Information processing method and system

Abstract

The embodiment of the invention provides an information processing method and system.The information processing method comprises the steps that to-be-processed information submitted by a user is received, and processing rule information submitted by the user for the to-be-processed information is determined; constructing an information processing task corresponding to the to-be-processed information based on the processing rule information, and receiving task description information submitted by the user for the information processing task; and executing at least one processing subtask included in the information processing task according to the task description information, obtaining information processing data corresponding to the to-be-processed information, and feeding back the information processing data to the user.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of artificial intelligence technology, and in particular to information processing methods and systems. Background Technology

[0002] In the current internet content ecosystem, with the explosive growth of user-generated content, accurate identification and processing of information content has become a crucial aspect of ensuring platform compliance and user experience. Traditional information processing methods primarily rely on pre-set rule engines or general artificial intelligence models, whose processing logic and judgment standards are typically static, uniform, and platform-driven. While these methods possess a degree of automation, they struggle to adapt to the personalized processing standards required by different users in diverse task scenarios. Due to a lack of effective understanding and dynamic adaptation to user intent, existing technologies often result in processing outcomes deviating from user expectations, leading to a high misjudgment rate and requiring significant manual intervention for correction. This not only reduces overall processing efficiency but also impacts user experience and satisfaction. Therefore, a more flexible and intelligent information processing method is urgently needed, capable of dynamically adjusting judgment logic based on actual user needs to achieve efficient, accurate, and personalized information processing. Thus, a more effective information processing method is urgently required to address the aforementioned problems. Summary of the Invention

[0003] In view of the above, embodiments of this specification provide an information processing method. One or more embodiments of this specification also relate to an information processing system, an information processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this specification, an information processing method is provided, comprising: Receive pending information submitted by a user and determine the processing rule information submitted by the user for the pending information; Based on the processing rule information, an information processing task corresponding to the information to be processed is constructed, and the task description information submitted by the user for the information processing task is received; Execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the information to be processed, and feed back the information processing data to the user.

[0005] According to a second aspect of the embodiments of this specification, an information processing system is provided, including a client and a server, comprising: The client is used to receive pending information submitted by the user and send the pending information to the server. The server is configured to determine the processing rule information submitted by the user for the information to be processed; construct an information processing task corresponding to the information to be processed based on the processing rule information, and receive task description information submitted by the user for the information processing task; execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the information to be processed, and feed the information processing data back to the client. The client is used to display the information processing data to the user through the information processing page.

[0006] According to a third aspect of the embodiments of this specification, an information processing apparatus is provided, comprising: The receiving module is configured to receive pending information submitted by a user and determine the processing rule information submitted by the user for the pending information. The construction module is configured to construct an information processing task corresponding to the information to be processed based on the processing rule information, and to receive task description information submitted by the user for the information processing task; The execution module is configured to execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the information to be processed, and feed back the information processing data to the user.

[0007] According to a fourth aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the above-described information processing method.

[0008] According to a fifth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the information processing method described above.

[0009] According to a sixth aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions that, when executed by a processor, implement the steps of the information processing method described above.

[0010] This specification provides an embodiment of an information processing method that receives user-submitted information to be processed, which may be a multimodal resource. The method determines processing rule information submitted by the user for the information to be processed. Based on the processing rule information, it constructs an information processing task corresponding to the information to be processed, enabling the construction of the information processing task with user participation and guidance, ensuring that the information processing task meets the user's review requirements for the information to be processed. The method receives task description information submitted by the user for the information processing task. Based on the task description information, it executes at least one processing sub-task included in the information processing task to obtain information processing data corresponding to the information to be processed, thus completing the execution of the information processing task under the user's guidance, obtaining information processing data that meets the user's review requirements, and feeding back the information processing data to the user. User participation in the construction and execution of the information processing task improves user satisfaction with resource review. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating an information processing method provided in one embodiment of this specification; Figure 2 This is a schematic diagram illustrating resource review of an information processing method provided in one embodiment of this specification; Figure 3 This is an interactive schematic diagram of an information processing method provided in one embodiment of this specification; Figure 4 This is a schematic diagram of the audit architecture of an information processing method provided in one embodiment of this specification; Figure 5 This is a schematic diagram of the structure of an information processing system provided in one embodiment of this specification; Figure 6 This is an architecture diagram of an information processing system provided in one embodiment of this specification; Figure 7 This is a schematic diagram of the structure of an information processing device provided in one embodiment of this specification; Figure 8 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0012] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0013] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0014] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0015] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0016] The technical solutions provided in this application can employ deep learning models with relatively large parameter scales. However, this large model is merely an example; this application does not limit the number of model parameters supported by the deep learning model used, aiming to meet actual needs. The deep learning models involved in this application can be artificial intelligence-based language models (LM) or multimodal models (MM).

[0017] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0018] Multimodal review: The process of semantically understanding inputs such as text, images, audio, video, and forms, and making compliance, quality, or risk judgments based on rules / policies. It differs from content recognition alone in that it introduces rules, evidence, and adjudication strategies.

[0019] Requirements Clarification / DSL: Through conversational interaction, user natural language requirements are structured into a machine-executable requirement specification (Domain Specific Language / JSON), including task type, target metrics, constraints, and data sources; this serves as the sole basis for subsequent planning and execution. Clarifying requirements and defining acceptance criteria can significantly reduce rework and improve delivery quality.

[0020] Unified Planning Agent: Reads the DSL of requirements, decomposes, orchestrates and optimizes tasks based on the registered "capability map" (available modalities, input / output contracts, cost and accuracy profiles of each sub-agent / tool), and generates an executable multi-step plan.

[0021] Agent: An atomic execution unit with specific modality processing or general reasoning capabilities (such as OCR, ASR, VLM, retrieval rearrangement, structured extraction, evaluation, etc.) that plans and orchestrates by connecting standardized inputs / outputs with observable metrics.

[0022] Execution orchestrator: Schedules agents according to a plan, responsible for data and control flow, caching, retries, rollbacks, timeouts, resource isolation, and log auditing; supports dry-run simulation.

[0023] Human-in-the-Loop (HITL): Triggering human review and approval at key steps according to strategy, and making decisions on continuing, rerunning, or replacing the plan based on intermediate evidence and acceptance criteria.

[0024] API Key: A long string, typically generated by the service provider, serving as a unique identifier and password for accessing its API.

[0025] Token: A temporary, time-limited credential used to prove the identity and permissions of a user or application. It is usually dynamically generated through an authentication process.

[0026] API Gateway: A core component in modern software architecture, especially microservice architecture. It serves as the unified entry point for all client requests (such as mobile apps, web pages, and other services). It receives all requests from clients, routes them to the corresponding backend microservices based on the request content, and returns the microservices' responses to the clients.

[0027] DAG (Directed Acyclic Graph): A data structure used to represent a network of relationships with explicit dependencies and execution order. Its "directed" property defines the sequential relationship, while its "acyclic" property ensures logical coherence and avoids infinite loops.

[0028] OCR (Optical Character Recognition) is a technology that converts printed or handwritten text in different types of documents (such as scanned paper documents and PDF files) into machine-coded text. It enables computers to "read" and understand this originally uneditable text content, thereby achieving automated processing.

[0029] VQA (Visual Question Answering) is an artificial intelligence task that requires a model to simultaneously understand an image and a natural language question about that image, and then generate an accurate natural language answer.

[0030] URI (Uniform Resource Identifier): A URI is a unique string used to name or locate any resource on the Internet.

[0031] To address the aforementioned technical problems, this specification provides an information processing method. This specification also relates to an information processing system, an information processing apparatus, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0032] See Figure 1 , Figure 1 A flowchart of an information processing method according to an embodiment of this specification is shown, which specifically includes the following steps.

[0033] Step 102: Receive the pending information submitted by the user and determine the processing rule information submitted by the user for the pending information.

[0034] Specifically, users can be those with information processing needs, which may include information detection or resource review requirements. Users can submit information to be processed through the information processing platform for detection or review. The information to be processed can be resources requiring review, such as images, videos, audio, and text. It can also be a multimodal resource composed of at least two of these types. The information to be processed includes, but is not limited to, user-generated content, user credentials, and documents that require review in terms of compliance, consistency between text and images, and authenticity. Processing rule information represents the user's processing intent, detection intent, review intent, review preferences, acceptance criteria, review requirements, and / or review needs. Processing rule information can be determined through a clarifying dialogue between the agent and the user.

[0035] Based on this, the system receives pending information submitted by users through the resource review platform, determines the processing rules submitted by users for the pending information, and clarifies the users' review and / or processing requirements for the pending information, so as to facilitate the subsequent processing of the pending information according to the users' review and / or processing requirements.

[0036] Furthermore, considering that the user's direct request information regarding the information to be processed may be incomplete or unclear, a clarifying dialogue can be initiated with the user to clarify their processing needs and obtain processing rule information. The specific implementation is as follows: The process involves: determining the initial processing rule information submitted by the user for the information to be processed; invoking a dialogue agent to determine initial processing intent information based on the initial processing rule information; determining the dialogue data generated between the user and the dialogue agent based on the initial processing intent information, and extracting processing elements from the dialogue data; updating the initial processing rule information based on the processing elements to obtain the processing rule information.

[0037] Specifically, the initial processing rule information can be a description of the review points submitted along with the information to be processed, used to clarify the review points when reviewing the information. The dialogue agent can be an advanced autonomous AI system driven by a large language model, capable of conversing with the user, including but not limited to understanding the user's input such as text, images, and voice, and generating feedback text for the user's input to clarify the user's review needs, and displaying it to the user, or generating audio to play for the user. The initial processing rule information is the review points or review needs initially submitted by the user based on the initial processing rule information. The dialogue data contains at least one round of dialogue text between the dialogue agent and the user. The dialogue agent can be the initiator of the dialogue. By inquiring about the user's review needs, the dialogue agent obtains clear processing elements, which can be the review elements for reviewing the information to be processed.

[0038] Based on this, the initial processing rules submitted by the user for the information to be processed are determined. A dialogue agent is invoked to determine the user's initial processing intent based on the initial processing rules. If the initial processing intent is insufficient to support detailed review of the information, dialogue data generated between the user and the dialogue agent based on the initial processing intent is determined. The dialogue agent can use large language model-driven conversational AI technology to conduct intent-clarifying question-and-answer sessions with the user. The dialogue agent asks questions to clarify the user's review needs, and the user answers the questions. After obtaining dialogue data corresponding to at least one round of dialogue, processing elements are extracted from the dialogue data. The initial processing rules are updated based on these processing elements to obtain the final processing rules.

[0039] For example, in a resource review scenario, the information to be processed can be a resource to be reviewed. If the resource to be reviewed is an advertising image, a user uploads the advertising image through a resource review platform. To express the user's review needs, the user can submit a descriptive text of the review points along with the uploaded advertising image; this is the initial processing rule information. If the initial processing rule information is unclear, a dialogue agent can be invoked to further extract the user's review needs through multi-turn dialogues. The agent actively interacts with the user. It employs conversational AI technology driven by a Large Language Model (LLM) to initiate a dialogue with the user. The agent first analyzes the initial processing rule information. If the user's review needs are ambiguous (the user only says "review this advertising image"), the agent will proactively ask questions and conduct multi-turn dialogues to clarify the user's review needs. The agent will ask questions like a professional reviewer: "Do you want to focus on reviewing extreme words in the image (such as 'first')?", "Do you need to check for QR codes that redirect to third-party platforms?", "Which do you value more, review speed or accuracy?". After the user answers the questions, processing elements are extracted from the dialogue data with the user. By analyzing dialogue data, the Agent can accurately extract users' core review needs, acceptance criteria, and preferences: {Key Points: 'Exaggerated Advertising', Constraints: 'High Accuracy', Exclusions: 'QR Code Recognition'}. After obtaining the processing elements, the Agent updates the initial processing rule information to obtain clear review needs, i.e., processing rule information.

[0040] In summary, processing elements are extracted from dialogue data. Based on these elements, the initial processing rules are updated to obtain further processing rules, revealing the user's specific processing needs, and enabling targeted review of the information to be processed.

[0041] Furthermore, before processing the user-submitted pending information, considering the potential risks associated with user processing requests, it is necessary to detect both user information and processing request information before proceeding with further processing. The specific implementation is as follows: The user's user information and processing request information are determined; based on the user information, a user compliance check is performed on the user; if the check passes, the processing request information is processed; if the processing check passes, a processing task identifier is configured for the information to be processed, and the processing task identifier is used to query the information to be processed.

[0042] Specifically, user information can be user identity information, and user compliance checks can involve verifying whether a user is a compliant user, i.e., checking the validity of the API Key or Token. Request processing information can include the number of user requests for resource review, as well as information such as the request source IP and request frequency. The purpose of processing request information is to detect malicious calls or attacks. The review task can be a globally unique task ID assigned to the information to be processed, used for querying, tracking, and other operations related to the review of that information.

[0043] Based on this, user information and processing request information are determined. User compliance checks are performed on the user based on the user information. If the check passes, the user is considered compliant and the processing request information continues to be checked. If the processing check passes, it indicates that the risk of this resource review has been eliminated, and a processing task identifier can be configured for the pending information. The processing task identifier is used to query the pending information and track the review process.

[0044] Continuing with the previous example, after a user submits pending information, the API gateway performs authentication, quota, and risk control checks on the user and the current review task. Upon submission, the API gateway, acting as the first line of defense, performs the following operations: Authentication: Verifies the caller's identity, such as checking the validity of the API Key or Token to ensure a compliant user. Quota: Checks whether the user's daily call count or data processing volume exceeds the limits of their purchased plan.

[0045] Risk Control: This initial check examines the request source IP, request frequency, etc., to prevent DDoS attacks or malicious calls. Upon successful inspection, a globally unique task_id is generated for this information processing task. This task_id is the "identifier" for this task; it will be used throughout the entire process to track, record, and correlate data generated in all subsequent steps (from dialogue and planning to execution and auditing). It is the core identifier for achieving end-to-end monitoring and problem troubleshooting.

[0046] In summary, by conducting compliance checks on users based on their user information and by checking the information processed in order to avoid potential user compliance risks and malicious attacks, among other risks.

[0047] Step 104: Construct an information processing task corresponding to the information to be processed based on the processing rule information, and receive the task description information submitted by the user for the information processing task.

[0048] Specifically, after receiving the user-submitted information to be processed and determining the processing rules submitted by the user for that information, an intelligent agent can be invoked. Based on the processing rules, an information processing task corresponding to the information to be processed can be constructed, and the task description information submitted by the user for the information processing task can be received. This intelligent agent can be an advanced autonomous AI system driven by a large language model, capable of constructing and executing the information processing task corresponding to the information to be processed. The intelligent agent can perform resource review on the multimodal resources (information to be processed) submitted by the user, achieving multimodal review. The information processing task includes review nodes that can be executed in parallel or sequentially, with each review node corresponding to a review task to be executed. The information processing task can be an execution plan for reviewing the information to be processed. The task description information includes positive and negative feedback from the user regarding the information processing task; positive feedback indicates that the user approves of the information processing task, and negative feedback indicates that the user does not approve of the information processing task.

[0049] Based on this, after receiving the user-submitted information to be processed and determining the processing rules submitted by the user for the information to be processed, the intelligent agent is invoked to construct the information processing task corresponding to the information to be processed based on the processing rules. The agent receives the task description information submitted by the user for the information processing task, clarifying the user's evaluation or attitude of whether they approve of the information processing task.

[0050] Furthermore, considering that information processing tasks need to meet users' review requirements for the information to be processed, the user's processing rules information needs to be fully considered when constructing information processing tasks. The specific implementation is as follows: The intelligent agent is invoked to determine the processing objective information and processing constraint information based on the processing rule information; the processing objective information and the processing constraint information are converted into a processing node sequence containing at least one processing node, and an intelligent agent is matched for each of the at least one processing node; according to the node order corresponding to the processing node sequence, a node execution graph is constructed based on the intelligent agents corresponding to each of the at least one processing node, and an information processing task corresponding to the information to be processed is constructed based on the node execution graph.

[0051] Specifically, the processing objective information represents the user's review goal for the information to be processed, such as detecting whether there is exaggerated advertising. The processing constraint information represents the constraints imposed by the user on the review of the information to be processed, such as prioritizing high accuracy. The processing node sequence consists of at least one processing node, and there is a parallel or sequential execution relationship between these nodes. For example, a text recognition node and a sensitive word detection node have a sequential execution relationship. The intelligent agent is the AI ​​Agent, which matches the corresponding intelligent agent based on the task type of the node. The node order can be determined based on the parallel or sequential execution relationship between at least one processing node. The execution graph can be a directed acyclic graph (DAG), which clearly shows the execution order and dependencies between each review node.

[0052] Based on this, the processing purpose and constraints for reviewing the information to be processed can be determined using the processing rule information. A processing node sequence containing at least one processing node is constructed according to the review and acceptance criteria, processing purpose information, and processing constraints, and a smart agent is matched for each of the at least one processing node. When matching a smart agent to a review node, multiple candidate smart agents can be identified, and at least one candidate smart agent with a high degree of matching with the review node is selected as the matching smart agent by scoring these multiple candidate agents. Following the node order corresponding to the processing node sequence, a node execution graph is constructed based on the smart agents corresponding to each of the at least one processing node, clarifying the execution order and dependencies between each review node. Based on the node execution graph, the information processing task corresponding to the information to be processed is constructed.

[0053] Continuing with the previous example, after the intelligent agent (the planning agent) determines the user's processing rule information, it can use this information as a DSL (Specific Logic Controller) to begin intelligent planning, generating a better execution plan (DAG). The execution plan is represented in the form of information processing tasks. After determining the processing rule information, the review objective is parsed to determine the processing purpose information and processing constraint information; that is, the DSL is parsed to understand the core review objective ("detecting exaggerated claims") and review constraints ("high accuracy priority"). The complex review objective is decomposed into a series of atomic subtasks, each of which is a review node: the processing node sequence can be represented as: Optical Character Recognition (OCR) -> Limit Word Matching -> Semantic Understanding. The agent queries the "capability graph" to find all available AI agents (models or rules) for each review node. For OCR, two intelligent agents, ocr-fast (fast OCR agent) and ocr-highres (high-precision OCR agent), may be found. A multi-objective scoring function can be used to score each candidate agent. Based on the "high accuracy first" constraint in the DSL, it assigns high weight to ocr-highres, ultimately selecting ocr-highres as the intelligent agent matched with OCR. The selected agents are combined into a directed acyclic graph (DAG), clarifying their execution order and dependencies. Semantic understanding nodes must wait for the OCR nodes to complete before execution, while logo recognition can be executed in parallel with OCR. Finally, a complete DAG plan is output, where each node is a configured "plan node protocol," which will be submitted to the user for confirmation later.

[0054] In summary, a node execution graph is constructed based on the intelligent agents corresponding to at least one processing node, and an information processing task corresponding to the information to be processed is constructed based on the node execution graph, so that the information processing task can meet the user's review requirements for the information to be processed.

[0055] Step 106: Execute at least one processing subtask included in the information processing task according to the task description information, obtain the information processing data corresponding to the information to be processed, and feed the information processing data back to the user.

[0056] Specifically, after constructing the information processing task corresponding to the information to be processed based on the processing rule information, and receiving the task description information submitted by the user for the information processing task, at least one processing subtask contained in the information processing task can be executed according to the task description information to obtain the information processing data corresponding to the information to be processed, and the information processing data can be fed back to the user. Among the at least one processing subtask, there is a parallel relationship or a sequential execution relationship between the subtasks, and at least one processing subtask needs to be executed according to the parallel or sequential execution relationship. The information processing data is the review result obtained after reviewing the information to be processed.

[0057] Based on this, after constructing the information processing task corresponding to the information to be processed based on the processing rule information, and receiving the task description information submitted by the user for the information processing task, the system executes at least one processing subtask contained in the information processing task according to the task description information, obtains the information processing data corresponding to the information to be processed, and feeds back the information processing data to the user, informing the user of the review result obtained by reviewing the information to be processed.

[0058] Furthermore, considering that the task description information includes the user's evaluation or opinion on the information processing task, if the user's confirmation of the information processing task is determined based on the task description information, then the information processing task does not need to be modified and can be executed immediately. The specific implementation is as follows: If the user confirms the information processing task based on the task description information, at least one processing subtask included in the information processing task is executed to obtain the information processing data corresponding to the information to be processed.

[0059] Based on this, if the user confirms the information processing task according to the task description information, it indicates that the user is satisfied with the information processing task and has no suggestions for modification. Therefore, if the user confirms the information processing task, at least one processing subtask contained in the information processing task can be directly executed to obtain the information processing data corresponding to the information to be processed.

[0060] Continuing with the previous example, the task description information represents the user's feedback on the information processing task. If the task description information confirms the user's acceptance of the information processing task, it means the user has confirmed the DAG plan. In other words, the user has reviewed the DAG plan generated by the planning agent. (The DAG plan shown to the user: "I will use high-precision OCR, then rule-based matching of limiting words, and finally semantic judgment using a VQA large model. The estimated time is 2 seconds, and the cost is 0.05 yuan. Do you agree?") The user then clicks "Confirm." The DAG plan is considered the final solution and is executed by the execution orchestrator. After executing the text recognition (OCR), limiting word matching, and semantic understanding subtasks within the DAG plan, the execution result—the information processing data—is obtained.

[0061] In summary, when a user-confirmed information processing task is determined based on the task description information, at least one processing subtask contained in the information processing task is executed to obtain the information processing data corresponding to the information to be processed. This allows users to participate in the construction process of the information processing task, constantly considering their review needs for the information to be processed, thus improving the user's resource review experience.

[0062] Furthermore, considering that the task description information includes user evaluations or opinions on the information processing task, if it is determined from the task description information that the user has not directly confirmed the information processing task, the user's suggested modifications to the information processing task can be determined based on the task description information, and the information processing task can be updated according to the suggested modifications. The specific implementation is as follows: If, based on the task description information, it is determined that the user has submitted task modification information for the information processing task, the information processing task is updated to a target information processing task based on the task modification information; the target information processing task is then used as the information processing task, and the step of receiving the task description information submitted by the user for the information processing task is executed.

[0063] Specifically, task modification information represents the user's evaluation or suggestions for modifying the information processing task, while target modification information may also include the user's intention to modify the information processing task. The target information processing task is the new information processing task constructed under the guidance of the task modification information; it is an update to the original information processing task.

[0064] Based on this, if the task description information indicates that the user has submitted task modification information for the information processing task, it means that the user is dissatisfied with the information processing task and needs to modify it, and the user has expressed their intention to modify it. Based on the task modification information, the user's intention to modify the information processing task can be determined. Based on the task modification information, the information processing task is updated to the target information processing task, the target information processing task is adopted as the information processing task, and the step of receiving the task description information submitted by the user for the information processing task is executed.

[0065] Continuing with the previous example, if the task description information indicates that the user is dissatisfied with the information processing task and requests modification, the user's response could be: "The cost is too high. I'd rather sacrifice some accuracy; could it be faster and cheaper?" Here, the modification intention is to sacrifice accuracy to improve efficiency and reduce costs. This modification intention is then used as task modification information to update the information processing task. After updating to the target information processing task, the system continues to receive task description information submitted by the user for the target information processing task.

[0066] In summary, based on task modification information, the information processing task is updated to the target information processing task, the target information processing task is used as the information processing task, and the steps of receiving the task description information submitted by the user for the information processing task are executed. Under the guidance of the user, the information processing task is iteratively updated, improving the user's satisfaction with resource review and meeting the user's resource review needs.

[0067] Furthermore, considering that the task modification information includes the user's intention to modify the information processing task, it means that the processing intention corresponding to the processing rule information is not comprehensive enough. Therefore, the target processing rule information can be constructed by updating the processing rule information based on the task modification information, and then the target information processing task can be reconstructed based on the target processing rule information. The specific implementation is as follows: The processing rule information is updated based on the task modification information to obtain the target processing rule information; the intelligent agent is invoked to construct the target information processing task corresponding to the information processing task based on the target processing rule information.

[0068] Based on this, the processing rule information is updated according to the task modification information to obtain the target processing rule information. The target processing rule information can be the new processing rule information generated after modifying the processing rule information based on the task modification information. The target information processing task is then constructed based on the target processing rule information. Before updating the processing rule information based on the task modification information, a dialogue agent can be invoked to analyze the task modification information and understand the user's modification intent. If the modification intent is unclear, it can be clarified through dialogue with the user. Only after clarifying the user's modification intent can the processing rule information be updated.

[0069] Continuing with the previous example, task modification information can represent the user's proposed modifications to the DAG plan, and the user's modifications are sent back to the dialogue agent. For example... Figure 2 As shown, the dialogue agent (intelligent dialogue assistant) can clarify the user's modification requests through multiple rounds of dialogue. After understanding the new instruction, the dialogue agent updates the DSL (changing the preference from high accuracy to low latency), and then re-triggers the planning agent to generate a new DAG plan that better meets the user's expectations. This "planning-feedback-modification" cycle can be repeated multiple times until the user is satisfied with the generated DAG plan.

[0070] In summary, by updating the processing rule information based on task modification information to obtain target processing rule information, and constructing the target information processing task corresponding to the information processing task based on the target processing rule information, the generated new target information processing task can meet the user's personalized needs for resource review efficiency and cost.

[0071] Furthermore, considering that an information processing task contains at least one processing subtask, and each processing subtask needs to be executed during the execution of the information processing task, the execution data of each task needs to be integrated after the execution of each processing subtask is completed in order to obtain the information processing data. The specific implementation is as follows: Execute at least one processing subtask included in the information processing task according to the task description information, and obtain task execution data corresponding to the at least one processing subtask respectively; integrate the at least one task execution data to obtain the information processing data corresponding to the information to be processed.

[0072] Based on this, at least one processing subtask is executed according to the task description information to obtain task execution data corresponding to each of the at least one processing subtask. When integrating the at least one task execution data, the dependencies between the at least one processing subtask need to be considered. Based on the dependencies between the at least one processing subtask, the target task execution data to be integrated is selected from the at least one task execution data to obtain the information processing data corresponding to the information to be processed.

[0073] Following the previous example, after the user confirms the generated DAG plan, it can be handed over to the execution orchestrator for the actual image review task. The execution orchestrator strictly schedules and manages the actual review work according to the user-confirmed DAG plan. The execution orchestrator receives the final DAG plan and parses all task nodes (at least one processing subtask) and their dependencies. The orchestrator identifies all "starting nodes" in the DAG without prior dependencies and dispatches the tasks in parallel. For advertising image review, it simultaneously sends two requests to the model / capability layer: Request 1: Call the ocr-highres Agent to process the image. Request 2: Call the logo-detector (logo detection) Agent to process the same image. The orchestrator waits for the parallel tasks to return results. Once a task is completed, it checks if any downstream tasks depend on this result. ocr-highres returns the recognized text: "Effective in one month, guaranteed weight loss!" Upon receiving this, the orchestrator immediately triggers the downstream task that depends on this text: sending this text to the RAG / rule engine for extreme word matching. The rules engine returns a hit result: "The word 'guarantee' is in violation." Simultaneously, the logo-detector also returns a result. Once all upstream dependencies of a node are complete, the orchestrator immediately executes that node. At this point, both dependencies of the semantic understanding node (OCR and Logo recognition) are complete. The orchestrator packages the image, recognized text, and logo information, calls the VQA model in the model / capability layer, and performs a semantic judgment on "whether exaggerated claims exist." After all DAG nodes have been executed, the orchestrator summarizes all AI outputs (OCR text, rule hits, VQA judgments, confidence scores, etc.) to form a complete "evidence package," i.e., information processing data.

[0074] In summary, by selecting target task execution data to be integrated from at least one task execution data based on the dependencies between at least one processing subtask, information processing data corresponding to the information to be processed can be obtained, thereby improving the accuracy of information processing data.

[0075] Furthermore, after completing the review of the information to be processed and obtaining the information processing data, the processing rule information, information processing tasks, task description information, and information processing data generated during the review process can be persistently stored. The specific implementation is as follows: Based on the information to be processed, the processing rule information, the information processing task, the task description information, and the information processing data, information processing record data corresponding to the information to be processed is generated; the information processing record data is persistently stored and used to train the information processing model.

[0076] Based on this, the information to be processed, processing rules, information processing tasks, task descriptions, and information processing data are integrated to generate information processing record data corresponding to the information to be processed. This information processing record data is persistently stored and used to train the information processing model, improving the prediction accuracy of the resource audit model and optimizing it.

[0077] In summary, persistent storage of information processing records allows for the training of information processing models and the optimization of resource auditing models.

[0078] This specification provides an embodiment of an information processing method that receives user-submitted information to be processed, which may be a multimodal resource. The method determines processing rule information submitted by the user for the information to be processed. Based on the processing rule information, it constructs an information processing task corresponding to the information to be processed, enabling the construction of the information processing task with user participation and guidance, ensuring that the information processing task meets the user's review requirements for the information to be processed. The method receives task description information submitted by the user for the information processing task. Based on the task description information, it executes at least one processing sub-task included in the information processing task to obtain information processing data corresponding to the information to be processed, thus completing the execution of the information processing task under the user's guidance, obtaining information processing data that meets the user's review requirements, and feeding back the information processing data to the user. User participation in the construction and execution of the information processing task improves user satisfaction with resource review.

[0079] The following is in conjunction with the appendix Figure 3 Taking the application of the information processing method provided in this specification in the review of advertising resources as an example, the information processing method will be further explained. Figure 3 This specification illustrates an interactive diagram of an information processing method provided in one embodiment, which specifically includes the following steps.

[0080] Step 302: The client uploads the materials to be reviewed, and the material review request is submitted to the API gateway.

[0081] Step 304: The API gateway performs authentication, quota, and risk checks.

[0082] Step 306: After the API gateway check is passed, the materials to be reviewed are written to object storage.

[0083] Step 308: The dialogue clarifier clarifies the user's audit intent through multiple rounds of dialogue and generates DSL and acceptance criteria.

[0084] Step 310: Plan the Agent retrieval capability graph, generate the DAG plan corresponding to the DSL, and send the DAG plan to the execution orchestrator.

[0085] In practical applications, the initially generated DAG plan for reviewing advertising resources can be sent to the user for confirmation before execution. If the user confirms the DAG plan, subsequent processing steps can proceed. If the user does not confirm the DAG plan and suggests modifications, the DAG plan needs to be regenerated based on the user's suggestions until a DAG plan is confirmed by the user.

[0086] Step 312: The orchestrator executes the DAG plan and performs multimodal semantic parsing.

[0087] The orchestrator receives the final DAG plan and resolves all task nodes and their dependencies. It then performs concurrent task scheduling, identifying any "starting nodes" in the DAG without prior dependencies and dispatching them in parallel. The orchestrator waits for the parallel tasks to return results. Once a task completes, it checks if any downstream tasks depend on that result. When all upstream dependencies of a node are complete, the orchestrator immediately executes that node and finally aggregates the execution results. After all DAG nodes have been executed, the orchestrator aggregates all AI outputs (OCR text, rule hits, VQA judgments, confidence scores, etc.) to form a complete "evidence package."

[0088] Step 314: The knowledge base performs knowledge base retrieval based on the parsed content, constructs the context, and sends it to the strategy engine.

[0089] Step 316: The policy engine sends the hit terms to the execution orchestrator.

[0090] Step 318: Execute the orchestrator to perform confidence fusion and conflict resolution, obtain the execution results of the DAG plan, and send the execution results to the HITL auditor.

[0091] Scenario A (Submission for Review): If the rule engine detects the hard violation word "guarantee," or if the VQA model determines it as "exaggerated advertising" but the confidence level is below the set threshold, the orchestrator will determine that "submission for review is required." It will send the complete evidence package to the HITL review module, where a human reviewer will make the final decision on the review backend interface.

[0092] Scenario B (Automatic Decision): If all checks pass with high confidence, the orchestrator will directly conclude "Approved".

[0093] Step 320: The HITL auditor performs manual compliance and makes a ruling, sending the ruling result to the execution orchestrator.

[0094] Step 322: Execute the orchestrator to re-orchestrate, generate an audit report, and send it to Results and Audit.

[0095] Step 324: Write back the results and audit to perform a closed-loop learning process.

[0096] This records the final results, the decision-maker (AI or an auditor), and the complete chain of evidence. This high-quality labeled data may be used for retraining future models to achieve closed-loop learning.

[0097] Step 326: The client receives the callback and executes the processing strategy.

[0098] The API gateway's callback mechanism provides feedback on the final result and reason to the client / user who initially initiated the task.

[0099] In summary, the information processing method provided in this embodiment structures natural language audit requirements into a DSL (task / modality / rule set), while generating rollbackable acceptance thresholds and policy snapshots. It pre-positions "task semantics" as executable specifications, reducing rework and becoming a single, reliable source for planning and testing. This forms an interpretable closed loop of "conclusion + basis + evidence," satisfying strong compliance auditing and replayable accountability. It integrates model confidence, evidence weight, cross-modal consistency, and clause support for hierarchical threshold decisions; it triggers high-precision backtracking and HITL based on risk and uncertainty, and incorporates backtracking sets and degradation strategies. While stabilizing SLAs, it reduces false positives / false negatives and amortized costs, achieving adaptive auditing with a "light path as the primary approach and heavy path as a fallback." It uses RAG to retrieve clause context from policy / case libraries and aligns it with parsed cross-modal facts to generate interpretable basis; manual audit feedback and online sample feedback update capability profiles and rule weights. It elevates identification to auditing ("with clauses, evidence, and explanation"), and continuously improves robustness and iteration efficiency through learning.

[0100] System architecture of information processing methods, such as Figure 4As shown, the system architecture specifically includes the access layer, data and governance, HITL and operations, observation and security / platform, planning and scheduling, perception and computing, and knowledge and planning domains. A detailed explanation is provided using a practical "advertising image review" task as an example. An advertiser uploaded an image promoting health products, which reads "See results in one month, guaranteed weight loss!" and includes a brand logo. The system needs to review whether the image violates regulations.

[0101] Step 1: Access Request and Material Storage. The advertiser submits the ad image (e.g., ad.jpg) and related metadata (such as the advertiser ID) by calling the API gateway's upload interface. After authentication, the API gateway forwards the request to the import service. The import service stores the image file ad.jpg in the data platform of the data and governance domain and obtains a material URI (e.g., s3: / / bucket / ad_images / xyz.jpg). The import service then passes this URI and task information to the planning and orchestration domain to initiate the review process.

[0102] Step Two: Intelligent Planning and Task Orchestration. This is the brain / core of the review process. It receives tasks and formulates detailed execution plans. After receiving the review task (including the material URI), the planning service queries the capability graph to find all AI capabilities that can handle "image ad review". The capability graph returns several candidate capabilities: ocr-highres (high-precision OCR), logo-detector (logo recognition), sensitive-word-engine (sensitive word rules), and vlm-vqa-model (visual question answering model). The planning service selects and plans according to a preset "accuracy priority" strategy, generating an execution plan (DAG): Step A (parallel): Call ocr-highres to recognize text in the image; Step B (parallel): Call logo-detector to recognize the brand logo; Step C (dependent on A): Send the text recognized by OCR to sensitive-word-engine for rule matching; Step D (dependent on A, B): Send the original image, OCR text, and logo information together to vlm-vqa-model for deep semantic understanding to determine whether there is "exaggerated advertising". This plan, which includes specific steps and dependencies, is sent to the orchestrator for execution.

[0103] Step 3: Multi-path Concurrent Execution and Evidence Construction. This is the concrete execution layer. The orchestrator acts as a commander-in-chief, simultaneously issuing instructions to multiple computation modules. According to the DAG plan, the orchestrator concurrently issues tasks to the perception and computation domain and the knowledge and rules domain: Task A (OCR): Orchestrator -> Model Routing -> Model Service, execute ocr-highres. The input is the image URI, and the output is a structured text block ["See results in one month", "Guaranteed weight loss!"]. The knowledge and rules domain provides knowledge retrieval services. Task B (Logo): Orchestrator -> Model Routing -> Model Service, execute logo-detector. The input is the image URI, and the output is the logo location and name. After the OCR task is completed, the orchestrator immediately triggers Task C. Task C (Rule Matching): The orchestrator sends the text ["See results in one month", "Guaranteed weight loss!"] to the rule engine of the knowledge and rules domain. The rule engine loads a policy from the policy service and matches "guarantee" as a violation word. The output is a rule hit result {"rule_id":"PROMISE_GUARANTEE","hit":"guarantee"}. Throughout the process, the evidence construction module continuously collects the outputs of all the above tasks and begins to build the evidence package.

[0104] Step Four: In-depth Analysis and Final Decision. After the initial analysis, more complex reasoning is performed, and all evidence is summarized to make a final judgment. Once Task A and Task B are completed, the orchestrator triggers Task D. Task D (VQA): The orchestrator packages the image URI, text, and logo information, sends it to the model route, and attaches an internal prompt: "Please determine if this advertisement contains exaggerated or promiscuous claims." The model service performs reasoning and obtains the judgment conclusion: {"conclusion":"exaggerated claims","confidence":0.95,"reason":"contains the promiscuous word 'guarantee'"}. The evidence construction module also adds the result of Task D to the evidence package. The orchestrator sends this complete evidence package to the decision engine. The decision engine, based on its internal logic (e.g., IF(rule hits PROMISE_GUARANTEE) OR(VQA conclusion is exaggerated claims and confidence > 0.8) THEN review rejection), makes a final judgment: "Reject".

[0105] Step 5: Result Archiving and Feedback. This completes the entire review loop, stores the results, and notifies the caller. The decision engine returns the final result (rejection) and a complete evidence package to the orchestrator. The orchestrator sends the review result to the data platform in the data and governance domain for "result archiving" for future auditing and model training. Through the callback service in the access layer, the review result (rejection) and the reason ("contains prohibited promises") are asynchronously pushed to the advertiser. (If manual review is required) This review case will be pushed to the HITL and operations domain's review console for secondary confirmation by operations personnel. Through these steps, a complex ad image review is completed efficiently and transparently. The output of each step becomes the input for the next, ultimately forming an irrefutable "evidence package" that supports automated decision-making.

[0106] Figure 5 This specification shows a schematic diagram of the structure of an information processing system according to one embodiment. Figure 5 As shown, the information processing system 500 includes a client 510 and a server 520, comprising: The client 510 is configured to receive pending information submitted by a user and send the pending information to the server 520; the server 520 is configured to determine the processing rule information submitted by the user for the pending information; construct an information processing task corresponding to the pending information based on the processing rule information, and receive task description information submitted by the user for the information processing task; execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the pending information, and feed the information processing data back to the client 510; the client 510 is configured to display the information processing data to the user through an information processing page.

[0107] In practical applications, the client receives pending information submitted by the user, which can be multimodal resources. The client can send the pending information to the server, where the server determines the processing rules submitted by the user for the pending information. The server then invokes an intelligent agent to construct an information processing task corresponding to the pending information based on the processing rules. This allows for the construction of the information processing task with user participation and guidance, ensuring that the task meets the user's review requirements. The server receives the task description information submitted by the user for the information processing task. Based on the task description, it executes at least one processing subtask within the information processing task, obtaining the information processing data corresponding to the pending information. This completes the execution of the information processing task under the user's guidance, obtaining information processing data that meets the user's review requirements, and then feeding the information processing data back to the client. The client can then display the information processing data to the user through an information processing page. User participation in the construction and execution of the information processing task improves user satisfaction with resource review.

[0108] In practical applications, the architecture of an information processing system is as follows: Figure 6 As shown, the information processing system interacts with clients through an API gateway. The client includes modules for dialogue management, multi-turn dialogue, plan generation, process management, task management, knowledge management, capability management, and tenant management. The API gateway can perform user authentication, quotas, and risk checks. The information processing system also includes an AI-Generator unit, an operations management unit, a Runtime unit, a core capability layer, a data layer, and a model layer. The AI-Generator unit includes a requirement clarification module, a task planning module, and a task execution module. The requirement clarification module includes a requirement clarification chatbot that, through multi-turn dialogue with the user, gradually identifies the task scope / modality / metric / rule source, outputs the requirement DSL and acceptance criteria, and achieves requirement management. The task planning module contains a unified planning agent, used to analyze the user's requirement DSL, retrieve candidate agents and capability sets from the capability graph, and generate a DAG plan through agent combination and requirement constraints. The task execution module is responsible for executing tasks as needed, calling core capabilities according to the plan generated by the unified planning agent.

[0109] The Operations Management Unit includes a DAG visualization module, a plan template management module, a capability marketplace module, and a knowledge base. The DAG visualization module graphically presents the structure of a directed acyclic graph (DAG). The template and capability marketplace allow users to configure industry templates (customer service, advertising, invoices), agents, etc. The knowledge base stores industry knowledge, primarily for RAG (Regional Acyclic Graph), expanding the cognitive boundaries of the model. The plan template management module manages plans generated by a unified planning agent, allowing users to continuously iterate and optimize. The Runtime Unit includes an SDK, a workflow engine, an audit strategy engine, and flow control. The SDK provides interfaces for external systems, handling actual audit tasks and obtaining audit results. The workflow engine can be based on the traditional BPMN protocol to execute audit workflows. The audit strategy engine performs HITL (High-Integrity Tolerance) decisions based on rules such as confidence levels configured for the task. For example, it uses a fixed algorithm to provide a quantitative indicator of "confidence," such as specifying a confidence interval of [0-1], where a larger number indicates higher confidence. When the confidence level given by the large model is 0.2, the result can be considered basically unreliable, and the task can be transferred to manual processing.

[0110] The core capability layer includes OCR services, feature services, ASR services, testing services, rule engine, VLM services, and RAG capabilities. The data layer includes database storage, vector library, rules, industry knowledge, and trajectory logs. The model layer includes large models for processing resources such as plain text, images, audio, and video.

[0111] The information processing system uses a three-layer closed loop (requirements clarification → solution planning → execution / HITL) to connect "natural language requirements—rules—model capabilities—review and decision-making" into a configurable, traceable, and continuously optimizable pathway. A complete resource review architecture is formed by a multi-round dialogue-based requirements clarification agent, a capability marketplace (Agent / model / rule package), process templates (industry-specific DAGs), a strategy center (thresholds / grading / handling), and an operations and maintenance portal (monitoring / auditing / replay). Dialogue-based requirements clarification can generate requirement DSLs and acceptance criteria. The unified planning agent generates executable DAG plans based on capability graphs and multi-objective constraints, supporting in-dialogue editing and optimization of execution plans.

[0112] Corresponding to the above method embodiments, this specification also provides embodiments of an information processing apparatus. Figure 7 A schematic diagram of the structure of an information processing apparatus according to one embodiment of this specification is shown. Figure 7 As shown, the device includes: The receiving module 702 is configured to receive pending information submitted by a user and determine the processing rule information submitted by the user for the pending information. The construction module 704 is configured to construct an information processing task corresponding to the information to be processed based on the processing rule information, and to receive task description information submitted by the user for the information processing task; The execution module 706 is configured to execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the information to be processed, and feed back the information processing data to the user.

[0113] In an optional embodiment, the receiving module 702 is further configured to: Determine the initial processing rule information submitted by the user for the information to be processed; Invoke the dialogue agent to determine the initial processing intent information based on the initial processing rule information; Determine the dialogue data generated between the user and the dialogue agent based on the initial processing intent information, and extract processing elements from the dialogue data; The initial processing rule information is updated based on the processing elements to obtain the processing rule information.

[0114] In an optional embodiment, the receiving module 702 is further configured to: Determine the user's user information and processing request information; Based on the user information, a user compliance check is performed on the user. If the check passes, the processing request information is then processed and checked. If the processing detection passes, a processing task identifier is configured for the information to be processed, and the processing task identifier is used to query the information to be processed.

[0115] In an optional embodiment, the construction module 704 is further configured to: The intelligent agent is invoked to determine the processing objective information and processing constraint information based on the processing rule information; The processing objective information and the processing constraint information are converted into a sequence of processing nodes containing at least one processing node, and smart agents are matched for each of the at least one processing node. According to the node order corresponding to the processing node sequence, a node execution graph is constructed based on the intelligent agents corresponding to the at least one processing node, and an information processing task corresponding to the information to be processed is constructed based on the node execution graph.

[0116] In an optional embodiment, the execution module 706 is further configured to: If the user confirms the information processing task based on the task description information, at least one processing subtask included in the information processing task is executed to obtain the information processing data corresponding to the information to be processed.

[0117] In an optional embodiment, the execution module 706 is further configured to: If it is determined from the task description information that the user has submitted task modification information for the information processing task, the information processing task is updated to the target information processing task based on the task modification information. The target information processing task is taken as the information processing task, and the step of receiving the task description information submitted by the user for the information processing task is executed.

[0118] In an optional embodiment, the execution module 706 is further configured to: The processing rule information is updated based on the task modification information to obtain the target processing rule information; The intelligent agent is invoked to construct the target information processing task corresponding to the information processing task based on the target processing rule information.

[0119] In an optional embodiment, the execution module 706 is further configured to: Execute at least one processing subtask included in the information processing task according to the task description information, and obtain task execution data corresponding to each of the at least one processing subtask; Integrate at least one task execution data to obtain the information processing data corresponding to the information to be processed.

[0120] In an optional embodiment, the execution module 706 is further configured to: Based on the information to be processed, the processing rule information, the information processing task, the task description information, and the information processing data, generate information processing record data corresponding to the information to be processed; The information processing record data is persistently stored and used to train the information processing model.

[0121] An embodiment of this specification provides an information processing apparatus that receives user-submitted information to be processed, which may be multimodal resources. It determines processing rule information submitted by the user for the information to be processed. Based on the processing rule information, it constructs an information processing task corresponding to the information to be processed, enabling the construction of the information processing task with user participation and guidance, ensuring that the information processing task meets the user's review requirements for the information to be processed. It receives task description information submitted by the user for the information processing task. According to the task description information, it executes at least one processing sub-task included in the information processing task to obtain information processing data corresponding to the information to be processed, thus completing the execution of the information processing task under the user's guidance, obtaining information processing data that meets the user's review requirements, and feeding back the information processing data to the user. User participation in the construction and execution of the information processing task improves user satisfaction with resource review.

[0122] The above is an illustrative scheme of an information processing device according to this embodiment. It should be noted that the technical solution of this information processing device and the technical solution of the information processing method described above belong to the same concept. For details not described in detail in the technical solution of the information processing device, please refer to the description of the technical solution of the information processing method described above.

[0123] Figure 8 A structural block diagram of a computing device 800 according to one embodiment of this specification is shown. The components of the computing device 800 include, but are not limited to, a memory 810 and a processor 820. The processor 820 is connected to the memory 810 via a bus 830, and a database 850 is used to store data.

[0124] The computing device 800 also includes an access device 840, which enables the computing device 800 to communicate via one or more networks 860. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 840 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0125] In one embodiment of this specification, the above-described components of the computing device 800 and Figure 8 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 8 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0126] The computing device 800 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 800 can also be a mobile or stationary server.

[0127] The processor 820 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described information processing method.

[0128] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the information processing method described above belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the information processing method described above.

[0129] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described information processing method.

[0130] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the information processing method described above belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the information processing method described above.

[0131] An embodiment of this specification also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described information processing method.

[0132] The above is an illustrative scheme of a computer program product according to this embodiment. It should be noted that the technical solution of this computer program product and the technical solution of the information processing method described above belong to the same concept. For details not described in detail in the technical solution of the computer program product, please refer to the description of the technical solution of the information processing method described above.

[0133] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0134] The computer program / instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0135] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0136] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0137] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. An information processing method, comprising: Receive pending information submitted by a user and determine the processing rule information submitted by the user for the pending information; Based on the processing rule information, an information processing task corresponding to the information to be processed is constructed, and the task description information submitted by the user for the information processing task is received; Execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the information to be processed, and feed back the information processing data to the user.

2. The information processing method according to claim 1, wherein determining the processing rule information submitted by the user for the information to be processed includes: Determine the initial processing rule information submitted by the user for the information to be processed; Invoke the dialogue agent to determine the initial processing intent information based on the initial processing rule information; Determine the dialogue data generated between the user and the dialogue agent based on the initial processing intent information, and extract processing elements from the dialogue data; The initial processing rule information is updated based on the processing elements to obtain the processing rule information.

3. The information processing method according to claim 1, further comprising, before determining the processing rule information submitted by the user for the information to be processed: Determine the user's user information and processing request information; Based on the user information, a user compliance check is performed on the user. If the check passes, the processing request information is then processed and checked. If the processing detection passes, a processing task identifier is configured for the information to be processed, and the processing task identifier is used to query the information to be processed.

4. The information processing method according to claim 1, wherein constructing the information processing task corresponding to the information to be processed based on the processing rule information includes: Based on the processing rule information, the processing objective information and processing constraint information are determined; The processing objective information and the processing constraint information are converted into a sequence of processing nodes containing at least one processing node, and smart agents are matched for each of the at least one processing node. According to the node order corresponding to the processing node sequence, a node execution graph is constructed based on the intelligent agents corresponding to the at least one processing node, and an information processing task corresponding to the information to be processed is constructed based on the node execution graph.

5. The information processing method according to claim 1, wherein executing at least one processing subtask included in the information processing task according to the task description information to obtain information processing data corresponding to the information to be processed includes: If the user confirms the information processing task based on the task description information, at least one processing subtask included in the information processing task is executed to obtain the information processing data corresponding to the information to be processed.

6. The information processing method according to claim 1, wherein executing at least one processing subtask included in the information processing task according to the task description information to obtain information processing data corresponding to the information to be processed includes: If it is determined from the task description information that the user has submitted task modification information for the information processing task, the information processing task is updated to the target information processing task based on the task modification information. The target information processing task is taken as the information processing task, and the step of receiving the task description information submitted by the user for the information processing task is executed.

7. The information processing method according to claim 6, wherein updating the information processing task to the target information processing task based on the task modification information comprises: The processing rule information is updated based on the task modification information to obtain the target processing rule information; The intelligent agent is invoked to construct the target information processing task corresponding to the information processing task based on the target processing rule information.

8. The information processing method according to claim 1, wherein executing at least one processing subtask included in the information processing task according to the task description information to obtain information processing data corresponding to the information to be processed includes: Execute at least one processing subtask included in the information processing task according to the task description information, and obtain task execution data corresponding to each of the at least one processing subtask; Integrate at least one task execution data to obtain the information processing data corresponding to the information to be processed.

9. The information processing method according to claim 1, further comprising, after feeding back the processed information data to the user: Based on the information to be processed, the processing rule information, the information processing task, the task description information, and the information processing data, generate information processing record data corresponding to the information to be processed; The information processing record data is persistently stored and used to train the information processing model.

10. An information processing system, comprising a client and a server, including: The client is used to receive pending information submitted by the user and send the pending information to the server. The server is used to determine the processing rule information submitted by the user for the information to be processed; Based on the processing rule information, an information processing task corresponding to the information to be processed is constructed, and the task description information submitted by the user for the information processing task is received; Execute at least one processing subtask included in the information processing task according to the task description information, obtain information processing data corresponding to the information to be processed, and feed the information processing data back to the client; The client is used to display the information processing data to the user through the information processing page.

11. A computing device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 9.

12. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 9.

13. A computer program product comprising a computer program or instructions which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 9.

Images