Large model training method, data report development method, equipment and storage medium

Abstract

The embodiment of the invention discloses a large model training method, a data report development method, computer equipment and a computer readable storage medium. According to the large model training method, a training set is used for training an original large model for different training scenes and corresponding training records; and marking the trained original large model as a report development large model, wherein the report development large model is composed of a first agent, a second agent and a third agent. The data report development method comprises the following steps: inputting a natural language into a trained report development large model to process and generate a development script, executing report processing operation according to the development script to obtain an initial report, and finally generating a data report after display verification. Therefore, non-developers can also develop and maintain the data report.

Description

Technical Field

[0001] The present application belongs to the field of data processing technology, and in particular relates to a large model training method, a data report development method, a computer device, and a computer-readable storage medium. Background Art

[0002] Data report generation typically begins with the business department submitting a report requirement. This requirement is then handed over to technical development for code development. After a series of data processing steps, the business department verifies the report is successful and launches it online. The entire report development process is handled by database, back-end, and front-end engineers, who write the code. This process is invisible to the business side, like a black box: After providing the requirements, the report is delivered. After a period of waiting, the business side receives a report that doesn't meet their needs, requiring development and revisions. After another wait, they receive a half-baked report, which requires further development and revisions. This cycle repeats, ultimately resulting in a barely acceptable report. If the business changes again, the report must be redeveloped. This process fails to meet business needs in terms of both satisfaction and timeliness. Improving the user-friendliness of data reports for non-developers, enabling them to develop and maintain them, is a pressing technical challenge facing those skilled in the art.

[0003] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0004] Based on this, it is necessary to address the above problems and propose a large model training method, data report development method, computer equipment and computer-readable storage medium, which can enable non-developers to develop and maintain data reports.

[0005] The present application solves the technical problem by adopting the following technical solutions: The present application provides a large model training method, comprising the following steps: obtaining an original large model and a training record set, the training record set including multiple training records; determining a training scenario corresponding to each training record, the training scenario including at least one of a data extraction scenario, a data configuration scenario, and a report design scenario; training the original large model according to different training scenarios and corresponding training records; marking the trained original large model as a report development large model, the report development large model consisting of a first intelligent agent, a second intelligent agent, and a third intelligent agent, and used to generate a script based on the natural language input by the user to assist in completing data report development.

[0006] In an optional embodiment of the present application, each training record consists of a tuple of a training language and an output script, and the training language is a natural language; the original large model is trained according to different training scenarios and corresponding training records, including: taking the training language as input and the output script corresponding to the training language as output, forming a set of input-output pairs to fine-tune the original large model; the original large model trained by the training record corresponding to the data extraction scenario is marked as a first intelligent agent, and the first intelligent agent is used to generate a script for accessing the data source according to the input training language; the original large model trained by the training record corresponding to the data configuration scenario is marked as a second intelligent agent, and the second intelligent agent is used to generate a script for configuring the database table according to the input training language; the original large model trained by the training record corresponding to the report design scenario is marked as a third intelligent agent, and the third intelligent agent is used to generate a script for designing data reports according to the input training language.

[0007] The present application also provides a data report development method, comprising the following steps: when natural language input by a user is obtained, the natural language is input into a report development model, which is trained according to the method provided above; a development script obtained by the report development model based on natural language processing is obtained, and report processing operations are performed according to the development script, the report processing operations including data extraction operations, data configuration operations and report design operations; a display verification operation is performed on the initial report obtained after executing the report processing operation; the initial report that passes the display verification operation is marked as a display report and output.

[0008] In an optional embodiment of the present application, when natural language including user requirements is obtained, the method also includes: extracting target requirements from the natural language, the target requirements are used to characterize the data report that the user wants to build; searching from the historical script library whether there is a matching historical record based on the target requirements, the historical record consists of a tuple of historical requirements and historical scripts, the historical requirements are target requirements previously proposed by the user, and the historical scripts are development scripts generated when the report development model creates a corresponding display report based on the historical requirements; if there is a historical record matching the target requirement in the historical script library, the historical script in the corresponding historical record is marked as a development script to execute the report processing operation; if there is no historical record matching the target requirement in the historical script library, the natural language is input into the report development model.

[0009] In an optional embodiment of the present application, the data extraction operation includes: determining the target data source by processing natural language based on a report development model; obtaining an extraction script obtained by a first intelligent agent based on natural language processing, where the extraction script is a type of development script; using the extraction script to build a connection with the target data source, and extracting target data from the target data source; and storing the target data in an analysis library.

[0010] In an optional embodiment of the present application, the data configuration operation includes: obtaining target data from the analysis library, the target data being obtained according to the data extraction operation; obtaining the configuration script and table structure rule design obtained by the second intelligent agent according to natural language processing, the configuration script being a type of development script; processing the target data according to the configuration script and table structure rule design to obtain an initial table structure, the initial table structure records the relationship between fields in the target data, and the structure and rules between the tables that record the fields; and storing the target data in the analysis library in the form of the initial table structure.

[0011] In an optional embodiment of the present application, the report design operation includes: obtaining target data from the analysis library, the target data is obtained according to the data extraction operation, and stored in the form of an initial table structure obtained according to the data configuration operation; obtaining a design script obtained by a third intelligent agent based on natural language processing, and a design template determined according to the natural language, the design script is a type of development script; processing the target data according to the design script and the design template to obtain an initial report.

[0012] In an optional embodiment of the present application, the display verification operation includes: pre-executing the initial report to determine whether the initial report is configured successfully; if the configuration is successful, extracting the verification data set from the initial report according to the preset extraction rules, and displaying it to the user; if the configuration fails, or a cancel instruction is received from the user, deleting the initial report, and re-obtaining the development script obtained by the report development model based on natural language processing, and executing the report processing operation according to the development script; if a save instruction is received from the user, it is determined that the initial report has passed the display verification operation.

[0013] The present application also provides a computer device, comprising a processor and a memory: the processor is configured to execute a computer program stored in the memory to implement the aforementioned method.

[0014] The present application also provides a computer-readable storage medium storing a computer program, which implements the aforementioned method when the computer program is executed by a processor.

[0015] The embodiments of the present application have the following beneficial effects: This application can pre-train the big model to obtain a report development big model specifically for data report development. The report development big model is then used to convert the user's natural language into a development script for data report development, and verification is performed after generation and before display to ensure the availability of the data report. The entire development process does not require coding or reading code, which not only shortens the development cycle, but also allows business parties to participate in the entire process, and even allow the business to develop new reports on their own. The entire report development process is represented and stored in a visual, graphical, and templated manner, which is not only convenient and fast during development, but also intuitive and easy to understand during maintenance and updates, greatly improving efficiency.

[0016] The above description is only an overview of the technical solution of this application. In order to more clearly understand the technical means of this application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of this application more obvious and easy to understand, the following preferred embodiments are specifically described in detail with reference to the accompanying drawings. It should be understood that the above general description and the detailed description below are only exemplary and explanatory and do not limit this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 A flowchart of a large model training method provided in one embodiment.

[0019] Figure 2 A flowchart of a data report development method provided in one embodiment.

[0020] Figure 3 A flowchart of a data report development method provided in one embodiment.

[0021] Figure 4 The present invention is a schematic block diagram of the structure of a computer device provided by an embodiment. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0023] The development of existing data reports is a complete black box for business personnel without development experience. If business personnel need to change requirements or maintain tables, they need to wait for developers to develop again, which is inefficient. How to enable non-developers without development experience or development capabilities to develop and maintain data reports? This application proposes a large model training method and a data report development method. The report development large model obtained by the large model training method can be used in the data report development method to generate development scripts using the user's natural language to assist in completing data report development. In order to clearly describe the method provided in this embodiment, please refer to Figures 1 to 3 The large model training method includes steps S110~S130, and the data report development method includes steps S210~S230.

[0024] This application provides a method for enabling non-developers to develop and maintain data reports. Non-developers are people who have no development experience and cannot use programming languages ​​for development and maintenance. With the recent leap in large model (also known as large language model, Large Language Model, LLM) technology, a groundbreaking path has emerged in the interactive form of natural language. Unlike traditional translation models obtained through supervised learning training, large models have demonstrated remarkable emergent capabilities. This capability exists in ultra-large-scale models with more than 100 billion parameters, enabling the model to seamlessly translate natural language into programming language. However, in order to enable large models to achieve faster, more accurate and efficient conversion from natural language to programming language to assist in the completion of data report development, this application proposes a large model training method.

[0025] Step S110: Obtain the original large model and a training record set, where the training record set includes multiple training records.

[0026] In one embodiment, the original large model is an open source large model that can be deployed on a designated device, including but not limited to glm-4-9b-chat, internlm2_5-chat-7b, chatglm3-6b, Qwen2.5-7B-Instruct, etc. This application does not limit the specific open source large model to be used. Whether listed or not, as long as it can be deployed on a designated device and can complete natural language to programming language translation, it is considered an original large model in this application.

[0027] The training record set includes multiple training records. Each training record consists of a pair of training language and output script. The training language is a natural language, and the output script is compiled from the programming language translated according to the requirements in the training language. It can complete data configuration by running in a specific scenario.

[0028] Step S120: Determine a training scenario corresponding to each training record, where the training scenario includes at least one of a data extraction scenario, a data configuration scenario, and a report design scenario.

[0029] In one embodiment, data report development can be divided into three phases: data extraction, data configuration, and report design. Data extraction involves extracting the required target data from a specific data source; data configuration involves configuring the target data according to pre-set relationships; and report design involves displaying the configured target data in a specified manner. Therefore, the required operations vary across different scenarios, and the corresponding training records and scripts must also perform different operations.

[0030] Different training records are used for different scenarios. For example, some training records specify how to extract data, while others describe how to configure data and the relationships between tables in the target data. Therefore, it is necessary to distinguish training records so that different training records can be trained separately for different training scenarios. This will generate different sub-models and subsequently generate scripts required for different scenarios.

[0031] The process of determining the training scenario corresponding to the training record can be: extracting target keywords from natural language, matching corresponding configuration scenarios according to the target keywords, and the configuration scenarios include at least one of data extraction scenarios, data configuration scenarios and report design scenarios.

[0032] Step S130: Train the original large model according to different training scenarios and corresponding training records; mark the trained original large model as the report development large model.

[0033] In one embodiment, the original large model is trained according to different training scenarios and corresponding training records, including: taking the training language as input and the output script corresponding to the training language as output, forming a set of input-output pairs to fine-tune the original large model; marking the original large model trained by the training records corresponding to the data extraction scenario as a first intelligent agent, and the first intelligent agent is used to generate a script for accessing the data source according to the input training language; marking the original large model trained by the training records corresponding to the data configuration scenario as a second intelligent agent, and the second intelligent agent is used to generate a script for configuring the database table according to the input training language; marking the original large model trained by the training records corresponding to the data extraction scenario as a third intelligent agent, and the third intelligent agent is used to generate a script for designing data reports according to the input training language.

[0034] In one embodiment, the training process first uses the training language as input and the output script corresponding to the training language as output, forming a set of input-output pairs to fine-tune the original large model for training.

[0035] Secondly, as mentioned above, the training records correspond to different training scenarios. Therefore, during the training process of the original large model, the original large model needs to be trained separately according to different training scenarios and their corresponding training records to obtain multiple sub-models, including the first intelligent agent, the second intelligent agent and the third intelligent agent.

[0036] The original large model trained by the training records of the corresponding data extraction scenario is marked as the first intelligent agent. The first intelligent agent is used to generate a script for accessing the data source based on the input training language, which is later called the extraction script.

[0037] The original large model trained by the training records of the corresponding data configuration scenario is marked as the second intelligent agent. The second intelligent agent is used to generate a script for configuring the database table based on the input training language, which is later referred to as the configuration script.

[0038] The original large model trained by the training records of the corresponding data extraction scenario is marked as the third intelligent agent. The third intelligent agent is used to generate a script for designing data reports based on the input training language, which is later called a design script.

[0039] Complete the training of the original large model, especially focusing on specific prompt words in the training records and professional terminology in the rules. Verify and correct the results obtained by the original large model based on corresponding and similar prompt words to ensure that the large model can complete its corresponding functions efficiently and accurately.

[0040] During the training process of the large model, taking the data extraction scenario as an example, a dialog scenario can be pre-created using prompts to automatically generate a database driver connection script based on the user's natural language input. After constructing this dialog scenario, the original large model is fed the training record of the data extraction scenario, including the natural language required to connect to the database. The original large model then automatically generates a training script to connect to the inferred data source. The training script is then automatically loaded into the database driver engine and executed to complete the connection to the inferred database. Specifically, the original large model receives information such as the description text and link address of each database in advance, infers the required database connection based on the user's natural language input, and writes the database connection script instructions based on the given execution environment. Training is then repeated through fine-tuning until the training conditions are met, at which point the original large model is marked as the first agent.

[0041] The first agent, the second agent and the third agent are aggregated to obtain a large model for report development.

[0042] After the above method is used, the large model can be trained. The report development large model can process the input natural language into a script to assist in the development of data reports. Regarding data report development, this application proposes a data report development method. The equipment involved in this data report development method includes, but is not limited to, the following: 1) a computer; 2) a Java virtual machine; 3) a MySQL database; 4) a database link driver; and 5) V-Chart. The computer is used to run the Java virtual machine, a relational database management system, and configure PC programs; the Java virtual machine is used to compile and run Java programs; the MySQL database is used to store extracted data; and V-Chart is used to build visualizations.

[0043] Step S210: When the natural language input by the user is obtained, the natural language is input into the report development model.

[0044] In one embodiment, the natural language input by the user includes user requirements and the type of data the user needs to extract from the data source, how the relationships between the extracted target data should be configured, and how the configured target data is converted into a data report for presentation. Typically, developers are required to create multiple scripts for extraction, configuration, and design, making this infeasible for non-developers. However, this application utilizes the large report development model trained using the method provided above, converting the natural language into a script for development.

[0045] In one embodiment, step S210 generates reports for a report development model that needs to be trained based on natural language. In actual processing, the user may have performed similar operations before and has developed data reports for the same data. What the user is currently doing is just to maintain or modify the content in the data report. That is to say, there are corresponding historical records in the data report development model, which records the natural language that the user has used and the scripts used for report development based on the natural language. In this case, there is no need to generate additional scripts, but directly use the scripts that already exist in the historical records to develop data reports. For this embodiment, step S210 will be replaced by S310 to S340. To clearly describe the process of the data report development method under this embodiment, please refer to Figure 3 .

[0046] Step S310: When natural language including user requirements is acquired, target requirements are extracted from the natural language.

[0047] In one embodiment, the target requirement is used to represent the data report that the user wants to build. In other words, it may include what kind of data to extract from what data source, how the relationship between the extracted target data should be configured, and how the configured target data should be converted into a data report for presentation.

[0048] Step S320: searching the history script library to see if there is a matching history record according to the target requirement.

[0049] In one embodiment, historical records consist of a pair consisting of a historical requirement and a historical script. A historical requirement is a target requirement previously proposed by a user, and a historical script is a development script generated when the report development model creates a corresponding presentation report based on the historical requirement. The historical script library is a pre-configured database that stores all natural language used in all data report development processes prior to the current data development, along with the corresponding scripts that successfully completed the data report development.

[0050] The target requirement is retrieved from the historical records. The determination process for whether a historical record exists can be to calculate the similarity between the target requirement and each historical record. Based on the similarity, a determination is made as to whether a historical record matching the target requirement exists in the historical script library. If the target requirement and the historical record are above a preset similarity threshold, the target requirement and the historical record are considered identical, indicating a matching historical record exists. Otherwise, a matching historical record is considered not to exist.

[0051] If there is a historical record matching the target requirement in the historical script library, step S330 is executed: marking the historical script in the corresponding historical record as a development script to perform a report processing operation.

[0052] In one embodiment, if a matching historical record exists, it can be determined that the historical requirements in the historical record and the current target requirements are identical or similar, and the same script can be directly used for data report development. Therefore, there is no need to generate a separate script; instead, the historical script in the historical record is directly used as the development script for subsequent report processing operations. Step S330 directly skips step S220, and execution continues directly with step S230.

[0053] If there is no historical record matching the target requirement in the historical script library, step S340 is executed: inputting the natural language into the report development model.

[0054] In one embodiment, if it does not exist, this indicates that the current target demand is a new demand. Therefore, it is necessary to execute the original technical solution of this application, that is, input the natural language into the report development model in step S210, and then continue with the original steps S220~S230.

[0055] Step S220: Obtain a development script obtained by natural language processing of the report development model, and perform report processing operations according to the development script. The report processing operations include data extraction operations, data configuration operations, and report design operations.

[0056] In one embodiment, as described above, the report development equation is divided into different configuration scenarios, each of which corresponds to a sub-model. Each sub-model will generate different development scripts based on the input natural language. In the data extraction scenario, an extraction script will be generated based on the first agent; in the data configuration scenario, a configuration script will be generated based on the second agent; and in the report design scenario, a design script will be generated based on the third agent. Different report processing operations will be performed based on different development scripts. Specifically, the extraction script is used to perform data extraction operations; the configuration script is used to perform data configuration operations; and the design script is used to perform report design operations. The specific execution process of each operation will be described in detail later.

[0057] In one embodiment, the data extraction operation includes: determining the target data source by processing natural language based on a report development model; obtaining an extraction script obtained by a first intelligent agent based on natural language processing, where the extraction script is a type of development script; using the extraction script to build a connection with the target data source, and extracting target data from the target data source; and storing the target data in an analysis library.

[0058] In one embodiment, the data extraction operation primarily aggregates the target data from the target data source into wide tables or dimension tables according to a specific business model, and then stores the extracted data in an analysis repository. The first agent can determine, based on natural language input, the required connections and configurations for various homogeneous and heterogeneous target data sources. Based on the target data source's description text and connection address, and in a given execution environment, it then writes instructions for a database connection script to generate an extraction script. The extraction script automatically triggers a connection test for the corresponding database driver engine, saves the successful connection configuration, and converts the configuration information and the connection generated by the driver engine into source code for storage via a code engine. The driver engine can implement its functionality by repackaging the MySQL database driver Jar package: mysql-connector-java, or the Oracle database driver Jar package: oracle.jdbc.driver.OracleDriver. The code engine can utilize the open source tool JHipster to quickly create web applications based on Maven, Spring, and Vue. The specific choice is not limited; any engine that can implement the corresponding functionality is included within the driver or code engine. Furthermore, all manual operations can be retained to facilitate subsequent configuration and modification of data reports.

[0059] The extraction script is automatically loaded into the database driver engine for execution to complete the connection to the inference database. After the connection is established, the target data can be extracted from the target data source. The target data is stored in the analysis library according to the wide table and dimension table information designed by the user for direct or indirect use in subsequent reports. Wide table: A collection of all attributes of a type of entity. For example, an employee wide table, most of the information of an employee in the company can be placed in this table, which can include basic employee information (name, gender, etc.); company information (department, position, etc.); authority information (region, etc.); usage records (first, most recent, etc.). Dimension table: A collection of enumeration values ​​that will change accordingly with business development. For example, a table that records postal information; a table of provincial and municipal information; a table of product line information.

[0060] The big model provides configuration of various homogeneous and heterogeneous data sources according to the user's description for the user to modify and confirm, and the big model generates scripts. In one embodiment, the data configuration operation includes: obtaining target data from an analysis library, where the target data is obtained according to a data extraction operation; obtaining a configuration script and a table structure rule design obtained by a second intelligent agent according to natural language processing, where the configuration script is a type of development script; processing the target data according to the configuration script and the table structure rule design to obtain an initial table structure, where the initial table structure records the relationship between fields in the target data, and the structure and rules between tables that record fields; and storing the target data in the analysis library in the form of a converted initial table structure.

[0061] In one embodiment, a configuration script is used to link the data source table fields to the wide table fields. The table structure rules are designed to determine the table structure storage design requirements of the target data from the natural language input by the user. Data processing can also be performed according to business logic based on business needs and business fields for the data model, or an intermediate table can be redesigned between the wide table and the dimension table in the previous step to integrate the data in this field into one table and store it in the report library for statistical analysis and caliber maintenance. Both wide tables and dimension tables are intermediate tables. This is used to distinguish them. Aggregating and integrating data that meets business needs into one table can also be called a business report, which effectively improves the access efficiency of the report and enhances performance.

[0062] The second agent analyzes the natural language input from the user to determine the tables (wide tables and dimension tables) used to store data and configure simple information, such as the name, description, and design of table fields, types, and instructions. Based on these prompts, the second agent generates a configuration script and table structure rules, automatically completing the configuration. The report development system's API accepts the configuration script and automatically loads it into the configuration module to run, completing the configuration of the database table to be developed.

[0063] Furthermore, during data configuration, the conversion process for table relationship rule configuration—that is, the process of determining table structure rule design—retrieves the original index fields and wide table fields in the target data, extracts the primary table and detail tables of the source table's primary key, and visually displays them to the user. Users can connect lines and other operations based on the displayed forms to determine the relationships between forms, such as the primary table containing the corresponding primary key of the data source table and the detail table fields containing the foreign key, thereby determining the table structure rule design, that is, the configuration rules and constraints for the table.

[0064] Because this application method is intended for non-developers, it can be validated after the user manually configures the table structure rules. If the validation is successful, the configured relationship information is converted into rule scripts and code, and the newly generated relationships in the user-configured relationships are added to the front-end and displayed to the user. The corresponding relationships, constraints, rules, etc. are configured on the connection. If the validation fails, the user is notified of the configuration error and asked to make adjustments.

[0065] The data configuration operation is similar, except that wide tables and dimension tables are used as new data sources to extract data that meets business needs and store it in the report library to provide it for report queries. This ensures the efficiency of report queries and improves the performance of report access.

[0066] In one embodiment, the report design operation includes: obtaining target data from an analysis library, the target data is obtained according to a data extraction operation, and is stored in an initial table structure obtained according to a data configuration operation; obtaining a design script obtained by a third intelligent agent according to natural language processing, and a design template determined according to natural language, the design script being a type of development script; processing the target data according to the design script and the design template to obtain an initial report.

[0067] In one implementation, after completing the preceding operations, the data in the intermediate table needs to be presented to different users using various report formats. For users, data report development also requires only three steps: template selection, data source configuration, and preview and publish. However, because this design utilizes report templates and integrates data generated from the preceding operations, this operation also requires the use of a large report development model.

[0068] First, the user selects the design template, that is, selects the appropriate report template for presenting and carrying data.

[0069] The next step is to configure the data source: assign appropriate data sources to the corresponding layout areas, charts, and other components in the template. This data source can be a wide table or dimension table in the analysis library, or an intermediate table (business report) in the report library. Specifically, the user interface is divided into three areas: the left area for the wide table or dimension table and fields; the right area for the data source table and fields; and the center area for the rule configuration area. Users can first connect the relationships between the left and right tables and fields with lines, and then configure the rules and relationships online. Users can use predefined prompts (left table, right table, rules, and relationship descriptions) to inform the report development master model. The third agent in the report development master model generates a script to complete the user's original operation. Users can make adjustments to the results generated by the master model, and the master model will modify the script based on the user's adjustments. The third agent parses the user's natural language input and automatically generates a link script, also known as a design script, that links the data source table fields to the wide table fields. By running the design script, the table field relationships are connected. Specifically, the third agent must receive in advance the description information of the fields of the database table of the data source output by the first agent, as well as the configuration information of the database table to be created during the data configuration operation performed by the second agent, and the API interface of the design module. At the same time, it is given instructions to infer the required fields of the database table to be created based on the user's natural language, and the relationship between them and the fields in the database table of the data source, and to write a field link script based on the given execution environment.

[0070] Finally, in the preview and publishing stage, the template style and data source configuration information selected by the user are obtained, the required parameters in the report template are assigned values, the corresponding data retrieval interface generates code, a lazy loading interface is generated and the report data is preloaded, all the contents of the report are filled in, and the complete page code is generated for browser compilation, key information is cached in cookies, and the page loading is completed. The preview can display the corresponding stylish and substantial data report page, and finally the report is published to obtain the initial report.

[0071] Step S230: performing a display verification operation on the initial report obtained after the report processing operation is performed; marking the initial report that passes the display verification operation as a display report and outputting it.

[0072] In one embodiment, the display verification operation includes: pre-executing the initial report to determine whether the initial report is successfully configured; if the configuration is successful, extracting the verification data set from the initial report according to the preset extraction rules, and displaying it to the user; if the configuration fails, or a cancel instruction is received from the user, deleting the initial report, and re-obtaining the development script obtained by the report development model based on natural language processing, and executing the report processing operation according to the development script; if a save instruction is received from the user, it is determined that the initial report has passed the display verification operation.

[0073] In one embodiment, the data report obtained according to the series of steps S220 is an initial report. After all, since it is a data report created by non-developers using a large report development model, the actual presentation effect may need to be improved. Therefore, before actual release, it needs to be displayed and confirmed by the user again, and the functionality needs to be verified by the user, that is, the presentation verification operation needs to be performed.

[0074] The process consists of two phases: pre-execution and save-execution. First, the initial report is pre-executed to determine whether the report can be rendered properly. If the configuration is successful, the user will see that the system has extracted a validation dataset according to the pre-set extraction rules. The validation dataset includes a preset number of data items (for example, 100). This validation dataset is displayed to the user for confirmation. Based on the validation dataset display, the user can choose from three options: Cancel, Execute, or Save. If the command is canceled, or if the configuration fails as previously determined, the initial report is deleted, and the development script generated by the report development model based on natural language processing is retrieved. Report processing operations are executed based on the development script, and the report is regenerated. If the command is executed, the initial report is pre-executed again, and the validation dataset is retrieved and displayed again for user confirmation. If the command is confirmed, the user-configured parameters are converted into logic code using the rule engine and pushed to the timer for automatic backend data acquisition and storage in the analysis database. The initial report is marked as a display report for output, completing the report development.

[0075] Therefore, this application can pre-train the large model to obtain a report development large model specifically for data report development. The report development large model is then used to convert the user's natural language into a development script for data report development, and verification is performed after generation and before display to ensure the availability of the data report. The entire development process does not require coding or reading code, which not only shortens the development cycle, but also allows business parties to participate in the entire process, and even allow the business to develop new reports on their own. The entire report development process is represented and stored in a visual, graphical, and templated manner, which is not only convenient and fast during development, but also intuitive and easy to understand during maintenance and updates, greatly improving efficiency.

[0076] In order to lower the technical level and threshold of users, the report design process in this solution is represented and expressed in the form of lines, which intuitively and clearly shows the relationship between tables and fields; and the relationship is directly marked on the lines, which is not only convenient for design, but also conducive to subsequent maintenance and updates.

[0077] Figure 4 FIG1 shows an internal structure diagram of a computer device in an embodiment. The computer device can be a terminal or a server. Figure 4As shown, the computer device includes a processor, a memory and a network interface connected via a system bus. Among them, the memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement a large model training method and / or a data report development method. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement a large model training method and / or a data report development method. It can be understood by those skilled in the art that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0078] In one embodiment, the present application further proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in any of the aforementioned embodiments.

[0079] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0080] It should be noted that, in this document, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context of the specific embodiment.

[0081] It should be understood that although the terms "first," "second," "third," etc. may be used herein to describe various information, such information should not be limited to these terms. These terms are used solely to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the term "if," as used herein, may be interpreted as "upon," "when," or "in response to a determination." Furthermore, as used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It should be further understood that the terms "comprising" and "including" indicate the presence of the recited features, steps, operations, elements, components, items, types, and / or groups, but do not preclude the presence, occurrence, or addition of one or more other features, steps, operations, elements, components, items, types, and / or groups. The terms "or," "and / or," "including at least one of the following," etc., as used herein, may be interpreted as inclusive, meaning any one or any combination. For example, “comprising at least one of the following: A, B, C” means “any of the following: A; B; C; A and B; A and C; B and C; A and B and C”; and for another example, “A, B or C” or “A, B and / or C” means “any of the following: A; B; C; A and B; A and C; B and C; A and B and C”. An exception to this definition will occur only when a combination of elements, functions, steps or operations are inherently mutually exclusive in some manner.

[0082] It should be understood that, although the various steps in the flowchart in the embodiment of the present application are shown in sequence according to the indication of the arrows, these steps are not necessarily performed in sequence in the order indicated by the arrows. Unless clearly stated herein, the execution of these steps is not strictly limited in order, and they can be performed in other orders. Moreover, at least a portion of the steps in the figure may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and their execution order is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0083] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0084] It should be noted that in this article, step codes such as S10 and S20 are used for the purpose of expressing the corresponding content more clearly and concisely, and do not constitute a substantial limitation on the order. When implementing the step, those skilled in the art may execute S20 first and then S10, etc., but these should all be within the scope of protection of this application.

[0085] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0086] In the subsequent description, the use of suffixes such as "module", "component" or "unit" to represent elements is only for the purpose of facilitating the description of the present application and has no specific meaning. Therefore, "module", "component" or "unit" can be used interchangeably.

[0087] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A large model training method, characterized in that: The steps include: Obtaining an original large model and a training record set, wherein the training record set includes multiple training records; Determining a training scenario corresponding to each training record, wherein the training scenario includes at least one of a data extraction scenario, a data configuration scenario, and a report design scenario; Training the original large model according to different training scenarios and corresponding training records; The trained original large model is marked as a report development large model. The report development large model is composed of a first intelligent agent, a second intelligent agent and a third intelligent agent, and is used to generate scripts based on the natural language input by the user to assist in completing data report development.

2. The large model training method according to claim 1, characterized in that: Each training record consists of a pair of training language and output script, and the training language is a natural language; The training of the original large model according to different training scenarios and corresponding training records includes: Taking the training language as input and the output script corresponding to the training language as output, a set of input-output pairs is formed to fine-tune the original large model; Marking the original large model trained by the training record corresponding to the data extraction scenario as the first agent, wherein the first agent is used to generate a script for accessing a data source according to an input training language; Marking the original large model trained by the training record corresponding to the data configuration scenario as the second agent, wherein the second agent is used to generate a script for configuring a database table according to an input training language; The original large model trained by the training record corresponding to the report design scenario is marked as the third intelligent agent, and the third intelligent agent is used to generate a script for designing a data report based on the input training language.

3. A data report development method, characterized in that: The steps include: When natural language input by the user is obtained, the natural language is input into the report development model, wherein the report development model is trained according to the method according to any one of claims 1 to 2; Obtaining a development script obtained by the report development model according to the natural language processing, and performing report processing operations according to the development script, wherein the report processing operations include data extraction operations, data configuration operations, and report design operations; Performing a display verification operation on the initial report obtained after executing the report processing operation; marking the initial report that passes the display verification operation as a display report and outputting it.

4. The data report development method according to claim 3, wherein: When the natural language including the user's needs is obtained, the method further includes: Extracting target requirements from the natural language, where the target requirements are used to represent a data report that the user wants to build; Searching the historical script library for matching historical records based on the target requirement, wherein the historical records consist of a pair of historical requirements and historical scripts. The historical requirements are target requirements previously proposed by the user, and the historical scripts are development scripts generated when the report development model creates a corresponding presentation report based on the historical requirements. If there is a historical record matching the target requirement in the historical script library, the corresponding historical script in the historical record is marked as a development script to execute the report processing operation; If there is no historical record matching the target requirement in the historical script library, the natural language is input into the report development model.

5. The data report development method according to claim 3, wherein: The data extraction operation includes: Developing a large model based on the report to process the natural language and determine the target data source; Obtaining an extraction script obtained by the first agent according to the natural language processing, wherein the extraction script is a type of the development script; Using the extraction script to establish a connection with the target data source, and extracting target data from the target data source; The target data is stored in an analysis library.

6. The data report development method according to claim 3, wherein: The data configuration operation includes: Acquiring target data from an analysis library, wherein the target data is obtained according to the data extraction operation; Obtaining a configuration script and a table structure rule design obtained by the second agent according to the natural language processing, wherein the configuration script is a type of the development script; Processing the target data according to the configuration script and the table structure rule design to obtain an initial table structure, wherein the initial table structure records the relationship between fields in the target data and the structure and rules between tables recording the fields; The target data in the analysis library is changed to the initial table structure for storage.

7. The data report development method according to claim 3, wherein: The report design operation includes: Obtaining target data from an analysis database, the target data being obtained according to the data extraction operation, and storing the target data in a manner according to an initial table structure obtained according to the data configuration operation; Obtaining a design script obtained by the third agent based on the natural language processing and a design template determined based on the natural language, wherein the design script is a type of the development script; The target data is processed according to the design script and the design template to obtain the initial report.

8. The data report development method according to claim 3, wherein: The display verification operation includes: Pre-execute the initial report to determine whether the initial report is successfully configured; If the configuration is successful, the verification data set is extracted from the initial report according to the preset extraction rules and displayed to the user; If the configuration fails or a cancellation instruction is received from the user, the initial report is deleted, and the development script obtained by the report development model according to the natural language processing is re-acquired, and the report processing operation is performed according to the development script; If a save instruction is received from the user, it is determined that the initial report has passed the display verification operation.

9. A computer device, characterized in that: including processor and memory; The processor is configured to execute the computer program stored in the memory to implement the method according to any one of claims 1 to 2, or the method according to any one of claims 3 to 8.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 or the method according to any one of claims 3 to 8 is implemented.

Images