General application integration management method and system based on artificial intelligence model

Abstract

The application discloses a general application integration management method and system based on an artificial intelligence model, relates to the technical field of industrial intelligence integration, and comprises the following steps: collecting industrial data to be integrated, inputting the industrial data into an artificial intelligence model for application semantic analysis, and generating an application semantic element set; performing semantic screening and coding identification on the application semantic element set, and generating semantic identification; establishing corresponding integration relationships by using the semantic identification, and generating a general application integration resource library; acquiring real-time industrial integration requirements, performing requirement semantic extraction, and generating target semantics; performing candidate capability chain searching and parameter semantic connection judgment on the target semantics by using the general application integration resource library, and generating an executable integration path set; performing interface calling and sequential connection integration on application capability items, and generating a general application integration management result. The application identifies a capability group with execution possibility through candidate capability chain searching, and improves the accuracy, connection integrity and actual execution stability of an executable path screening result.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent integration technology, and in particular to a general application integration management method and system based on artificial intelligence models. Background Technology

[0002] As industrial software, business applications, and interface technologies continue to evolve, the integration methods for heterogeneous applications have gradually shifted from static interface docking to semantic and dynamic collaboration. Especially in industrial data processing scenarios, application capability descriptions, interface parameter expressions, runtime context awareness, and call relationship organization are beginning to be combined with artificial intelligence models to support unified access, semantic recognition, and integrated management of multi-source application capabilities. This has achieved application function collaboration and data transfer to a certain extent, but the overall efficiency and flexibility are limited.

[0003] However, in existing technologies, there is a lack of a unified semantic association mechanism for capability descriptions, interface input and output expressions, and constraint information between different general applications. This results in incomplete paths, mismatched parameters, and untraceable execution states of general applications, thereby limiting the efficiency, accuracy, and adaptability stability of general application integration. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a general application integration management method based on an artificial intelligence model to solve the problem of incomplete paths in general applications.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a general application integration management method based on an artificial intelligence model, which includes collecting industrial data to be integrated for general applications, inputting it into an artificial intelligence model for application semantic parsing, and generating an application semantic element set, wherein the application semantic element set includes application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints.

[0008] Semantic filtering and encoding of the application semantic element set are performed to generate semantic identifiers;

[0009] By using semantic identifiers, a correspondence and integration relationship is established between multiple application capability items and interface input parameter attributes and interface output parameter attributes, thereby generating a general application integration resource library;

[0010] Acquire real-time industrial integration requirements and use artificial intelligence models to extract requirement semantics, generating target semantics;

[0011] The target semantics are searched for candidate capability chains and the connection of parameter semantics are determined by using a general application integration resource library to generate an executable integration path set.

[0012] The executable integration path set is used to perform interface calls and sequential integration of application capability items, generating a general application integration management result.

[0013] As a preferred embodiment of the general application integration management method based on artificial intelligence model described in this invention, the industrial data to be integrated for the general application includes application capability data, interface status data, historical call data, and runtime context data.

[0014] As a preferred embodiment of the general application integration management method based on artificial intelligence model described in this invention, the application semantic parsing is a process of obtaining an application semantic element set by performing capability decomposition, interface behavior recognition, parameter semantic extraction, context dependency recognition, and call element extraction on the industrial data to be integrated.

[0015] As a preferred embodiment of the general application integration management method based on artificial intelligence models described in this invention, the specific steps for generating semantic identifiers are as follows:

[0016] The application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints required to establish the corresponding integration relationship based on the application semantic element set are used as the filtered semantic elements.

[0017] Semantic elements are written with type tags, length tags, and value tags respectively, and arranged in TLV tag order to form a sequence of encoded semantic elements;

[0018] The correspondence between the encoded combinations in the sequence of encoded semantic elements is verified to generate semantic identifiers.

[0019] As a preferred embodiment of the general application integration management method based on an artificial intelligence model described in this invention, the specific steps for generating the general application integration resource library are as follows:

[0020] Write semantic identifiers into application capability items, interface input parameter attributes, and interface output parameter attributes, respectively;

[0021] Based on the semantic identification correspondence, the interface input parameter attributes and interface output parameter attributes corresponding to each application capability item are targeted and associated to form a corresponding integration relationship;

[0022] After aggregating the corresponding integration relationships, a general application integration resource library is generated.

[0023] As a preferred embodiment of the general application integration management method based on artificial intelligence model described in this invention, the real-time industrial integration requirements are the application integration goals to be achieved in the general application, the target input conditions, the target output conditions, and the target constraints.

[0024] As a preferred embodiment of the general application integration management method based on artificial intelligence models described in this invention, the specific steps for generating target semantics are as follows:

[0025] Extract initial requirement information from real-time industrial integration requirements;

[0026] The initial demand information is input into the artificial intelligence model for semantic segmentation, and functional segments, input condition segments, output condition segments and constraint condition segments are identified as preliminary semantics;

[0027] Semantic recognition is performed on the preliminary semantics to generate target functional semantics, target input semantics, target output semantics, and target constraint semantics, which are then aggregated into target semantics.

[0028] As a preferred embodiment of the general application integration management method based on artificial intelligence models described in this invention, the specific steps for generating the executable integration path set are as follows:

[0029] Utilize target semantic recognition to identify candidate application capability items in the general application integration resource library;

[0030] By using context constraints and invocation constraints, it is determined whether the output parameter attributes of the candidate application capability item match the input parameter attributes of the previous candidate application capability item, and the successfully matched candidate application capability item is connected with the previous candidate application capability item to form a candidate capability chain.

[0031] Semantic alignment analysis is performed on the candidate capability chains to determine whether the types, attributes, context constraints, and calling constraints of the input and output parameters meet the parameter semantic connection conditions. After marking whether the connection is successful or not, the results are integrated into the parameter semantic connection result.

[0032] The candidate capability chains are filtered by the parameter semantic connection results to obtain the executable integration path set.

[0033] As a preferred embodiment of the general application integration management method based on an artificial intelligence model described in this invention, the specific steps for generating the general application integration management result are as follows:

[0034] Based on the order of capability items in the executable integration path set, the target semantics of candidate application capability items are selected sequentially.

[0035] Parameter binding is performed on the target input conditions of the target semantics to obtain the input parameter attributes required for the call;

[0036] Make an API call on the current application capability item to obtain the output parameter attributes and execution status of the application capability item;

[0037] The output parameter attributes and execution status are matched with the input conditions of the next application capability item to determine whether the parameter semantic connection conditions are met, and the parameter semantic connection result of the current application capability item and the input parameter attributes required for the next application capability item to be called are obtained.

[0038] The system processes all application capability items in the executable integration path set sequentially. After integrating and normalizing the input parameter attributes, output parameter attributes, and execution status required for calling all application capability items, it generates a general application integration management result.

[0039] Secondly, this invention provides a general application integration management system based on an artificial intelligence model, including:

[0040] The acquisition module is used to collect industrial data to be integrated for general applications and input it into an artificial intelligence model for application semantic parsing to generate an application semantic element set. The application semantic element set includes application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints.

[0041] The encoding module is used to perform semantic filtering and encoding on the application semantic element set, and generate semantic identifiers.

[0042] The resource module is used to establish the corresponding integration relationship between multiple application capability items and interface input parameter attributes and interface output parameter attributes using semantic identifiers, and to generate a general application integration resource library.

[0043] The extraction module is used to acquire real-time industrial integration requirements and use artificial intelligence models to extract the semantics of the requirements, generating target semantics.

[0044] The filtering module is used to perform candidate capability chain search and parameter semantic connection determination on the target semantics using a general application integration resource library, and generate an executable integration path set.

[0045] The integration module is used to perform interface calls and sequential integration of application capability items using executable integration path sets, and generate general application integration management results.

[0046] The beneficial effects of this invention are as follows: by identifying capability groups with execution potential through candidate capability chain search, invalid splicing and parameter breakage in the subsequent interface call stage are avoided. At the same time, candidate capability chain search can simultaneously confirm the input-output connection relationship and call constraint relationship in industrial data processing, ensuring the consistency, feasibility and reusability of each application capability item, thereby improving the accuracy, connection integrity and actual execution stability of the executable path screening results. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a general application integration management method based on an artificial intelligence model.

[0049] Figure 2 This is a schematic diagram of a general application integration management system based on an artificial intelligence model.

[0050] Figure 3 A flowchart for generating semantic identifiers.

[0051] Figure 4 A flowchart for generating a universal application integration resource library. Detailed Implementation

[0052] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0053] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0054] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0055] Reference Figures 1-4 This is one embodiment of the present invention, which provides a general application integration management method based on an artificial intelligence model, including the following steps:

[0056] S1. Collect the industrial data to be integrated for general applications and input it into the artificial intelligence model for application semantic parsing to generate an application semantic element set. The application semantic element set includes application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints.

[0057] S1.1 The industrial data to be integrated for general applications includes application capability data, interface status data, historical call data, and runtime context data, specifically:

[0058] By utilizing the application access connector, interface resolver, log collection component, and runtime context collection component, application capability data, interface status data, historical call data, and runtime context data are extracted from the application registration source, interface definition source, call log source, and runtime status source of the general application, respectively.

[0059] It should be noted that the application capability data includes the executable actions, input parameter requirements, output result types, and function triggering conditions of the functional operations in general applications; the interface status data includes the current availability of the interface, response time, interface call error information, and interface version information; the historical call data includes the call records of each application capability data in the past time, parameter usage, call frequency, and call success rate; and the runtime context data includes the execution environment conditions, related resource status, parallel call status, and time constraints.

[0060] S1.2 Application semantic parsing is the process of using artificial intelligence models to perform capability decomposition, interface behavior recognition, parameter semantic extraction, context dependency recognition, and call element extraction on the industrial data to be integrated, resulting in an application semantic element set. Specifically,

[0061] Artificial intelligence models are structured around multi-layer neural networks, specifically including an input layer, a context encoding layer, a multi-application branch layer, and an output layer.

[0062] The input layer receives the industrial data to be integrated and converts it into word vector sequences and position vector sequences. The context encoding layer performs context encoding on the word vector sequences and position vector sequences to generate semantic encoding sequences, which are used to capture the relationships between application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints. The multi-application branch layer marks the start and end positions of actions on the semantic encoding sequences, extracts the application capability items corresponding to each action from the application capability data, identifies the interface input parameter attributes and interface output parameter attributes within the application capability items, and determines the relationship between the interface input parameter attributes and interface output parameter attributes and the corresponding application capability items. At the same time, it generates call constraints and context constraints by combining historical call data and context information. The output layer integrates the application capability items, interface input parameter attributes, interface input parameter attributes, context constraints, and call constraints to output a complete set of application semantic elements.

[0063] The training process for an artificial intelligence model is as follows:

[0064] Collect historical application capability data, interface status data, call data, and runtime context data, and use manually annotated application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints as supervision signals to construct a training sample set.

[0065] The training sample set is derived from the correspondence between historical application registration text, interface definition text, call records, and manual verification records.

[0066] During training, the training sample set is input into the artificial intelligence model. After the input layer generates vector representations, the shared semantic features are formed through the context encoding layer and then passed to the multi-application branch layer. Each branch is responsible for predicting application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints.

[0067] An application capability loss term is constructed based on the difference between the predicted and labeled results of the application capability terms. An interface input parameter loss term is constructed based on the difference between the predicted and labeled results of the interface input parameter attributes. An interface output parameter loss term is constructed based on the difference between the predicted and labeled results of the interface output parameter attributes. A context constraint loss term is constructed based on the difference between the predicted and labeled results of the context constraint. A call constraint loss term is constructed based on the difference between the predicted and labeled results of the call constraint.

[0068] The loss terms of each application are combined according to preset loss weights to form a joint loss function for multiple applications. The weight values ​​can be set based on the proportion of each application in the training sample set. For example: application capability loss weight 0.3, input parameter loss weight 0.2, output parameter loss weight 0.2, context constraint loss weight 0.15, and call constraint loss weight 0.15.

[0069] The parameters of the artificial intelligence model are iteratively updated based on the multi-application joint loss function. After the update, the parameters of the input layer, context encoding layer, multi-application branch layer and output layer are adjusted synchronously using a gradient backpropagation method.

[0070] During continuous training, training stops when the change in the joint loss function across multiple applications falls below a preset convergence threshold. The convergence threshold is set to 10% of the initial change, meaning the joint loss decreases to a fixed percentage of the initial change, resulting in the trained AI model.

[0071] It should be noted that the interface input parameter attributes and interface output parameter attributes include the input parameter name, input parameter type, input parameter meaning, output parameter name, output parameter type, and output parameter meaning. The call constraints are the order, dependency, successful call conditions, and failure restrictions of the interface input parameter attributes and interface output parameter attributes in the actual call. The context constraints are the parallel call relationship, resource usage, and time constraints of each application capability item.

[0072] S2. Perform semantic filtering and encoding on the application semantic element set to generate semantic identifiers.

[0073] S2.1. Based on the application semantic element set, the application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints required to establish the corresponding integration relationship are used as the filtered semantic elements, specifically:

[0074] From the set of application semantic elements, the application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints are filtered to identify the semantic elements that participate in the integration relationship, and the call relationship established between the input parameter attributes and output parameter attributes corresponding to each application capability item is clarified.

[0075] At the same time, the context constraints and call constraints of the call relationship are marked to form the filtered semantic elements.

[0076] It should be noted that the selection criteria are as follows:

[0077] Traverse each application capability item in the application semantic element set and extract the corresponding interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints.

[0078] The application capability items are sequentially checked for logical correspondence, interface parameter connection, constraint integrity, and annotation call. Only the application capability items that pass all of the above checks are considered as the filtered semantic elements.

[0079] In this embodiment, the current application capability item is used as an example for filtering.

[0080] This includes checking whether the input parameter attributes of the current application capability item are complete, and determining whether the output parameter attributes are consistent with the action logic of the current application capability item.

[0081] If the action logic of the current application capability item is inconsistent, then the application capability item is excluded.

[0082] If the input parameter attributes of the current application capability item are complete and the action logic is consistent, then the current application capability item is retained and the next step of interface parameter connection check is performed.

[0083] Check whether the output parameter attributes of the current application capability item match the interface input parameter attributes of any of the current and subsequent application capability items;

[0084] If a match cannot be found, it will be marked as "temporarily unconnectable" and reserved for subsequent verification.

[0085] If a match is successful, the current application capability item is retained and the constraint integrity is checked.

[0086] Check whether the context constraints and invocation constraints of the current application capability items exist and whether the fields are complete;

[0087] If any constraints are missing, the current application capability item will be excluded.

[0088] If all constraints are complete, the current application capability item is used as a semantic element participating in the integration relationship.

[0089] S2.2. Using TLV tag encoding technology, semantic elements are written into type tags, length tags, and value tags respectively, and arranged in the order of TLV tags to form a sequence of encoded semantic elements. Specifically,

[0090] TLV tag encoding technology is used to assign corresponding type tags to semantic elements. For example, application capability items are marked as "01", interface input parameter attributes are marked as "02", interface output parameter attributes are marked as "03", context constraints are marked as "04", and call constraints are marked as "05".

[0091] It should be noted that TLV tag encoding technology is a "type-length-value" encoding technology, where type represents the category of semantic element, such as application capability items or interface input parameter attributes; length represents the length of the value content; and value represents the actual semantic element content.

[0092] The length marker is determined based on the specific content of the semantic elements, such as by counting the byte length of the value content and writing the length information into the length marker.

[0093] Write the actual element content of the semantic element into the value tag.

[0094] After all semantic elements have completed the TLV tag writing, the TLV tag combination of each semantic element is arranged in the logical order of its appearance in the filtered semantic elements to form a sequence of encoded semantic elements.

[0095] It should be noted that the TLV tag order refers to arranging the type tag, length tag, and value tag of each semantic element in the logical order in which they appear in the integration relationship.

[0096] S2.3, Perform correspondence verification on the encoded combinations in the encoded semantic element sequence and generate semantic identifiers, specifically,

[0097] Check sequentially whether the type tag and length tag of each TLV tag combination in the encoded semantic element sequence match, and confirm whether the content of the value tag is complete and meets the requirements of the corresponding type tag.

[0098] Check whether there is a logical correspondence between the input parameter attributes and output parameter attributes of each application capability item in the sequence of encoded semantic elements. Ensure that the input parameter attributes and output parameter attributes of each application capability item form a logical correspondence in the sequence, that is, the input parameter attributes are the input parameter attributes required for the execution of the application capability item, and the output parameter attributes are the output parameter attributes generated after the execution of the application capability item.

[0099] At the same time, verify whether the context constraints and invocation constraints are correctly connected to ensure that the context constraints and invocation constraints can correctly guide the execution of application capabilities.

[0100] A consistency check is performed on the combination of tags for all semantic elements in the sequence, and element records that do not conform to the correspondence are found to be abnormal and excluded.

[0101] After verifying all the encoding combinations, the correct correspondences of each semantic element in the sequence are organized and collected to obtain the semantic identifier.

[0102] It should be noted that the logical correspondence refers to whether the input parameter attributes and output parameter attributes belong to the same application capability item and conform to the actual calling and processing logic of the same application capability item. That is, whether the input parameter attributes are the parameters that the application capability item actually needs to receive when it is executed, whether the output parameter attributes are the parameters that the application capability item can actually generate after it is executed, and whether the input-to-output process conforms to the processing logic and calling constraints of the application capability item itself.

[0103] For example, if the application capability is "temperature prediction", the input parameter attribute is "historical temperature data", and the output parameter attribute is "predicted temperature result", then there is a logical correspondence between the input and the output.

[0104] If the input parameter attribute is "device number" and the output parameter attribute is "image recognition result", but the application capability item itself is not "image recognition", then there is no logical correspondence.

[0105] S3. Use semantic identifiers to establish the corresponding integration relationship between multiple application capability items and interface input parameter attributes and interface output parameter attributes, and generate a general application integration resource library.

[0106] S3.1. Write semantic identifiers into application capability items, interface input parameter attributes, and interface output parameter attributes respectively; according to the correspondence of semantic identifiers, assign the corresponding interface input parameter attributes and interface output parameter attributes of each application capability item to form corresponding integration relationships; after aggregating the corresponding integration relationships, a general application integration resource library is generated, specifically as follows:

[0107] Semantic identifiers are written into the application capability items, interface input parameter attributes, and interface output parameter attributes, respectively.

[0108] For each application capability item identifier, the input parameter attributes and output parameter attributes are matched, and the semantic identifier corresponding to each application capability item is compared one by one with the semantic identifier corresponding to the interface input parameter attribute and the semantic identifier corresponding to the interface output parameter attribute.

[0109] When semantic identifiers are inconsistent, no directional association is established between the current application capability item and the interface input parameter attribute and the interface output parameter attribute. Instead, the next set of semantic identifiers is compared until the semantic identifiers are consistent.

[0110] When the semantic identifiers are consistent, the interface input parameter attributes are associated with the input positions of the corresponding application capability items, and the interface output parameter attributes are associated with the output positions of the corresponding application capability items. Then, the context constraints and call constraints are written into the corresponding associated positions to form a corresponding integration relationship between the application capability items and the interface input parameter attributes and interface output parameter attributes.

[0111] It should be noted that the correspondence and integration relationship between application capability items and interface input parameter attributes and interface output parameter attributes refers to the interface input parameter attributes corresponding to each application capability item, the interface output parameter attributes output by the application capability item, and the input-output connection relationship formed between the interface input parameter attributes and interface output parameter attributes and the application capability item during the calling process.

[0112] The correspondence between interface input attributes, interface output attributes, and calling constraints in semantic identifiers refers to the verified association between interface input parameter attributes, interface output parameter attributes, and calling constraints under the same semantic identifier.

[0113] The correspondence is specifically manifested as follows: the interface input parameter attribute represents the parameters that need to be received when the application capability item is called; the interface output parameter attribute represents the parameters that can be generated after the application capability item is called; and the call constraint represents the correspondence between the interface input parameter attribute and the interface output parameter attribute in terms of call order, call conditions, and call restrictions.

[0114] S3.2. After aggregating the corresponding integration relationships, the generated general application integration resource library is as follows:

[0115] The established integration relationships are categorized and organized according to the order of application capability items to obtain the interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints gathered under the same application capability item.

[0116] Duplicate integration relationships are merged, and integration relationships for different application capability items are grouped and stored.

[0117] The corresponding integration relationships after grouping are unified into a general application integration resource library.

[0118] S4. Obtain real-time industrial integration requirements and use artificial intelligence models to extract requirement semantics and generate target semantics.

[0119] S4.1 Real-time industrial integration requirements are the application integration goals, target input conditions, target output conditions, and target constraints to be achieved in general applications, specifically:

[0120] The API request receiving technology and message subscription receiving technology are used to collect the application integration goals, target input conditions, target output conditions and target constraints that the general application is currently waiting to complete.

[0121] If input parameter attributes, output parameter attributes, and constraints from unknown sources appear, a one-to-one correspondence is made between the current runtime context data of the general application and the existing corresponding integration relationships in the general application integration resource library to obtain the required application integration target, target input conditions, target output conditions, and target constraints.

[0122] Furthermore, if input parameter attributes, output parameter attributes, and constraints of unknown origin are found, these attributes will be compared item by item with the current runtime context data of the general application and the existing corresponding integration relationships in the general application integration resource library.

[0123] When the comparison results are consistent, the corresponding input parameter attributes, output parameter attributes, and constraints will be added to the application integration target, target input conditions, target output conditions, and target constraints.

[0124] When the comparison results are inconsistent, input parameter attributes, output parameter attributes, and constraints with unclear sources are left as unmatched and the comparison is repeated.

[0125] All application integration objectives, target input conditions, target output conditions, and target constraints are uniformly organized to form a complete real-time industrial integration requirement.

[0126] Among them, the application integration target represents the functional items that need to be completed in this integration, the target input conditions represent the input parameter attributes, input sources and input states that need to be available before the application integration target is executed, the target output conditions represent the output parameter attributes, output format and output destination obtained after the application integration target is completed, and the target constraint conditions represent the context constraints and calling constraints that need to be satisfied when the application integration target is executed.

[0127] S4.2 Extract initial demand information from real-time industrial integration requirements; input the initial demand information into an artificial intelligence model for semantic segmentation, identifying functional segments, input condition segments, output condition segments, and constraint condition segments as preliminary semantics, specifically,

[0128] The application integration objectives, target input conditions, target output conditions, and target constraints are read from real-time industrial integration requirements to form initial requirement information.

[0129] The initial requirements information is organized into a series of statements that retain the function name, input parameter name, output parameter name, constraint name, and constraint value, arranged in the order of the fields.

[0130] When an artificial intelligence model performs segmentation recognition on continuous statements, it takes the action expression corresponding to the function name as the function segment, the statement corresponding to the input parameter name, input source, and input state as the input condition segment, the statement corresponding to the output parameter name, output form, and output destination as the output condition segment, and the statement corresponding to the context constraint and calling constraint as the constraint condition segment.

[0131] After the artificial intelligence model completes segment recognition, it organizes the functional segments, input condition segments, output condition segments, and constraint condition segments into preliminary semantics.

[0132] S4.3 Perform semantic recognition on the preliminary semantics to generate target functional semantics, target input semantics, target output semantics, and target constraint semantics, which are then aggregated into target semantics. Specifically,

[0133] The initial semantics are vectorized according to the function name, input parameter name and source, output parameter name and destination, and constraint name and value, while preserving field information and paragraph order.

[0134] Through the multi-application branch layer of the artificial intelligence model, semantic features are extracted from the encoded preliminary semantics, and the extracted paragraphs are classified and sequence labeled. Functional segments are mapped to target functional semantics, input condition segments are mapped to target input semantics, output condition segments are mapped to target output semantics, and constraint condition segments are mapped to target constraint semantics.

[0135] The target function semantics, target input semantics, target output semantics, and target constraint semantics are organized according to the same real-time industrial integration requirements to form the target semantics.

[0136] It should be noted that the specific process of training an artificial intelligence model is as follows:

[0137] Historical industrial integration requirements and their manually annotated results are collected. The function name, input parameters, output parameters, and constraints of each requirement are labeled as function segments, input condition segments, output condition segments, and constraint condition segments, forming requirement training samples. The sources of these training samples include historical requirement records, call logs, and manually verified requirement documents.

[0138] During training, the required training samples are input into the artificial intelligence model, and shared features are formed through the input layer and the context encoding layer. These features are then passed to the multi-application branch layer, and each branch outputs the functional segment prediction result, the input condition segment prediction result, the output condition segment prediction result, and the constraint condition segment prediction result, respectively.

[0139] Functional segment loss terms are constructed based on the differences between functional segment prediction results and functional segment annotation results; input condition loss terms are constructed based on the differences between input condition segment prediction results and input condition segment annotation results; output condition loss terms are constructed based on the differences between output condition segment prediction results and output condition segment annotation results; and constraint condition loss terms are constructed based on the differences between constraint condition segment prediction results and constraint condition segment annotation results.

[0140] The functional segment loss term, input condition loss term, output condition loss term, and constraint condition loss term are combined according to preset weights to form a multi-application joint loss function. The weight values ​​can be set based on the proportion of each segment in the training sample set, for example, functional segment loss weight 0.35, input condition loss weight 0.25, output condition loss weight 0.25, and constraint condition loss weight 0.15.

[0141] The parameters in the multi-application learning model are iteratively updated based on the multi-application joint loss function. The parameters of the input embedding layer, context encoding layer and each paragraph branch layer are adjusted synchronously using a gradient backpropagation-based approach.

[0142] During continuous training, training stops when the change in the joint loss function between adjacent training rounds is lower than a preset convergence threshold. The convergence threshold can be 10% of the initial change, that is, when the joint loss drops to a fixed proportion of the initial change, the artificial intelligence model is trained.

[0143] S5. Utilize the general application integration resource library to perform candidate capability chain search and parameter semantic connection determination on the target semantics, and generate an executable integration path set.

[0144] S5.1 Utilize target semantic recognition to identify candidate application capability items in the general application integration resource library, specifically,

[0145] The target function semantics are compared item by item with the application capability items in the general application integration resource library, and application capability items with consistent functional meanings are selected.

[0146] The target input semantics and target output semantics are compared with the interface input parameter attributes and interface output parameter attributes corresponding to the filtered application capability items, and application capability items with consistent input and output are retained.

[0147] At the same time, verify the semantics of the target constraints with the context constraints and the calling constraints, and retain the application capability items that are consistent with the constraints.

[0148] Application capability items that simultaneously satisfy the following criteria—consistency in functional meaning, input / output, and constraints—will be considered as candidate application capability items.

[0149] S5.2. Using context constraints and invocation constraints, determine whether the output parameter attributes of a candidate application capability item match the input parameter attributes of the previous candidate application capability item, and connect the successfully matched candidate application capability items with the previous candidate application capability item to form a candidate capability chain. Specifically,

[0150] Select two adjacent application capability items from the candidate application capability items, and check whether the interface output parameter attribute name, parameter type, and parameter meaning of the former candidate application capability item are consistent with the interface input parameter attribute name, parameter type, and parameter meaning of the latter candidate application capability item.

[0151] When the interface output parameter attributes are consistent with the interface input parameter attributes, check whether the context constraints and call constraints corresponding to the two application capability items can be continuously satisfied.

[0152] If the interface output parameter attributes, interface input parameter attributes, context constraints, and call constraints are all consistent, the two candidate application capability items will be connected in sequence to form a candidate capability chain.

[0153] If there is any inconsistency between the interface input / output parameter attributes and the context constraints or calling constraints, a candidate capability chain will not be formed.

[0154] S5.3 Perform semantic alignment analysis on the candidate capability chains, determine whether the types, attributes, context constraints, and calling constraints of the input and output parameters meet the parameter semantic connection conditions, and mark whether the connection is successful or not, then integrate them into the parameter semantic connection result, specifically,

[0155] Following the connection order of the candidate capability chain, the interface output parameter attributes of the previous candidate application capability item are compared with the interface input parameter attributes of the next candidate application capability item one by one.

[0156] When the interface output parameter attributes and interface input parameter attributes are consistent in parameter type, parameter meaning, and parameter destination, and the context constraints and call constraints between adjacent application capability items do not conflict, the connection is marked as successful.

[0157] When the interface output parameter attributes and interface input parameter attributes are inconsistent in parameter type, parameter meaning, parameter destination, or when there are conflicts in context constraints and call constraints between adjacent application capability items, it is marked as a connection failure.

[0158] The markers of successful and unsuccessful connections in the candidate capability chain are uniformly organized to form the parameter semantic connection result.

[0159] S5.4. The candidate capability chains are filtered based on the parameter semantic connection results to obtain the executable integration path set, specifically as follows:

[0160] Check the parameter semantic connection results corresponding to each candidate capability chain; when all adjacent connections in a candidate capability chain are marked as successful, the candidate capability chain is retained as an executable integration path; when any connection failure is marked in a candidate capability chain, the candidate capability chain is removed; the retained executable integration paths are organized to form an executable integration path set.

[0161] S6. Use the executable integration path set to perform interface calls and sequential integration of application capability items to generate a general application integration management result.

[0162] S6.1. Based on the order of capability items in the executable integration path set, the target semantics of candidate application capability items are selected sequentially, specifically as follows:

[0163] Following the order of the executable integration paths, determine the current application capability item one by one. Read the target functional semantics, target input semantics, target output semantics, and target constraint semantics corresponding to the current application capability item, and use them as the target semantics of the current application capability item.

[0164] Proceed to the next application capability item in sequence until all application capability items in the executable integration path set have been selected.

[0165] S6.2 Parameter binding is performed on the target input conditions of the target semantics to obtain the input parameter attributes required for the call. Specifically,

[0166] The target input semantics corresponding to the current application capability item are mapped to the interface input parameter attributes corresponding to the current application capability item item one by one.

[0167] When the parameter name, parameter type, and parameter meaning of the interface input port are all consistent, the input parameter in the target input semantics is written into the corresponding interface input parameter attribute position to form the input parameter attribute required for the call.

[0168] If any of the parameter names, parameter types, or parameter meanings of the interface input port are inconsistent, the current application capability item will not be entered into the interface call.

[0169] S6.3. Make an interface call to the current application capability item to obtain the output parameter attributes and execution status of the application capability item. Specifically,

[0170] Write the required input parameter attributes to the interface request corresponding to the current application capability item; after sending the interface request, read the interface return result; identify the output parameter name, output parameter type and output parameter value from the interface return result and use them as the output parameter attributes of the application capability item; at the same time, use the success flag, failure flag and error message in the interface return result as the execution status.

[0171] S6.4 Match the output parameter attributes and execution status with the input conditions of the next application capability item to determine whether the parameter semantic connection conditions are met. Obtain the parameter semantic connection result of the current application capability item and the input parameter attributes required for the next application capability item call. Specifically,

[0172] The output parameter attributes of the current application capability item are compared item by item with the target input semantics and interface input parameter attributes of the next application capability item.

[0173] If the execution status of the current application capability item is successful, and the output parameter name, output parameter type, and output parameter meaning are consistent with the input parameter name, input parameter type, and input parameter meaning of the next application capability item, then the parameter semantic connection result of the current application capability item is marked as successful, and the corresponding output parameter attribute is written into the input parameter attribute required for the next application capability item to call.

[0174] If the execution status of the current application capability item is failed, or if there is any inconsistency between the output parameter attribute and the input condition of the next application capability item, then the parameter semantic connection result of the current application capability item is marked as connection failure, and the continued invocation of the current executable integration path is stopped.

[0175] S6.5. Process all application capability items in the executable integration path set sequentially. After integrating and normalizing the required input parameter attributes, output parameter attributes, and execution status of all application capability items, generate a general application integration management result. Specifically,

[0176] After completing the interface calls for all application capability items according to the connection order of the executable integration path set, the input parameter attributes, output parameter attributes, and execution status required for each application capability item call are uniformly organized.

[0177] Input and output parameter attributes with different names but the same meaning are uniformly named, and execution status is uniformly marked as success or failure.

[0178] The unified and organized input parameter attributes, output parameter attributes, and execution status are summarized according to the same executable integration path to form a general application integration management result.

[0179] This embodiment also provides a general application integration management system based on an artificial intelligence model, including:

[0180] The data acquisition module is used to collect industrial data to be integrated for general applications and input it into the artificial intelligence model for application semantic parsing to generate an application semantic element set. The application semantic element set includes application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints.

[0181] The encoding module is used to perform semantic filtering and encoding on the application semantic element set, and generate semantic identifiers.

[0182] The resource module is used to establish the corresponding integration relationship between multiple application capability items and interface input parameter attributes and interface output parameter attributes using semantic identifiers, and to generate a general application integration resource library.

[0183] The extraction module is used to acquire real-time industrial integration requirements and use artificial intelligence models to extract the semantics of the requirements, generating target semantics.

[0184] The filtering module is used to perform candidate capability chain search and parameter semantic connection determination on the target semantics using a general application integration resource library, and generate an executable integration path set.

[0185] The integration module is used to perform interface calls and sequential integration of application capability items using executable integration path sets, and generate general application integration management results.

[0186] This embodiment also provides a computer device applicable to the general application integration management method based on an artificial intelligence model, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the general application integration management method based on an artificial intelligence model as proposed in the above embodiment.

[0187] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0188] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the general application integration management method based on an artificial intelligence model as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0189] In summary, this invention identifies capability groups with execution potential through candidate capability chain search, avoiding invalid splicing and parameter breakage in subsequent interface call stages. At the same time, candidate capability chain search can simultaneously confirm the input-output connection relationship and call constraint relationship in industrial data processing, ensuring the consistency, feasibility, and reusability of each application capability item, thereby improving the accuracy, connection integrity, and actual execution stability of the executable path screening results.

[0190] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A general application integration management method based on an artificial intelligence model, characterized in that, include, Collect industrial data to be integrated for general applications and input them into an artificial intelligence model for application semantic parsing to generate an application semantic element set. The application semantic element set includes application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints. Semantic filtering and encoding of the application semantic element set are performed to generate semantic identifiers; By using semantic identifiers, a correspondence and integration relationship is established between multiple application capability items and interface input parameter attributes and interface output parameter attributes, thereby generating a general application integration resource library; Acquire real-time industrial integration requirements and use artificial intelligence models to extract requirement semantics, generating target semantics; The target semantics are searched for candidate capability chains and the connection of parameter semantics are determined by using a general application integration resource library to generate an executable integration path set. The executable integration path set is used to perform interface calls and sequential integration of application capability items, generating a general application integration management result.

2. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The industrial data to be integrated for the general application includes application capability data, interface status data, historical call data, and runtime context data.

3. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The application semantic parsing is a process of obtaining an application semantic element set by performing capability decomposition, interface behavior recognition, parameter semantic extraction, context dependency recognition, and call element extraction on the industrial data to be integrated.

4. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The specific steps for generating semantic identifiers are as follows: The application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and call constraints required to establish the corresponding integration relationship based on the application semantic element set are used as the filtered semantic elements. Semantic elements are written with type tags, length tags, and value tags respectively, and arranged in TLV tag order to form a sequence of encoded semantic elements; The correspondence between the encoded combinations in the sequence of encoded semantic elements is verified to generate semantic identifiers.

5. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The specific steps for generating the universal application integration resource library are as follows: Write semantic identifiers into application capability items, interface input parameter attributes, and interface output parameter attributes, respectively; Based on the semantic identification correspondence, the interface input parameter attributes and interface output parameter attributes corresponding to each application capability item are targeted and associated to form a corresponding integration relationship; After aggregating the corresponding integration relationships, a general application integration resource library is generated.

6. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The real-time industrial integration requirements are the application integration goals, target input conditions, target output conditions, and target constraints to be achieved in general applications.

7. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The specific steps for generating the target semantics are as follows: Extract initial requirement information from real-time industrial integration requirements; The initial demand information is input into the artificial intelligence model for semantic segmentation, and functional segments, input condition segments, output condition segments and constraint condition segments are identified as preliminary semantics; Semantic recognition is performed on the preliminary semantics to generate target functional semantics, target input semantics, target output semantics, and target constraint semantics, which are then aggregated into target semantics.

8. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The specific steps for generating the executable integration path set are as follows: Utilize target semantic recognition to identify candidate application capability items in the general application integration resource library; By using context constraints and invocation constraints, it is determined whether the output parameter attributes of the candidate application capability item match the input parameter attributes of the previous candidate application capability item, and the successfully matched candidate application capability item is connected with the previous candidate application capability item to form a candidate capability chain. Semantic alignment analysis is performed on the candidate capability chains to determine whether the types, attributes, context constraints, and calling constraints of the input and output parameters meet the parameter semantic connection conditions. After marking whether the connection is successful or not, the results are integrated into the parameter semantic connection result. The candidate capability chains are filtered by the parameter semantic connection results to obtain the executable integration path set.

9. The general application integration management method based on an artificial intelligence model as described in claim 1, characterized in that: The specific steps for generating the general application integration management result are as follows: Based on the order of capability items in the executable integration path set, the target semantics of candidate application capability items are selected sequentially. Parameter binding is performed on the target input conditions of the target semantics to obtain the input parameter attributes required for the call; Make an API call on the current application capability item to obtain the output parameter attributes and execution status of the application capability item; The output parameter attributes and execution status are matched with the input conditions of the next application capability item to determine whether the parameter semantic connection conditions are met, and the parameter semantic connection result of the current application capability item and the input parameter attributes required for the next application capability item to be called are obtained. The system processes all application capability items in the executable integration path set sequentially. After integrating and normalizing the input parameter attributes, output parameter attributes, and execution status required for calling all application capability items, it generates a general application integration management result.

10. A general application integration management system based on an artificial intelligence model, based on the general application integration management method based on an artificial intelligence model as described in any one of claims 1 to 9, characterized in that: include, The acquisition module is used to collect industrial data to be integrated for general applications and input it into an artificial intelligence model for application semantic parsing to generate an application semantic element set. The application semantic element set includes application capability items, interface input parameter attributes, interface output parameter attributes, context constraints, and calling constraints. The encoding module is used to perform semantic filtering and encoding on the application semantic element set, and generate semantic identifiers. The resource module is used to establish the corresponding integration relationship between multiple application capability items and interface input parameter attributes and interface output parameter attributes using semantic identifiers, and to generate a general application integration resource library. The extraction module is used to acquire real-time industrial integration requirements and use artificial intelligence models to extract the semantics of the requirements, generating target semantics. The filtering module is used to perform candidate capability chain search and parameter semantic connection determination on the target semantics using a general application integration resource library, and generate an executable integration path set. The integration module is used to perform interface calls and sequential integration of application capability items using executable integration path sets, and generate general application integration management results.

Images