109 results about "Social platform" patented technology

The invention provides a social robot detection method and system, computer equipment and a storage medium, and belongs to the technical field of social robot detection.The method comprises the steps that user multi-source data on a social platform and the social relation between users are collected; a hybrid encoder is adopted to capture local dependence and global time sequence dynamic states of behaviors, and user behavior characteristics are generated; aggregating structure attention features between the initial node features of the target account and the initial node features of the social relation account to obtain multi-modal relation aggregated structure features; introducing a multi-modal adversarial training strategy, adaptively adjusting the disturbance intensity of each modal based on gradient sensitivity, and aligning the characterization of the clean sample and the adversarial sample in combination with a content discriminator and a behavior discriminator; and outputting a target user classification result through the multi-task target function optimization model. According to the method, the problems of insufficient modal fusion and insufficient adversarial robustness of an existing method can be effectively solved, and the accuracy and stability of social robot detection in a complex and adversarial scene are improved.

The invention discloses a social group discovery system based on fused semantic information. The system comprises a presentation layer, a presentation layer, a service layer and a data layer. The presentation layer supports checking of a group division result; the display layer displays a user structure network, group dynamic division and group characteristic indexes through a visual tool; the service layer is composed of a semantic tag generation module, a user structure analysis module, a social group division module and a social group analysis module, semantic and structural features are fused through a graph attention network, and accurate group division is realized in combination with an adaptive embedded graph aggregation algorithm; and the data layer stores the original data set of the social platform and the processed data. The system inputs user texts into a large language model to generate semantic tags, node embedding is generated by calculating node association weights, group division is completed by combining an adaptive embedding graph aggregation algorithm, and finally group characteristics are quantitatively analyzed from the aspects of cohesion, conductivity and the like. The system is suitable for scenes such as vertical domain community mining.

The invention belongs to the technical field of network space security and social platform governance, and particularly provides an agent-driven active interactive social robot detection method, electronic equipment and a storage medium, and the method comprises the steps: obtaining original data of a to-be-detected account, and forming a structured data packet; constructing a preliminary screening evaluation model, and outputting a detection task state based on the structured data packet; generating an initial detection state; obtaining an interaction record; carrying out evidence extraction on the interaction record to obtain an evidence vector of the round; and performing sequential updating on the initial risk score and the uncertainty based on the evidence vector, and outputting a detection result when a stop condition is met. According to the method, an active interactive evidence obtaining mechanism of a detection agent is introduced, and on the basis of passive preliminary screening, probe interaction is automatically selected and initiated to obtain incremental evidence, so that the recognition capability of a high-simulation social robot is improved, and an auditing report containing key evidence and an interactive link is output.

The invention provides a doctor influence comprehensive evaluation method based on multi-modal data fusion, and the method comprises the steps: collecting media reports, clinical information, social platforms and other multi-source heterogeneous data, introducing timestamp marks, and achieving the time sequence semantic vector representation of a doctor entity through a medical field pre-training model and sine function embedding; in combination with BiLSTM-CRF and an attention mechanism, doctor attributes and key events are identified, and an LSTM and a dynamic clustering algorithm are adopted to extract and divide doctor attribute evolution trajectories in stages; an influence score is calculated through an exponential decay function, stage knowledge graph nodes are generated, and incremental updating and node merging and splitting are supported; a graph convolutional network and a graph attention mechanism are adopted, a doctor influence evolution chain is constructed, information dynamic association and trend prediction are achieved, and the time sequence precision and data comprehensiveness of doctor influence evaluation and the dynamic evolution ability of a knowledge graph are improved.

Systems and methods translate and govern directives across heterogeneous platforms. In some embodiments, directives are normalized into a common intermediate representation (CIR) for validation, conflict handling, and deployment. In other embodiments, a direct semantic mapping engine converts source directives directly into target-native, machine-executable directives without a CIR and without normalization, consulting a platform capability database and synthesizing target-compatible scripts, profiles, workflows, or rules where native primitives are absent. Ephemeral triggers (e.g., time window, incident, risk, posture, geofence) cause generation and atomic activation of temporary directives that supersede baseline directives for a defined scope and interval, with withdrawal on expiry to restore the baseline. Adapters actuate across families including cloud control planes, operating systems (Windows / macOS / Linux), IoT / OT, Kubernetes, identity, email / DLP, browser, and SaaS / social platforms. Both approaches share validation, simulation, and governance / deployment engines supporting approvals, audit logging, staged rollout, rollback, and continuous feedback.

The invention discloses a multi-modal content generation method based on a diffusion model and multi-dimensional feature fusion, which comprises the following steps: S1, collecting image-text data from a plurality of social platforms, converting the image-text data into structured data, and performing multi-dimensional attribute labeling on the basis of a pre-training language model; s2, carrying out feature coding on an input text and a label as well as an image and a label thereof by adopting a CLIP encoder and an Embedding and MLP module respectively, and carrying out multi-modal fusion on all feature codes; s3, constructing a pseudo Prompt embedding vector; on the basis of a diffusion generation model SDXL and in combination with a LoRA fine tuning technology, personalized fine tuning of generated contents in a social scene is realized, and a high-quality image generation model is constructed; s4, based on the trained student model and a sampling acceleration strategy, finally quickly generating multiple groups of images with consistent semantics and different styles; and S5, establishing a multi-dimensional automatic image generation quality evaluation system, and performing reliability verification on the automatic image generation quality evaluation system based on expert review. According to the invention, high-quality rapid generation of the multi-modal content is realized.

The invention relates to the technical field of artificial intelligence and social media public opinion monitoring, in particular to an image recognition method and system based on machine learning, and the method comprises the steps: carrying out the OCR text extraction and regional visual feature extraction of an input image, generating a text emotion embedding vector in combination with a preset brand emotion dictionary, and carrying out the recognition of the emotion embedding vector; and visual and text features are fused through a cross-modal adversarial encoder to obtain joint semantic embedding representation, and whether hidden negative public opinions are formed or not is judged through a brand deviation degree discriminator. According to the method, semantic antagonism image contents including derogatory characters, competitive product comparison or sensitive symbols and the like can be effectively identified, the detection rate of recessive negative public opinions is improved, and the method is suitable for real-time monitoring of a large-scale social platform.

The application discloses a kind of important node identification methods of multi-view based graph neural network, obtain user information from social platform;According to user information, all users are represented as nodes, the relationship between users is represented as edge, and the relevant attributes of users are represented as features, to establish network graph structure;Extension is multi-view, data is described from different angles, and the embedding of nodes is obtained using the method of graph neural network on different views;The relationship between nodes is represented using high-order topological structure matrix, and the final representation of nodes is obtained by using attention method to fuse multi-view information, to predict the importance of nodes, to obtain end-to-end training model, to identify important nodes;The application not only can improve the expression ability of model, but also can be generalized to data without obvious graph structure, extract new relationship, establish graph structure, use graph neural network model, and improve the performance of important node identification in network.

Systems and methods for providing opportunities for interactions and communication between vehicles and passengers. In particular, a vehicle socialization platform is provided for facilitating a wide range of social and utilitarian exchanges between vehicles and their occupants, thus providing a community of interconnected vehicles. Vehicle interactions can include passive communication, including vehicles from a particular fleet acknowledging other fleet vehicles without intervention from a passenger or user. Vehicle interactions can include active communication, in which passengers are engaged with and have some control over vehicle communications through dedicated experiences provided in a ridehail application or on an in-vehicle tablet. In particular, some experiences can become dynamically available based on the fleet vehicle's proximity to another fleet vehicle. The vehicle socialization platform is a smart socialization platform that enables two-way communication between vehicles and passengers, facilitating a wide range of social and utilitarian exchanges between vehicles and vehicle occupants.

The application discloses a multi-round multi-modal toxicity detection method and device, and relates to the technical field of natural language processing and computer vision. The method comprises the following steps: obtaining multi-round dialogue data of a Chinese social platform, constructing a multi-modal dialogue data set containing text and images and performing preprocessing to obtain a processed multi-modal data set; designing a multi-dimensional labeling system, performing multi-dimensional toxicity type labeling on each dialogue, and calculating a labeled real cumulative toxicity score; constructing a systematic evaluation framework, and setting a toxicity content evaluation task; based on a test sample data set and the set toxicity content evaluation task, performing toxicity detection on the test sample data set by using a pre-trained visual language model, outputting a cumulative toxicity score predicted by the pre-trained visual language model, and calculating a cumulative toxicity detection accuracy; and based on the cumulative toxicity detection accuracy, performing evaluation, and outputting a model evaluation result. The application can improve the evaluation of original visual content understanding ability.

The application provides a trust evaluation method of a multi-layer relationship-aware graph neural network, belongs to the technical field of social network data processing and graph learning, and aims to solve the problems of mixed multi-source trust information, insufficient modeling of nonlinear interaction between nodes, and insufficient utilization of multi-layer graph convolution features of existing trust evaluation methods in sparse social networks. The method constructs a directed trust graph, uses a relationship-aware encoder to model the node's own features and neighborhood features, combines an adaptive propagation module to represent the directional and conditional dependence relationship between nodes, and obtains the final node pair representation through cross-layer attention fusion to realize trust evaluation between target node pairs. The method can be applied to social platform risk control, recommendation system credibility analysis, transaction network reputation evaluation and other scenes.

The invention provides a Chinese intention recognition method based on a cue word template, and relates to the field of network security in computer science and technology. The method comprises the following steps: segmenting a to-be-recognized text to obtain a token sequence; replacing attitude word tokens in the token sequence with special tokens [MASK] through a cue word template, and respectively adding special tokens [CLS] and [SEP] at the left end and the right end to obtain an input token sequence; inputting the input token sequence into a multi-layer Transform encoder to obtain a text token vector, a [MASK] special token vector and a [CLS] special token vector; averagely pooling the text token vector and the [MASK] special token vector to obtain a text pooling vector and a [MASK] pooling vector; sequentially splicing the [CLS] special token vector, the text pooling vector and the [MASK] pooling vector to obtain a final representation vector of the to-be-recognized text; and inputting the final representation vector into the MLP, and identifying the Chinese intention of the to-be-identified text. According to the method, accurate recognition of the Chinese text intention in the social platform is realized.

The invention relates to the technical field of network security, and discloses a social robot collaborative group detection method based on a multi-order collaborative knowledge graph. According to the method, user data in a Chinese social platform is collected through a crawler, a multi-order collaborative interaction graph is constructed based on the interaction relation of users of the social platform and then input into a graph encoder module, node feature vectors are obtained and input into a time perception module and a graph perception module respectively, and a time sequence prediction task and a side link prediction task are executed. Meanwhile, the model is constructed based on a time-graph semantic consistency perception module of a self-attention mechanism. And finally, inputting a node feature vector obtained by training under the guidance of double tasks into a group detection module constructed based on a Gaussian mixture model and a K-means clustering method for detection. According to the method, the complex interaction relationship among the users in the social robot group and the abnormal mode in the time sequence activity can be effectively captured, and efficient detection of the social robot collaborative group in the social network platform is realized.

The invention provides an image tampering positioning method and device based on an improved SAM model. The image tampering positioning method based on the improved SAM model comprises the steps that a corresponding relation between image features and a predicted tampering positioning result and a corresponding relation between prompt features and the predicted tampering positioning result are established through the SAM-Adapter model; obtaining a to-be-detected image, and extracting a current image feature of the to-be-detected image; determining a target tampering pixel coordinate in the to-be-detected image based on a preset tampering prompt point acquisition strategy, and generating a current prompt feature according to the target tampering pixel coordinate; and determining a target prediction tampering positioning result corresponding to the current image feature and the current prompt feature through the corresponding relationship. The feature extraction capability of tampering traces is improved through an SAM-Adapter model, the problem of cross-dataset performance reduction is solved, and the tampering positioning requirement of social platform images in a real scene is met; through a prompt point acquisition strategy, a tampering area does not need to be known in advance, a pain point without a prior prompt point is solved, and the method can be directly used for an actual detection scene.

The present application relates to a kind of based on platform recommendation utility characterization implementation social platform information propagation prediction method, comprising the following steps: generating social content representation and social user representation;Social platform recommendation vector is built;According to network structure channel, the propagation process is predicted;Fusion platform recommendation channel;Decoding obtains subsequent time infected user.The present application also relates to a kind of based on platform recommendation utility characterization implementation social platform information propagation prediction device, processor and its computer readable storage medium.The momentum updating mode is used to iteratively update platform recommendation vector in the present application based on platform recommendation utility characterization implementation social platform information propagation prediction method, device, processor and its computer readable storage medium, can keep the stability of feature semantic distribution, using the idea of gated network, platform recommendation mechanism is regarded as parallel channel with friend propagation, when carrying out propagation prediction, the influence of platform recommendation mechanism is introduced.

This invention discloses a method for optimizing the allocation of epidemic prevention and control resources based on complex networks. The method includes: acquiring big data from online social platforms to measure participants' attitudes and opinions on epidemics and corresponding prevention and control measures; designing and executing specific decision-making experiments to measure participants' feedback to different types of incentives and information; analyzing experimental results and data to extract individual behavioral rules in different research scenarios and constructing a new generation of agent-based epidemic transmission models; conducting multi-layer community analysis based on deep learning to achieve accurate community discovery within a multi-layer framework; combining community discovery with impact maximization research to identify key user sets in the multi-layer system, allocate prevention and control resources, explore the evolutionary patterns of epidemics, and achieve optimized control of the epidemic transmission process under multi-dimensional uncertainties in parameters and models. This invention contributes to the rational allocation of epidemic prevention and control resources, providing effective protection for the lives and health of the people.

The application discloses a red envelope flow control method and system based on a social platform, and relates to the technical field of computers.The application dynamically divides user levels based on a social graph, so that high-value users in a star fan group can obtain priority access rights, the requests of core fans can still be stably reached in a flow peak, and the risk of loss of high-value users is reduced.Rigid concurrent control is converted into dynamic probability decision-making, the passing rate under a node overload scenario is automatically inhibited through a load-sensitive function, and global lock competition caused by atomic operation is avoided.A random decision mechanism disperses request peak pressure, maintains millisecond-level response capability, and guarantees the stability of the system under the condition of instantaneous concurrent access of ten thousand people.Meanwhile, a delay retry channel associated with a social level is enabled for limited requests, buffer opportunities are provided for core propagators, a timeout compensation strategy is combined, negative emotions of users caused by direct rejection are reduced, the integrity of a red envelope fission path is maintained, and positive circulation of social propagation is strengthened.

The application discloses a kind of federal social recommendation method and device based on adaptive diffusion denoising, belong to personalized recommendation technical field in electronic commerce and social platform. For the problem of serious noise interference, server load is too high and slow response in existing social federal recommendation, the application constructs interaction graph and social graph by collecting user interaction and social data, initializes user and project embedding representation, carries out adaptive diffusion denoising process in client local, wherein noise scheduling is dynamically adjusted according to user social structure characteristics, and through federal training framework, client local model updating and server global parameter aggregation are realized, finally, denoised social embedding and user interaction embedding are fused to carry out recommendation score prediction and list generation. The application effectively suppresses social noise interference, improves recommendation accuracy and robustness, reduces server load and communication overhead, and is suitable for personalized recommendation service of electronic commerce and social platform.

The invention provides a social media agent simulation method based on dynamic role information, and the method comprises the steps: employing a large language model LLM to expand a role template to generate basic attributes and advanced attributes, and enabling the basic attributes and advanced attributes to serve as a role file of an agent; knowledge matched with the role file is captured from the general knowledge data set and serves as long-term memory; the intelligent agent generates a dynamic plan according to the role file and the action history; the intelligent agent obtains the posts from the social platform according to the dynamic plan, retrieves the advanced attributes according to the post content, judges whether the post content accords with the advanced attributes or not, and executes interaction actions including like, forwarding and / or commenting if the post content accords with the advanced attributes; and when the agent posts on the social platform, querying long-term memory according to a posting theme, calculating text similarity between a query result and personalized knowledge, and selecting the query result of which the similarity is greater than a preset value to generate posting content. Therefore, the intelligent agent can make reasonable actions according to the role, and the posting and comment content is highly consistent with the role information.

The present application provides a social user implicit relationship recommendation method and device, the method comprises the following steps: obtaining user data of multiple social platforms; according to the user data, a plurality of groups of heterogeneous information graphs based on user social information are constructed; according to the heterogeneous information graph, clustering processing is carried out respectively through a double information based multi-view attribute graph clustering method, a rule based clustering method and an attention mechanism based graph neural network recommendation clustering method, three kinds of clustering results are obtained; the three kinds of clustering results are fused to obtain a target recommendation result, which can improve the efficiency of recommendation in the user social network.

The embodiment of the invention provides a vehicle-mounted interaction system, method and device, a vehicle end, a cloud end and a vehicle. The vehicle-mounted interaction system comprises an interaction layer used for receiving demand information sent by a user side through a server of a social platform; wherein the demand information is information input by the user on the social platform; the understanding layer is used for performing semantic recognition on the demand information and portrait data of the user based on a large model service cluster of the cloud to obtain a user intention of the user; the decision-making layer is used for performing task splitting on the user intention of the user based on the cloud to obtain at least one sub-task instruction; and the execution layer is used for executing the subtask corresponding to the subtask instruction based on the vehicle side and / or the cloud side. According to the vehicle-mounted interaction system, an efficient, safe and personalized interaction link can be realized, and the use experience of a user is improved.

The application relates to a kind of social platform robot detection methods based on multi-agent, comprising: creating social platform API connection instance to collect target user comment history data to obtain original data set;Build the orchestrator-multi-agent collaborative detection model driven by large language model;Data processing agent constructs comment level relationship to input data, and carries out cleaning, standardization and quality detection, forms the structured data set after preprocessing and writes into shared memory;Feature engineering agent reads structured data from shared memory, calculates user-level multi-dimensional behavior feature vector and writes into shared memory;Rule discrimination agent reads feature data, executes rule discrimination, judges whether feature hits piece by piece, thereby outputs rule risk score and writes into shared memory;Modeling agent reads feature data, trains anomaly detection and supervised classification model, executes multi-model fusion to obtain robot risk determination result and writes into shared memory;Explanation generation agent reads model detection result and rule data, and fusion outputs the final interpretable social robot detection result.The application reduces the dependence on single rule or single model by adopting multi-agent cooperation to complete data processing, feature construction, rule discrimination and multi-model fusion detection, greatly improves the accuracy and robustness of social platform robot detection.

The application provides a kind of online social platform topic propagation topology fingerprint graph construction method, comprising: constructing topic propagation network graph, determining source node, extracting maximum connected subgraph, calculating subgraph each node forwarding volume, node hop count, adjacent edge forwarding intensity;From each subgraph source node, with forwarding intensity as weight, the shortest weighted path from source node as filter function, forward search out connected component sequence, record each connected component start time step and duration, the same connected component is merged;From each subgraph recorded search end node, start reverse search, get ring sequence from reverse search, record each ring structure start time step and duration, the same ring structure is merged;According to connected component sequence set, ring sequence set constructs topic propagation fingerprint graph.The application can effectively extract the significant topology of connected path and loop formed in the process of online social platform topic propagation, and quantify its duration.

The invention belongs to the technical field of unreal information detection, and particularly relates to an unsupervised unreal information detection method based on group intelligence and content cross inspiration. Comprising the steps that blog articles and comment data are acquired from a social platform, and a blog article-comment interaction network is constructed; constructing prompt words according to the blog article-comment interaction network; according to the cue word, processing the blog article and the comment data by adopting a large language model to obtain an attitude division result of the comment; according to an attitude division result, quantifying an attitude vector of the blog article; semantic embedding is extracted from the blog article; extracting a semantic center vector from the comments corresponding to the blog article according to the attitude division result of the comments; fusing the attitude vector of the blog article, the semantic embedding and the semantic center vector of the comment corresponding to the blog article to obtain comprehensive feature representation of the blog article; performing clustering processing on all the blog articles according to the comprehensive feature representation of each blog article to obtain an unreal information detection result; the method can effectively overcome the limitation of a supervised learning method on unknown unreal information identification, and improves the accuracy of an unreal information detection result.

The application relates to a conversation message processing method and device, computer equipment and a storage medium. The method comprises the following steps: in a first social application, a first conversation window of a conversation between a first conversation object and a second conversation object is displayed; in the first conversation window, a message card of a media conversation message generated by the second conversation object forwarding a target media object belonging to a media social platform into the conversation through a second social application is displayed; the message card displays thumbnail description information of the target media object; in the case that the media conversation message is archived after authorization of the second conversation object, in response to an archiving query instruction for the conversation, text format conversation archive information of the conversation is displayed; the conversation archive information comprises message archive information of the media conversation message, and the message archive information comprises object description information matched with the thumbnail description information and used for describing the target media object. The method can widen the archiving mode of the conversation message.

The application discloses a short video abnormal behavior recognition method based on user behavior sequence, relates to the risk control technical field, and specifically comprises the following steps: obtaining a comprehensive behavior sequence of a user, containing various attribute features of the user on a short video social platform; adopting a deep interest network DIN as a training model, and dividing the user behavior sequence into continuous value behavior features and discrete value behavior features; the continuous value behavior features and the discrete value behavior features are sorted in time sequence, the id and the device of the user are taken as conditional input, and then the deep interest network DIN makes accurate classification according to different users under the behavior sequence features; obtaining the continuous value behavior features and the discrete value behavior features of the user, based on the short video user behavior sequence recognition network structure of the deep interest network DIN, performing associated learning between the user behavior sequence features through an Attention structure and a full connection FC structure, and identifying water army users through behavior similarity and characteristics.

The invention discloses a content delivery method and device of a social platform, a storage medium and equipment, and belongs to the technical field of deep learning. Acquiring a node to be scheduled in the strategy map, wherein the node is generated according to the delivery target and is associated with a plurality of delivery strategies; according to the platform data of the social platform, generating a platform ecological snapshot and a target crowd coverage degree; generating an account adaptation degree according to the release record of the user account; according to the platform ecological snapshot, the target population coverage degree and the account adaptation degree, screening the most adaptive n triads, wherein each triad comprises a delivery strategy, a social platform and a user account; and performing content delivery by using a corresponding user account in a corresponding social platform by adopting a delivery strategy corresponding to the triple. According to the method, the optimal'strategy-platform-account 'combination can be dynamically screened out, the content triggering precision and the interactive conversion efficiency are improved, the scheduling weight corresponding to the delivery effect is updated to the node, the node is scheduled according to the scheduling weight, and autonomous optimization of delivery is realized.

The invention relates to the technical field of computers, and discloses an intelligent design and propagation adaptation method and system for non-perpetual element digital cultural and creative products. The method comprises the following steps: digitally extracting non-abandoned patterns, processes and stories through image recognition and natural language processing, and constructing a structured database; based on user preferences and market trends, utilizing a generative adversarial network to generate a multivariate cultural and creative design scheme; according to platform characteristics of short videos, e-commerce, social contact and the like, a multi-modal content package is automatically adapted through a scenarized propagation model; and user feedback data is combined to reinforce a learning mechanism to optimize a design and propagation strategy. The system comprises a non-residual element database, a generative design module, a propagation adaptation model and a feedback optimization unit. By adopting the technical scheme, the design threshold can be reduced, the non-immature transformation efficiency and propagation precision can be improved, and digital inheritance and innovative propagation of non-immature culture can be assisted.

The embodiment of the invention discloses a social platform content publishing method and device. According to the method, an end-to-end automatic publishing link of content shooting, processing and publishing is realized through a voice interaction function of the intelligent glasses. Meanwhile, according to the method, voiceprint authentication and equipment holding authentication (the intelligent glasses are bound with a social platform account of the user) are combined to form double-factor authentication, so that an efficient, convenient and safe intelligent glasses content ecological closed loop is created, and seamless and non-inductive'zero operation 'experience is provided for the user. The social platform content publishing device disclosed by the embodiment of the specification also has the above beneficial effects.

The invention provides a social event discovery method based on hierarchical multi-relation graph and text fusion, and relates to the technical field of data mining, and the method comprises the steps: firstly collecting social platform message flow data to construct a multi-relation message graph; then hierarchical graph coding is carried out to output node structure representation; performing two-way alignment on the sentence-level semantic representation and a text sequence, and extracting sentence-level semantic representation; then event awareness contrast learning is carried out, and model parameters are optimized through overall loss; and finally, clustering by using a DBSCAN algorithm, updating parameters according to an increment period, and reserving historical knowledge. According to the method, multi-dimensional message data of a social platform is integrated, a multi-relation message graph is constructed and hierarchically coded, and structure and semantic features are fused; through cooperation of event perception comparative learning and a dynamic clustering strategy, sample distinction degree is enhanced, dynamic change is adapted by adopting an incremental maintenance mechanism, event discovery timeliness and accuracy are improved, and complex scene requirements of a social platform are met.