Generative retrieval method and device for SOA service

Abstract

The invention relates to a generative retrieval method and device for an SOA service, and belongs to the technical field of artificial intelligence. The method comprises the steps that a unique URL identifier is determined according to the service content of the SOA service, and a constraint condition unique identifier is constructed according to the constraint condition of the SOA service; generating a plurality of groups of query training sample sets according to the URL identifier and the unique identifier of the constraint condition, and training a generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; obtaining a user query, inputting the user query into the fully trained generative retrieval model, and generating a target URL identifier and a target constraint condition unique identifier corresponding to the user query; and determining a target SOA service according to the target URL identifier and the target constraint condition unique identifier. According to the method and the device, the technical problem that the SOA service matched with the query cannot be accurately obtained in the SOA service retrieval task is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a generative retrieval method and device for SOA services. Background Art

[0002] With the rapid development of intelligent and connected vehicles, traditional distributed electronic control units (ECUs) and electronic and electrical architectures (EEAs) are unable to meet the rapid iteration requirements of intelligent functions in terms of isolated functions and data transmission performance. SOA (service-oriented architecture) achieves flexible division, reuse and interoperability of various functional modules through low coupling and standardized interfaces, thereby improving data collection efficiency, reducing development complexity and costs, and promoting cross-platform reuse, supporting the development of intelligent connected vehicles with high real-time and high bandwidth requirements.

[0003] The development of automobile intelligence integrates the progress of advanced sensor technology, Internet of Things, artificial intelligence and big data analysis, enabling vehicles to achieve more complex functions. At the same time, consumers' demand for intelligent and personalized services is increasing, which has driven automobile manufacturers to develop more SOA services to meet market expectations. When automobile developers search and select SOA services, they usually find registered services through service catalogs or registration centers, filter them by keywords and tags, and check service documents to understand interface definitions and usage instructions. Subsequently, developers will evaluate the quality indicators of services (such as response time and call frequency) to select the services that best meet their needs, and verify the compatibility and stability of services through test calls before integrating them into the system. With the surge in the number of SOA services and the complexity of service content, it has become increasingly difficult to find suitable SOA services. An accurate retrieval mechanism can quickly identify and obtain SOA services that meet specific needs, which can not only significantly shorten the development cycle and improve the success rate of projects, but also ensure that the final product can meet the market's high requirements for intelligence and personalization. In addition, this mechanism can help developers select high-quality options from a large number of services, reduce the risk of system failures, and promote innovation and technological progress.

[0004] SOA service retrieval is to retrieve relevant SOA services based on the developer's query. Existing retrieval methods are mainly divided into sparse retrieval and dense retrieval. Sparse retrieval refers to retrieval methods that use keyword matching, such as inverted index. This type of method will sort according to the degree of match between the keywords in the user query and the service description text. Although this method is simple and efficient, it has obvious disadvantages, such as difficulty in handling synonyms and contextual relevance, resulting in low accuracy of retrieval results, especially when the service description does not contain the exact keywords of the user query. Dense retrieval uses a deep learning model to generate embedded representations of queries and documents, and then performs retrieval by calculating the similarity between the embeddings. This method can capture semantic information and handle the synonymy problem of different expressions. However, dense retrieval cannot accurately handle the constraints in the query, and requires high computing resources. The retrieval performance may depend on the quality and diversity of the model training data. Summary of the invention

[0005] In view of this, it is necessary to provide a generative retrieval method and device for SOA services to solve the technical problems that sparse retrieval may not accurately match the developer's query in the SOA service retrieval task, especially when the service document description is not completely consistent with the query language, and although dense retrieval can provide better semantic matching, it cannot accurately handle the constraints in the query and has high computational and storage overheads.

[0006] In order to solve the above problems, the present invention provides a generative retrieval method for SOA services, comprising: Determine a unique URL identifier based on the service content of the SOA service, and construct a constraint unique identifier based on the constraints of the SOA service; Generate multiple sets of query training sample sets according to the URL identifier and the constraint condition unique identifier, and train the generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; Obtaining a user query, inputting the user query into a well-trained generative retrieval model, and generating a target URL identifier and a target constraint unique identifier corresponding to the user query; The target SOA service is determined based on the target URL identifier and the target constraint unique identifier.

[0007] In a possible implementation, determining a unique URL identifier according to the service content of the SOA service includes: Determine a unique URL identifier based on the access address of the SOA service.

[0008] In a possible implementation, constructing a constraint unique identifier according to the constraint of the SOA service includes: Build the initial index tree; Constraints on SOA services are divided into different levels according to categories; According to the subordinate relationship between the various levels of the constraint conditions, the content of each level is filled into the nodes of the initial index tree to create a constraint condition index tree; Encoding the constraint condition nodes in the constraint condition index tree to generate codes for each constraint condition node; The constraint node codes at each level are combined to form a unique constraint identifier.

[0009] In a possible implementation manner, encoding the constraint nodes in the constraint index tree includes: Setting a fixed code length for each constraint level; Set empty codes for levels where no constraints exist or where constraints are inappropriate.

[0010] In a possible implementation, generating multiple groups of query training sample sets according to the URL identifier and the constraint condition unique identifier includes: Generate multiple groups of query keywords based on the URL identifier and the constraint condition unique identifier; Input multiple groups of query keywords into a well-trained large language model to generate multiple natural language queries and construct multiple groups of query training sample sets.

[0011] In a possible implementation, the training of the generative retrieval model according to the query training sample set to obtain a fully trained generative retrieval model includes: Taking the query training sample set and SOA service content as input, inputting them into the initial generative retrieval model, generating a URL identifier and a constraint condition unique identifier corresponding to each query training sample; According to the difference between the URL identifier and constraint unique identifier corresponding to each query training sample and the target URL identifier and target constraint unique identifier, the loss function of the initial generative retrieval model is modified until the difference disappears, thereby obtaining a fully trained generative retrieval model.

[0012] In a possible implementation, determining the target SOA service according to the target URL identifier and the target constraint unique identifier includes: Determine the initial SOA service based on the target URL; The initial SOA service is verified according to the target constraint unique identifier to determine the target SOA service.

[0013] In a second aspect, the present invention further provides a generative search device for SOA services, comprising: An identifier determination module, used to determine a unique URL identifier according to the service content of the SOA service, and to construct a constraint condition unique identifier according to the constraint conditions of the SOA service; A retrieval model training module is used to generate multiple query training sample sets according to the URL identifier and the constraint condition unique identifier, and train the generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; An identifier acquisition module, used to acquire a user query, input the user query into a well-trained generative retrieval model, and generate a target URL identifier and a target constraint unique identifier corresponding to the user query; The SOA service determination module is used to determine the target SOA service according to the target URL identifier and the target constraint condition unique identifier.

[0014] In a third aspect, the present invention further provides an electronic device, comprising: a processor and a memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the steps in the generative retrieval method for SOA services as described above are implemented.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium storing one or more programs, wherein the one or more programs can be executed by one or more processors to implement the steps in the generative retrieval method for SOA services as described above.

[0016] The beneficial effects of the present invention are: using the SOA service URL and the constraint unique identifier to construct a unique ID number, the uniqueness and resolvability are guaranteed, and efficient matching of services and queries is supported. The construction structure of the constraint unique identifier finely models various constraints, which helps the generation model to accurately learn the constraints in the service. Compared with traditional methods, the present invention has significant advantages in accuracy and retrieval efficiency. It can not only perform semantic understanding of the SOA service URL identifier, but also accurately search according to the constraints of the service. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A method flow chart of an embodiment of a generative retrieval method for SOA services provided by the present invention; Figure 2 for Figure 1 A method flow chart of step S101 in an embodiment; Figure 3is a schematic diagram of an embodiment of a generative search device for SOA services provided by the present invention; Figure 4 It is a schematic diagram of the operating environment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0018] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0019] A specific embodiment of the present invention discloses a generative search method for SOA services, see Figure 1 ,include: S101, determining a unique URL identifier according to the service content of the SOA service, and constructing a constraint condition unique identifier according to the constraint conditions of the SOA service; S102, generating multiple groups of query training sample sets according to the URL identifier and the constraint condition unique identifier, and training a generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; S103, obtaining a user query, inputting the user query into a well-trained generative retrieval model, and generating a target URL identifier and a target constraint unique identifier corresponding to the user query; S104: Determine the target SOA service according to the target URL identifier and the target constraint unique identifier.

[0020] In this embodiment, a unique ID number is constructed by combining the SOA service URL and the constraint unique identifier, which ensures uniqueness and resolvability and supports efficient matching of services and queries. The construction structure of the constraint unique identifier finely models various constraints, helping the generation model to accurately learn the constraints in the service. Compared with traditional methods, the present invention has significant advantages in accuracy and retrieval efficiency. It can not only perform semantic understanding of the SOA service URL identifier, but also accurately search according to the constraints of the service.

[0021] It should be noted that when SOA services are registered in the system, they usually contain the following information: Access address (URL): The network access point of the service, providing a unique identifier for the service.

[0022] Port type: The logical port number that distinguishes different services, combined with the access address to complete the access to the corresponding ECU.

[0023] Operation: defines a specific function of a service, including input parameters, output parameters, and possible exceptions.

[0024] Message: A data structure that describes the operation, including input and output messages.

[0025] Binding: Specifies the protocol and message format for accessing a service.

[0026] When a user queries, it may include a description of the service functions, and may also include a description of the service constraints, such as output format, protocol requirements, efficiency indicators, etc.

[0027] In a specific embodiment, when a SOA service is registered in the system, the service manager will assign it a URL based on the service content. The URL is usually composed of multiple keywords separated by " / " and is unique. The keywords are a concise summary of the service, contain high-quality semantic information, and can be used directly as the identity number of the SOA service.

[0028] Further, a unique URL identifier is determined based on the access address of the SOA service. It is understandable that other unique information can also be used as the URL identifier.

[0029] In some embodiments, the constraint condition unique identifier is constructed according to the constraint condition of the SOA service. Figure 2 ,include: S201, constructing an initial index tree; S202, classifying the constraints of SOA services into different levels according to categories; S203, according to the subordinate relationship between the various levels of the constraint conditions, filling the content of each level into the nodes of the initial index tree to create a constraint condition index tree; S204, encoding the constraint condition nodes in the constraint condition index tree to generate codes for each constraint condition node; S205: Combining the constraint condition node codes at each level to form a constraint condition unique identifier.

[0030] In this embodiment, the constraint condition hierarchy is first defined. The root node of the index tree represents a general "service constraint" concept and does not contain any specific conditions. The first-level nodes are divided according to the common constraint types in SOA services, such as protocol constraints: describing the access protocols supported by the service (such as HTTP, SOAP, gRPC, etc.); output format constraints: defining the data format returned by the service (such as XML, JSON); efficiency indicator constraints: covering service performance indicators (such as maximum response time, throughput); security constraints: specifying access permission requirements (such as authentication methods, encryption requirements).

[0031] Furthermore, the nodes at the second level and below further subdivide each constraint type. For example, under "protocol constraints", it can be divided into specific protocol types, while under "efficiency indicator constraints", it can be refined into specific numerical ranges or thresholds. Then build an index tree structure, define each constraint condition as a node, and include the following information in the node: Node name: represents the specific constraint type (such as "HTTP protocol", "response time <200ms"); Node attributes: can include descriptive attributes, such as constraint type, applicable scenarios, priority, etc.; Parent-child relationship: build a tree structure based on the subordinate relationship between conditions, the parent node can represent a broader condition, and the child node represents a more detailed constraint.

[0032] Finally, encode each constraint node in the index tree so that each service can obtain a unique code based on all its constraints. Assign a fixed code length to each level. Each constraint level contains a finite number of conditions, and map each condition to a unique number. For example, the first-level nodes can use two-digit codes (such as "01" for protocol constraints, "02" for output format constraints, etc.), and the second-level nodes can use two-digit codes to further refine (such as "0101" for HTTP protocol, "0201" for XML format). At the same time, set an empty code "00" at the second level to mark the situation where this condition does not exist for the service. Finally, index the SOA service, and connect the codes of all index nodes that meet the conditions at the second level (the constraints not included are filled with empty codes) to form the index tree identity number of the service.

[0033] In some embodiments, generating multiple query training sample sets according to the URL identifier and the constraint condition unique identifier includes: Generate multiple groups of query keywords based on the URL identifier and the constraint condition unique identifier; Input multiple groups of query keywords into a well-trained large language model to generate multiple natural language queries and construct multiple groups of query training sample sets.

[0034] In this embodiment, a mapping relationship has been established between the SOA service and the URL identity number and the index tree identity number. It is necessary to construct a query corresponding to the service and assign the service identity number to the query, thereby establishing a mapping relationship between the query and the URL identity number and the index tree identity number. The embodiment of the present invention utilizes the generation capability of a large language model and generates high-quality training data based on a prompt learning method.

[0035] The specific steps are as follows: 1) Prompt design: Based on the identity number of each SOA service and the functions and constraints described in it, a set of rich prompts is designed to guide the large language model to generate corresponding query examples. These prompts include the functional description of the service, performance requirements, protocol constraints, etc., to cover the diverse expressions that users may make when querying. 2) Generate queries: Input the designed prompts into the large language model and let the model generate multiple natural language queries, which cover the different expressions and detailed descriptions that users may propose. Each generated natural language query will be automatically associated with the corresponding SOA service identity number to form a training sample. 3) Quality screening: In order to ensure that the generated queries are accurate and useful, a combination of automatic and manual screening is adopted. First, the generated queries are preliminarily screened using rules or auxiliary models to remove redundant, irrelevant or misleading content. Then, some samples are manually reviewed to further optimize the data quality. Finally, a high-quality, rich and diverse training dataset is constructed, which not only saves labor costs, but also improves the understanding and retrieval performance of the generated model for SOA service queries, and can more accurately match and return relevant services. In some embodiments, the training of the generative retrieval model according to the query training sample set to obtain a fully trained generative retrieval model includes: Taking the query training sample set and SOA service content as input, inputting them into the initial generative retrieval model, generating a URL identifier and a constraint condition unique identifier corresponding to each query training sample; According to the difference between the URL identifier and constraint unique identifier corresponding to each query training sample and the target URL identifier and target constraint unique identifier, the loss function of the initial generative retrieval model is modified until the difference disappears, thereby obtaining a fully trained generative retrieval model.

[0036] In this embodiment, the goal of generating model training is to enable the model to have the ability to generate accurate identification numbers based on user queries or service descriptions. For training input and output: the input of training data includes natural language queries and SOA service descriptions. These input texts contain different expressions that users may use, functional descriptions of services, and constraints of services, aiming to simulate the user's real retrieval needs. The target output corresponding to each input sample is the identification number of the service, including a combination based on URL and index tree. The URL part ensures the uniqueness of the identification number, while the index tree part represents the constraints related to the service, helping the model understand the mapping relationship between the service and its constraints.

[0037] It should be noted that the base model can adopt advanced pre-trained generative models, such as GPT, T5, or variants of LLM.

[0038] Since the pre-trained models have been pre-trained on a large amount of natural language data, they have rich language understanding and generation capabilities and can capture the complex mapping relationship between queries and ID numbers. An autoregressive generative model is used as the model training task, that is, the model predicts the next character or token based on the current generated content until the entire ID number generation is completed. Using cross entropy loss as the optimization target, the error between the ID number sequence generated by the model and the real ID number sequence is calculated, and the model parameters are gradually optimized. By minimizing the loss function, the model continuously improves the accuracy of the generated ID number. Through the above training process, the generative model can learn the mapping relationship between complex queries and service ID numbers.

[0039] In some embodiments, determining the target SOA service according to the target URL identifier and the target constraint unique identifier includes: Determine the initial SOA service based on the target URL; The initial SOA service is verified according to the target constraint unique identifier to determine the target SOA service.

[0040] In this embodiment, the retrieval system first receives the user's natural language query, and parses it to extract key information contained in the query, such as the description of service functions, performance requirements, protocol requirements, etc.

[0041] Then, based on the parsed information, a prompt word is constructed and input into the generation model, which generates an identity number that matches the user query. This identity number consists of two parts: one is a unique identifier for the SOA service URL, and the other is a code generated from the service constraint index tree, representing the specific constraints associated with the service. The generation process ensures that the model not only understands the functional description of the service, but also takes into account the various constraints specified by the user.

[0042] The model then uses the generated ID to search for pre-indexed SOA services. The uniqueness and semantic structure of the ID allows the system to quickly locate service items that meet user requirements. This mapping process can effectively filter out services that do not meet the requirements and retain only the most relevant candidate services. At the same time, the constraint index tree ID is used to verify the user's service constraint requirements to ensure that the search results meet the user's requirements.

[0043] Finally, the model organizes the matched SOA service information into an easy-to-understand format and returns it to the user, including the service access address, functional description, related constraints, etc. Based on the complexity of the query, the accuracy and confidence of the match, the retrieval system adaptively chooses to provide a single or multiple related services so that the user can choose the most suitable solution.

[0044] The embodiment of the present invention realizes end-to-end retrieval of SOA services through generative retrieval technology, which significantly improves the accuracy and efficiency of retrieval. The identity number is constructed by combining the SOA service URL and the constraint index tree to ensure uniqueness and resolvability, and support efficient matching of services and queries. The service constraint index tree structure finely models various constraints, helping the generation model to accurately learn the constraints in the service. In addition, the invention uses a large language model to automatically build high-quality training data, which greatly saves labor costs and enhances the generation model's ability to understand user queries. Compared with traditional methods, the present invention has significant advantages in accuracy and retrieval efficiency. It can not only perform semantic understanding, but also accurately search according to the constraints of the service.

[0045] Based on the above-mentioned generative search method for SOA services, the embodiment of the present invention also provides a generative search device for SOA services, see Figure 3 ,include: The identifier determination module 310 is used to determine a unique URL identifier according to the service content of the SOA service, and to construct a constraint condition unique identifier according to the constraint conditions of the SOA service; A retrieval model training module 320 is used to generate multiple query training sample sets according to the URL identifier and the constraint condition unique identifier, and train the generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; The identifier acquisition module 330 is used to acquire a user query, input the user query into a well-trained generative retrieval model, and generate a target URL identifier and a target constraint unique identifier corresponding to the user query; The SOA service determination module 340 is used to determine the target SOA service according to the target URL identifier and the target constraint unique identifier.

[0046] like Figure 4 As shown, based on the above-mentioned generative retrieval method for SOA services, the present invention also provides an electronic device, which can be a computing electronic device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The electronic device includes a processor 410, a memory 420, and a display 430. Figure 4 Only some components of the electronic device are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0047] In some embodiments, the memory 420 may be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device. In other embodiments, the memory 420 may also be an external storage electronic device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the memory 420 may also include both an internal storage unit of the electronic device and an external storage electronic device. The memory 420 is used to store application software and various types of data installed in the electronic device, such as program codes installed in the electronic device. The memory 420 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a generative retrieval program 440 for SOA services is stored on the memory 420, and the generative retrieval program 440 for SOA services can be executed by the processor 410, thereby realizing the generative retrieval method for SOA services in each embodiment of the present application.

[0048] In some embodiments, the processor 410 may be a central processing unit (CPU), a microprocessor or other data processing chip, configured to run program codes stored in the memory 420 or process data, such as executing a generative retrieval method for SOA services.

[0049] In some embodiments, the display 430 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 430 is used to display information on the generative retrieval electronic device for SOA services and to display a visual user interface. The components 410-430 of the electronic device communicate with each other via a system bus.

[0050] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0051] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A generative retrieval method for SOA services, characterized in that: include: Determine a unique URL identifier based on the service content of the SOA service, and construct a constraint unique identifier based on the constraints of the SOA service; Generate multiple sets of query training sample sets according to the URL identifier and the constraint condition unique identifier, and train the generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; Obtaining a user query, inputting the user query into a well-trained generative retrieval model, and generating a target URL identifier and a target constraint unique identifier corresponding to the user query; The target SOA service is determined based on the target URL identifier and the target constraint unique identifier.

2. The generative retrieval method for SOA services according to claim 1, characterized in that: Determining a unique URL identifier according to the service content of the SOA service includes: Determine a unique URL identifier based on the access address of the SOA service.

3. The generative retrieval method for SOA services according to claim 1, characterized in that: The step of constructing a constraint unique identifier according to the constraint of the SOA service includes: Build the initial index tree; Constraints on SOA services are divided into different levels according to categories; According to the subordinate relationship between the various levels of the constraint conditions, the content of each level is filled into the nodes of the initial index tree to create a constraint condition index tree; Encoding the constraint condition nodes in the constraint condition index tree to generate codes for each constraint condition node; The constraint node codes at each level are combined to form a unique constraint identifier.

4. The generative retrieval method for SOA services according to claim 3, characterized in that: The encoding of the constraint condition nodes in the constraint condition index tree comprises: Setting a fixed code length for each constraint level; Set empty codes for levels where no constraints exist or where the constraints are inappropriate.

5. The generative retrieval method for SOA services according to claim 1, characterized in that: The step of generating multiple query training sample sets according to the URL identifier and the constraint condition unique identifier includes: Generate multiple groups of query keywords based on the URL identifier and the constraint condition unique identifier; Input multiple groups of query keywords into a well-trained large language model to generate multiple natural language queries and construct multiple groups of query training sample sets.

6. The generative retrieval method for SOA services according to claim 1, characterized in that: The step of training the generative retrieval model according to the query training sample set to obtain a fully trained generative retrieval model includes: Taking the query training sample set and SOA service content as input, inputting them into the initial generative retrieval model, generating a URL identifier and a constraint condition unique identifier corresponding to each query training sample; According to the difference between the URL identifier and constraint unique identifier corresponding to each query training sample and the target URL identifier and target constraint unique identifier, the loss function of the initial generative retrieval model is modified until the difference disappears, thereby obtaining a fully trained generative retrieval model.

7. The generative retrieval method for SOA services according to claim 1, characterized in that: Determining the target SOA service according to the target URL identifier and the target constraint condition unique identifier includes: Determine the initial SOA service based on the target URL; The initial SOA service is verified according to the target constraint unique identifier to determine the target SOA service.

8. A generative search device for SOA services, characterized in that: include: An identifier determination module, used to determine a unique URL identifier according to the service content of the SOA service, and to construct a constraint condition unique identifier according to the constraint conditions of the SOA service; A retrieval model training module is used to generate multiple query training sample sets according to the URL identifier and the constraint condition unique identifier, and train the generative retrieval model according to the query training sample sets to obtain a fully trained generative retrieval model; An identifier acquisition module, used to acquire a user query, input the user query into a well-trained generative retrieval model, and generate a target URL identifier and a target constraint unique identifier corresponding to the user query; The SOA service determination module is used to determine the target SOA service according to the target URL identifier and the target constraint condition unique identifier.

9. An electronic device, characterized in that: include: Processor and memory; The memory stores a computer-readable program executable by the processor; When the processor executes the computer-readable program, the processor implements the steps in the generative retrieval method for SOA services according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the generative retrieval method for SOA services according to any one of claims 1 to 7.

Images