AI agent real-time data verification method, system, medium, program and terminal based on dynamic mapping

Abstract

The application provides constructing a dynamic mapping table by using historical data, collecting historical input data and output data, and establishing a one-to-one correspondence with the mapping identifier. When real-time input data is obtained, the system can determine which category set the real-time input data belongs to by matching the mapping identifier, and judge whether the real-time input data meets the characteristic mode of the category by combining the characteristic range of the historical input data corresponding to the category set. When real-time input data is obtained, the system can generate corresponding real-time output data, and determine which category set the output should belong to by matching the mapping identifier. Subsequently, the system compares the real-time output data with the result range or mode of the historical output data in the category set to determine whether it meets the output expectation of the category. Thus, the input-output accuracy, stability and business controllability of the AI agent during operation are improved.

Description

Technical Field

[0001] The present application relates to the field of AI agents, and in particular to a method, system, medium, program and terminal for real-time data verification of AI agents based on dynamic mapping. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, AI agents have been widely applied in fields such as natural language processing and decision support. However, the consistency and reliability of their input and output are becoming increasingly prominent. AI agents typically receive multimodal input from users, such as text and speech. However, their output is susceptible to various factors, including limitations in model training data, flaws in algorithm design, and uncertainties in the external environment. These factors, combined, can lead to biased, inconsistent, and even confusing outputs.

[0003] Currently, verification methods for AI agents primarily focus on the training phase, while verification of runtime input and output is significantly insufficient. Verification methods for the runtime phase primarily include static rule matching and fixed dataset comparison. Static rule matching verifies input and output based on preset rules, but these rules are often too rigid to adapt to complex and changing real-world application scenarios. Fixed dataset verification compares current output with historical data. While this can ensure output consistency to a certain extent, it lacks the ability to respond to real-time changes and cannot promptly detect emerging anomalies. Both methods have significant limitations when processing dynamic input, making them difficult to meet the requirements for reliable AI applications in complex environments.

[0004] Furthermore, existing technologies have serious deficiencies in the traceability and explainability of AI outputs, both during training evaluation and during operational phases of rule matching and dataset verification. This makes it difficult for users to determine whether AI-generated results meet expectations, and it's impossible to trace the specific rationale behind AI decisions. This not only reduces AI's credibility in high-risk scenarios but also hinders its further application in broader areas.

[0005] Therefore, an efficient, flexible and dynamically adaptable input and output verification mechanism is needed to improve the stability and security of AI in practical applications. Summary of the Invention

[0006] In view of the shortcomings of the prior art described above, the purpose of this application is to provide an AI agent real-time data verification method, system, medium, program and terminal based on dynamic mapping to solve the above problems.

[0007] To achieve the above-mentioned purpose and other related purposes, the first aspect of the present application provides a real-time data verification method for an AI agent based on dynamic mapping, comprising: obtaining a number of historical data of the AI ​​agent, wherein the historical data includes historical input data and its corresponding historical output data; performing feature extraction on each historical input data to obtain its corresponding input feature set; using a clustering algorithm to classify historical input data and its corresponding historical output data with similar input feature sets into a category set, and assigning a unique mapping identifier to each category set; establishing a dynamic mapping table based on all category sets and their corresponding mapping identifiers; counting and analyzing the feature range of historical input data and the result range or pattern of historical output data in each category set; obtaining real-time input data of the AI ​​agent; extracting the features of the real-time input data to obtain its corresponding input feature set; searching for a matching mapping identifier in the dynamic mapping table based on the input feature set corresponding to the features of the real-time input data, and judging whether the real-time input data meets the feature range of the input data in the category set corresponding to the mapping identifier. If not, the real-time input data is judged to be abnormal.

[0008] In one embodiment of the first aspect of the present application, based on the real-time output data corresponding to the acquired real-time input data output, it is determined whether the real-time output data conforms to the result range or pattern of the output data within the category set corresponding to the matching mapping identifier in the dynamic mapping table; if not, the real-time output data is determined to be abnormal.

[0009] In one embodiment of the first aspect of the present application, when determining whether the real-time input data meets the feature range of the input data within the category set corresponding to the mapping identifier, if the input feature set of the real-time input data does not belong to any category set in the existing dynamic mapping table, a new category set is created for the real-time data and its corresponding real-time output data, a unique mapping identifier is assigned, and the category set and its corresponding mapping identifier are included in the dynamic mapping table.

[0010] In one embodiment of the first aspect of the present application, the result range or pattern of the output data is comprehensively modeled through rule templates, vector space ranges, and multimodal matching logic to support unified verification of various types of output results such as structured text, natural language text, and images.

[0011] In an embodiment of the first aspect of the present application, the input feature set includes numerical features, and / or text features, and / or categorical features.

[0012] In one embodiment of the first aspect of the present application, when feature extraction is performed on historical input data and / or real-time input data respectively, feature enhancement is performed in combination with the context information of the historical input data and / or real-time input data, and / or interaction history, and / or user portrait information.

[0013] To achieve the above-mentioned purpose and other related purposes, the second aspect of the present application provides an AI agent real-time data verification system based on dynamic mapping, including: a dynamic mapping table establishment module, used to obtain a number of historical data of the AI ​​agent, wherein the historical data includes historical input data and its corresponding historical output data; feature extraction is performed on each historical input data to obtain its corresponding input feature set; a clustering algorithm is used to classify historical input data and its corresponding historical output data with similar input feature sets into a category set, and a unique mapping identifier is assigned to each category set; a dynamic mapping table is established based on all category sets and their corresponding mapping identifiers; the feature range of historical input data and the result range or pattern of historical output data in each category set are counted and analyzed; a real-time data analysis module is used to obtain the real-time input data of the AI ​​agent; the features of the real-time input data are extracted to obtain its corresponding input feature set; according to the input feature set corresponding to the features of the real-time input data, a matching mapping identifier is searched in the dynamic mapping table, and it is determined whether the real-time input data meets the feature range of the input data in the category set corresponding to the mapping identifier. If not, the real-time input data is determined to be abnormal.

[0014] To achieve the above-mentioned purpose and other related purposes, the third aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which implements any of the aforementioned methods when executed by a processor.

[0015] To achieve the above-mentioned purpose and other related purposes, the fourth aspect of the present application provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer implements any of the methods described above.

[0016] To achieve the above-mentioned purpose and other related purposes, the fifth aspect of the present application provides an electronic terminal, including a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to implement any of the above-mentioned methods.

[0017] As described above, the present application has the following beneficial effects:

[0018] This application uses historical data to construct a dynamic mapping table, aggregates historical input data and output data, and establishes a one-to-one correspondence with mapping identifiers. When real-time input data is acquired, the system can determine which category set the real-time input data belongs to by matching the mapping identifier. This is combined with the feature range of the historical input data corresponding to the set to determine whether the real-time input data conforms to the characteristic pattern of the category. If it conforms, the input is reasonable and within the expected range, and the system can process it normally. If it does not conform, potential abnormal inputs can be discovered in a timely manner, thereby improving the data stability, security, and response reliability of the AI ​​agent during operation. Furthermore, after acquiring real-time input data, the system can generate corresponding real-time output data and determine which category set the output belongs to by matching the mapping identifier. The system then compares the real-time output data with the result range or pattern of the historical output data in the category set to determine whether it conforms to the output expectations of the category. If it conforms, the current output is semantically or logically reasonable and can be considered a valid response. If it does not conform, potential output anomalies or model deviations can be identified in a timely manner, thereby improving the output accuracy, stability, and business controllability of the AI ​​agent during operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Shown is a flow chart of a method for real-time data verification of an AI agent based on dynamic mapping in one embodiment of the present application.

[0020] Figure 2 Shown is a flow chart of a method for real-time data verification of an AI agent based on dynamic mapping in one embodiment of the present application.

[0021] Figure 3 Shown is a structural diagram of an AI agent real-time data verification system based on dynamic mapping in one embodiment of the present application.

[0022] Figure 4 Shown is a structural schematic diagram of an electronic terminal in one embodiment of the present application. DETAILED DESCRIPTION

[0023] The following describes the embodiments of the present application through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0024] In the embodiments of this application, terms such as "first" and "second" are used to distinguish between identical or similar items with substantially the same function or effect. For example, the first XX and the second XX are merely used to distinguish between different XXs and do not limit their order. Those skilled in the art will understand that terms such as "first" and "second" do not limit the quantity or order of execution, and that terms such as "first" and "second" do not necessarily define differences.

[0025] It should be noted that in the embodiments of this application, words such as "exemplary" or "for example" represent examples, illustrations, or descriptions. Any embodiment or design described in this application as "exemplary" or "for example" should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0026] In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc or abc, where a, b, c can be single or multiple.

[0027] like Figure 1-2 As shown, the first aspect of the present application provides an AI agent real-time data verification method based on dynamic mapping, comprising:

[0028] S1: Obtain some historical data of the AI ​​agent, wherein the historical data includes historical input data and its corresponding historical output data.

[0029] Preferably, historical data can be derived from user interaction logs, system call logs, or other forms of input and output samples recorded by the AI ​​agent during actual deployment. Historical input data includes natural language input and structured instructions from the user, while output data is the response or operation result generated by the AI ​​agent based on this input.

[0030] Preferably, historical data covering the main input types and output modes of the target application scenario is collected to ensure that subsequent modeling is fully representative and comprehensive. In practical applications, at least thousands to tens of thousands of input and output sample pairs can be collected, depending on the complexity of the specific task and the diversity of the inputs. When the additional data does not significantly improve the quality of the clustering or the stability of the mapping, the data volume can be considered to have basically met the modeling requirements.

[0031] S2: Perform feature extraction on each historical input data to obtain its corresponding input feature set.

[0032] It should be understood that when the historical input data and / or real-time input data is text data, feature extraction is performed directly. When it is voice data, the voice data is first converted into text data using ASR automatic speech recognition technology before feature extraction is performed.

[0033] In one embodiment of the first aspect of the present application, when feature extraction is performed on historical input data and / or real-time input data respectively, feature enhancement is performed in combination with the context information of the historical input data and / or real-time input data, and / or interaction history, and / or user portrait information.

[0034] It should be understood that by introducing contextual enhancement features such as contextual information of input data, user interaction history, and user portrait information, the limitations of semantic expression of a single input data can be effectively compensated. In practical applications, the user's current input is often closely related to their previous multi-round conversations, operational behaviors, or long-term preferences. Feature extraction based solely on the current input content may not accurately reflect their true intentions. By integrating the contextual conversation content, semantic coherence can be retained; introducing interaction history helps identify user behavior patterns; and combining user portrait information (such as domain preferences, common vocabulary styles, etc.) can enhance the system's adaptability to individual differences. The fusion processing of the above information can significantly enhance the expressive power of input features and improve the accuracy of similar merging in the clustering stage. It also helps to improve the accuracy and robustness of matching real-time input with historical categories, especially when facing semantically ambiguous or boundary input.

[0035] Preferably, after obtaining the input feature set of the historical input data, the input feature set is preprocessed. The preprocessing process includes data cleaning and feature scaling. Data cleaning includes detecting, correcting, or removing missing values ​​and outliers that may exist in the input feature set to ensure the integrity and validity of the input data. Feature scaling is to normalize or standardize the original numerical features. For example, the feature values ​​can be scaled to the interval [0, 1] or standardized to a distribution with a mean of 0 and a variance of 1 to prevent features of different scales from interfering with the matching calculation.

[0036] S3: Use a clustering algorithm to classify historical input data with similar input feature sets and their corresponding historical output data into a category set, and assign a unique mapping identifier to each category set.

[0037] Preferably, the clustering algorithm can use K-means, DBSCAN, hierarchical clustering, etc. to group and classify historical input data with similar characteristics and associate the corresponding historical output results. A unique mapping identifier is assigned to each clustered category set, which serves as an index basis for subsequent input and output matching and verification.

[0038] S4: Create a dynamic mapping table based on all category sets and their corresponding mapping identifiers.

[0039] It should be understood that the dynamic mapping table is used to record historical input and output data and their mapping relationship with category identifiers. This mapping table can be stored in a database, cache, or embedded data structure, and is queryable and updateable. It is the core data structure of the entire verification process.

[0040] S5: Count and analyze the feature range of the historical input data and the result range or pattern of the historical output data within each category set.

[0041] It should be understood that by establishing the value boundaries and semantic distribution of each feature of historical input data, the feature range is determined, and typical expressions or behavioral patterns of the corresponding output results are extracted, forming the basis for subsequent comparison. The result range of historical output data can be modeled using rule templates, vector space ranges, example sets, etc. to adapt to different output forms.

[0042] S6: Obtain real-time input data of the AI ​​agent; extract features of the real-time input data to obtain its corresponding input feature set.

[0043] It should be understood that when the AI ​​agent receives new input, the system extracts its features and adopts the same feature extraction (and enhancement strategy) as the historical input data, thereby maintaining the consistency and comparability of the input features.

[0044] S7: Querying a matching mapping identifier in the dynamic mapping table according to the input feature set corresponding to the feature of the real-time input data.

[0045] In an embodiment of the first aspect of the present application, the input feature set includes numerical features, and / or text features, and / or categorical features.

[0046] Preferably, corresponding matching strategies are adopted for different types of features:

[0047] When the system uses numerical features as input feature sets, matching is achieved by comparing the numerical features in the real-time input data with the numerical feature ranges in the category set corresponding to the mapping identifier. For example, the mapping table specifies that the age range of 18-30 years old is an input feature set, corresponding to mapping identifier A. If the age feature value in the new data is 22 years old, it meets this range and is initially determined to be a possible match with identifier A. The system predefines the value range of the numerical features covered by each category set in the dynamic mapping table as the basis for matching. For example, the mapping table specifies that input data with the "age" feature between 18 and 30 years old is classified into a certain input feature set and corresponds to mapping identifier A. When the system receives a new real-time input data with the "age" field value of 22 years old, it compares this value with the corresponding value range in the mapping table. Since 22 years old falls within the range of 18-30 years old, the numerical feature of the real-time input data meets the range criteria set by category set A. Based on this, the system can initially determine that the data is a possible match with the category set represented by mapping identifier A.

[0048] Preferably, in order to enhance the flexibility and accuracy of matching, the system can also be set according to business needs: closed interval or open interval matching rules (such as whether to include boundary values); allowed numerical tolerance range (such as ±5% floating); and priority judgment strategy (when multiple category sets are matched, how to select the best match).

[0049] When categorical features are used as input feature sets, the matching process is based on an exact match of categorical values. That is, the categorical feature values ​​carried in the real-time input data are directly compared with the categorical values ​​contained in each category set in the dynamic mapping table to determine their attribution. For example, if a category set is defined in the mapping table as containing color feature values ​​of "red", "blue", and "green", and the set is associated with mapping identifier B. At this time, if the color feature value in the real-time input data is "blue", it can be determined that the value is within the categorical input feature set already in the mapping table, and therefore matches the category set represented by identifier B, thereby establishing an association relationship. In practical applications, categorical features can include but are not limited to: discrete attributes (such as gender: male / female, user type: ordinary / member / enterprise); label fields (such as product categories, fault types, behavioral labels); multiple-choice or enumerated information (such as interests, hobbies, industry, functional modules, etc.).

[0050] Preferably, the present invention supports both matching of single category feature values ​​and matching of multiple feature combinations. In a multi-feature scenario, the system will simultaneously examine whether multiple category fields simultaneously meet the matching conditions of a certain category set. For example, if a category set stipulates that "gender is female, user type is member" corresponds to mapping identifier C, the system will only determine its belonging identifier C under the premise that these two fields of the new data are matched at the same time. Furthermore, in order to improve flexibility, the system can support: partial matching (such as matching any value contained in a multiple-choice field); matching priority rules (selecting the best when multiple category sets can be matched); and matching fault tolerance mechanisms (such as supporting synonymous value mapping: "girl" can be regarded as "female").

[0051] When using text-based features as the input feature set, a text similarity algorithm is usually used to match the text content in the real-time input data with the existing text feature set in the mapping table. Due to the characteristics of openness, expression diversity and semantic ambiguity of text-based features, it is impossible to directly perform accurate matching like numerical or categorical features. Therefore, it is necessary to use a similarity algorithm to quantitatively compare texts with similar semantics or lexical features. Preferably, text matching methods such as cosine similarity algorithm, word embedding similarity, and edit distance can be used to calculate the similarity between real-time text and historical text. For example, if the text feature corresponding to a certain category set in the mapping table is "unable to start the system", and the real-time input text is "cannot start the machine", the two have different words on the surface but similar semantics. The similarity calculated by the BERT model is 0.87, which exceeds the matching threshold of 0.8 set by the system. Therefore, it can be determined that the real-time input should be classified as "unable to start the system" and associated with its corresponding mapping identifier.

[0052] It should be understood that in some scenarios, the mapping identifier (i.e., category number) is not simply determined by feature similarity, numerical range, or exact match, but is generated by a specific business logic or rule system definition. These rules are usually formulated based on domain knowledge, business processes, policy constraints, etc., and have clear judgment logic. Therefore, preferably, matching rules are established, that is, for the case where the generation method of some mapping identifiers is driven by specific business rules, the system needs to preset a set of executable matching logic rules and apply them to the second feature set composed of real-time input data. Such rules are usually formulated based on business needs, industry experience, or strategies, and have a clear conditional judgment structure, such as "if condition A and condition B are met, then classify it as mapping identifier X". This type of rule-based mapping method is suitable for scenarios where input features are difficult to accurately classify through a single similarity or range judgment, and is particularly widely applicable in systems with fixed business hierarchies, label divisions, behavioral rules, etc.

[0053] Preferably, the categorical features include keyword features, and the numerical features include semantic vector features.

[0054] It should be understood that keyword features are used to capture representative and important words in the input data. They usually extract the core expression content in the input through technical means such as word frequency statistics, TF-IDF, and named entity recognition (NER), which helps to reflect the explicit semantic information of the input. Semantic vector features map the input content to a low-dimensional semantic space through deep semantic modeling methods such as word embedding and sentence vectors, thereby capturing the implicit semantic associations and contextual dependencies in the text. Keyword features focus on interpretability and keyword coverage, while semantic vector features emphasize overall semantic understanding and generalization capabilities.

[0055] S8: Determine whether the real-time input data conforms to the characteristic range of the input data within the category set corresponding to the mapping identifier; if not, determine that the real-time input data is abnormal.

[0056] It should be understood that the input feature set of real-time input data is compared with the feature range of historical input data that has been modeled in the corresponding category set. The feature range may include keyword distribution, semantic vector space boundaries, contextual semantic coverage, etc. When the feature set of real-time input data deviates significantly from this range in multiple dimensions, or cannot be effectively matched with any existing category set, it means that the input is not covered by the known patterns of the current system and may be an abnormal input, an out-of-bounds request, or a new semantic scenario. At this time, the system can determine that the input has potential risks, such as misleading questions, illegal content, requests outside the model's capability boundaries, etc., and trigger a preset security response mechanism, such as an alarm prompt, rejection processing, transfer to manual review, or recording samples to be learned, to ensure the robustness and controllability of the AI ​​system during operation. This mechanism can effectively prevent the AI ​​model from generating incorrect responses or out-of-control behavior when facing uncertain input, and is a key component to improve system security and robustness.

[0057] In an embodiment of the first aspect of the present application, the following steps are further included:

[0058] S9: Outputting the real-time output data corresponding to the acquired real-time input data, determining whether the real-time output data conforms to the result range or pattern of the output data within the category set corresponding to the matching mapping identifier in the dynamic mapping table; if not, determining that the real-time output data is abnormal.

[0059] It should be understood that after generating real-time output data, the system compares it with the result range or pattern of historical output data within the category set corresponding to the matching mapping identifier in the dynamic mapping table. This result range or pattern can be modeled based on the semantic vector distribution, format template, keyword set, logical structure characteristics, and even statistical models of multimodal expressions such as images and audio. If the current output result significantly deviates from the output pattern of the category in terms of semantic expression, content structure, or output form, or contains unexpected content such as logical conflicts or semantic errors, the system may determine that the output is an anomaly. Such anomalies may indicate that the model has generated an unreasonable, misleading, or low-reliability response in the current context, or that the input has exceeded the model's output stability range. The system can then trigger an exception handling mechanism, such as logging the anomalous output, refusing to display it, outputting a warning message, or directing manual processing, thereby avoiding the transmission of erroneous information and the initiation of potential risks. This mechanism not only improves the controllability and response quality of the model's output, but also provides fundamental support for the model's traceability and credibility, playing a crucial role in ensuring the safe and stable operation of AI agents.

[0060] Preferably, the result range or pattern of the output data is comprehensively modeled through rule templates, vector space range, and multimodal matching logic to support unified verification of various types of output results such as structured text, natural language text, and images.

[0061] In an embodiment of the first aspect of the present application, the following steps are further included:

[0062] S10: When determining whether the real-time input data meets the feature range of the input data within the category set corresponding to the mapping identifier, if the input feature set of the real-time input data does not belong to any category set in the existing dynamic mapping table, a new category set is created for the real-time data and its corresponding real-time output data, a unique mapping identifier is assigned, and the category set and its corresponding mapping identifier are included in the dynamic mapping table.

[0063] It should be understood that during the matching and judgment process of real-time input data, if its input feature set cannot form an effective match with the input feature range of any existing category set in the current dynamic mapping table, it means that the input represents a new input pattern or scenario that the system has not yet covered. To prevent such new input from being misjudged as anomaly and to improve the system's adaptability to dynamically changing environments, the system can register the input and its corresponding output as a new data sample, automatically create a new category set, and assign a unique mapping identifier. The feature range and output pattern of the newly created category set can be automatically generated based on the initial sample and can be continuously improved through continuous data supplementation during subsequent operations. By incorporating this new category set and its mapping identifier into the dynamic mapping table, the system can achieve self-expansion and learning, with the ability to continuously absorb new scenarios and new semantic expressions, enhance its adaptability to input diversity and complexity, help the system maintain the coverage and effectiveness of the verification mechanism during long-term operation, and reflect the dynamic and evolvable nature of the method.

[0064] like Figure 3 As shown, the second aspect of the present application provides an AI agent real-time data verification system based on dynamic mapping, including: a dynamic mapping table establishment module, used to obtain a number of historical data of the AI ​​agent, wherein the historical data includes historical input data and its corresponding historical output data; feature extraction is performed on each historical input data to obtain its corresponding input feature set; a clustering algorithm is used to classify historical input data and its corresponding historical output data with similar input feature sets into a category set, and a unique mapping identifier is assigned to each category set; a dynamic mapping table is established according to all category sets and their corresponding mapping identifiers; the feature range of historical input data and the result range or pattern of historical output data in each category set are counted and analyzed; a real-time data analysis module is used to obtain the real-time input data of the AI ​​agent; the features of the real-time input data are extracted to obtain its corresponding input feature set; according to the input feature set corresponding to the features of the real-time input data, a matching mapping identifier is searched in the dynamic mapping table, and it is determined whether the real-time input data meets the feature range of the input data in the category set corresponding to the mapping identifier. If not, the real-time input data is determined to be abnormal.

[0065] It should be understood that the specific process of each module executing the above corresponding steps has been described in detail in the above method embodiment, and for the sake of brevity, it will not be repeated here.

[0066] It should also be understood that the division of modules in the embodiments of the present application is illustrative and is merely a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the present application may be integrated into a single processor, or may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0067] A third aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements any of the aforementioned methods when executed by a processor.

[0068] A fourth aspect of the present application provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer is enabled to implement any of the aforementioned methods.

[0069] like Figure 4 As shown, the fifth aspect of the present application provides an electronic terminal, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the aforementioned methods. The electronic terminal includes: at least one processor 401, a memory 402, at least one network interface 403, and a user interface 405. The various components in the device are coupled together via a bus system 404. It will be understood that the bus system 404 is used to implement connection and communication between these components. In addition to the data bus, the bus system 404 also includes a power bus, a control bus, and a status signal bus.

[0070] The user interface 405 may include a display, a keyboard, a mouse, a trackball, a click gun, keys, buttons, a touch pad or a touch screen.

[0071] It will be appreciated that the memory 402 may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM) or a programmable read-only memory (PROM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memory described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0072] The memory 402 in the embodiment of the present invention is used to store various categories of data to support the operation of the electronic terminal 400. Examples of such data include: any executable program for operating on the electronic terminal 400, such as an operating system 4021 and an application 4022; the operating system 4021 includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application 4022 can include various applications, such as a media player (Media Player), a browser (Browser), etc., for implementing various application services. The method provided by the embodiment of the present invention can be included in the application 4022.

[0073] The methods disclosed in the above embodiments of the present invention can be applied to or implemented by processor 401. Processor 401 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 401 or by software instructions. The above processor 401 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 401 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor 401 may be a microprocessor or any conventional processor. The steps of the accessory optimization method provided in conjunction with the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium located in a memory. The processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0074] In an exemplary embodiment, the electronic terminal 400 may be configured to execute the aforementioned method using one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs).

[0075] As used in this specification, the terms "component," "module," "system," and the like are used to refer to computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. By way of illustration, both an application running on a computing device and a computing device can be a component. One or more components can reside in a process and / or an execution thread, and a component can be located on one computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures stored thereon. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component across a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0076] Those skilled in the art will appreciate that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0077] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0078] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0079] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0080] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0081] In the above embodiments, the functions of each functional unit can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (program) are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. Available media may be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., high-density digital video discs (DVDs), or semiconductor media (e.g., solid state disks (SSDs)).

[0082] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other media that can store program codes.

[0083] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0084] In summary, the present invention effectively overcomes various shortcomings of the prior art and has high industrial utilization value.

[0085] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical concepts disclosed in this application shall be covered by the claims of this application.

Claims

1. A real-time data verification method for AI agents based on dynamic mapping, characterized in that: include: Acquire some historical data of the AI ​​agent, wherein the historical data includes historical input data and its corresponding historical output data; Perform feature extraction on each historical input data to obtain its corresponding input feature set; Use a clustering algorithm to group historical input data and their corresponding historical output data with similar input feature sets into a category set, and assign a unique mapping identifier to each category set; Establish a dynamic mapping table based on all category sets and their corresponding mapping identifiers; Count and analyze the characteristic range of historical input data and the result range or pattern of historical output data within each category set; Obtain real-time input data for AI agents; Extracting features of the real-time input data to obtain a corresponding input feature set; searching the dynamic mapping table for a matching mapping identifier based on the input feature set corresponding to the feature of the real-time input data, and determining whether the real-time input data meets the feature range of input data within the category set corresponding to the mapping identifier; if not, determining that the real-time input data is abnormal; Outputting real-time output data corresponding to the acquired real-time input data, determining whether the real-time output data conforms to a result range or pattern of output data within a category set corresponding to a matching mapping identifier in a dynamic mapping table; if not, determining that the real-time output data is abnormal; The input feature set includes text features, and a text similarity algorithm is used to match the text content in the real-time input data with the existing text feature set in the mapping table. The feature range includes keyword distribution, semantic vector space boundaries, and context semantic coverage.

2. The method for real-time data verification of an AI agent based on dynamic mapping according to claim 1, characterized in that: The result range or pattern of the output data is comprehensively modeled through rule templates, vector space ranges, and multimodal matching logic to support unified verification of various types of output results including structured text, natural language text, and images.

3. The method for real-time data verification of an AI agent based on dynamic mapping according to claim 1, characterized in that: When extracting features from historical input data and / or real-time input data respectively, feature enhancement is performed in combination with contextual information of the historical input data and / or real-time input data, and / or interaction history, and / or user portrait information.

4. A real-time data verification system for AI agents based on dynamic mapping, characterized in that: include: A dynamic mapping table establishment module is used to obtain a number of historical data of the AI ​​agent, wherein the historical data includes historical input data and its corresponding historical output data; feature extraction is performed on each historical input data to obtain its corresponding input feature set; a clustering algorithm is used to classify historical input data and its corresponding historical output data with similar input feature sets into a category set, and a unique mapping identifier is assigned to each category set; a dynamic mapping table is established based on all category sets and their corresponding mapping identifiers; statistics and analysis are performed on the feature range of the historical input data and the result range or pattern of the historical output data in each category set; the input feature set includes text features, and a text similarity algorithm is used to match and judge the text content in the real-time input data with the existing text feature set in the mapping table, and the feature range includes keyword distribution, semantic vector space boundary, and context semantic coverage. A real-time data analysis module is used to obtain real-time input data of an AI agent; extract features of the real-time input data to obtain its corresponding input feature set; query a matching mapping identifier in the dynamic mapping table according to the input feature set corresponding to the features of the real-time input data, and determine whether the real-time input data conforms to the feature range of the input data within the category set corresponding to the mapping identifier; if not, determine that the real-time input data is abnormal; output the real-time output data corresponding to the acquired real-time input data, and determine whether the real-time output data conforms to the result range or pattern of the output data within the category set corresponding to the matching mapping identifier in the dynamic mapping table; if not, determine that the real-time output data is abnormal.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.

6. A computer program product, characterized in that The computer program product includes computer program code, and when the computer program code is run on a computer, the computer is enabled to implement the method according to any one of claims 1 to 3.

7. An electronic terminal comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the method according to any one of claims 1 to 3.

Images