Government affair service field business object information extraction method and system based on large language model

Abstract

The invention discloses a government affair service field business object information extraction method and system based on a large language model, belongs to the technical field of deep learning, and aims to solve the technical problem of how to comprehensively, accurately and completely realize business object extraction. Comprising the following steps: receiving business data submitted by a business handling system through a business object management system, and determining related government affair service items and business object definitions associated with the government affair service items based on the business data; a cue word is constructed through the business object management system, and the cue word and the business data are sent to the large language model for processing; receiving and analyzing JSON-format data returned by the large language model through the business object management system, and extracting business object information based on the JSON-format data; and storing the extracted business object information in a database, and establishing an association relationship between the business data and the business object information.

Description

Technical Field

[0001] The present invention relates to the technical field of deep learning, and more specifically to a method and system for extracting business object information in the field of government services based on large language models. Background Art

[0002] Currently, in the field of government services, the extraction of business object information usually adopts the method of pre-defining business objects and then taking values from information fields at specified locations (forms, application information, etc.) according to data extraction rules through hard coding. This method has the following problems:

[0003] (1) Poor maintainability: Whenever a new business object needs to be extracted or the extraction rules of existing objects need to be modified, it is necessary to modify the code or adjust the information field configuration in the business process, which increases the complexity and workload of system maintenance;

[0004] (2) Lack of flexibility: It is unable to quickly respond to new business scenarios or newly added object types, and the adaptability of the system is poor;

[0005] (3) High risk of human error: Since it is necessary to manually write code and maintain information fields, omissions or errors are likely to occur, affecting the accuracy and integrity of data;

[0006] (4) Low efficiency: For a large number of different types of business objects, a large number of extraction rules need to be written and maintained, which is not only time-consuming and laborious, but also difficult to ensure the consistency and effectiveness of all rules;

[0007] (5) Difficulty in processing multi-modal data: For some non-standard format text information and data such as pictures and videos, traditional methods are difficult to effectively extract the required business object information.

[0008] How to comprehensively, accurately and completely implement the extraction of business objects is a technical problem that needs to be solved. Summary of the Invention

[0009] The technical task of the present invention is to address the above deficiencies and provide a method and system for extracting business object information in the field of government services based on large language models to solve the technical problem of how to comprehensively, accurately and completely implement the extraction of business objects.

[0010] In a first aspect, a method for extracting business object information in the field of government services based on large language models according to the present invention is applied among a client, an object management system and a database, and includes the following steps:

[0011] Configure a business object management system. Through the configured business object management system, define business objects, maintain the government service items to which the business objects belong, and the persistence strategy for associating business object information under the government service items, provide a method for obtaining business objects, and manage the extracted business object information. Among them, the business object definition includes an object description, which is used to explain the object meaning to the large language model and dynamically construct prompt words;

[0012] Receive business data submitted by the business processing system through the business object management system, and determine the relevant government service items and the business object definitions associated with the government service items based on the business data;

[0013] Based on the business object definition and business data, construct prompt words through the business object management system, and send the prompt words and business data to the large language model for processing;

[0014] Receive and parse the data in JSON format returned by the large language model through the business object management system, and extract business object information based on the JSON format data;

[0015] Store the extracted business object information in the database, and establish an association relationship between the business data and the business object information.

[0016] Preferably, by configuring the business object management system, deploy an object definition function, an item management function, and an object information management function in the business object management system, and configure a data interface for the business object management system. The data interface is a RESTful API that supports JSON format data transmission;

[0017] The object definition function is used to define business objects, including defining the basic field information and object description of the business objects;

[0018] The item management function is used to maintain the government service items to which the business objects belong, and maintain the persistence strategy for managing business object information under the current item; the object information management function is used to provide a method for obtaining business object information and manage the extracted business object information;

[0019] The data interface is used to receive business data from the business processing system. The business data is in readable Chinese field names, semantically clear English field names, or Chinese pinyin.

[0020] Preferably, the prompt is dynamically constructed based on the description of the business object, attribute requirements, data format description, and preset object description. The main meaning of the prompt is to require the large language model to extract business object information from the input text and output the extracted business object information in a structured manner. In addition to the fixed object information, the output JSON-formatted data also needs to include the attribution information of the business object, where the attribution information includes the attribution subject and the attribution object.

[0021] Preferably, after receiving the JSON-formatted data returned by the large language model, the business object management system determines the data storage method according to the persistence strategy of the business object under the current matter, and persists the deserialized business object information based on the persistence strategy. The data storage methods include saving, updating, and deleting.

[0022] In a second aspect, an information extraction system for business objects in the field of government services based on a large language model of the present invention is applied between a client, an object management system, and a database, and is used to extract business object information through an information extraction method for business objects in the field of government services based on a large language model as described in any item of the first aspect. The system includes a function configuration module, a matter association module, a prompt construction and large model call module, a data processing module, and a data storage module;

[0023] The function configuration module is used to perform the following: configure the business object management system, define business objects through the configured business object management system, maintain the government service matters to which the business objects belong, and the persistence strategy of the business object information associated under the government service matters, provide a business object acquisition method, and manage the extracted business object information. The object description in the business object definition is used to explain the object meaning to the large language model and dynamically construct prompts;

[0024] The matter association module calls the business object management system and is used to perform the following: receive business data submitted by the business handling system through the business object management system, and determine the relevant government service matters and the business object definitions associated with the government service matters based on the business data;

[0025] The prompt construction and large model call module calls the business object management system and the large language model and is used to perform the following: construct a prompt based on the business object definition and the business data through the business object management system, and send the prompt and the business data to the large language model for processing;

[0026] The data processing module calls the business object management system and the large language model and is used to perform the following: receive and parse the JSON-formatted data returned by the large language model through the business object management system, and extract business object information based on the JSON-formatted data;

[0027] The data storage module calls the business object management system to perform the following: storing the extracted business object information in a database and establishing an association relationship between business data and business object information.

[0028] Preferably, when configuring the business object management system, the function configuration module is used to deploy object definition functions, matter management functions, and object information management functions in the business object management system, and configure a data interface for the business object management system. The data interface is a RESTful API that supports data transmission in JSON format;

[0029] The object definition function is used to define business objects, including defining basic field information and object descriptions of business objects;

[0030] The matter management function is used to maintain the government service matters to which business objects belong and maintain the persistence strategy for managing business object information under the current matter; the object information management function is used to provide methods for obtaining business object information and manage the extracted business object information;

[0031] The data interface is used to receive business data from the business processing system. The business data is readable Chinese field names, English field names with clear semantics, or Chinese pinyin.

[0032] Preferably, the prompt is dynamically constructed based on the description, attribute requirements, data format description of the business object, and a preset object description. The main meaning of the prompt is to require the large language model to extract business object information from the input text and output the extracted business object information in a structured manner. In addition to the fixed object information, the output JSON format data also needs to additionally include the attribution information of the business object. The attribution information includes the attribution subject and the attribution object.

[0033] Preferably, after the business object management system receives the JSON format data returned by the large language model, it determines the data storage method according to the persistence strategy of the business object under the current matter, and persistently stores the deserialized business object information based on the persistence strategy. The data storage methods include saving, updating, and deleting.

[0034] The method and system for extracting business object information in the government service field based on the large language model of the present invention have the following advantages:

[0035] 1. By introducing a large language model to extract business object information in the government service field, the implementation of the extraction logic is converted from developers to the large language model, completely avoiding hard coding, improving maintainability, and reducing the probability of human errors;

[0036] 2. Since only the model definition needs to be maintained, the need to maintain the value relationships of object fields is avoided, greatly improving flexibility to handle real-time changes.

[0037] 3. When using a large language model that supports multimodality, various types of unstructured data can be extracted, broadening the object extraction channels and reducing the formatting requirements for source data. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0039] The present invention will be further described below in conjunction with the drawings.

[0040] Figure 1 FIG. is a flowchart of a method for extracting business object information in the field of government affairs services based on a large language model for Embodiment 1;

[0041] Figure 2 FIG. is a flowchart of the data exchange process between the object management system and the large language model in a method for extracting business object information in the field of government affairs services based on a large language model for Embodiment 1;

[0042] Figure 3 FIG. is a flowchart of the data processing process of the business object management system for the JSON format data returned by the large model in a method for extracting business object information in the field of government affairs services based on a large language model for Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The present invention will be further described below in conjunction with the drawings and specific embodiments, so that those skilled in the art can better understand the present invention and implement it. However, the embodiments cited do not limit the present invention. Without conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0044] The embodiments of the present invention provide a method for extracting business object information in the field of government affairs services based on a large language model, which is used to solve the technical problem of how to comprehensively, accurately, and completely implement business object extraction.

[0045] Embodiment 1:

[0046] A method for extracting business object information in the field of government affairs services based on a large language model of the present invention includes five steps: function configuration, matter association, large model call, data processing, and data storage.

[0047] Step S100 Function Configuration: Configure the business object management system. Through the configured business object management system, define business objects, maintain the government service items to which the business objects belong, and the persistence strategy for the associated business object information under the government service items, provide a method for obtaining business objects, and manage the extracted business object information. The object description is included in the business object definition, and the object description is used to explain the object meaning to the large language model and dynamically construct prompt words.

[0048] In step S100 of this embodiment, by configuring the business object management system, deploy an object definition function, an item management function, and an object information management function in the business object management system, and configure a data interface for the business object management system. The data interface is a RESTful API that supports JSON format data transmission.

[0049] The object definition function is used to define business objects, including defining the basic field information and object description of the business objects; the item management function is used to maintain the government service items to which the business objects belong and maintain the persistence strategy for managing business object information under the current item; the object information management function is used to provide a method for obtaining business object information and manage the extracted business object information; the data interface is used to receive business data from the business processing system, and the business data is in readable Chinese field names or semantically clear English field names or Chinese pinyin.

[0050] In this embodiment, through step S100, three basic functions are configured in the object management system to support the large language model to extract objects from business information, namely the object definition function, the item management function, and the object information management function.

[0051] The object definition function is used to define business objects. In addition to the basic field information, the definition of business objects also needs to include an object description for explaining the object meaning to the large language model and constructing prompt words; the item management function is used to maintain the government service items to which the business objects belong and maintain the persistence strategy for the associated object data under the current item; the object information management function is used to manage the extracted business object information and provide a method for obtaining object data. At the same time, the object management system provides an interface for receiving the complete business processing data to obtain the original data extracted by the business object. Since these data will ultimately be handed over to the large language model for recognition, it is required that these data are in readable Chinese field names or semantically clear English field names or Chinese pinyin. There is no structured requirement for the input parameters.

[0052] Step S200 Item Association: Receive the business data submitted by the business processing system through the business object management system, and determine the relevant government service items and the business object definitions associated with the government service items based on the business data.

[0053] Step S300 Large Model Invocation: Based on the business object definition and business data, construct a prompt through the business object management system, and send the prompt and business data to the large language model for processing.

[0054] In this embodiment, the prompt is dynamically constructed based on the description, attribute requirements, data format description of the business object, and the preset object description. The main meaning of the prompt is to require the large language model to extract business object information from the input text and output the extracted business object information in a structured manner. In addition to the fixed object information, the output JSON format data also needs to include the attribution information of the business object, and the attribution information includes the attribution subject and the attribution object.

[0055] Send the constructed prompt together with the business processing information to the large language model (taking the API specification of OpenAI's gpt-4 model as an example, the prompt is sent through the system role, and the business information is sent through the user role, and the non-streaming output is set) and wait for the return result.

[0056] Step S400 Data Processing: Receive and parse the JSON format data returned by the large language model through the business object management system, and extract business object information based on the JSON format data.

[0057] In this embodiment, after the business object management system receives the JSON format data returned by the large language model, it determines the data storage method according to the persistence strategy of the business object under the current matter, and persists the deserialized business object information based on the persistence strategy. The data storage methods include saving, updating, and deleting.

[0058] Step S500 Data Storage: Store the extracted business object information in the database, and establish an association relationship between the business data and the business object information.

[0059] Based on the method disclosed in this embodiment, a specific case is given: Suppose there is a government service matter called "Enterprise Business License Application", and it is necessary to extract the business object "Enterprise Information". The specific operations are as follows:

[0060] (1) In the object management system, the business object "Enterprise Information" is predefined, including the following attributes: enterprise name, unified social credit code, legal representative, registered capital, enterprise type, business scope, establishment date, business term, and domicile;

[0061] (2) In the object management system, associate the "Enterprise Information" object with the "Enterprise Business License Application" matter;

[0062] (3) After the enterprise completes the business license registration, the business processing system transmits the complete registration information to the object management system. This information may include application forms, review opinions, approval documents, etc.;

[0063] (4) Based on the received information, the object management system determines that this is an "enterprise business license registration" matter and retrieves the object definition of the "enterprise information" to be extracted from the associated table;

[0064] (5) The system constructs the following prompt: "Please extract the enterprise object information and the attribution information of this enterprise object information from the following business processing information. The enterprise object information includes enterprise name, unified social credit code, legal representative, registered capital, enterprise type, business scope, establishment date, business term, and domicile. The attribution information includes the attribution subject and the attribution object. Please return the result in JSON format, with the key names being the above attribute names and the key values being the corresponding extracted information.";

[0065] (6) Send the prompt and the business processing information to the large language model together;

[0066] The large language model processes the information and returns JSON data in the following format:

[0067]

[0068]

[0069] (7) The object management system parses this JSON data, stores the information in the object information database, and establishes an association relationship with this business processing.

[0070] Through the above operations, the system can automatically extract structured business object information from various formats of business processing information without writing specific extraction rules for each case.

[0071] The method disclosed in this embodiment utilizes the generality and flexibility of the large language model, without writing specific extraction rules for each business object. By adjusting the prompt, the system can adapt to various new business scenarios and object types, greatly improving the maintainability and scalability of the system. At the same time, this method can effectively process multi-modal data, improving the accuracy and integrity of information extraction.

[0072] Embodiment 2:

[0073] A business object information extraction system in the field of government affairs services based on a large language model of the present invention is applied among a client, an object management system, and a database, and includes a function configuration module, a matter association module, a prompt construction and large model call module, a data processing module, and a data storage module.

[0074] The function configuration module is used to perform the following: configure the business object management system, define business objects through the configured business object management system, maintain the government service items to which the business objects belong, and the persistence strategy of the associated business object information under the government service items, provide a business object acquisition method, and manage the extracted business object information. The object description is included in the business object definition, and the object description is used to explain the object meaning to the large language model and dynamically construct prompt words.

[0075] In this embodiment, by configuring the business object management system, an object definition function, an item management function, and an object information management function are deployed in the business object management system, and a data interface is configured for the business object management system. The data interface is a RESTful API and supports data transmission in JSON format.

[0076] The object definition function is used to define business objects, including defining the basic field information and object description of the business objects; the item management function is used to maintain the government service items to which the business objects belong and maintain the persistence strategy of managing business object information under the current item; the object information management function is used to provide a business object information acquisition method and manage the extracted business object information; the data interface is used to receive business data from the business processing system, and the business data is readable Chinese field names or semantically clear English field names or Chinese pinyin.

[0077] The function configuration module configures three basic functions in the object management system to support the large language model to extract objects from business information, namely the object definition function, the item management function, and the object information management function.

[0078] The object definition function is used to define business objects. In addition to the basic field information, the definition of business objects also needs to include an object description to explain the object meaning to the large language model and construct prompt words; the item management function is used to maintain the government service items to which the business objects belong and maintain the persistence strategy of the associated object data under the current item; the object information management function is used to manage the extracted business object information and provide an object data acquisition method. At the same time, the object management system provides a receiving interface for the complete business processing data to obtain the original data extracted by the business objects. Since these data will ultimately be handed over to the large language model for recognition, it is required that these data are readable Chinese field names or semantically clear English field names or Chinese pinyin. There is no structured requirement for the input parameters.

[0079] The item association module calls the business object management system and is used to perform the following: receive the business data submitted by the business processing system through the business object management system, and determine the relevant government service items and the business object definitions associated with the government service items based on the business data.

[0080] The prompt construction and large model call module calls the business object management system and the large language model to perform the following: Based on the business object definition and business data, construct a prompt through the business object management system, and send the prompt and business data to the large language model for processing.

[0081] In this embodiment, the prompt is dynamically constructed based on the description, attribute requirements, data format description of the business object, and the preset object description. The main meaning of the prompt is to require the large language model to extract business object information from the input text and output the extracted business object information in a structured manner. In addition to the fixed object information, the output JSON format data also needs to include the attribution information of the business object, and the attribution information includes the attribution subject and the attribution object.

[0082] Send the constructed prompt together with the business processing information to the large language model (taking the API specification of the gpt-4 model of OpenAI as an example, the prompt is sent through the system role, and the business information is sent through the user role, and the non-stream output is set) and wait for the return result.

[0083] The data processing module calls the business object management system and the large language model to perform the following: Receive and parse the JSON format data returned by the large language model through the business object management system, and extract business object information based on the JSON format data.

[0084] In this embodiment, after the business object management system receives the JSON format data returned by the large language model, it decides the data storage method according to the persistence strategy of the business object under the current matter, and persists the deserialized business object information based on the persistence strategy. The data storage methods include saving, updating, and deleting.

[0085] The data storage module calls the business object management system to perform the following: Store the extracted business object information in the database and establish an association relationship between the business data and the business object information.

[0086] The system disclosed in this embodiment can implement the extraction of business object information by implementing the method disclosed in Embodiment 1.

[0087] The above has introduced in detail the method and system for extracting business object information in the field of government affairs services based on large language models provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for extracting business object information in the field of government services based on a large language model, characterized in that: Applied between the client, object management system and database, it includes the following steps: Configure a business object management system, through which business objects are defined, government service items to which business objects belong and persistence strategies for associated business object information under government service items are maintained, business object acquisition methods are provided, and extracted business object information is managed. The business object definition includes an object description, which is used to explain the meaning of the object to the large language model and dynamically construct prompt words; Receive business data submitted by the business processing system through the business object management system, and determine the relevant government service items and the business object definitions associated with the government service items based on the business data; Based on the business object definition and business data, the prompt words are constructed through the business object management system, and the prompt words and business data are sent to the large language model for processing; Receive and parse the JSON format data returned by the large language model through the business object management system, and extract business object information based on the JSON format data; The extracted business object information is stored in the database, and an association relationship between the business data and the business object information is established.

2. The method for extracting business object information in the field of government services based on a large language model according to claim 1 is characterized in that: By configuring the business object management system, deploying the object definition function, the event management function and the object information management function in the business object management system, and configuring the data interface for the business object management system, the data interface is a RESTful API that supports data transmission in JSON format; The object definition function is used to define business objects, including defining the basic field information and object description of the business object; The item management function is used to maintain the government service items to which the business object belongs, and to maintain the persistence strategy for managing the business object information under the current item; the object information management function is used to provide a method for obtaining business object information and manage the extracted business object information; The data interface is used to receive business data from the business processing system. The business data is a readable Chinese field name or a semantically clear English field name or Chinese pinyin.

3. The method for extracting business object information in the field of government services based on a large language model according to claim 1 is characterized in that: The prompt words are dynamically constructed based on the description of the business object, attribute requirements, data format description and pre-set object description. The main meaning of the prompt words is to require the large language model to extract business object information from the input text and output the extracted business object information in a structured manner. In addition to fixed object information, the output JSON format data also needs to contain additional business object ownership information, which includes the ownership subject and the ownership object.

4. The method for extracting business object information in the field of government services based on a large language model according to claim 1 is characterized in that: After the business object management system receives the data in JSON format returned by the large language model, it determines the data storage method according to the persistence strategy of the business object under the current matter, and stores the deserialized business object information persistently based on the persistence strategy. The data storage methods include saving, updating and deleting.

5. A business object information extraction system in the government service field based on a large language model, characterized in that: Applied between a client, an object management system and a database, and used to extract business object information through a business object information extraction method in the government service field based on a large language model as described in any one of claims 1 to 4, the system comprising a function configuration module, a matter association module, a prompt word construction and large model calling module, a data processing module and a data storage module; The function configuration module is used to perform the following: configure the business object management system, define the business object through the configured business object management system, maintain the government service items to which the business object belongs and the persistence strategy of the business object information associated with the government service items, provide a business object acquisition method and manage the extracted business object information, wherein the business object definition includes an object description, and the object description is used to explain the meaning of the object to the large language model and dynamically construct prompt words; The item association module calls the business object management system to perform the following: receiving business data submitted by the business processing system through the business object management system, and determining the relevant government service items and the business object definitions associated with the government service items based on the business data; The prompt word construction and large model calling module calls the business object management system and the large language model to perform the following: based on the business object definition and business data, construct the prompt word through the business object management system, and send the prompt word and business data to the large language model for processing; The data processing module calls the business object management system and the large language model to perform the following: receiving and parsing the JSON format data returned by the large language model through the business object management system, and extracting business object information based on the JSON format data; The data storage module calls the business object management system to execute the following: storing the extracted business object information in a database, and establishing an association relationship between the business data and the business object information.

6. The system for extracting business object information in the field of government services based on a large language model according to claim 5 is characterized in that: When configuring the business object management system, the function configuration module is used to deploy the object definition function, the event management function and the object information management function in the business object management system, and configure the data interface for the business object management system. The data interface is a RESTful API that supports data transmission in JSON format. The object definition function is used to define business objects, including defining the basic field information and object description of the business object; The item management function is used to maintain the government service items to which the business object belongs, and to maintain the persistence strategy for managing the business object information under the current item; the object information management function is used to provide a method for obtaining business object information and manage the extracted business object information; The data interface is used to receive business data from the business processing system. The business data is a readable Chinese field name or a semantically clear English field name or Chinese pinyin.

7. The system for extracting business object information in the field of government services based on a large language model according to claim 5 is characterized in that: The prompt words are dynamically constructed based on the description of the business object, attribute requirements, data format description and pre-set object description. The main meaning of the prompt words is to require the large language model to extract business object information from the input text and output the extracted business object information in a structured manner. In addition to fixed object information, the output JSON format data also needs to contain additional business object ownership information, which includes the ownership subject and the ownership object.

8. The system for extracting business object information in the field of government services based on a large language model according to claim 5 is characterized in that: After the business object management system receives the data in JSON format returned by the large language model, it determines the data storage method according to the persistence strategy of the business object under the current matter, and stores the deserialized business object information persistently based on the persistence strategy. The data storage methods include saving, updating and deleting.

Images