Method and system for realizing intelligent brief report based on large model

Abstract

The invention discloses a method and a system for realizing intelligent brief report based on a large model, and belongs to the field of data configuration technology and large model technology, the realization of the method comprises the following steps: generating business indexes, compiling and executing SQL statements, screening index data from an original data table, and inserting the screened data into the index table; problem template configuration: inputting and selecting time placeholders according to needs to generate practical problems; similar problem configuration: setting similar problems, and enriching the knowledge base; and configuring an answer template, selecting a digital or JSON type index placeholder, and customizing answer content according to requirements. The method solves the problems that in a traditional mode, knowledge base maintenance needs manual retrieval and modification, cost is high, and data lags behind, data real-time performance is ensured, reasoning deviation is avoided, accuracy and timeliness of brief report generation are improved, an efficient and intelligent brief report generation scheme is provided for all fields, and industrial technology upgrading is promoted.

Description

Technical Field

[0001] The present invention relates to the field of data configuration technology and large model technology, and in particular to a method and system for realizing intelligent briefing based on a large model. Background Art

[0002] Currently, large model knowledge bases rely primarily on manual maintenance. Any changes to the relevant data used in the knowledge base require professionals to retrieve the corresponding content from the knowledge base and then manually modify it. This maintenance method is not only labor-intensive and cumbersome, but also difficult to achieve real-time updates, resulting in data lags. This not only prevents users from obtaining real-time information, impacting the user experience, but over time, the accumulation of outdated information in the knowledge base can even lead to deviations in the large model's inference and output, impairing its usability and accuracy. Summary of the Invention

[0003] The technical task of the present invention is to address the above shortcomings and provide a method and system for realizing intelligent briefings based on large models, which significantly innovates the large-model intelligent briefing generation technology, solves the problem that the knowledge base maintenance in the traditional model requires manual retrieval and modification, which is costly and has data lags, ensures the real-time nature of data, avoids reasoning bias, improves the accuracy and timeliness of briefing generation, provides efficient and intelligent briefing generation solutions for various fields, and promotes industry technology upgrades.

[0004] The technical solution adopted by the present invention to solve its technical problem is:

[0005] A method for implementing intelligent briefings based on a large model integrates data configuration technology with large model technology, combines it with an automated timed task mechanism, establishes a question-and-answer configuration template system, and implements dynamic updating and maintenance of knowledge base content. The implementation of this method includes the following steps:

[0006] Generate business indicators by writing and executing SQL statements to filter indicator data from the original data table and insert the filtered data into the indicator table;

[0007] Configure question templates, input and select time placeholders as needed to generate practical questions;

[0008] Similar question configuration: set similar questions. Similar questions will share the same answer template with the main question. In the knowledge base, the system will generate multiple question-answer pairs to enrich the knowledge base.

[0009] Configure the answer template, choose digital or JSON indicator placeholders, and customize the answer content as needed: when generating knowledge base answers, digital indicators will be automatically replaced with actual real data to ensure the accuracy and timeliness of the answer content; for JSON indicators, they will be displayed in a loop and multi-threaded data will be spliced when generating answers.

[0010] This approach combines cutting-edge data configuration technology with big model technology, deeply integrating various indicator data maintained by customers to build a large-model knowledge base Q&A configuration template system. By establishing an automated scheduled task mechanism, the content of the big-model knowledge base can be dynamically updated and maintained, creating an innovative implementation of intelligent briefings based on big models, providing intelligent and precise briefing generation solutions for related fields.

[0011] Furthermore, the business indicators are generated and scheduled tasks are set to specify the SQL execution time.

[0012] Furthermore, the answer template configuration, for JSON indicators, is specifically implemented as follows:

[0013] First, analyze the map structure in the JSON indicator and split it into multiple subtasks based on the data volume and system resources.

[0014] Submit the split subtasks to the ForkJoinPool, which assigns a worker thread to each subtask for processing. During the processing, each thread independently loops and splices the assigned map structure.

[0015] During the execution process, if a thread completes its task ahead of time, it can obtain tasks from the task queue of other threads to execute, thereby making full use of thread resources to avoid the situation where some threads are idle while others are busy;

[0016] When all subtasks are completed, the processing results of each subtask are merged into the final display content.

[0017] Furthermore, the implementation of this method also includes the configuration of scheduled tasks.

[0018] Configure the task execution cycle through a visual interface, and support Cron expressions to customize complex scheduling rules to meet diverse business needs.

[0019] Furthermore, during the task execution process, a full-link monitoring function is provided to track the task status in real time to achieve traceability of historical execution records, support task details viewing and execution result export.

[0020] Furthermore, the implementation of this method also includes the construction of an intelligent agent.

[0021] It supports users to independently build personalized intelligent agent processes and flexibly select the generated knowledge base. When users ask questions to the large model, the intelligent agent accurately retrieves matching content from the user-selected knowledge base based on established process rules, and then conducts in-depth analysis, integration and refinement of the retrieved information to provide users with accurate feedback on corresponding briefing information in a form that meets user needs and briefing format specifications.

[0022] Furthermore, the large model technology implements large model processing based on vector model technology, as follows:

[0023] The BERT pre-trained language model acquires rich linguistic knowledge and semantic information through unsupervised learning on large-scale text data. In the construction of intelligent agents, BERT encodes user questions and documents in the knowledge base into vector representations. The BERT-based bidirectional encoding mechanism fully considers contextual information, allowing the generated vectors to better reflect the semantics of the text.

[0024] Calculate the similarity between the user question vector and the document vector in the knowledge base, and select the most relevant documents based on the similarity ranking. Cosine similarity is a commonly used vector similarity calculation method. It measures the similarity between two vectors by calculating the cosine value of the angle between them. The closer the cosine value is to 1, the more similar the directions of the two vectors are, that is, the closer the textual semantics they represent are. The calculation formula is as follows:

[0025]

[0026] Where A and B are two vectors, “·” represents the dot product of the vectors, and ||A|| and ||B|| represent the modulus of the vectors respectively.

[0027] The entire process is efficient and intelligent, leveraging the knowledge base's real-time dynamic update mechanism to continuously inject the latest information, significantly improving the accuracy and timeliness of briefing generation. This not only significantly improves user information acquisition efficiency and provides a smoother and more convenient experience, but also provides a reliable basis for user decision-making at a deeper level, helping them quickly resolve problems, efficiently advance business operations, and comprehensively enhance user execution effectiveness in all aspects of their work.

[0028] The present invention also claims a system for implementing intelligent briefing based on a large model, comprising:

[0029] Generate business indicator module, which is used to filter indicator data from the original data table by writing and executing SQL statements, and insert the filtered data into the indicator table;

[0030] Question template configuration module, used to input and select time placeholders to generate practical questions as needed;

[0031] The similar question configuration module is used to set similar questions. Similar questions will share the same answer template with the main question and generate multiple question-answer pairs in the knowledge base to enrich the knowledge base.

[0032] The answer template configuration module is used to select numeric or JSON indicator placeholders to customize the answer content as needed;

[0033] The system specifically implements intelligent briefing based on large models through the above method.

[0034] The present invention also claims a device for implementing intelligent briefing based on a large model, comprising: at least one memory and at least one processor;

[0035] The at least one memory is configured to store a machine-readable program;

[0036] The at least one processor is configured to call the machine-readable program to implement the above method.

[0037] The present invention also claims protection for a computer-readable medium having computer instructions stored thereon, which are capable of implementing the above method when executed by a processor.

[0038] Compared with the prior art, the method and system for realizing intelligent briefing based on a large model of the present invention have the following beneficial effects:

[0039] The present invention integrates data configuration and large model technology, and uses an automated timed task mechanism to achieve dynamic updating of the knowledge base, significantly innovating the large-model intelligent briefing generation technology. It solves the problem that knowledge base maintenance under the traditional model requires manual retrieval and modification, which is costly and has data lags. It ensures real-time data, avoids reasoning bias, improves the accuracy and timeliness of briefing generation, provides efficient and intelligent briefing generation solutions for various fields, and promotes industry technology upgrades. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a flowchart of a method for implementing intelligent briefing based on a large model provided by an embodiment of the present invention;

[0041] Figure 2 It is a flowchart of the process of constructing an intelligent agent provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The present invention will be further described below with reference to specific embodiments.

[0043] At present, large-model knowledge bases mainly rely on manual maintenance. Once the relevant data used in the knowledge base changes, professionals are required to retrieve the corresponding content in the knowledge base and then modify it manually. The manual maintenance method is not only labor-intensive and cumbersome, but also difficult to achieve real-time updates, resulting in data lag. This will not only prevent users from obtaining information in real time, affecting the user experience, but over time, a large amount of lagged information will accumulate in the knowledge base, and may even cause deviations in large-model reasoning and output, weakening the usability and accuracy of the large model. Based on this, an embodiment of the present invention provides a method for realizing intelligent briefings based on large models, integrating data configuration technology with large-model technology, combining with an automated timed task mechanism, establishing a question-and-answer configuration template system, and realizing dynamic updates and maintenance of knowledge base content.

[0044] The implementation of this method includes the following:

[0045] For metric maintenance, write and execute SQL statements to filter metric data from the original data table, set up scheduled tasks to specify the SQL execution time, and insert the filtered data into the metric table. Enter and select time placeholders as needed to generate practical questions. Set up similar questions to enrich the knowledge base. Choose between numeric or JSON metric placeholders to customize the answers.

[0046] Configure scheduled tasks and the task execution cycle through a visual interface. Cron expressions are also supported to customize complex scheduling rules to meet diverse business needs. During task execution, the system provides full-link monitoring capabilities, real-time tracking of task status, traceable historical execution records, and support for viewing task details and exporting execution results. Users can independently build personalized intelligent agent processes and flexibly select generated knowledge bases. When users ask questions to the large model, the intelligent agent accurately retrieves matching content from the user-selected knowledge base based on established process rules, providing users with accurate feedback on the corresponding briefing information.

[0047] This method helps generate accurate and timely intelligent briefings, ensures real-time data, avoids reasoning bias, provides efficient and intelligent briefing generation solutions for various fields, and promotes industry technology upgrades.

[0048] The method is further described in detail below with reference to the accompanying drawings.

[0049] This method deeply integrates various business indicator data maintained by customers, organically integrates data configuration technology and big model technology, and introduces an automated timed task mechanism to build a complete question-and-answer configuration template system. This enables real-time dynamic updating and efficient maintenance of knowledge base content, and ultimately relies on big model technology to accurately and efficiently generate intelligent briefings. The specific implementation steps are as follows:

[0050] 1. Generate business indicators:

[0051] This module provides users with a convenient, efficient, and flexible indicator data management solution. Within this module, users can easily manage and manage various indicator data. Specifically, users can write and execute SQL statements to perform in-depth screening of raw data tables. Raw data tables typically contain a wealth of data. Based on their specific needs, users can leverage SQL's powerful filtering capabilities to accurately extract the required indicator data from massive amounts of data.

[0052] To further optimize the data processing process, the module also supports scheduling tasks. Users can set the appropriate SQL statement execution time based on business needs and data update frequency. The system also implements SQL parsing and validation based on Apache Calcite. This automatically executes SQL filtering operations at the preset time, ensuring timely and accurate indicator data updates. Finally, the filtered indicator data is inserted into a dedicated indicator table. This indicator table acts as a data warehouse, providing stable and reliable data support for the subsequent question-and-answer configuration module.

[0053] 2. Question template configuration:

[0054] When maintaining question templates, users can create templates based on their needs and freely select time dimension placeholders such as "yearly," "monthly," "weekly," "daily," and "hourly." When generating knowledge base questions, the system automatically replaces these placeholders with the current real time, making the questions more relevant and timely.

[0055] 3. Similar problem configuration:

[0056] To maintain similar questions, users can set up similar questions related to the main question. These similar questions will share the same answer template with the main question. In the knowledge base, the system will generate multiple pairs of questions and answers, greatly enriching the content of the knowledge base and improving the efficiency of users' access to information.

[0057] 4. Answer template configuration:

[0058] When maintaining answer templates, users also enjoy full autonomy and flexibility. On the one hand, users can freely input the required answer content according to specific needs to accurately fit different question and answer scenarios. On the other hand, users can also choose indicator placeholders. These placeholders correspond to the indicators maintained in the first step, covering two categories: numerical indicators and JSON indicators. When generating knowledge base answers, numerical indicators will be automatically replaced with actual real data to ensure the accuracy and timeliness of the answer content. For JSON indicators, since there may be multiple map structures inside them, the system will display and splice these data in a loop when generating answers, and use ForkJoinPool to implement multi-threaded splicing. The specific steps are:

[0059] (1) First, analyze the map structure in the JSON type indicator and split it into multiple subtasks based on the data volume and system resources.

[0060] (2) Submit the split subtasks to the ForkJoinPool. The ForkJoinPool assigns a worker thread to each subtask for processing. During the processing, each thread independently loops and splices the assigned map structure.

[0061] (3) During the execution process, if a thread completes its task ahead of time, it can "steal" tasks from the task queue of other threads to execute, thereby making full use of thread resources and avoiding the situation where some threads are idle while some threads are busy.

[0062] (4) After all subtasks are completed, the processing results of each subtask are merged into the final display content.

[0063] Furthermore, users can also customize the loop body and add the text they wish to display within it. This feature further expands the customization of answer templates, allowing users to flexibly adjust the presentation of answers based on their specific circumstances, providing the knowledge base with richer, more detailed, and more tailored answers.

[0064] 5. Scheduled task configuration: Provide efficient and stable task scheduling management functions for the intelligent briefing system.

[0065] This module utilizes a flexible parameter mapping mechanism, allowing users to associate dynamic variables in the Q&A configuration (such as "This Month" and "Latest Indicators") with task parameters, automatically injecting real-time data during system execution. This module allows users to configure task execution cycles (such as daily at midnight or every Monday morning) through a visual interface. It also supports customizing complex scheduling rules using Cron expressions to meet diverse business needs.

[0066] During task execution, the system provides full-link monitoring, tracking task status, resource usage, and execution logs in real time. In the event of an anomaly, a retry mechanism and multi-level alerts (email / SMS notifications) are automatically triggered to ensure the reliable operation of critical tasks. Historical execution records are traceable, allowing for detailed task viewing and export of execution results.

[0067] 6. Intelligent agent construction:

[0068] Users can independently build personalized agent processes and flexibly select the knowledge base generated in the above steps. When users pose questions to the large model, the agent responds quickly. Based on established process rules, it accurately retrieves matching content from the user's selected knowledge base. The system then conducts in-depth analysis, integration, and refinement of this retrieved information, providing users with precise presentation information in a format that meets user needs and briefing format specifications.

[0069] The vector model technology is used in the large model processing process described in this method. The specific implementation is as follows:

[0070] The BERT pre-trained language model acquires rich linguistic knowledge and semantic information through unsupervised learning on large-scale text data. In the construction of the intelligent agent, BERT encodes the user's question and the documents in the knowledge base into vector representations. BERT's bidirectional encoding mechanism fully considers contextual information, allowing the generated vectors to better reflect the semantics of the text.

[0071] Calculate the similarity between the user question vector and the document vectors in the knowledge base, and sort by similarity to select the most relevant documents. Cosine similarity is a commonly used vector similarity calculation method. It measures their similarity by calculating the cosine value of the angle between two vectors. The closer the cosine value is to 1, the more similar the directions of the two vectors are, that is, the closer the textual semantics they represent are. The calculation formula is:

[0072]

[0073] Where A and B are two vectors, “·” represents the dot product of the vectors, and ||A|| and ||B|| represent the modulus of the vectors respectively.

[0074] The entire process is efficient and intelligent, leveraging the knowledge base's real-time dynamic update mechanism to continuously inject the latest information, significantly improving the accuracy and timeliness of briefing generation. This not only significantly improves user information acquisition efficiency and provides a smoother and more convenient experience, but also provides a reliable basis for user decision-making at a deeper level, helping them quickly resolve problems, efficiently advance business operations, and comprehensively enhance user execution effectiveness in all aspects of their work.

[0075] An embodiment of the present invention further provides a system for implementing intelligent briefing based on a large model. The system implements intelligent briefing based on a large model through the method for implementing intelligent briefing based on a large model described in the above embodiment.

[0076] Write and execute SQL statements to filter indicator data from the original data table, set up a scheduled task to specify the SQL execution time, and insert the filtered data into the indicator table. Enter and select time placeholders as needed to generate practical questions. Set up similar questions to enrich the knowledge base. Choose between numeric or JSON indicator placeholders to customize the answers.

[0077] The task execution cycle is configured through a visual interface, and Cron expressions are supported to customize complex scheduling rules to meet diverse business needs. During task execution, the system provides full-link monitoring capabilities, real-time tracking of task status, traceability of historical execution records, and support for viewing task details and exporting execution results. Users can independently build personalized intelligent agent processes and flexibly select generated knowledge bases; users can ask questions to the large model, and the intelligent agent, based on established process rules, accurately retrieves matching content from the user-selected knowledge base and provides users with accurate feedback on the corresponding briefing information.

[0078] The system includes:

[0079] 1. Generate a business indicator module, which is used to filter indicator data from the original data table by writing and executing SQL statements, and insert the filtered data into the indicator table.

[0080] This module provides users with a convenient, efficient, and flexible indicator data management solution. Within this module, users can easily manage and manage various indicator data. Specifically, users can write and execute SQL statements to perform in-depth screening of raw data tables. Raw data tables typically contain a wealth of data. Based on their specific needs, users can leverage SQL's powerful filtering capabilities to accurately extract the required indicator data from massive amounts of data.

[0081] To further optimize the data processing process, the module also supports scheduling tasks. Users can set the appropriate SQL statement execution time based on business needs and data update frequency. The system also implements SQL parsing and validation based on Apache Calcite. This automatically executes SQL filtering operations at the preset time, ensuring timely and accurate indicator data updates. Finally, the filtered indicator data is inserted into a dedicated indicator table. This indicator table acts as a data warehouse, providing stable and reliable data support for the subsequent question-and-answer configuration module.

[0082] 2. Question template configuration module, used to input and select time placeholders as needed to generate practical questions.

[0083] When maintaining question templates, users can create templates based on their needs and freely select time dimension placeholders such as "yearly," "monthly," "weekly," "daily," and "hourly." When generating knowledge base questions, the system automatically replaces these placeholders with the current real time, making the questions more relevant and timely.

[0084] 3. Similar question configuration module is used to set similar questions. Similar questions will share the same answer template with the main question, and generate multiple question-answer pairs in the knowledge base to enrich the knowledge base.

[0085] To maintain similar questions, users can set up similar questions related to the main question. These similar questions will share the same answer template with the main question. In the knowledge base, the system will generate multiple pairs of questions and answers, greatly enriching the content of the knowledge base and improving the efficiency of users' access to information.

[0086] 4. The answer template configuration module is used to select numeric or JSON indicator placeholders and customize the answer content as needed.

[0087] When maintaining the answer template, on the one hand, users can freely input the required answer content according to specific needs to accurately fit different question and answer scenarios. On the other hand, users can also choose indicator placeholders. These placeholders correspond to the indicators maintained in the first step, covering two categories: numerical indicators and JSON indicators. When generating knowledge base answers, numerical indicators will be automatically replaced with actual real data to ensure the accuracy and timeliness of the answer content. For JSON indicators, since there may be multiple map structures inside them, the system will display and splice these data in a loop when generating answers, and use ForkJoinPool to implement multi-threaded splicing. The specific steps are:

[0088] (1) First, analyze the map structure in the JSON type indicator and split it into multiple subtasks based on the data volume and system resources.

[0089] (2) Submit the split subtasks to the ForkJoinPool. The ForkJoinPool assigns a worker thread to each subtask for processing. During the processing, each thread independently loops and splices the assigned map structure.

[0090] (3) During the execution process, if a thread completes its task ahead of time, it can "steal" tasks from the task queue of other threads to execute, thereby making full use of thread resources and avoiding the situation where some threads are idle while some threads are busy.

[0091] (4) After all subtasks are completed, the processing results of each subtask are merged into the final display content.

[0092] Furthermore, users can also customize the loop body and add the text they wish to display within it. This feature further expands the customization of answer templates, allowing users to flexibly adjust the presentation of answers based on their specific circumstances, providing the knowledge base with richer, more detailed, and more tailored answers.

[0093] 5. Scheduled task configuration module provides efficient and stable task scheduling management functions for the intelligent briefing system.

[0094] This module utilizes a flexible parameter mapping mechanism, allowing users to associate dynamic variables in the Q&A configuration (such as "This Month" and "Latest Indicators") with task parameters, automatically injecting real-time data during system execution. This module allows users to configure task execution cycles (such as daily at midnight or every Monday morning) through a visual interface. It also supports customizing complex scheduling rules using Cron expressions to meet diverse business needs.

[0095] During task execution, the system provides full-link monitoring, tracking task status, resource usage, and execution logs in real time. In the event of an anomaly, a retry mechanism and multi-level alerts (email / SMS notifications) are automatically triggered to ensure the reliable operation of critical tasks. Historical execution records are traceable, allowing for detailed task viewing and export of execution results.

[0096] 6. The intelligent agent construction module is used to enable users to independently build personalized intelligent agent processes and flexibly select the knowledge base generated in the above steps.

[0097] When users pose questions to the large model, the intelligent agent responds quickly. Based on established process rules, it accurately retrieves matching content from the user's selected knowledge base. The system then conducts in-depth analysis, integration, and refinement of this retrieved information, providing users with precise feedback in a format that meets user needs and briefing format specifications.

[0098] Vector model technology is used in the process of large model processing. The specific implementation is as follows:

[0099] The BERT pre-trained language model acquires rich linguistic knowledge and semantic information through unsupervised learning on large-scale text data. In the construction of the intelligent agent, BERT encodes the user's question and the documents in the knowledge base into vector representations. BERT's bidirectional encoding mechanism fully considers contextual information, allowing the generated vectors to better reflect the semantics of the text.

[0100] Calculate the similarity between the user question vector and the document vectors in the knowledge base, and sort by similarity to select the most relevant documents. Cosine similarity is a commonly used vector similarity calculation method. It measures their similarity by calculating the cosine value of the angle between two vectors. The closer the cosine value is to 1, the more similar the directions of the two vectors are, that is, the closer the textual semantics they represent are. The calculation formula is:

[0101]

[0102] Where A and B are two vectors, “·” represents the dot product of the vectors, and ||A|| and ||B|| represent the modulus of the vectors respectively.

[0103] An embodiment of the present invention further provides a device for implementing intelligent briefing based on a large model, comprising: at least one memory and at least one processor;

[0104] The at least one memory is configured to store a machine-readable program;

[0105] The at least one processor is configured to call the machine-readable program to implement the method for realizing intelligent briefing based on a large model as described in the above embodiment.

[0106] Embodiments of the present invention further provide a computer-readable medium having computer instructions stored thereon. When executed by a processor, the computer instructions cause the processor to execute the method for implementing intelligent presentations based on a large model as described in the above embodiments. Specifically, a system or device equipped with a storage medium can be provided. The storage medium stores software program code that implements the functions of any of the above embodiments, and the computer (or CPU or MPU) of the system or device can read and execute the program code stored in the storage medium.

[0107] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute part of the present invention.

[0108] Examples of storage media for providing program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, the program code can be downloaded from a server computer via a communication network.

[0109] In addition, it should be clear that the functions of any of the above embodiments can be achieved not only by executing the program code read by the computer, but also by enabling the operating system operating on the computer to complete part or all of the actual operations based on the instructions of the program code.

[0110] In addition, it can be understood that the program code read from the storage medium is written into the memory provided in the expansion board inserted into the computer or into the memory provided in the expansion unit connected to the computer, and then based on the instructions of the program code, the CPU installed on the expansion board or expansion unit is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above embodiments.

[0111] The present invention has been shown and described in detail above through the accompanying drawings and preferred embodiments. However, the present invention is not limited to these disclosed embodiments. Based on the above multiple embodiments, those skilled in the art can know that the code review methods in the above different embodiments can be combined to obtain more embodiments of the present invention, and these embodiments are also within the scope of protection of the present invention.

Claims

1. A method for realizing intelligent briefing based on a large model, characterized in that: By integrating data configuration technology with large model technology and combining it with an automated timed task mechanism, a question-and-answer configuration template system is established to achieve dynamic updating and maintenance of knowledge base content. The implementation of this method includes the following steps: Generate business indicators by writing and executing SQL statements to filter indicator data from the original data table and insert the filtered data into the indicator table; Configure question templates, input and select time placeholders as needed to generate practical questions; Similar question configuration: set similar questions. Similar questions will share the same answer template with the main question. In the knowledge base, the system will generate multiple question-answer pairs to enrich the knowledge base. Configure the answer template, choose digital or JSON indicator placeholders, and customize the answer content as needed: when generating knowledge base answers, digital indicators will be automatically replaced with actual real data to ensure the accuracy and timeliness of the answer content; for JSON indicators, they will be displayed in a loop and multi-threaded data will be spliced when generating answers.

2. The method for realizing intelligent presentation based on a large model according to claim 1, characterized in that: The business indicators are generated and scheduled tasks are set to specify the SQL execution time.

3. The method for realizing intelligent presentation based on a large model according to claim 1, characterized in that: The answer template configuration, for JSON indicators, is specifically implemented as follows: First, analyze the map structure in the JSON indicator and split it into multiple subtasks based on the data volume and system resources. Submit the split subtasks to the ForkJoinPool, which will assign a worker thread to each subtask for processing; During the processing, each thread independently performs loop display and splicing operations on the assigned map structure; During the execution process, if a thread completes its task ahead of time, it can obtain tasks from the task queue of other threads to execute, thereby making full use of thread resources to avoid the situation where some threads are idle while others are busy; When all subtasks are completed, the processing results of each subtask are merged into the final display content.

4. The method for realizing intelligent presentation based on a large model according to claim 1, characterized in that: The implementation of this method also includes the configuration of scheduled tasks. Configure the task execution cycle through a visual interface, and support Cron expressions to customize complex scheduling rules to meet diverse business needs.

5. The method for realizing intelligent presentation based on a large model according to claim 1 or 4, characterized in that: During the task execution process, a full-link monitoring function is provided to track the task status in real time to make historical execution records traceable, and support task details viewing and execution result export.

6. The method for realizing intelligent presentation based on a large model according to claim 1, characterized in that: The implementation of this method also includes the construction of intelligent agents, It supports users to independently build personalized intelligent agent processes and flexibly select the generated knowledge base. When users ask questions to the large model, the intelligent agent accurately retrieves matching content from the user-selected knowledge base based on established process rules, and then conducts in-depth analysis, integration and refinement of the retrieved information to provide users with accurate feedback on corresponding briefing information in a form that meets user needs and briefing format specifications.

7. The method for realizing intelligent presentation based on a large model according to claim 1 or 6, characterized in that: The large model technology is based on vector model technology to achieve large model processing, as follows: The BERT pre-trained language model acquires rich linguistic knowledge and semantic information through unsupervised learning on large-scale text data. In the construction of intelligent agents, BERT encodes user questions and documents in the knowledge base into vector representations. The BERT-based bidirectional encoding mechanism fully considers contextual information, allowing the generated vectors to better reflect the semantics of the text. Calculate the similarity between the user question vector and the document vector in the knowledge base, and filter out the most relevant documents based on the similarity ranking; the calculation formula is as follows: Where A and B are two vectors, "·" represents the dot product of the vectors, and ||A|| and ||B|| represent the moduli of the vectors, respectively.

8. A system for realizing intelligent briefing based on a large model, characterized in that: include: Generate business indicator module, which is used to filter indicator data from the original data table by writing and executing SQL statements, and insert the filtered data into the indicator table; Question template configuration module, used to input and select time placeholders to generate practical questions as needed; The similar question configuration module is used to set similar questions. Similar questions will share the same answer template with the main question and generate multiple question-answer pairs in the knowledge base to enrich the knowledge base. The answer template configuration module is used to select numeric or JSON indicator placeholders to customize the answer content as needed; The system specifically implements intelligent briefing based on a large model through the method described in any one of claims 1 to 7.

9. A device for realizing intelligent briefing based on a large model, characterized in that: include: at least one memory and at least one processor; The at least one memory is configured to store a machine-readable program; The at least one processor is configured to call the machine-readable program to implement the method according to any one of claims 1 to 7.

10. A computer-readable medium, characterized in that The computer readable medium stores computer instructions, which, when executed by a processor, can implement the method according to any one of claims 1 to 7.

Images