Generative content processing method and device, equipment, medium and program product

Abstract

The invention discloses a generative content processing method and device, equipment, a medium and a program product, and relates to the technical field of data processing.The method comprises the steps that target generative content issued by a player in a game and a content identification code corresponding to the target generative content are obtained; identifying the content type of the target generative content, and performing feature extraction on the target generative content of the content type to obtain feature data corresponding to the target generative content; and based on a content verification strategy corresponding to the content identification code, performing exception verification on the feature data to obtain exceptional content corresponding to the target generation type content. By implementing the technical scheme of the invention, automatic anomaly detection of the generative content published by the players in the game can be realized, and the content security of the game community is guaranteed.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to methods, apparatus, equipment, media, and program products for processing generative content. Background Technology

[0002] Game content security governance is a core element in maintaining a healthy ecosystem for virtual communities. With the increasing openness of gaming platforms, the amount of multimodal content generated by players, including text, images, audio, and video, is increasing dramatically, placing extremely high demands on the real-time performance and accuracy of content security governance systems. Currently, the industry generally adopts a "review before release" mechanism, which involves using a feature extraction engine to obtain content features and then submitting them to a strategy engine for risk decision-making to achieve rapid interception of illegal content.

[0003] However, current content verification solutions typically employ a one-size-fits-all approach or require deploying separate verification services for different content types or business scenarios. This approach lacks a flexible and unified routing mechanism, failing to quickly match and execute the most suitable verification strategy based on the specific attributes of the content and its source context. This results in scalability, limited granular operation, and a significant increase in the complexity and cost of strategy management when facing complex and ever-changing business needs. Summary of the Invention

[0004] In view of this, this application provides a method, apparatus, device, medium, and program product for processing generative content, in order to solve the problems of poor flexibility and low efficiency of current content security verification methods.

[0005] In a first aspect, this application provides a method for processing generative content, including: obtaining target generative content published by a player in a game, and the content identifier code corresponding to the target generative content; identifying the content type of the target generative content, extracting features from the target generative content of the content type, and obtaining feature data corresponding to the target generative content; and performing anomaly verification on the feature data based on the content verification strategy corresponding to the content identifier code, and obtaining abnormal content corresponding to the target generative content.

[0006] In one optional implementation, based on the content verification strategy corresponding to the content identifier code, anomaly verification is performed on the feature data to obtain the abnormal content corresponding to the target generative content, including: determining the content verification strategy corresponding to the content identifier code based on the scenario corresponding to the content identifier code; and performing anomaly verification on the feature data according to the content verification strategy to obtain the abnormal content corresponding to the target generative content.

[0007] In one optional implementation, the content verification strategy is generated by: obtaining strategy configuration information corresponding to each content identifier code; and generating a content verification strategy corresponding to each content identifier code based on the filling result of the strategy configuration information in a preset strategy template.

[0008] In one optional implementation, policy update information is obtained; a preset policy template is populated according to the policy update information to generate an updated content verification policy.

[0009] In one optional implementation, obtaining the policy configuration information corresponding to each content identifier code includes: generating the policy configuration information corresponding to the content identifier code in response to a policy configuration event triggered on the game server for the content identifier code.

[0010] In one optional implementation, based on the filling result of the policy configuration information in the preset policy template, a content verification policy corresponding to each content identifier code is generated, including: for any content identifier code, filling the policy configuration information corresponding to the content identifier code into the preset policy template to obtain an initial policy file corresponding to the content identifier code; pre-compiling the initial policy file to generate a policy object corresponding to the content identifier code; constructing a mapping relationship between each content identifier code and each policy object, and determining the content verification policy corresponding to the content identifier code based on the mapping relationship.

[0011] In one optional implementation, the memory usage of policy objects and the usage frequency of each policy object are monitored; when the memory usage exceeds a preset threshold, the policy object with the lowest usage frequency is eliminated.

[0012] In one optional implementation, feature extraction is performed on the target generative content of the content type to obtain feature data corresponding to the target generative content, including: determining a content analysis model corresponding to the content type; using the content analysis model to perform semantic understanding of the expressive intent of the target generative content to obtain predictive information characterizing the existence of anomalies in the target generative content; wherein, the feature data includes predictive information and anomaly type.

[0013] In one optional implementation, the target verification strategy that the abnormal content hits is obtained; when the number of hits of the target verification strategy reaches a preset circuit breaker value, the circuit breaker result is obtained; the hit statistics of the target verification strategy, the circuit breaker result, and the hit target verification strategy are encapsulated to obtain the abnormal judgment result of the abnormal content.

[0014] Secondly, this application provides a generative content processing apparatus, comprising: a first acquisition module, used to acquire target generative content published by a player in a game, and a content identifier code corresponding to the target generative content; an extraction module, used to identify the content type of the target generative content, perform feature extraction on the target generative content of the content type, and obtain feature data corresponding to the target generative content; and a verification module, used to perform anomaly verification on the feature data based on the content verification strategy corresponding to the content identifier code, and obtain abnormal content corresponding to the target generative content.

[0015] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the generative content processing method of the first aspect or any corresponding embodiment described above.

[0016] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the generative content processing method of the first aspect or any corresponding embodiment described above.

[0017] Fifthly, this application provides a computer program product, including computer instructions for causing a computer to execute a processing method for generative content of the first aspect or any corresponding embodiment described above.

[0018] The generative content processing method provided in this application, by acquiring the target generative content and its content identifier code, provides clear context and classification basis for subsequent processing, ensuring the relevance and traceability of content identification. By identifying content types and extracting features, the target generative content is transformed into standardized feature data, objectifying and quantifying subjective and complex content information, providing a computational basis for subsequent judgment, and significantly improving the standardization and efficiency of processing. By performing anomaly verification on the feature data based on the content verification strategy corresponding to the content identifier code, the most suitable verification strategy can be flexibly adapted according to different content sources or business scenarios, rather than using a single judgment standard. This achieves high configurability and refined management of the verification process, ultimately efficiently outputting the judgment results of abnormal content. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of this application, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application; Figure 2 This is a schematic flowchart of a first method for processing generative content according to an embodiment of this application; Figure 3 This is a second flowchart illustrating a method for processing generative content according to an embodiment of this application; Figure 4 This is a schematic diagram of a strategy configuration interface according to an embodiment of this application; Figure 5 This is a schematic diagram of a third process for processing generative content according to an embodiment of this application; Figure 6 This is a structural block diagram of a generative content processing apparatus according to an embodiment of this application; Figure 7 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0023] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0024] As one optional application scenario in this application embodiment, Figure 1 This diagram illustrates an application scenario of a generative content processing system. For example... Figure 1 As shown, the system may include at least one terminal device and at least one server. Figure 1The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0025] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0026] Taking casual games as an example, the terminal device has the game's client application installed. In the game's virtual world, players can not only control virtual characters to move and interact through operation controls, but also freely create and publish maps in the game, and publish various forms of generative content such as text, images, voice, and even video in chat channels and personal spaces to achieve personalized expression and social interaction.

[0027] However, this highly free environment for content creation and publishing also presents significant challenges to content security governance in games. Player-generated content exhibits diverse modalities and complex scenarios, and related technologies typically employ single or fixed strategies for content verification. This makes it difficult to implement differentiated and refined risk identification for different scenarios (such as public chat versus private spaces) and different content types (such as text versus images). This "one-size-fits-all" verification approach easily leads to high false positive rates and missed detections of risky content, impacting not only the experience of regular players but also hindering the effective maintenance of a clean and healthy gaming community.

[0028] The generative content processing method provided in this application, by assigning specific content identification codes to player-published content and dynamically matching corresponding content verification strategies based on these identification codes, achieves refined and scenario-based governance of user-generated content within the game. It can automatically invoke the most suitable verification strategy according to the source, type, and business scenario of the content, without the need to deploy independent services for different scenarios. While improving the accuracy and efficiency of risk identification, it greatly enhances the flexibility and scalability of the system, providing an effective solution for the security governance of large-scale, multimodal UGC (User Generated Content).

[0029] According to an embodiment of this application, a method for processing generative content is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0030] This embodiment provides a method for processing generative content, which can be used in electronic devices, such as computer 101, mobile terminal 102, etc. Figure 2 This is a flowchart of a generative content processing method according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps: Step S201: Obtain the target generative content published by the player in the game, and the content identifier code corresponding to the target generative content.

[0031] Targeted generated content refers to user-generated content (UGC) published by players in the game, including multimodal content such as text, images, audio, and video. This content is freely created by players and may contain inappropriate information. A content identifier is a unique code used to identify the business context of the content; it may contain metadata such as vendor information, game product type, and scene (e.g., chat channel, user profile page). Specifically, in the game environment, players can publish generated content through various channels, such as chat boxes, image uploads, voice messages, or video sharing. This content is captured by the game client and transmitted to the game server over the network. The process of acquiring targeted generated content is automated. The game server monitors player activity in real time, and when it detects a player attempting to publish content, it immediately captures the content data (such as text strings, image files, or audio streams) and marks it as targeted generated content to be processed. Simultaneously, to distinguish content from different sources and scenes, a unique content identifier is assigned to each piece of content. This content identifier is generated based on the game's business logic and may include vendor information, game product type, specific scene, and other contextual parameters. For example, on a multi-game platform, content identifiers can help identify which game, server, and interaction scenario the content originates from, thus ensuring that the content is correctly routed to subsequent processing flows. The generation of content identifiers relies on the game's backend configuration management; operators can predefine the rules for the identifiers, enabling the system to quickly match appropriate security policies.

[0032] Step S202: Identify the content type of the target generative content, extract features from the target generative content of the content type, and obtain the feature data corresponding to the target generative content.

[0033] Content type refers to the modality or form of the target generative content, such as text, image, audio, video, etc. Feature data refers to the feature values ​​extracted from the generative content, such as a feature vector or a set of probability scores, representing the confidence level of the content belonging to various risk categories. Specifically, it can be automatically classified by analyzing the content's metadata or format. For example, text content may be identified as plain text or rich text based on file extensions or header information, while images and videos are distinguished by file header tags as JPEG, PNG, or MP4 formats. Once the type is determined, a multimodal feature extraction engine is invoked for deep analysis. The feature extraction process deconstructs and interprets the target generative content, transforming it from unstructured raw data into a series of structured indicators that can be understood and processed by computers. These indicators aim to quantify the performance of the target generative content across different risk dimensions. For example, a piece of content may be parsed into multiple feature items, each representing its strength or probability on a specific attribute. Ultimately, these feature items are organized into a structured feature vector or dataset, which no longer contains the original, directly readable content, but rather carries the abstract results after analysis and computation. This feature dataset is essentially a "risk profile" of target-generated content. It unifies complex and diverse user-generated content into a standardized data format that allows for high-speed and accurate logical judgment.

[0034] Step S203: Based on the content verification strategy corresponding to the content identifier code, perform anomaly verification on the feature data to obtain the abnormal content corresponding to the target generative content.

[0035] Content verification strategies refer to predefined rule sets used to determine whether content is abnormal based on feature data. Abnormal content refers to content identified as illegal or high-risk through content verification strategies. Specifically, corresponding verification rules are loaded from a pre-configured strategy library based on the content identifier code. These rules are configured by operations personnel and exist in the form of logical conditions, clearly defining under what circumstances various feature data will be judged as abnormal. During the verification process, feature data is used as input, and its internal rule matching algorithm (such as the Rete algorithm) compares the data with the rule conditions in real time. Rules can support threshold judgments for single feature items or complex logical combinations of multiple feature items. Once feature data meets the conditions set by one or more rules, the rule is "hit," and a judgment result is generated. This result is encapsulated into a standard response. Based on this response, the identified abnormal content is marked, and subsequent processing procedures such as content blocking and reporting are automatically triggered, thereby efficiently and automatically completing the security governance of game content.

[0036] The generative content processing method provided in this application, by acquiring the target generative content and its content identifier code, provides clear context and classification basis for subsequent processing, ensuring the relevance and traceability of content identification. By identifying content types and extracting features, the target generative content is transformed into standardized feature data, objectifying and quantifying subjective and complex content information, providing a computational basis for subsequent judgment, and significantly improving the standardization and efficiency of processing. By performing anomaly verification on the feature data based on the content verification strategy corresponding to the content identifier code, the most suitable verification strategy can be flexibly adapted according to different content sources or business scenarios, rather than using a single judgment standard. This achieves high configurability and refined management of the verification process, ultimately efficiently outputting the judgment results of abnormal content.

[0037] This embodiment provides a method for processing generative content, which can be used in electronic devices, such as computer 101, mobile terminal 102, etc. Figure 3 This is a flowchart of a generative content processing method according to an embodiment of this application, such as... Figure 3 As shown, the process includes the following steps: Step S301: Obtain the target generative content published by the player in the game, and the content identifier code corresponding to the target generative content. For details, please refer to [link to relevant documentation]. Figure 2 Step S201 of the illustrated embodiment will not be described again here.

[0038] Step S302: Identify the content type of the target generative content, extract features from the target generative content of each content type, and obtain the feature data corresponding to the target generative content. For details, please refer to [link to details]. Figure 2 Step S202 of the illustrated embodiment will not be described again here.

[0039] Step S303: Based on the content verification strategy corresponding to the content identifier code, perform anomaly verification on the feature data to obtain the abnormal content corresponding to the target generative content.

[0040] Specifically, step S303 includes: Step S3031: Based on the scenario corresponding to the content identifier code, determine the content verification strategy corresponding to the content identifier code.

[0041] When processing specific generated content, the system first parses its accompanying content identifier, which essentially defines the business context of the content, such as which game, server partition, or content distribution channel it originates from. Then, this content identifier is used as a query key to send a request to the policy configuration center. Based on pre-configured mapping logic, the policy configuration center quickly retrieves and returns all valid content verification policies that match the context represented by the identifier. These policies constitute the rule set for security verification of that specific content.

[0042] Step S3032: Perform anomaly verification on the feature data according to the content verification strategy to obtain the abnormal content corresponding to the target generative content.

[0043] After obtaining the set of content verification strategies applicable to the current scenario, these strategy rules are loaded into the execution environment of the rule engine. Next, the extracted structured feature data is injected into the rule engine as input. The rule engine iterates through all strategy rules in parallel, quickly matching and logically calculating the values ​​in the feature data against the conditions set by each rule. If the feature data meets the conditions of one or more strategy rules, it means a violation judgment has been triggered, and the rule is "hit." The rule engine then outputs a verification result containing a hit rule identifier, risk level, and suggested remedial measures. Based on this result, the target generated content can be clearly marked as abnormal content.

[0044] The generative content processing method provided in this application significantly enhances the accuracy and adaptability of the content verification process by introducing a scenario-based strategy matching mechanism. Specifically, this application does not treat content verification strategies as fixed or uniform rules, but emphasizes dynamically determining the applicable strategy based on the specific scenario corresponding to the content identifier code. This design allows the same method to flexibly cope with diverse application scenarios within the game, customizing verification logic for each scenario, thereby significantly improving the targeting and accuracy of abnormal content identification. Furthermore, strongly associating strategy selection with scenarios also enables operators to implement refined strategy configuration and management for different scenarios according to actual governance needs, achieving a dual optimization of governance efficiency and effectiveness.

[0045] In some optional implementations, the content verification strategy is generated in the following ways: Step a1: Obtain the policy configuration information corresponding to each content identifier code.

[0046] Policy configuration information refers to the raw data set by operations personnel through the management backend to define content validation rules. This data may include rule conditional logic, rule identifiers, and the scope of rule effectiveness. Specifically, the policy configuration information corresponding to each content identifier is centrally stored in a unified policy management backend or configuration database. Operations personnel create and define a series of rules for different content scenarios (corresponding to different content identifiers) through a graphical interface or API. The specific content of these rules, such as conditional expressions, risk tags, and effective status, constitutes the policy configuration information. By periodically polling, receiving event notifications, or directly querying, all policy definition data associated with each content identifier that is in effect can be read from the policy management backend or configuration database, thus completing the acquisition of policy configuration information.

[0047] In some optional implementations, step a1 includes: generating policy configuration information corresponding to the content identifier in response to a policy configuration event for the content identifier triggered on the game server.

[0048] Strategy configuration events refer to actions initiated by operations personnel in the game server's management system, involving the creation, modification, enabling, disabling, or deletion of strategy configuration information. Specifically, such as... Figure 4 As shown, when operators perform operations on the game server's management system, such as adding, modifying, or deleting verification rules under a content identifier, a policy configuration event is triggered. The system captures this event and extracts its core parameters, such as the content identifier being manipulated, details of the rule changes, and the operation type. Subsequently, based on these parameters, it either directly generates a new policy configuration information document or updates existing policy configuration information records. This process ensures that the operators' intentions are accurately and promptly transformed into identifiable and processable structured policy configuration information.

[0049] In the above implementation, an event-driven agile response model is established by directly linking the generation of strategy configuration information with strategy configuration events triggered on the game server. When new security risks arise during game operation or when strategy pre-configuration for specific scenarios is required, operators can directly trigger the corresponding event on the server. The system then responds and generates strategy configuration information for that content identifier. This design ensures that strategy creation is closely synchronized with changes in game business scenarios, enabling precise and rapid responses to the actual security needs of front-end business, and achieving deep integration of strategy configuration and business operation. Furthermore, since the configuration action originates from explicit events on the server, the authority and traceability of the strategy source are guaranteed, laying a solid foundation for closed-loop strategy management.

[0050] Step a2: Based on the results of filling the policy configuration information in the preset policy template, generate the content verification policy corresponding to each content identifier code.

[0051] A predefined policy template is a predefined rule framework with a fixed structure and general syntax. Its function is to quickly transform unstructured policy configuration information into policy description files that can be recognized and executed by the rule engine, according to a standardized format. Specifically, a policy template is pre-built, which is a framework containing a fixed logical structure and general syntax placeholders. When generating content validation policies, the acquired policy configuration information is used, and its variable parts (such as specific features, comparison operators, thresholds, etc.) are filled into the corresponding placeholders in the predefined template. Through this filling mechanism, unstructured configuration information is transformed into standard-formatted, syntactically complete rule description text. These texts undergo a subsequent compilation process, ultimately becoming content validation policies that can be executed quickly in the rule engine.

[0052] In the above implementation, by introducing a preset strategy template as a core component, the strategy generation process is decomposed into two standardized steps: obtaining configuration information and template filling. This design eliminates the need for developers to modify program code when creating and modifying strategies. Operations personnel only need to adjust the highly readable strategy configuration information to quickly generate or update highly customized content verification strategies applicable to different content identifier codes, greatly reducing the maintenance cost and operational threshold of strategies. At the same time, the template-based generation mechanism ensures the structural consistency and standardization of all strategies, avoiding logical errors that may be introduced by manual writing, and providing a solid technical foundation for batch management, version control, and automated deployment of strategies. Thus, while improving the flexibility of strategy management, it ensures the reliability and maintainability of the entire verification system.

[0053] In some alternative implementations, step a2 includes: Step a21: For any content identifier, fill the policy configuration information corresponding to the content identifier into the preset policy template to obtain the initial policy file corresponding to the content identifier.

[0054] The initial policy file is an intermediate file generated by filling specific policy configuration information into a preset policy template. Specifically, for any content identifier, all policy rule entries belonging to that identifier are first identified from the acquired policy configuration information set. Then, each policy rule is traversed, and its specific judgment logic (e.g., "feature value A is greater than threshold B") is transformed and embedded according to the format and syntax specified by the preset policy template. This embedding process is programmatic and automated, integrating multiple discrete rule configurations to generate a complete text file containing all the validation rules applicable to that content identifier, such as a .drl file using the Drools rule engine. This file is the initial policy file; it is readable but not yet optimized for direct execution.

[0055] Step a22: Pre-compile the initial policy file to generate policy objects corresponding to the content identifier codes.

[0056] A policy object is a data structure that is directly and quickly executed in memory, formed from an initial text-based policy file after compilation by the rule engine. Specifically, the initial policy file is in text format, which has relatively low execution efficiency. To improve performance, the rule engine's compilation function can be invoked to pre-compile this text file. The compilation process checks the syntax of the file and converts it into a highly optimized, directly operable, and executable binary memory object—the policy object. This policy object contains all the necessary matching logic and data structures, enabling the rule engine to quickly load it for real-time data matching, significantly improving the efficiency of rule execution.

[0057] Step a23: Construct the mapping relationship between each content identifier code and each policy object, and determine the content verification policy corresponding to the content identifier code based on the mapping relationship.

[0058] A mapping relationship refers to a correspondence table established and maintained in memory or a cache. Specifically, after a policy object is generated, an index table or "dictionary" is established and maintained in memory (usually in a distributed cache or local cache). This data structure uses the content identifier code as the key and its corresponding, compiled policy object as the value. In this way, each content identifier code is directly associated with its own dedicated, immediately usable policy object. When it is necessary to determine the content validation policy corresponding to a certain content identifier code, a quick lookup operation is performed in the above mapping relationship using that content identifier code as the key to directly locate and obtain the corresponding policy object. This policy object itself is the executor of the content validation policy ultimately used for validation.

[0059] In the above implementation, a highly automated and optimized policy production pipeline is constructed by clearly dividing the generation process into three core steps: template filling, pre-compilation, and relation mapping. Filling policy configuration information into a preset template generates the initial policy file, achieving standardization and automation of policy definition and ensuring consistency in policy syntax and structure. Furthermore, the initial file is pre-compiled; this crucial step transforms the readable rule file into policy objects that can be efficiently executed by the rule engine, fundamentally eliminating the overhead of real-time parsing and compilation on each request, thereby significantly improving policy execution speed and runtime efficiency. By constructing and maintaining the mapping relationship between content identifiers and pre-compiled policy objects, an efficient policy index library is established. This enables the system to immediately locate and activate the corresponding, ready-to-use verification policy through a single fast hash lookup when handling massive concurrent requests, ensuring low latency and high performance in the decision-making process.

[0060] In some optional implementations, step a2 further includes: monitoring the memory usage of policy objects and the usage frequency of each policy object; when the memory usage exceeds a preset threshold, eliminating the policy object with the lowest usage frequency.

[0061] Memory usage refers to the amount of memory occupied by a policy object after it is loaded into the server's runtime memory. Usage frequency is a metric that measures the number of times a policy object is accessed per unit of time. Specifically, for monitoring memory usage, data on the memory size occupied by each policy object loaded into memory can be collected periodically or through event hooks and accumulated to obtain the overall and individual object memory usage. For monitoring usage frequency, an access counter can be maintained for each policy object. Each time the policy object is invoked to perform content validation, its corresponding counter increments. By continuously recording the values ​​of these counters and potentially combining them with time windows (such as the most recent minute or hour) to calculate the access frequency per unit of time, the popularity of each policy object can be accurately grasped. All this monitoring data is aggregated to provide data support for subsequent resource scheduling decisions.

[0062] The preset threshold is a safety limit set for system resources such as memory usage; it's a pre-configured value. Specifically, when the total memory usage of policy objects exceeds the preset safety threshold, a memory cleanup mechanism is automatically triggered. The core logic of this mechanism is intelligent eviction based on the usage frequency of policy objects. First, all policy objects in memory are sorted by their usage frequency, identifying those that have been accessed the least recently, i.e., have the lowest usage frequency. Then, starting with these least frequently used objects, they are removed from memory one by one, releasing the memory space they occupy. This process continues until the total memory usage falls back below the preset safety threshold. This frequency-based eviction strategy prioritizes the operation of frequently used, core business policy objects within limited memory resources, thereby maximizing resource efficiency. For example, a memory eviction mechanism can be predefined, setting the KieBase node cache memory to 1GB. Objects gain weight as they are used; when memory usage exceeds 1GB, the object with the lowest weight is automatically evicted. The memory eviction mechanism is designed to limit the size of the cache. It uses a total weight cap, meaning each object added to the cache is marked with a weight based on its byte capitalization. When the accumulated weight exceeds the total weight, entries are evicted according to the Least Recently Used (LRU) strategy. The entire cache block may also have a time-based expiration policy, such as 24 hours.

[0063] In the above implementation, an intelligent, data-driven memory management strategy is established by continuously monitoring the memory usage of policy objects and tracking their usage frequency. When memory usage approaches a preset threshold, the strategy objects with the lowest usage frequency can be automatically and accurately identified and eliminated. This frequency-based elimination mechanism essentially achieves an adaptive cache optimization, ensuring that limited and valuable memory resources are always occupied by the most frequently accessed hot policies, thereby maintaining a very high cache hit rate overall. This not only effectively prevents the risk of program memory overflow caused by unlimited memory growth and ensures the long-term stable operation of the system, but also indirectly improves the average processing performance and response efficiency of the entire content verification system by optimizing resource allocation so that core high-frequency business policies can always be accessed quickly.

[0064] In some optional implementations, the content verification strategy is generated in the following ways: Step b1: Obtain policy update information.

[0065] Policy update information refers to any changes made to existing policy configurations. Specifically, policy update information can be obtained through two collaborative mechanisms. When operations personnel modify and submit policy configurations in the policy management backend, the backend system generates a policy update event and proactively notifies all nodes of the content security governance system via message queues or event buses. Additionally, at fixed time intervals (e.g., every minute), queries are proactively initiated to the policy management backend or a unified configuration version service to check if the policy version loaded is up-to-date. If not, differentiated update information is retrieved. These two mechanisms can be used individually or in combination to ensure the system can detect policy changes promptly and accurately.

[0066] Step b2: Fill the preset policy template with the policy update information to generate the updated content verification policy.

[0067] Upon receiving policy update information, the processing flow is similar to that of the initial policy generation, but with a greater emphasis on incremental changes and replacements. The update information is parsed to identify the specific policy rules that need to be added, modified, or deleted. Then, these updated rule configurations are re-populated into the preset policy template, generating a new initial policy file containing the latest logic. This new initial policy file is then sent to the rule engine for pre-compilation, producing updated and optimized policy objects. Finally, this newly generated policy object replaces the old, corresponding policy objects in the memory mapping relationship. At this point, the updated content validation policy officially takes effect, and all subsequent content validation requests will be executed according to this new set of rule logic, thus achieving dynamic and seamless policy updates.

[0068] In the above implementation, by establishing a standardized update process, the policy update behavior is reduced to a clear path: obtaining policy update information and then re-generating the policy using a preset template. This design allows policy iteration and optimization to proceed without modifying the core system code or redeploying services, greatly improving the agility of policy maintenance. When business scenarios change or new violation types emerge, operations personnel can quickly complete online adjustments and releases of policies by submitting update information, thereby ensuring that content verification capabilities can adapt to dynamically changing security requirements in a timely manner. Simultaneously, because the entire update process is based on a mature template-filling mechanism, it effectively avoids formatting errors or logical inconsistencies that may arise from direct manual modification of policy files, ensuring the accuracy of policy updates and system stability.

[0069] This embodiment provides a method for processing generative content, which can be used in electronic devices, such as computer 101, mobile terminal 102, etc. Figure 5 This is a flowchart of a generative content processing method according to an embodiment of this application, such as... Figure 5 As shown, the process includes the following steps: Step S501: Obtain the target generative content published by the player in the game, and the content identifier code corresponding to the target generative content. For details, please refer to [link to relevant documentation]. Figure 3 Step S301 of the illustrated embodiment will not be described again here.

[0070] Step S502: Identify the content type of the target generative content, extract features from the target generative content of the content type, and obtain the feature data corresponding to the target generative content.

[0071] Specifically, step S502 above includes: Step S5021: Determine the content analysis model corresponding to the content type.

[0072] A content analysis model is a trained computational model used to understand and parse the core semantics of generative content. Specifically, once the specific type of the target generative content is identified, this type is used as the lookup key to query a pre-established model registry or model routing table. This table pre-establishes mapping relationships between different content types (such as text, images, and audio) and specific content analysis models. Based on the pre-defined mapping logic, it returns one or more analysis model identifiers or access endpoints specifically designed to process that content type. This process ensures that appropriate analysis tools are automatically and accurately assigned to different types of content—for example, assigning models skilled in semantic analysis to text content and models skilled in visual recognition to image content—thus laying the foundation for subsequent deep feature extraction.

[0073] Step S5022: Use the content analysis model to perform semantic understanding of the expressive intent of the target generative content, and obtain predictive information that indicates that the target generative content is abnormal.

[0074] The feature data includes prediction information and anomaly types.

[0075] Predictive information is a quantitative risk assessment result output by a content analysis model after analyzing the target generated content. Specifically, after determining the corresponding content analysis model, the raw data of the target generated content is submitted to the model for calculation. Based on its massive internally learned knowledge and understanding of complex patterns, the model performs in-depth analysis of the input content, attempting to understand its underlying expressive intent or semantic category. The model outputs not a simple yes or no, but a series of structured, quantitative predictive information. This information includes, but is not limited to: the probability that the content is classified as belonging to various preset anomaly types (e.g., {anomaly type A: 0.98, anomaly type B: 0.1, anomaly type C: 0.01}), and a comprehensive risk score. This predictive information is essentially a quantitative assessment report of the content security status by the model, providing core data for subsequent strategy verification.

[0076] The generative content processing method provided in this application goes beyond traditional keyword matching or simple feature matching. By introducing a dedicated content analysis model that matches the content type, it achieves a deep semantic understanding of the expressive intent of the target generative content. This enables the system to discern the underlying meaning and user intent behind the content, thereby significantly improving its ability to identify hidden, variant, or context-related violations. Furthermore, the feature data output by this application is not conventional numerical features, but directly includes the predicted information derived from the model's semantic understanding and its corresponding anomaly types. This provides a highly condensed and directly logically applicable decision-making basis for subsequent policy verification. This structured feature data output greatly simplifies the complexity of policy rules and makes anomaly judgment results more accurate and interpretable, laying a solid data foundation for achieving refined and intelligent content security governance.

[0077] Step S503: Based on the content verification strategy corresponding to the content identifier code, perform anomaly verification on the feature data to obtain the abnormal content corresponding to the target generated content. For details, please refer to [link to relevant documentation]. Figure 3 Step S303 of the illustrated embodiment will not be described again here.

[0078] In some alternative implementations, the generative content processing apparatus includes: Step c1: Obtain the target verification strategy for the abnormal content.

[0079] Target validation strategies refer to the specific content validation strategies that are actually triggered (i.e., "hit") by the feature data during anomaly validation. Specifically, during the validation process executed by the rule engine, when the input feature data meets the triggering conditions of one or more content validation strategies, these strategies are considered hit. After completing pattern matching and logical operations, the rule engine explicitly records the unique identifiers of all hit strategies, such as strategy IDs or strategy names, in its output validation results. By parsing the rule engine's response, these identifiers can be directly extracted, thus accurately identifying which specific strategy rules (i.e., target validation strategies) caused the current content to be judged as an anomaly.

[0080] Step c2: When the number of hits of the target verification strategy reaches the preset circuit breaker value, obtain the circuit breaker result.

[0081] A preset circuit breaker threshold is a pre-defined threshold for a specific verification strategy, representing the maximum number of times it is allowed to be hit within a certain time window. The circuit breaker result refers to the protective actions taken and the output status when the number of hits by the target verification strategy reaches the preset circuit breaker threshold. Specifically, a circuit breaker monitoring mechanism is set up for important verification strategies to continuously track their hit count within a specific time window. When it is detected that the real-time hit count of a target verification strategy has reached or exceeded its preset threshold (i.e., the preset circuit breaker threshold), the circuit breaker logic is automatically triggered. This logic executes predefined protective actions, typically including: immediately setting the strategy's status to temporary dormancy or forced bypass to prevent it from being triggered again and potentially causing a cascading failure effect on the system; and simultaneously generating a standard circuit breaker result notification. This result notification contains the circuit breaker status, the triggered strategy identifier, and a possible reason code. By listening to or querying this status change, the required circuit breaker result information can be obtained.

[0082] Step c3 encapsulates the hit statistics of the target verification strategy, the circuit breaker result, and the hit target verification strategy to obtain the anomaly judgment result of the abnormal content.

[0083] The hit statistics are a quantitative indicator for a target validation strategy, used to record the number of times the strategy is successfully triggered within a specific time period. Specifically, after completing all validation and monitoring logic, a result assembly process is initiated. It queries a centralized statistical data store for the cumulative hit count of the currently hit target validation strategy in the recent period (e.g., the last hour), i.e., the hit statistics. Then, it standardizes and encapsulates three key information elements: the core cause of the anomaly (i.e., the identifier of the hit target validation strategy), the rule's operational activity and status (i.e., the hit statistics and circuit breaker results), and the final handling basis (the anomaly conclusion derived from the previous two), forming a structured data object. This final data object is the anomaly determination result, providing a complete and comprehensive information view of the anomaly event, facilitating precise handling operations (such as interception, logging, and alarming) and subsequent analysis by upstream systems.

[0084] In the above implementation, by continuously tracking the number of hits of a specific target verification strategy and triggering processing logic when it reaches a preset circuit breaker threshold, an effective system self-protection and risk warning mechanism is constructed. This effectively prevents the subsequent processing flow from becoming overloaded due to a large number of triggers of a single strategy in a short period of time, thus improving the overall robustness of the system. Furthermore, this application integrates three key information categories—hit statistics, circuit breaker results, and the hit strategy itself—into a single encapsulation as the final anomaly determination result. This encapsulation provides downstream systems or operators with an extremely rich and comprehensive decision-making context, not only indicating what the anomaly is but also revealing the frequency of the anomaly and the current handling status of the system. This greatly assists in subsequent precise hierarchical handling and in-depth data analysis, thereby achieving an efficient closed loop from anomaly detection to handling decision-making.

[0085] This embodiment also provides a generative content processing apparatus for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0086] This embodiment provides a generative content processing apparatus, such as... Figure 6 As shown, it includes: The first acquisition module 601 is used to acquire the target generative content published by the player in the game, and the content identifier code corresponding to the target generative content; The extraction module 602 is used to identify the content type of the target generative content, extract features from the target generative content of the content type, and obtain the feature data corresponding to the target generative content. The verification module 603 is used to perform anomaly verification on feature data based on the content verification strategy corresponding to the content identifier code, and obtain the abnormal content corresponding to the target generative content.

[0087] In some alternative implementations, the verification module 603 includes: The first determination submodule is used to determine the content verification strategy corresponding to the content identifier code based on the scenario corresponding to the content identifier code. The verification submodule is used to perform anomaly verification on the feature data according to the content verification strategy, and obtain the abnormal content corresponding to the target generative content.

[0088] In some optional implementations, the verification submodule includes: The first acquisition unit is used to acquire the policy configuration information corresponding to each content identifier code; The generation unit is used to generate content verification strategies corresponding to each content identifier code based on the population results in the preset strategy template according to the strategy configuration information.

[0089] In some optional implementations, the verification submodule further includes: The second acquisition unit is used to acquire policy update information; The fill unit is used to fill the preset policy template according to the policy update information and generate the updated content verification policy.

[0090] In some optional implementations, the first acquisition unit includes: The generation subunit is used to generate the strategy configuration information corresponding to the content identifier code in response to the strategy configuration event triggered on the game server for the content identifier code.

[0091] In some optional implementations, the generation unit includes: The fill sub-unit is used to fill the policy configuration information corresponding to any content identifier code into the preset policy template to obtain the initial policy file corresponding to the content identifier code; The pre-compilation subunit is used to pre-compile the initial policy file and generate policy objects corresponding to the content identifier codes; The construction sub-unit is used to build the mapping relationship between each content identifier code and each strategy object, and to determine the content verification strategy corresponding to the content identifier code based on the mapping relationship.

[0092] In some optional implementations, the generation unit further includes: The monitoring subunit is used to monitor the memory usage of policy objects and the usage frequency of each policy object; The eviction subunit is used to evict the least frequently used policy object when the memory usage exceeds a preset threshold.

[0093] In some alternative implementations, the extraction module 602 includes: The second determination submodule is used to determine the content analysis model corresponding to the content type; The prediction submodule is used to perform semantic understanding of the expressive intent of the target generative content using a content analysis model, and to obtain predictive information that indicates that the target generative content is abnormal. The feature data includes prediction information and anomaly types.

[0094] In some alternative implementations, the generative content processing apparatus further includes: The second acquisition module is used to acquire the target verification strategy for abnormal content; The third acquisition module is used to acquire the circuit breaker result when the number of hits of the target verification strategy reaches the preset circuit breaker value; The encapsulation module is used to encapsulate the hit statistics of the target verification strategy, the circuit breaker result, and the hit target verification strategy to obtain the anomaly judgment result of the abnormal content.

[0095] The generative content processing apparatus provided in this application embodiment can execute the generative content processing method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0096] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0097] The following is a detailed reference. Figure 7 The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 701, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 702 or a program loaded from memory 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0098] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic devices to exchange data via wireless or wired communication with other devices. Although Figure 7 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0099] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 709, or installed from memory 708, or installed from ROM 702. When the computer program is executed by processor 701, it performs the functions defined in the generative content processing method of embodiments of this application.

[0100] Figure 7 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0101] This application also provides a computer-readable storage medium. The methods described in this application can be implemented in hardware or firmware, or implemented as recordable on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the generative content processing method shown in the above embodiments is implemented.

[0102] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0103] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A method of processing generative content, characterized by, The method comprises: obtaining target generation content published by a player in a game, and a content identification code corresponding to the target generation content; identifying a content type of the target generation content, and performing feature extraction on the target generation content of the content type to obtain feature data corresponding to the target generation content; performing abnormality checking on the feature data based on a content checking strategy corresponding to the content identification code to obtain abnormal content corresponding to the target generation content.

2. The method of claim 1, wherein, The abnormality checking on the feature data based on the content checking strategy corresponding to the content identification code to obtain the abnormal content corresponding to the target generation content comprises: determining the content checking strategy corresponding to the content identification code based on a scenario corresponding to the content identification code; performing abnormality checking on the feature data according to the content checking strategy to obtain the abnormal content corresponding to the target generation content.

3. The method of claim 2, wherein, The generation mode of the content checking strategy comprises: obtaining strategy configuration information corresponding to each content identification code; generating the content checking strategy corresponding to each content identification code based on a filling result of the strategy configuration information in a preset strategy template.

4. The method of claim 3, wherein, The method further comprises: obtaining strategy update information; filling the preset strategy template according to the strategy update information to generate an updated content checking strategy.

5. The method of claim 3, wherein, The obtaining of the strategy configuration information corresponding to each content identification code comprises: generating the strategy configuration information corresponding to the content identification code in response to a strategy configuration event triggered by a game server for the content identification code.

6. The method according to claim 3 or 4, characterized in that, The generation of the content checking strategy corresponding to each content identification code based on the filling result of the strategy configuration information in the preset strategy template comprises: for any one content identification code, filling the strategy configuration information corresponding to the content identification code into a preset strategy template to obtain an initial strategy file corresponding to the content identification code; pre-compiling the initial strategy file to generate a strategy object corresponding to the content identification code; constructing a mapping relationship between each content identification code and each strategy object, and determining the content checking strategy corresponding to the content identification code based on the mapping relationship.

7. The method of claim 6, wherein, The method further comprises: monitoring a memory occupation space of the strategy object and a use frequency of each strategy object; when the memory occupation space exceeds a preset threshold, eliminating the strategy object with the lowest use frequency.

8. The method of claim 1, wherein, The feature extraction on the target generation content of the content type to obtain the feature data corresponding to the target generation content comprises: determining a content analysis model corresponding to the content type; performing semantic understanding on an expression intention of the target generation content by using the content analysis model to obtain prediction information representing that the target generation content is abnormal; wherein the feature data comprises the prediction information and an abnormal type.

9. The method of claim 1, wherein, The method further comprises: obtaining a target checking strategy hit by the abnormal content; when a hit frequency of the target checking strategy reaches a preset fuse value, obtaining a fuse result; The hit statistics of the target verification policy, the fusing result and the target verification policy that hits are packaged to obtain an exception determination result of the exception content.

10. A processing apparatus of generative content, characterized by, The device comprises: A first obtaining module is configured to obtain target generative content published by a player in a game and a content identification code corresponding to the target generative content. An extracting module is configured to identify a content type of the target generative content, and perform feature extraction on the target generative content of the content type to obtain feature data corresponding to the target generative content. A verification module is configured to perform exception verification on the feature data based on a content verification policy corresponding to the content identification code to obtain exception content corresponding to the target generative content.

11. An electronic device, comprising: The device comprises: A memory and a processor, which are in communication connection with each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the processing method of the generative content according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are used to make the computer execute the processing method of the generative content according to any one of claims 1 to 9.

13. A computer program product, characterised in that, The computer instructions are used to make the computer execute the processing method of the generative content according to any one of claims 1 to 9.

Images