Adaptive data table structure generation and data storage method and system

Abstract

This invention discloses an adaptive data table structure generation and data storage method and system, belonging to the field of data processing technology. The technical problem it addresses is: how to provide a solution that can automatically identify the information type in any image and intelligently and flexibly store it in a structured and persistent manner. The method includes: extracting unstructured text information from the image to be processed using optical character recognition; extracting key attribute information from the unstructured text information based on data type and through structural recombination, and organizing the key attribute information into structured data objects for output; based on the pattern information of the structured data objects and the table name annotation table of the target database, calling the second largest language model (LLM) for semantic-level matching analysis, storing the structured data objects into the matching existing data table according to the semantic correspondence of the fields, or executing a table creation command to create a new data table and storing the structured data objects in the new data table.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to an adaptive data table structure generation and data storage method and system. Background Technology

[0002] Currently, users generate a large amount of fragmented image information through devices such as mobile phones and computers, such as screenshots of meeting whiteboards, poster photos, product descriptions, and personal notes. Traditional data recording methods mainly fall into two categories: 1. Manual input: Users view the pictures and manually input key information into a memo, Excel or a specific app. This method is inefficient, time-consuming, labor-intensive and prone to errors. 2. Fixed template applications: such as business card scanning apps and recipe recording apps. These applications can only handle specific types of data. Their backend database table structure is fixed in advance. When users need to record a new type of data (such as an insect encyclopedia picture, a bill, or an application form), these applications are powerless and cannot meet users' diverse and personalized data persistence needs.

[0003] How to provide a solution that can automatically identify the information type in any image and intelligently and flexibly store it in a structured and persistent manner is a technical problem that urgently needs to be solved in the current technology field. Summary of the Invention

[0004] The technical objective of this invention is to address the above-mentioned shortcomings by providing an adaptive data table structure generation and data storage method and system, thereby solving the technical problem of how to provide a solution that can automatically identify the information type in any image and intelligently and flexibly perform structured and persistent storage.

[0005] In a first aspect, the present invention provides an adaptive data table structure generation and data storage method, comprising the following steps: Data acquisition: Acquiring the image to be processed; Information extraction: Extracting unstructured text information from the image to be processed using optical character recognition methods; Data Conversion: Input instructions are constructed based on unstructured text information and preset Prompt instructions. Based on the input instructions, the first major language model is called to perform type recognition and structure recombination. The data type is determined by type recognition. Based on the data type, key attribute information is extracted from the unstructured text information through structure recombination. The key attribute information is organized into structured data objects according to the format specified by the Prompt instructions and output. The key attribute information includes name, parameters and features. Matching analysis: Extract the pattern information of structured data objects, and based on the pattern information of structured data objects and the table name annotation table of the target database, call the second largest language model LLM to perform semantic matching analysis to search whether there are any existing data tables in the target database that are suitable. Data storage: If a matching existing data table exists in the target database, the structured data object is stored in the matching existing data table according to the semantic correspondence of the fields. If no matching existing data table is found in the target database, a table creation instruction for the matching database is generated based on the structured data table and the third language model is called. The table creation instruction is executed to create a new data table and the structured data object is stored in the new data table.

[0006] Preferably, when acquiring images to be processed, the image sources include images captured in real time by the camera of an electronic device, images already stored in the photo album of the electronic device, and screenshots.

[0007] Preferably, the structured data object is in JSON format and includes a data type identifier, a main description, and attribute fields, wherein the attribute fields correspond to key attribute information in the unstructured text.

[0008] Preferably, the schema information includes the key names, data types, and field meanings of the structured data objects; In matching analysis, semantic matching analysis refers to comparing the semantic similarity between the field meanings of structured data objects and the fields of existing data tables, rather than simply comparing field names.

[0009] As a preferred option, the standard table creation command is the CREATE TABLE statement in Structured Query Language. The third large language model automatically matches the field data type based on the field attributes of the structured data object and sets the primary key, index, and comment information. In the table creation command, the field data types are adapted to the syntax specifications of the target database, and the primary key is selected from the key attribute information of the structured data object to ensure data uniqueness.

[0010] Secondly, the present invention provides an adaptive data table structure generation and data storage system, comprising a data acquisition module, an information extraction module, a data conversion module, a matching analysis module, and a data storage module; The data acquisition module is used to perform the following: acquire the image to be processed; The information extraction module is used to perform the following: extract unstructured text information from the image to be processed using optical character recognition methods; The data conversion module is used to perform the following: construct input instructions based on unstructured text information and preset Prompt instructions; based on the input instructions, call the first language model to perform type recognition and structure recombination; determine the data type through type recognition; extract key attribute information from the unstructured text information based on the data type and through structure recombination; and organize the key attribute information into structured data objects according to the format specified by the Prompt instructions and output them. The key attribute information includes name, parameters, and features. The matching analysis module is used to perform the following: extract the pattern information of structured data objects, and based on the pattern information of structured data objects and the table name annotation table of the target database, call the second largest language model LLM to perform semantic matching analysis and search whether there is a suitable existing data table in the target database. The data storage module performs the following actions: If a matching existing data table exists in the target database, the structured data object is stored in the matching existing data table according to the semantic correspondence of the fields. If no matching existing data table is found in the target database, a table creation instruction for the matching database is generated based on the structured data table and the third language model is called. The table creation instruction is executed to create a new data table and the structured data object is stored in the new data table.

[0011] Preferably, when acquiring images to be processed, the image sources include images captured in real time by the camera of an electronic device, images already stored in the photo album of the electronic device, and screenshots.

[0012] Preferably, the structured data object is in JSON format and includes a data type identifier, a main description, and attribute fields, wherein the attribute fields correspond to key attribute information in the unstructured text.

[0013] Preferably, the schema information includes the key names, data types, and field meanings of the structured data objects; Semantic matching analysis refers to comparing the semantic similarity between the field meanings of a structured data object and the field semantics of an existing data table, rather than simply comparing field names.

[0014] As a preferred option, the standard table creation command is the CREATE TABLE statement in Structured Query Language. The third large language model automatically matches the field data type based on the field attributes of the structured data object and sets the primary key, index, and comment information. In the table creation command, the field data types are adapted to the syntax specifications of the target database, and the primary key is selected from the key attribute information of the structured data object to ensure data uniqueness.

[0015] The adaptive data table structure generation and data storage method and system of the present invention have the following advantages: 1. High degree of automation and efficiency improvement: It realizes full-process automation from unstructured image information to structured data entry. Users only need to take a picture or screenshot once to complete complex data sorting and storage, completely freeing users from time-consuming, tedious and error-prone manual entry work, and greatly improving the efficiency of data recording. 2. Extremely high flexibility and versatility: It breaks away from the strong dependence of traditional solutions on preset data templates and can dynamically process any unknown type of data. Whether it is meeting notes, product parameters or knowledge cards in any field, this solution can adaptively understand and create a suitable storage structure for it, with unprecedented versatility and scalability. 3. Intelligent database self-management capability: It has the ability to intelligently manage database schema. By introducing a large language model for semantic judgment, it can autonomously decide whether to reuse existing similar tables or create new tables, realizing the "self-organization" and "self-evolution" of the database, effectively avoiding data redundancy, and ensuring the long-term rationality and consistency of data storage structure. 4. Optimized user interaction experience: The complex technical process is completely encapsulated behind simple user operations. Users do not need to care about the underlying logic such as how data should be classified or where it should be stored, providing a seamless and intuitive "instant capture and storage" experience, which greatly reduces the user's threshold and mental burden. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0017] The invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of an adaptive data table structure generation and data storage method according to Example 1. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments are not intended to limit the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0020] This invention provides an adaptive data table structure generation and data storage method and system to solve the technical problem of how to provide a solution that can automatically identify the information type in any image and intelligently and flexibly store it in a structured and persistent manner. Example

[0021] This invention provides an adaptive data table structure generation and data storage method, comprising five steps: data acquisition, information extraction, data transformation, matching analysis, and data storage.

[0022] Step S100 Data Acquisition: Acquire the image to be processed.

[0023] In a specific implementation, the images to be processed can be obtained from sources including real-time images captured by the camera of an electronic device, images stored in the photo album of an electronic device, and screenshots, such as a user taking a picture of a page in an encyclopedia of insects.

[0024] Step S200 Information Extraction: Extract unstructured text information from the image to be processed using optical character recognition methods.

[0025] The acquired images are processed through information extraction, and a mature OCR engine is used to recognize all the text in the images, forming one or more unstructured text segments. At this point, the text is unordered, for example, "Asian migratory locust Locustamigratoria, Orthoptera, Acrididae...".

[0026] Step S300 Data Conversion: Construct input instructions based on unstructured text information and preset Prompt instructions. Based on the input instructions, call the first major language model to perform type recognition and structure reorganization. Determine the data type through type recognition. Based on the data type, extract key attribute information from the unstructured text information through structure reorganization. Organize the key attribute information into structured data objects according to the format specified by the Prompt instructions and output them. The key attribute information includes name, parameters, and features.

[0027] The structured data object is in JSON format and includes a data type identifier, a main description, and attribute fields. The attribute fields correspond to key attribute information in the unstructured text.

[0028] As a specific implementation of data transformation, the unstructured text extracted by OCR is input into the LLM, and the LLM completes two tasks through preset prompts: (1) Identify the type: Determine the core theme of the text description, such as "insects", "recipe", "meeting minutes", etc.; (2) Generate structure: Based on the identified type, extract key information and organize it into a logically clear structured data object. The most common format is JSON.

[0029] Step S400 Matching Analysis: Extract the pattern information of the structured data objects. Based on the pattern information of the structured data objects and the table name annotation table of the target database, call the second major language model LLM to perform semantic-level matching analysis and search whether there are any suitable existing data tables in the target database.

[0030] In this embodiment, the schema information includes the key name, data type, and field meaning of the structured data object; during matching analysis, semantic matching analysis refers to comparing the semantic similarity between the field meaning of the structured data object and the field semantics of the existing data table, rather than just comparing the field names.

[0031] As a specific implementation of matching analysis, the generated structured data objects (or their schema information, such as all the key names) are input into the LLM. Through a preset prompt, the LLM performs a semantic-level check on the target database. This check goes beyond simply verifying the existence of tables with the same names; it interprets the meaning of the fields. For example, it determines whether fields like `commonName` and `scientificName` in the new data semantically match fields like `name_cn` and `name_latin` in a table named `insects` or `animal_records`. The LLM outputs a result: "Match" or "No Match".

[0032] Step S500 Data storage: If a matching existing data table exists in the target database, the structured data object is stored in the matching existing data table according to the semantic correspondence of the fields. If no matching existing data table is found in the target database, a table creation instruction for the matching database is generated based on the structured data table and the third language model is called. The table creation instruction is executed to create a new data table and the structured data object is stored in the new data table.

[0033] In this embodiment, the standard table creation command is the CREATE TABLE statement in Structured Query Language. The third large-scale language model automatically matches the field data type based on the field attributes of the structured data object and sets the primary key, index, and comment information. In the table creation command, the field data type adapts to the syntax specifications of the target database, and the primary key is selected from the key attribute information of the structured data object to ensure data uniqueness.

[0034] As a specific implementation of data storage, if the LLM determines that a usable data table exists, it stores the data from the generated JSON object into that existing data table according to the field mappings. Then the process ends. If no matching existing data table is found, the process of creating and storing the data is executed, specifically as follows: (1) Call LLM again, instructing it to generate a standard database instruction, such as the CREATE TABLE statement in SQL, based on the key name and data type (string, number, boolean, etc.) of the JSON object generated in step S103. (2) Execute this instruction generated by LLM to create a brand new data table in the database that is tailored for this type of data; (3) Store the data in the generated JSON object into the newly created data table, and the process ends.

[0035] The workflow of the method disclosed in this embodiment will be explained in detail through a specific example (insect encyclopedia image processing).

[0036] Step 1: Acquire Image Data: The user takes a photo of the page describing the "Asian Migratory Locust" in an insect encyclopedia using a smartphone. The image acquisition module 201 captures the photo.

[0037] Step 2: OCR Extraction of Unstructured Text Information: The photo is processed, and the identified text may be messy, for example: "Asian migratory locust, scientific name: Locusta migratoria, order: Orthoptera, family: Acrididae, length approximately 55 mm, diet: grasses, habitat: grasslands, farmland, agricultural pest."

[0038] Step 3: LLM identifies the type and generates structured data: Send the above text string, along with a preset prompt, to the LLM. This prompt can be designed as follows: "You are a data analysis expert. Please analyze the following text, determine its core entity type, and extract all its attributes into a JSON object with the following structure:" { "itemType": "The most likely subject of the image, in English", "describe": "A brief description of the subject, in Chinese". "attribute": "JSON structure, attributes of the image related to this subject." }

[0039] The text is as follows: '[Insert text extracted from S102 here]'.

[0040] After receiving the instruction, LLM analyzes it and returns a structured JSON object, as shown below: { "itemType": "insect", "describe": "Insects are the general term for animals belonging to the class Insecta within the phylum Arthropoda." "attribute": { "commonName": "Asian Migratory Locust", "scientificName": "Locusta migratoria", "order": "Orthoptera", "family": "Actidae" "size_mm": 55, "diet": ["Gramineae plants"], "habitat": ["grassland", "farmland"], "isPest": true }

[0041] Step 4: Does the LLM semantic search match the data table? Extract the pattern from the above JSON (i.e., all key names: itemType, attribute, ...), and send a second request to the LLM. This prompt can be designed as follows: "You are a database architect. In a database, the following tables exist: [contacts, recipes, meeting_notes]. Now I have a new data object with attribute fields [commonName, scientificName, order, family, ...]. Please determine if there is a suitable table to store this object. Please only answer with the most matching table name, or answer 'None'." Because the fields in the contacts table and other tables do not match the insect information at all, the LLM will return "None" after semantic judgment.

[0042] Step 5: Create a new table and store data: Since the previous step returned "None", the process enters the branch of creating a new table.

[0043] First, send a third request to the LLM, with the prompt designed as follows: "You are an SQL expert. Please generate a CREATE TABLE statement for your MySQL database based on the structure of the following JSON object. Please choose appropriate data types for the fields and use commonName as the primary key. The JSON is as follows: '[Insert JSON generated by S103 here]'" After receiving the instruction, the LLM generates an SQL statement, for example: CREATE TABLE `insects` ( `id` BIGINT UNSIGNED NOT NULL AUTO_INCREMENT COMMENT 'Unique Identifier ID', `common_name` VARCHAR(255) NOT NULL COMMENT 'common name', `scientific_name` VARCHAR(255) NOT NULL COMMENT 'scientific name', `order_name` VARCHAR(255) COMMENT 'Order (Biological Classification)', `family_name` VARCHAR(255) COMMENT 'Family (Biotaxation)', `size_mm` DECIMAL(10, 2) COMMENT 'Size (mm)', `diet` JSON COMMENT 'diet (stored as a JSON array)', `habitat` JSON COMMENT 'Habitat (stored as a JSON array)', `is_pest` BOOLEAN DEFAULT FALSE COMMENT 'Is it a pest?' `created_at` TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP COMMENT 'Record creation time', `updated_at` TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATECURRENT_TIMESTAMP COMMENT 'Record update time', PRIMARY KEY (`id`), UNIQUE KEY `uk_scientific_name` (`scientific_name`) ) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ciCOMMENT='Insect Information Table'; Then, the above SQL statement is received and executed in the target database DB, successfully creating a new table named inserts.

[0044] Finally, the generated JSON data is converted into an SQL INSERT statement, and the SQL INSERT statement is executed to store the complete information of the Asian migratory locust into the newly created inserts table.

[0045] The method in this embodiment leverages the powerful semantic understanding and instruction generation capabilities of large-scale language models to achieve fully automated processing of arbitrary image information. It not only intelligently transforms unstructured information in images into structured data, but its core lies in its ability to intelligently manage the database itself, autonomously determining through semantic analysis whether to reuse existing data tables or dynamically create new ones for storage. Example

[0046] This invention discloses an adaptive data table structure generation and data storage system, comprising a data acquisition module, an information extraction module, a data conversion module, a matching analysis module, and a data storage module.

[0047] The data acquisition module is used to perform the following: acquire the image to be processed.

[0048] In a specific implementation, the images to be processed can be obtained from sources including real-time images captured by the camera of an electronic device, images stored in the photo album of an electronic device, and screenshots, such as a user taking a picture of a page in an encyclopedia of insects.

[0049] The information extraction module is used to perform the following: extract unstructured text information from the image to be processed using optical character recognition methods.

[0050] The information extraction module uses a mature OCR engine to recognize all text in an image, forming one or more unstructured text segments. At this point, the text is unordered, for example, "Locusta migratoria, Orthoptera, Acrididae...".

[0051] The data conversion module performs the following actions: constructing input instructions based on unstructured text information and preset Prompt instructions; based on the input instructions, calling the first major language model for type recognition and structure reorganization; determining the data type through type recognition; extracting key attribute information from the unstructured text information based on the data type and through structure reorganization; and organizing the key attribute information into structured data objects according to the format specified by the Prompt instructions, whereby the key attribute information includes name, parameters, and features.

[0052] The structured data object is in JSON format and includes a data type identifier, a main description, and attribute fields. The attribute fields correspond to key attribute information in the unstructured text.

[0053] As a specific implementation of the data transformation module, this module is used to input the unstructured text extracted by OCR into the LLM. Through preset prompt instructions, the LLM completes two tasks: (1) Identify the type: Determine the core theme of the text description, such as "insects", "recipe", "meeting minutes", etc.; (2) Generate structure: Based on the identified type, extract key information and organize it into a logically clear structured data object. The most common format is JSON.

[0054] The matching analysis module is used to perform the following: extract the pattern information of structured data objects; based on the pattern information of structured data objects and the table name annotation table of the target database, call the second major language model LLM to perform semantic-level matching analysis; and search whether there are any suitable existing data tables in the target database.

[0055] In this embodiment, the schema information includes the key name, data type, and field meaning of the structured data object; during matching analysis, semantic matching analysis refers to comparing the semantic similarity between the field meaning of the structured data object and the field semantics of the existing data table, rather than just comparing the field names.

[0056] As a specific implementation of the matching analysis module, this module is used to input the generated structured data objects (or their schema information, such as all the "key" names) into the LLM. Through a preset prompt, the LLM performs a "semantic-level" check on the target database. This check goes beyond simply checking for the existence of tables with the same names; it interprets the meaning of the fields. For example, it determines whether fields such as commonName and scientificName in the new data semantically match fields such as name_cn and name_latin in a table named inserts or animal_records. The LLM outputs a result: "match" or "no match".

[0057] The data storage module performs the following actions: If a matching existing data table exists in the target database, the structured data object is stored in the matching existing data table according to the semantic correspondence of the fields. If no matching existing data table is found in the target database, a table creation instruction for the matching database is generated based on the structured data table and the third language model is called. The table creation instruction is executed to create a new data table and the structured data object is stored in the new data table.

[0058] In this embodiment, the standard table creation command is the CREATE TABLE statement in Structured Query Language. The third large-scale language model automatically matches the field data type based on the field attributes of the structured data object and sets the primary key, index, and comment information. In the table creation command, the field data type adapts to the syntax specifications of the target database, and the primary key is selected from the key attribute information of the structured data object to ensure data uniqueness.

[0059] As a specific implementation of the data storage module, if the LLM determines that a usable data table exists, it stores the data from the generated JSON object into that existing data table according to the field mappings. Then the process ends. If no matching existing data table is found, the process of creating and storing the data is executed, specifically as follows: (1) Call LLM again, instructing it to generate a standard database instruction, such as the CREATE TABLE statement in SQL, based on the key name and data type (string, number, boolean, etc.) of the JSON object generated in step S103. (2) Execute this instruction generated by LLM to create a brand new data table in the database that is tailored for this type of data; (3) Store the data in the generated JSON object into the newly created data table, and the process ends.

[0060] The adaptive data table structure generation and data storage method and system provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An adaptive data table structure generation and data storage method, characterized in that, Includes the following steps: Data acquisition: Acquiring the image to be processed; Information extraction: Extracting unstructured text information from the image to be processed using optical character recognition methods; Data Conversion: Input instructions are constructed based on unstructured text information and preset Prompt instructions. Based on the input instructions, the first major language model is called to perform type recognition and structure recombination. The data type is determined by type recognition. Based on the data type, key attribute information is extracted from the unstructured text information through structure recombination. The key attribute information is organized into structured data objects according to the format specified by the Prompt instructions and output. The key attribute information includes name, parameters and features. Matching analysis: Extract the pattern information of structured data objects, and based on the pattern information of structured data objects and the table name annotation table of the target database, call the second largest language model LLM to perform semantic matching analysis to search whether there are any existing data tables in the target database that are suitable. Data storage: If a matching existing data table exists in the target database, the structured data object is stored in the matching existing data table according to the semantic correspondence of the fields. If no matching existing data table is found in the target database, a table creation instruction for the matching database is generated based on the structured data table and the third language model is called. The table creation instruction is executed to create a new data table and the structured data object is stored in the new data table.

2. The adaptive data table structure generation and data storage method according to claim 1, characterized in that, When acquiring images to be processed, the image sources include images captured in real time by the camera of the electronic device, images already stored in the photo album of the electronic device, and screenshots.

3. The adaptive data table structure generation and data storage method according to claim 1, characterized in that, The structured data object is in JSON format and contains a data type identifier, a main description, and attribute fields. The attribute fields correspond to key attribute information in the unstructured text.

4. The adaptive data table structure generation and data storage method according to claim 1, characterized in that, Schema information includes the key names, data types, and field meanings of structured data objects; In matching analysis, semantic matching analysis refers to comparing the semantic similarity between the field meanings of structured data objects and the fields of existing data tables, rather than simply comparing field names.

5. The adaptive data table structure generation and data storage method according to claim 1, characterized in that, The standard table creation command is the CREATE TABLE statement in Structured Query Language. The third large language model automatically matches the field data type based on the field attributes of the structured data object and sets the primary key, index, and comment information. In the table creation command, the field data types are adapted to the syntax specifications of the target database, and the primary key is selected from the key attribute information of the structured data object to ensure data uniqueness.

6. An adaptive data table structure generation and data storage system, characterized in that, It includes a data acquisition module, an information extraction module, a data conversion module, a matching analysis module, and a data storage module; The data acquisition module is used to perform the following: acquire the image to be processed; The information extraction module is used to perform the following: extract unstructured text information from the image to be processed using optical character recognition methods; The data conversion module is used to perform the following: construct input instructions based on unstructured text information and preset Prompt instructions; based on the input instructions, call the first language model to perform type recognition and structure recombination; determine the data type through type recognition; extract key attribute information from the unstructured text information based on the data type and through structure recombination; and organize the key attribute information into structured data objects according to the format specified by the Prompt instructions and output them. The key attribute information includes name, parameters, and features. The matching analysis module is used to perform the following: extract the pattern information of structured data objects, and based on the pattern information of structured data objects and the table name annotation table of the target database, call the second largest language model LLM to perform semantic matching analysis and search whether there is a suitable existing data table in the target database. The data storage module performs the following actions: If a matching existing data table exists in the target database, the structured data object is stored in the matching existing data table according to the semantic correspondence of the fields. If no matching existing data table is found in the target database, a table creation instruction for the matching database is generated based on the structured data table and the third language model is called. The table creation instruction is executed to create a new data table and the structured data object is stored in the new data table.

7. The adaptive data table structure generation and data storage system according to claim 6, characterized in that, When acquiring images to be processed, the image sources include images captured in real time by the camera of the electronic device, images already stored in the photo album of the electronic device, and screenshots.

8. The adaptive data table structure generation and data storage system according to claim 6, characterized in that, The structured data object is in JSON format and contains a data type identifier, a main description, and attribute fields. The attribute fields correspond to key attribute information in the unstructured text.

9. The adaptive data table structure generation and data storage system according to claim 6, characterized in that, Schema information includes the key names, data types, and field meanings of structured data objects; Semantic matching analysis refers to comparing the semantic similarity between the field meanings of a structured data object and the field semantics of an existing data table, rather than simply comparing field names.

10. The adaptive data table structure generation and data storage system according to claim 6, characterized in that, The standard table creation command is the CREATE TABLE statement in Structured Query Language. The third large language model automatically matches the field data type based on the field attributes of the structured data object and sets the primary key, index, and comment information. In the table creation command, the field data types are adapted to the syntax specifications of the target database, and the primary key is selected from the key attribute information of the structured data object to ensure data uniqueness.

Images