and irrelevant". According to the results of the text association analysis, the present invention adds corresponding edges between text pairs containing "support" and "opposition" relationships, thereby enriching the structural information of the heterogeneous graph.

[0069] In view of the problem that existing social robot detection methods lack heterogeneous graph modeling and analysis capabilities, the inventors of the present invention proposed a social robot detection framework based on heterogeneous graphs. The framework integrates different types of nodes (users, content, topics, etc.) and different types of edges (follow, forward, reply, participate, support, etc.) into the same heterogeneous network, and adopts a meta-path-based heterogeneous graph modeling method to better identify social robot account features, thereby achieving high-precision social robot detection.

[0070] Specifically, in order to achieve the above technical effects, the present invention includes the following key technical points:

[0071] Key point 1: To address the problem that existing social robot detection methods ignore implicit associations on the content side, a text association analysis method based on a large language model is proposed. This method can automatically identify the associations between texts, thereby enriching the nodes and edges in the heterogeneous graph.

[0072] Key point 2: In view of the shortcomings of existing social robot detection methods that only use homogeneous graphs, a social robot detection framework based on meta-paths is proposed. This framework supports business users to design meta-paths based on business experience, improves the model's ability to utilize multi-source heterogeneous information, and achieves more accurate social robot detection results.

[0073] Key point 3: A social robot detection system based on heterogeneous graphs has been built. A social robot detection system based on large heterogeneous graphs has been implemented, including content association structure mining, heterogeneous graph construction, meta-path design, account feature extraction, social robot detection, etc., which supports business analysts to comprehensively use multi-source information such as accounts, structures and content to achieve better social robot detection results.

[0074] In order to make the above features and effects of the present invention more clearly and understandably described, the following embodiments are specifically cited and described in detail with reference to the accompanying drawings. This specification discloses one or more embodiments that contain key technical points of the present invention. The disclosed embodiments are only for illustration. The scope of protection of the present invention is not limited to the disclosed embodiments, and the present invention is defined by the attached claims.

[0075] like Figure 1 As shown, the present invention proposes a social robot detection method based on heterogeneous graph, comprising:

[0076] Step 1. For texts related to specific events, use a large language model to analyze the implicit associations between texts. Implicit associations are logical implications between texts, including support, opposition, and irrelevance.

[0077] Step 2. Integrate account, content and structural features to construct a heterogeneous graph under a specific event. The nodes include users, texts and topics, and the relationships include following, forwarding, commenting, participating, publishing, supporting, opposing, etc.

[0078] Step 3. Combine the experience of business personnel to design meta-path information for social robot detection;

[0079] Step 4. Build a new graph structure based on the meta-path, and use graph neural network modeling to obtain user node representations under different relationships;

[0080] Step 5. Fuse the user nodes under different relationships to obtain the final user representation;

[0081] Step 6: Build a social robot classifier and train it using the labeled social robot account data;

[0082] Step 7: Input the user representation to be classified into the classifier to obtain the final social robot detection result.

[0083] Specific process reference Figure 1 , described as:

[0084] 1. Text association mining: According to the needs of the text association mining task, a large language model prompt module is designed. The template consists of two parts: the first part is the system instruction (system), which is used to inform the large model of the input and output format requirements of the text association mining task. The large language model can use an existing open source large model, such as the MiniCPM-2B-sft-bf16 open source model; the second part is the input part, which assembles the two texts to be mined into the template. A typical text association mining task prompt template is as follows: [

[0086] {

[0087] "role":"system",

[0088] "content":"You are a text association analysis model. You need to analyze the input text pair and analyze the logical relationship between the two texts. The types of logical relationships include support, opposition and neutrality. Please give the classification results directly."

[0089] },

[0090] {

[0091] "role":"user",

[0092] "content":"Text 1: $sentence1$; Text 2: $sentence2$."

[0093] } ]

[0095] In the template, $sentence1$ and $sentence2$ are placeholders, representing the text pairs to be analyzed. The large model will return the logical relationship between the text pairs based on the input text pairs.

[0096] 2. Heterogeneous graph construction: All accounts, text content and topics under the event are uniformly modeled into a heterogeneous graph. Among them, the account node is the user (account) node of all users participating in the event, the content node is all the post information published under the event, and the topic node is all the topic label hashtag information under the event. When processing the content node, the present invention will pre-process the text, identify the URL, user mention (@) and other information therein and delete them, and remove the content nodes with a length less than 5 after pre-processing. When processing the topic node, the relevance of the topic and the event will be analyzed, and the hashtag label nodes that are not related to the event will be deleted. In terms of edges, the edge types between user-user nodes include attention, forwarding, and commenting; the edge type between user-content is a publishing relationship; the edge relationship between user-topic is a participation relationship, the edge relationship between content-content includes a support relationship and an opposition relationship, the relationship between content-topic is a subordinate relationship, and the relationship between topic-content is an inclusion relationship.

[0097] 3. Meta-path design: Based on the experience and knowledge of business personnel, meta-path information between different users can be defined from the above heterogeneous graph. The design of meta-path needs to consider the support or opposition relationship between different nodes. Meta-path information is used to characterize whether there is a certain relationship between two users. Meta-path information cannot determine the characteristics of users, but it determines whether there is a certain similarity between users. The characteristics of each user can be improved and strengthened by integrating other similar and mutual characteristics.

[0098] The user-oriented meta-path design will be given below.

[0099] <user, publication, text, affiliation, topic, inclusion, text, user,>: This meta-path describes the connection between pairs of users who publish content on the same topic;

[0100] <user, post, text, support, text, post, user>: This meta-path captures the connection between pairs of users holding the same position.

[0101] <user, post, text, objection, text, post, user>: This meta-path captures the connection between pairs of users holding opposite positions.

[0102] <user, follow, user>: This meta-path describes pairs of users that are connected by a social relationship.

[0103] <user, follow, user, forward, user>: This path describes the connection between pairs of users with high-order associations.

[0104] In addition to the above meta-paths, business personnel can also define new meta-paths based on business experience and add them to the analysis process.

[0105] 4. User node representation: A homogeneous user association graph is obtained based on the custom meta-path extracted from the constructed heterogeneous graph. Each node in the graph represents a user, and two nodes reachable by the meta-path are connected by an edge. The user association graph is modeled using a graph neural network to obtain user representation under a specific meta-path. Homogeneity means that in the user association graph, each node represents a user, and each edge represents the same similarity relationship, so they have the same properties.

[0106] When modeling graph neural networks, the user's original expression contains the following numerical information, classification information, and text information:

[0107] Numerical information including the number of followers, number of tweets, number of followers, username length, etc. are all converted into vectors in a unified semantic space through a multi-layer perceptron;

[0108] Classification information: including whether it is protected, whether it is authenticated, whether the profile contains a personal homepage, etc., which is converted into a vector in a unified semantic space through a multi-layer perceptron;

[0109] Text information: including all tweet information and self-introduction information, which is converted into vectors through pre-trained models (such as RoBERTa), and then mapped to a unified semantic space through a multi-layer perceptron.

[0110] The above vectors are concatenated to form the initial representation of the user, and the user association graph is modeled using a graph neural network. The graph neural network can aggregate the representations of other users reachable by the meta-path to enhance and improve the representation of the current user, and obtain the user representation under a specific meta-path.

[0111] 5. User representation fusion: After obtaining the user representations under different meta-paths, the self-attention mechanism is used to fuse the user representations under different meta-paths to obtain the complete representation of the user.

[0112] 6. Social robot classifier training: The classifier consists of a multi-layer perceptron network and is trained using labeled social robot account data. During the training process, for the labeled accounts, follow steps 1 to 5 to obtain their representation vectors as the input of the classifier; calculate the cross entropy loss based on the labeling results and the classifier's prediction results, and adjust the model parameters until the model converges.

[0113] 7. Social robot account inference: Input the user's complete representation into the classifier, and mark the account with a classifier output probability greater than 0.5 as a social robot account. The classifier needs to be trained in advance on a labeled dataset.

[0114] A social robot detection system based on heterogeneous graphs. Its business logic is shown in the attached Figure 2 , including at least the following modules:

[0115] Heterogeneous graph construction module: According to the analysis tasks specified by the user, including event titles, keywords and other information, the corresponding tweets and users are queried from the database, and a heterogeneous user relationship network is constructed based on the query results.

[0116] Feature-based detection model: To improve the operating efficiency of the system, for users with rich user features, the feature-based social robot detection model is directly called to infer the user type. This model uses the user's initial representation containing numerical information, classification information, and text information as user features, and is obtained after training using an annotated dataset.

[0117] Detection model based on user graph relations: For users with scarce user features, we combine the constructed heterogeneous graph and meta-path to obtain their representation and use a classifier to classify them.

[0118] Social robot detection database: contains multi-source information such as event information, account information, text content, structural features, etc., and also contains social robot detection results for each account.

[0119] The following is a system embodiment corresponding to the above method embodiment. This embodiment can be implemented in conjunction with the above embodiment. The relevant technical details mentioned in the above embodiment are still valid in this embodiment. In order to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied in the above embodiment.

[0120] like Figure 4 As shown, the present invention also proposes a social robot detection device based on a heterogeneous graph, which includes:

[0121] The initial module obtains user texts of specified events in social networks and uses a large language model to analyze implicit associations between texts;

[0122] The analysis module uses the account, content and topic of the user text as nodes to construct a heterogeneous graph under the specified event; extracts the feature representation of each account in the heterogeneous graph according to the preset meta-path; and uses the graph neural network to model the feature representation to obtain the user node representation under the meta-path; and fuses all the user node representations of the account to obtain the final representation of the account;

[0123] The classification module inputs the final representation into the classifier to obtain the detection result of whether the account belongs to a social robot.

[0124] The social robot detection device based on heterogeneous graph, wherein the initial module includes: inputting the user text into the large language model through a prompt template; the prompt template includes a system instruction for informing the large language model of specific task requirements and a pair of user texts of a specified event in the social network;

[0125] The edge types between account-account nodes in this heterogeneous graph include following, forwarding, and commenting; the edge type between account-content is publishing relationship; the edge relationship between account-topic is participation relationship; the edge relationship between content-content includes support relationship, opposition relationship, and irrelevant relationship; the relationship between content-topic is affiliation relationship; the relationship between topic-content is inclusion relationship;

[0126] The social robot detection device based on heterogeneous graph, wherein the meta-path includes: a meta-path for describing the connection between pairs of users who publish content on the same topic; and / or a meta-path for describing the connection between pairs of users holding the same position; and / or a meta-path for describing the connection between pairs of users holding opposite positions; the social robot detection method based on heterogeneous graph, wherein the feature representation includes numerical information, classification information and text information; wherein the numerical information includes the number of account followers, the number of tweets posted, the number of followers and the length of the user name; the classification information is identity authentication information; the text information includes: the account's self-introduction and tweets.

[0127] like Figure 5 As shown, the present invention further proposes a first electronic device A in another embodiment, which includes the social robot detection device described above, and the electronic device may be connected to an information display device.

[0128] like Figure 6 As shown, the first electronic device A can also be connected to the data acquisition device C and the information display device D through a wired or wireless information transmission scheme. The data acquisition device C is used to collect and obtain user text to be identified and classified, and the information display device D is used to display the detection results obtained by the analysis of the present invention.

[0129] The information display device D can sort and process the data output by the first electronic device A based on the information display mechanism to improve the readability of the data output by the first electronic device A. The information display mechanism can be manually preset, for example, the data output by the first electronic device A is visually displayed, which can be based on the display parameters and / or attributes set by the user. The display parameters can be, for example, the display data range, and the display attributes can be, for example, the display font, color, whether to scroll and play, etc. The user is presented with the key information specified by the user, such as whether the key account belongs to the detection result of the social robot, etc. The user can understand this information more timely without having to access the secondary page or scroll the page, saving the user's operation. Or the information display mechanism can be an artificial intelligence AI display model, which can learn the user's key information based on the user's previous usage habits, such as viewing time, number of clicks, number of edits, etc., and then automatically present rich and necessary key information to the user.

[0130] The present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a readable storage medium. When the computer program is executed by a processor, the computer can execute the social robot detection method based on heterogeneous graphs provided by the above-mentioned methods.

[0131] The present invention also proposes a storage medium VIII in another embodiment for storing a computer program for executing the social robot detection method based on heterogeneous graphs. It should be understood that the storage medium in the embodiment of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DR RAM).

[0132] Figure 7 A schematic block diagram of a second electronic device 1000 that can be used to implement an embodiment of the present invention is shown. The second electronic device 1000 electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The second electronic device 1000 can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples, and are not intended to limit the implementation of the present invention described and / or required herein. The second electronic device 1000 may be the same or different from the first electronic device A.

[0133] The second electronic device 1000 includes a computing unit I, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory II (ROM) or a computer program loaded from a storage medium VIII into a random access memory (RAM) III. In RAM III, various programs and data required for the operation of the device 1000 can also be stored. The computing unit I, ROM II, and RAM III are connected to each other via a bus IV. An input / output (I / O) interface V is also connected to the bus IV.

[0134] A plurality of components in the second electronic device 1000 are connected to the I / O interface V, including: an input unit VI, such as a keyboard, a mouse, etc.; an output unit VII, such as various types of displays, speakers, etc.; a storage medium VIII, such as a disk, an optical disk, etc.; and a communication unit IX, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit IX allows the second electronic device 1000 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0135] The computing unit I may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of computing unit I include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit I performs the various methods and processes described above, such as method steps S1-S3. For example, in some embodiments, the method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage medium VIII. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1000 via ROM II and / or a communication unit IX. When the computer program is loaded into RAM III and executed by the computing unit I, one or more steps of the method described above may be performed. Alternatively, in other embodiments, the computing unit I may be configured to execute the method in any other appropriate manner (e.g., by means of firmware).

[0136] 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 social robot detection method based on heterogeneous graph, characterized in that: include: In the initial step, we obtain user texts of specified events in social networks and use large language models to analyze implicit associations between texts; The analysis step is to construct a heterogeneous graph under the specified event with the account, content and topic of the user text as nodes; extract the feature representation of each account in the heterogeneous graph according to the preset meta-path; and use the graph neural network to model the feature representation to obtain the user node representation under the meta-path; and fuse all the user node representations of the account to obtain the final representation of the account; In the classification step, the final representation is input into the classifier to obtain the detection result of whether the account belongs to a social robot.

2. The social robot detection method based on heterogeneous graph according to claim 1, characterized in that: The initial step includes: inputting the user text into the large language model through a prompt template; the prompt template includes a system instruction for informing the large language model of specific task requirements and a pair of user texts of a specified event in the social network.

3. The social robot detection method based on heterogeneous graph according to claim 1 or 2, characterized in that: The edge types between account-account nodes in this heterogeneous graph include following, forwarding, and commenting; the edge type between account-content is publishing relationship; the edge relationship between account-topic is participation relationship; the edge relationship between content-content includes support relationship, opposition relationship, and irrelevant relationship; the relationship between content-topic is affiliation relationship; the relationship between topic-content is inclusion relationship; The meta-path includes: a meta-path for describing the connection between pairs of users who publish content on the same topic; and / or a meta-path for describing the connection between pairs of users holding the same position; and / or a meta-path for describing the connection between pairs of users holding opposite positions.

4. The social robot detection method based on heterogeneous graph according to claim 2, characterized in that: The feature representation includes numerical information, categorical information, and text information; the numerical information includes the number of followers of the account, the number of tweets posted, the number of followers, and the length of the username; This classified information is identity authentication information; The text information includes: the account’s self-introduction and tweets.

5. A social robot detection device based on heterogeneous graph, characterized in that: include: The initial module obtains user texts of specified events in social networks and uses a large language model to analyze implicit associations between texts; The analysis module uses the account, content and topic of the user text as nodes to construct a heterogeneous graph under the specified event; extracts the feature representation of each account in the heterogeneous graph according to the preset meta-path; and uses the graph neural network to model the feature representation to obtain the user node representation under the meta-path; and fuses all the user node representations of the account to obtain the final representation of the account; The classification module inputs the final representation into the classifier to obtain the detection result of whether the account belongs to a social robot.

6. The social robot detection device based on heterogeneous graph according to claim 5, characterized in that: The initial module includes: inputting the user text into the large language model through a prompt template; the prompt template includes a system instruction for informing the large language model of specific task requirements and a pair of user texts of a specified event in the social network; The edge types between account-account nodes in this heterogeneous graph include follow, forward, and comment; the edge type between account and content is a publishing relationship; the edge relationship between account and topic is a participation relationship, and the edge relationship between content and content includes a support relationship, an opposition relationship, and an irrelevant relationship. The relationship between content and topic is a subordinate relationship, and the relationship between topic and content is an inclusion relationship.

7. The social robot detection device based on heterogeneous graph according to claim 5 or 6, characterized in that: The meta-path includes: a meta-path for describing the connection between pairs of users who publish content on the same topic; and / or a meta-path for describing the connection between pairs of users holding the same position; and / or a meta-path for describing the connection between pairs of users holding opposite positions; the social robot detection method based on heterogeneous graphs is characterized in that the feature representation contains numerical information, classification information and text information; the numerical information includes the number of account followers, the number of tweets posted, the number of followers and the length of the user name; the classification information is identity authentication information; the text information includes: the account's self-introduction and tweets.

8. An electronic device, characterized in that: It includes a social robot detection device as described in claims 5-7, and the electronic device is connected to an information display device, and the information display device is used to display the detection result according to display parameters and attributes set by the user or through an artificial intelligence model.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the social robot detection method described in any one of claims 1-4.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the social robot detection method described in any one of claims 1-4 are implemented.

Images