Intelligent report generation method based on large language model

Abstract

The invention discloses an intelligent report generation method based on a big language model, which comprises the following steps of: obtaining report data interpretation by utilizing the big language model and combining industry knowledge according to original data for generating a report; according to report data interpretation, a report is intelligently generated in combination with mode requirements; and adjusting the content and / or mode of the report according to the user input. According to the method, highly automatic and intelligent data processing is realized by utilizing the large language model, manual intervention is not needed in the whole process from data collection to report generation, time and cost are remarkably reduced, and data analysis depth and accuracy are improved. And performing context analysis in combination with industry knowledge, identifying key business driving factors and potential risk points, and generating a high-quality report. Through a user input function, report content and format can be adjusted in real time, and user experience and decision effectiveness are greatly enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent report generation, and particularly relates to an intelligent report generation method based on a large language model. Background Art

[0002] With the rapid development of information technology, enterprises have accumulated a large amount of data in their daily operations. This data contains rich information and potential value, which can provide strong support for enterprise decision-making. However, how to effectively extract valuable information from the vast amount of data and transform it into an easily understandable report has become an important challenge in enterprise management.

[0003] Traditional data analysis methods usually rely on manual operations, including basic interactive functions such as manual querying, filtering, and visual display. Although these methods can help enterprise managers quickly understand the business status to a certain extent, they often have problems such as low efficiency and dependence on professional skills when dealing with complex data analysis tasks. In recent years, the development of artificial intelligence (AI) and natural language processing (NLP) technologies has provided new ideas and tools for solving these problems.

[0004] Currently, there are already some enterprise management platforms and tools based on artificial intelligence and data analysis technologies. Although these methods have made significant progress in data display, they still have the following deficiencies: existing methods usually require users to manually operate to analyze data, increasing the learning cost for non-professionals and limiting the data processing efficiency. Moreover, existing methods mainly rely on users' experience and intuition for data interpretation, lacking automated professional analysis capabilities and making it difficult to extract valuable business insights from a large amount of data.

[0005] In addition, the interaction modes of existing methods are relatively single, lacking intelligent interaction and automated data processing capabilities, and unable to meet users' needs for efficient and intelligent interaction. Moreover, most methods fail to provide personalized data analysis and interpretation customized according to the specific needs of enterprises or users, limiting the applicability and flexibility of the methods. Summary of the Invention

[0006] The present invention provides an intelligent report generation method based on a large language model to solve the problems of many deficiencies in existing methods in terms of automation level, data professional interpretation ability, diverse interaction modes, and personalized analysis.

[0007] The technical solution adopted by the present invention is as follows:

[0008] An intelligent report generation method based on a large language model, comprising:

[0009] Based on the original data used to generate the report, using a large language model and combining industry knowledge, obtain an interpretation of the report data;

[0010] Based on the interpretation of the report data and combining with the pattern requirements, intelligently generate a report;

[0011] According to the user input, adjust the content and / or pattern of the report.

[0012] The intelligent report generation method based on a large language model in the present invention further includes the following additional technical features:

[0013] Based on the interpretation of the report data and combining with the pattern requirements, intelligently generate a report. Specifically,

[0014] Based on the interpretation of the report data and combining with the user's historical behavior, intelligently generate a personalized report; and / or,

[0015] Based on the interpretation of the report data and combining with the user's language questions, intelligently generate a personalized report; and / or,

[0016] Based on the interpretation of the report data and combining with the user's preset template selection, intelligently generate a personalized report; and / or,

[0017] Based on the interpretation of the report data and combining with the report output device, intelligently generate a personalized report.

[0018] Intelligently generate a report. Specifically,

[0019] Using a large language model, based on the interpretation of the report data, automatically write the report content,

[0020] Combining with the pattern requirements, adjust the report pattern and intelligently generate a report.

[0021] According to the user input, adjust the content and / or pattern of the report. Specifically,

[0022] Parse the user input, adjust the report content and / or pattern, and output the report;

[0023] Parse the user's re-input of the output report, repeatedly adjust the report content and / or pattern until the user has no input or confirms satisfaction;

[0024] Analyze the user satisfaction information to form user historical behavior data.

[0025] Based on the original data used to generate the report, using a large language model and combining industry knowledge, obtain an interpretation of the report data. Specifically,

[0026] Combining industry knowledge and historical report generation data, using a large language model, perform context analysis on the original data and extract key information;

[0027] Among them, the key information at least includes business driving factors and potential risk points.

[0028] The intelligent report generation method based on a large language model further includes

[0029] collecting the original data for generating the report, where the original data includes basic data and detailed data

[0030] Among them, the basic data at least includes any one of the functional positioning, title, and introduction of the report, and the detailed data at least includes multiple types of indicators and dimensions to interpret the detailed data.

[0031] The acquisition of the original data at least includes any one of the data in the enterprise internal database, external programming interface services, and the Internet.

[0032] Performing preprocessing on the original data, at least including any one of data cleaning, data standardization and normalization, data type conversion, and data enrichment

[0033] Among them, data cleaning includes removing duplicate data, handling missing values, and handling outliers; data standardization and normalization include format standardization and normalization processing; data enrichment includes derivative variables and data merging.

[0034] The present invention also provides an intelligent report generation system based on a large language model, including:

[0035] A data collection module for acquiring the original data for generating the report;

[0036] A data preprocessing module for performing preprocessing on the original data;

[0037] A large language model processing module for obtaining an interpretation of the report data according to the original data for generating the report and combining industry knowledge;

[0038] A report generation module for intelligently generating a report according to the interpretation of the report data and combining the pattern requirements, and adjusting the content and / or pattern of the report according to user input.

[0039] The present invention further provides an electronic device, including:

[0040] A memory for storing computer instructions;

[0041] A processor for implementing the intelligent report generation method based on a large language model when executing the computer instructions.

[0042] Due to the adoption of the above technical solutions, the beneficial effects obtained by the present invention are:

[0043] 1. In the present invention, based on the original data for generating a report, by using a large language model and combining industry knowledge, a report data interpretation is obtained. The present invention realizes highly automated and intelligent data processing by using a large language model, significantly reducing the need for manual intervention.

[0044] Conventional data analysis methods usually require a large amount of manual operations, including steps such as manual querying, filtering, and visual display. This not only increases the workload but also requires users to have relatively high professional skills, thus limiting the efficiency and accuracy of data analysis. In contrast, the large language model adopted in the present invention can automatically process the original data, and no manual intervention is required throughout the whole process from data collection to final report generation, greatly improving the work efficiency.

[0045] In addition, the large language model can not only understand complex data patterns but also perform context analysis by combining industry knowledge, identify key business drivers and potential risk points, provide in-depth data insights, and be able to generate high-quality and insightful reports in a short time. Therefore, the present invention not only significantly reduces the time and cost of manual operations but also improves the depth and accuracy of data analysis, providing strong support for enterprise decision-making.

[0046] While improving the efficiency of data analysis, the present invention ensures the depth and precision of the analysis results, solving the problems of relying on manual operations and lacking in-depth analysis capabilities in the prior art.

[0047] 2. In the present invention, according to the report data interpretation and in combination with the mode requirements, a report is intelligently generated; according to user input, the content and / or mode of the report are adjusted. The present invention significantly improves the user interaction experience and the precision of the report through personalized report generation and interaction functions. It can dynamically generate highly personalized report content according to user input. This flexible adjustment mechanism ensures that each report can accurately meet the actual needs of users, thereby improving the effectiveness and pertinence of decision-making.

[0048] In addition, users can use simple user input, and the updated report content will be immediately responded to and displayed. This intelligent interaction method not only greatly enhances the user experience but also enables non-professional users to easily obtain the required in-depth analysis results without the need to have complex data processing skills.

[0049] By combining personalized report generation and interaction functions, the present invention not only improves the customization degree of the report but also simplifies the user operation process, making data analysis and report generation more efficient and convenient, solving the problems of single interaction method and insufficient flexibility in the prior art, and significantly enhancing the usability of the present method and user satisfaction.

[0050] 3. As a preferred embodiment of the present invention, the user input is parsed, the report content and / or mode is adjusted, and the report is output; the user's re-input to the output report is parsed, and the report content and / or mode is repeatedly adjusted until the user has no input or confirms satisfaction; the user satisfaction information is analyzed to form user historical behavior data. The present invention significantly enhances the user experience and satisfaction through the real-time preview and confirmation function and the multiple iteration optimization mechanism. Each time the user adjusts the report, a real-time preview is provided, enabling the user to immediately see the change effect of the report. This immediate feedback mechanism allows the user to repeatedly confirm and adjust the content, style, and layout of the report before generating the final version, ensuring that each report can accurately meet the actual needs of the user.

[0051] In addition, the user is supported to make multiple inquiries and adjustments until the report fully meets their expectations. This interactive inquiry and adjustment function not only enhances the personalization and flexibility of the report but also enables non-professional users to easily obtain the required in-depth analysis results, improving the usability and practicality of the present method.

[0052] Through this iteration optimization mechanism, the user can continuously refine and improve the report until its content, style, and layout fully meet their expectations. This highly customized process ensures that the finally generated report can meet the specific needs of the user to the greatest extent, thereby enhancing the user's satisfaction and trust.

[0053] Overall, the present invention solves the problems of single interaction mode and insufficient flexibility in the prior art through real-time preview and confirmation and multiple iteration optimization.

[0054] 4. As a preferred embodiment of the present invention, the acquisition of the original data includes at least any one of the data in the enterprise internal database, external programming interface service, and the Internet. The present invention has a flexible data source integration ability, can extract and integrate information from multiple data sources, and ensure the comprehensiveness and accuracy of the data. The present invention can integrate the data from the enterprise internal database, external API service, and the Internet. This multi-source data integration ability enables the present method to adapt to various complex data environments and extract valuable information from them.

[0055] This flexibility not only broadens the data source channels but also improves the data quality and coverage, making the generated report more comprehensive and in-depth. Whether it is structured data (such as numerical values and texts in tables) or semi-structured or unstructured data (such as JSON files, social media comments, etc.), it can be effectively processed and analyzed. In addition, by carefully organizing and classifying these multi-source data, the efficiency and accuracy of data processing are further improved.

[0056] Through flexible data source integration, the present invention solves the problems of single data source and limited coverage in the prior art, and significantly enhances the adaptability and scalability of the method. In the face of multi-domain data, it can effectively integrate data from different sources and provide comprehensive and accurate analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0058] Figure 1 It is a schematic flow chart of the intelligent report generation method based on the large language model according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] In order to more clearly illustrate the overall concept of the present invention, the following will be described in detail by way of examples in conjunction with the drawings of the specification.

[0060] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0061] As Figure 1 shown, an intelligent report generation method based on a large language model includes:

[0062] S100: According to the original data for generating a report, using a large language model and combining industry knowledge, obtain an interpretation of the report data.

[0063] The main purpose of this step is to generate an insightful interpretation of the report data through in-depth analysis of the original data, combined with industry knowledge and the powerful processing capabilities of large language models (LLMs). This process can not only identify key business drivers and potential risk points, but also provide users with comprehensive and in-depth data analysis results to support them in making more accurate decisions.

[0064] Among them, use a large language model and combine industry knowledge to perform context analysis of the original data. By combining industry knowledge, enhance the depth and accuracy of data analysis and provide more targeted insights.

[0065] Specifically, use a large language model to perform in-depth insight extraction on the data to identify key business drivers and potential risk points; combine industry background and historical data for context-related analysis, considering broader market conditions, competitor dynamics and other factors.

[0066] This step enhances the comprehensiveness and depth of data analysis, helping users not only see the data but also understand the story behind it.

[0067] In addition, use large language models to generate report data explanations. Through advanced natural language processing technology, convert complex data patterns into easy-to-understand text descriptions, providing detailed report data explanations.

[0068] Specifically, use large language models to automatically write text content, including data explanations, analysis insights, and conclusion summaries; generate highly personalized report content according to users' preferences and historical behaviors.

[0069] This step simplifies the process of understanding complex data, enabling non-professional users to easily obtain valuable business insights, improving the readability and practicality of the report.

[0070] It should be noted that a real-time update and feedback mechanism is set up. Through an instant feedback and adjustment mechanism, ensure that the report content always meets the needs and expectations of users.

[0071] Specifically, users can use natural language questions or instructions to adjust the content and format of the report in real time. This method responds immediately and displays the updated report content; provide a real-time preview function, allowing users to immediately see the effect of the report changes and generate the final version after confirmation of satisfaction.

[0072] This step enhances the user experience and satisfaction, enabling each report to best meet the actual needs of users.

[0073] Generally speaking, through the above steps, the present invention can efficiently extract valuable information from raw data and generate detailed report data explanations in combination with industry knowledge. This automated and intelligent processing method not only reduces manual intervention, improves work efficiency, but also significantly enhances the depth and accuracy of data analysis, providing strong support for enterprise decision-making, and solving the problems of relying on manual operations and lacking in-depth analysis capabilities in the prior art.

[0074] S200: According to the report data explanation, combined with the pattern requirements, intelligently generate a report.

[0075] The main purpose of this step is to intelligently generate a business report with a reasonable structure and rich content according to the previously generated report data explanation, combined with the specific needs of users and the preset pattern requirements. This process can not only ensure the high quality and personalized customization of the report, but also improve the generation efficiency and meet the diverse needs of different users.

[0076] It is understandable that after obtaining a detailed report data interpretation, the appropriate template and format will be automatically selected according to the user's preferences and preset mode requirements to generate the final report. Through this intelligent method, the content and style of the report can be dynamically adjusted to adapt to different application scenarios and user needs.

[0077] Among them, determining the mode requirements, clarifying the specific requirements of the user for the report format and content, provides guidance for subsequent generation.

[0078] Specifically, extract the mode requirements from the user input or preset templates, such as report types (business analysis, market trend prediction, etc.), chart types (bar charts, pie charts, etc.), and text styles (formal, concise, etc.). Analyze the user's preferences and historical behaviors to further refine the mode requirements.

[0079] This step ensures that the generated report meets the user's expectations and specific needs, improving the applicability and flexibility of the report.

[0080] Secondly, select and adjust the template. According to the mode requirements, select the appropriate template and make necessary adjustments to adapt to the specific report needs.

[0081] Specifically, select a design style and layout that best meets the current needs from multiple preset templates. Dynamically adjust the elements in the template, such as chart types, color themes, etc., according to the user's feedback and real-time instructions.

[0082] This step provides a highly customizable report generation solution, enabling each report to precisely match the specific needs of the user.

[0083] Thirdly, dynamically generate charts and visual elements, presenting the data analysis results in the form of intuitive charts and visual elements to enhance the readability and comprehensibility of the report.

[0084] Specifically, according to the report data interpretation, dynamically generate various charts and visual elements (such as bar charts, line charts, pie charts, etc.) and embed them in the report. Provide interactive chart functions that allow users to explore data details by clicking, dragging, etc.

[0085] This step makes complex data relationships and trends clear at a glance, enhancing the professional level and user experience of the report.

[0086] Finally, automatically generate the text content, using large language models to automatically generate high-quality text content, including data interpretation, analysis insights, and conclusion summaries.

[0087] Specifically, use a large language model to automatically generate a detailed text description based on the report data interpretation, ensuring that the language is fluent and accurately reflects the data analysis results. Adjust the level of detail and style of the text content according to the user's preferences and historical behavior.

[0088] This step simplifies the process of understanding complex data, enabling non-professional users to easily obtain valuable business insights and improving the readability and usability of the report.

[0089] It should be noted that this method supports multi-format output and distribution. By supporting multiple formats of output, it can adapt to different usage scenarios and distribution requirements.

[0090] Specifically, export the generated report into multiple formats such as PDF, HTML, and Word, which is convenient for users to view and share on different devices and platforms. Provide batch generation and automated distribution functions, which are suitable for business scenarios that require regular generation of a large number of reports.

[0091] This step enhances the flexibility and convenience of the report, meets the diverse distribution requirements, and improves work efficiency.

[0092] Generally speaking, through the above steps, the present invention can intelligently generate high-quality and personalized business reports according to the report data interpretation and pattern requirements. This process not only reduces manual intervention and improves the generation efficiency, but also significantly enhances the depth and accuracy of the report, providing strong support for enterprise decision-making and solving the problems of relying on manual operations and lacking personalized customization in the existing technology.

[0093] S300: Adjust the content and / or pattern of the report according to the user input.

[0094] The main purpose of this step is to dynamically adjust the content and format of the generated report by responding to the user's real-time input and feedback, ensuring that the final report can best meet the user's personalized needs. This interactive adjustment mechanism not only improves the user experience, but also enhances the flexibility and applicability of the report.

[0095] It can be understood that after generating the preliminary report, the user is allowed to further adjust the content and pattern of the report by asking questions or giving instructions in natural language. This method will parse the user's input and make corresponding adjustments according to its intention, providing instant feedback and preview functions, allowing the user to repeatedly confirm and optimize the report until satisfied.

[0096] Among them, first receive the user input, obtain the user's feedback and specific modification requirements, providing a basis for subsequent adjustment.

[0097] Specifically, the user puts forward specific modification requirements through natural language questions or instructions, such as "adding a comparative analysis of the sales data in the second quarter" or "changing the color theme of the report to blue". This method receives and parses the user's input, and identifies the specific content and patterns that need to be adjusted.

[0098] This step simplifies the user's operation process, enabling non-professional users to easily make complex report adjustments, and improving the usability of this method.

[0099] Secondly, parse the user input, accurately understand the user's intention, and convert it into executable operation instructions.

[0100] Specifically, use a large language model to parse the user's natural language input, identify the specific parts (such as charts, text paragraphs, etc.) that the user hopes to adjust and their adjustment methods (such as adding, deleting, modifying). Combine context information and historical behavior data to further refine the user's intention and ensure that the adjustment meets the user's actual needs.

[0101] This step improves the intelligence level of this method, reduces the possibility of misunderstandings and misoperations, and enhances the user experience.

[0102] Thirdly, dynamically adjust the report content and patterns. According to the user's input and parsing results, dynamically adjust the content and patterns of the report to ensure that the report always meets the user's expectations.

[0103] Specifically, according to the parsing results, dynamically adjust the charts, text, and other elements in the report. For example, add new data analysis modules, modify the chart types, or update the text descriptions. Provide a real-time preview function, and the user can immediately see the effect of the adjusted report and further modify it as needed.

[0104] This step enhances the flexibility and personalized customization ability of the report, enabling each report to accurately match the specific needs of the user, and improving the quality and practicality of the report.

[0105] It can be understood that through an instant feedback mechanism, it is ensured that the user can quickly confirm the adjustment effect and decide whether to continue modifying or generate the final version.

[0106] Specifically, after each adjustment, provide a real-time preview, and the user can immediately see the change effect of the report. After the user confirms satisfaction, they can choose to generate the final version of the report; if further adjustment is needed, they can continue to put forward new modification requirements.

[0107] This step provides a highly interactive user experience, enabling the user to repeatedly confirm and optimize the report before generating the final version, and ensuring the quality and satisfaction of the final report.

[0108] Secondly, support users to ask follow-up questions and make adjustments multiple times until the report fully meets their expectations. Users can put forward modification requirements multiple times, and this method will respond quickly each time and display the updated report content. Through multiple iterations of optimization, ensure that the finally generated report can meet the actual needs of users to the greatest extent.

[0109] This step enhances the practicality and user satisfaction of this method, and solves the problems of single interaction mode and insufficient flexibility in the prior art.

[0110] Generally speaking, through the above steps, the present invention can dynamically adjust the generated report content and mode according to the real-time input and feedback of users, ensuring that the final report can meet the personalized needs of users to the greatest extent. This intelligent interactive adjustment mechanism not only improves the user experience, but also enhances the flexibility and applicability of the report, and solves the problems of relying on manual operations and lack of personalized customization in the prior art.

[0111] As a preferred embodiment of the present invention, according to the report data interpretation, combined with the mode requirements, generate a report intelligently. Specifically,

[0112] According to the report data interpretation, combined with the user's historical behavior, generate a personalized report; and / or,

[0113] According to the report data interpretation, combined with the user's language questions, generate a personalized report; and / or,

[0114] According to the report data interpretation, combined with the user's preset template selection, generate a personalized report; and / or,

[0115] According to the report data interpretation, combined with the report output device, generate a personalized report.

[0116] The main purpose of this embodiment is to generate highly personalized business reports by combining mode requirements (such as historical behavior, natural language questions, preset template selection, and report output device), using large language models and report data interpretation. This multi-dimensional personalized customization can not only meet the specific needs of different users, but also significantly improve the user experience and the quality of the report.

[0117] The specific form of the mode requirements in this embodiment is not limited, and any one of the following embodiments can be adopted, or any combination of the following embodiments can be used.

[0118] Embodiment 1: Combine the user's historical behavior. Generate a personalized report that conforms to their habits and needs according to the user's historical behavior and preferences.

[0119] Specifically, analyze historical behavior data, extract information from the user's historical queries, feedback, and report usage records, and identify the user's preferences and common patterns. Adjust the report content, and dynamically adjust the content structure of the report according to the user's preferences, such as paying more attention to certain key performance indicators (KPIs) or specific types of charts.

[0120] Specifically, user A often views the cost analysis section in the financial statements. This method will automatically highlight the relevant data when generating a new report and add more cost analysis details.

[0121] This embodiment improves the relevance and practicality of the report, enabling each report to better meet the specific needs of users.

[0122] Embodiment Two: Combine with the user's language questions. Through natural language processing technology, parse the user's real-time questions and dynamically adjust the report content to respond to specific needs.

[0123] Specifically, receive and parse the questions. The user puts forward specific modification requirements through natural language questions or instructions, such as "Compare the main cost changes between this year and last year". Dynamically generate the corresponding content, parse the user's questions, generate the corresponding analysis and charts, and embed them in the report.

[0124] Specifically, user B asks "How about the sales growth in the second quarter?" This method will automatically generate relevant sales data analysis charts and add them to the report.

[0125] This embodiment simplifies the user's operation process, enabling non-professional users to easily obtain the required in-depth analysis results and enhancing the usability of this method.

[0126] Embodiment Three: Combine with the user's preset template selection. Generate a personalized report that conforms to the user's style and layout preferences according to the preset template selected by the user.

[0127] Specifically, provide a template selection interface, provide the user with multiple preset template options, and each template has a different design style and layout. Apply the selected template. After the user selects a suitable template, this method generates the report content according to the format requirements of the template.

[0128] Specifically, user C selects a template with a simple style. The report generated by this method will adopt a simple design style, reduce redundant information, and highlight key data.

[0129] This embodiment provides a highly customized report generation solution, enabling each report to accurately match the specific needs of users.

[0130] Example 4: Integrated report output device. Optimize the format and layout of the report according to the characteristics of the output device used by the user to ensure the best reading experience.

[0131] Specifically, detect the output device type, automatically detect the device type used by the user (such as desktop computer, tablet, mobile phone, etc.). Adaptive layout adjustment, according to the device type, adjust the chart and text layout of the report to ensure the best reading experience on any device.

[0132] Specifically, when user D views the report on a mobile phone, this method automatically adjusts the layout of the report, shrinks the chart and displays it completely within the screen width to ensure that all content is clearly visible.

[0133] This embodiment enhances the flexibility and applicability of the report, enabling users to conveniently view and share the report on different devices.

[0134] Generally speaking, through the mode requirements, the flexibility and practicality of the present invention in different application scenarios are improved, reflecting its significant advantages in improving user experience and report quality.

[0135] As a preferred embodiment under this implementation manner, generate the report intelligently. Specifically,

[0136] Utilize the large language model to automatically write the report content according to the report data interpretation.

[0137] Combine the mode requirements, adjust the report mode, and generate the report intelligently.

[0138] The main purpose of this embodiment is to deeply analyze the report data interpretation by using the large language model (LLMs) and automatically generate high-quality and personalized report content according to the user's mode requirements. This process not only reduces manual intervention, improves the generation efficiency, but also significantly enhances the depth and accuracy of the report.

[0139] After obtaining the detailed report data interpretation, the large language model will be used to automatically write the report content and dynamically adjust the format and style of the report according to the user's mode requirements. This intelligent method ensures that the report not only meets the specific needs of the user, but also has high-quality content and a professional presentation form.

[0140] Among them, use the large language model to automatically write the report content. Based on the report data interpretation, automatically generate detailed and insightful text content.

[0141] Specifically, for data parsing and insight extraction, a large language model is used to deeply analyze the interpretation of report data to identify key business drivers, trends, and potential risk points. For automatic text writing, based on the analysis results, detailed text content including data interpretation, analysis insights, and conclusion summaries is automatically generated.

[0142] Specifically, User I needs a report on sales trends. This method analyzes sales data through a large language model and automatically generates a description: "In the past four quarters, sales have shown a steady growth trend. Especially in the second quarter, sales increased by 15%, mainly due to the launch of new products."

[0143] This step simplifies the process of understanding complex data, enabling non-professional users to easily obtain valuable business insights and improving the readability and usability of the report.

[0144] Secondly, adjust the report format according to the style requirements. Dynamically adjust the format and style of the report according to the user's style requirements to ensure that the report meets the user's expectations.

[0145] Specifically, determine the style requirements. Extract style requirements from user input or preset templates, such as report type (business analysis, market trend prediction, etc.), chart type (bar chart, pie chart, etc.), and text style (formal, concise, etc.).

[0146] Adjust the template and format. Select a suitable template and make necessary adjustments according to the style requirements. For example, if the user selects a concise-style template, this method will reduce redundant information and highlight key data.

[0147] Specifically, User J selects a "market trend prediction" type of report and hopes to present it in a concise style. This method selects a concise-style template and automatically generates a report that highlights market trend changes, avoiding excessive technical details.

[0148] This step provides a highly customized report generation solution, enabling each report to accurately match the specific needs of the user.

[0149] Finally, generate the report intelligently. Integrating the above steps, generate a final report with a reasonable structure, rich content, and standardized format.

[0150] Specifically, integrate the content and format, embed the automatically generated text content and charts into the selected template to form a complete report. Provide real-time preview. Provide a real-time preview function, allowing users to immediately see the effect of the report and make further modifications as needed. Output reports in multiple formats, supporting the export of the generated report to multiple formats such as PDF, HTML, and Word, facilitating users to view and share on different devices and platforms.

[0151] Specifically, user K needs to submit a report to the senior management team. This method generates a PDF report containing detailed data analysis and charts, and automatically adds the company logo and standard format to ensure the professionalism and consistency of the report.

[0152] This step enhances the flexibility and convenience of the report, meets diverse distribution needs, and improves work efficiency.

[0153] Generally speaking, through the above steps, the present invention can automatically generate high-quality report content based on the interpretation of report data by a large language model, and dynamically adjust the format and style of the report according to the user's pattern requirements, ensuring that the final report can meet the user's personalized needs to the greatest extent. This intelligent generation mechanism not only reduces manual intervention, improves generation efficiency, but also significantly enhances the depth and accuracy of the report, providing strong support for enterprise decision-making.

[0154] As a preferred embodiment of the present invention, according to user input, adjust the content and / or pattern of the report, specifically,

[0155] Parse the user input, adjust the report content and / or pattern, and output the report;

[0156] Parse the user's re-input of the output report, repeat adjusting the report content and / or pattern until the user has no input or confirms satisfaction;

[0157] Analyze the user satisfaction information to form user historical behavior data.

[0158] The main purpose of this embodiment is to dynamically adjust the content and pattern of the report by parsing the user's real-time input, and through multiple iterations of optimization, ensure that the finally generated report fully meets the user's expectations. At the same time, the user's feedback and satisfaction information will be recorded to form user historical behavior data to further improve the quality of personalized services.

[0159] After generating the preliminary report, allow the user to further adjust the content and pattern of the report through natural language questions or instructions. Parse the user's input and make corresponding adjustments according to their intentions, providing instant feedback and preview functions, allowing the user to repeatedly confirm and optimize the report until satisfied. Finally, analyze the user's satisfaction information to form user historical behavior data for reference in future personalized services.

[0160] Among them, parse the user input, accurately understand the user's intention, and convert it into an executable operation instruction.

[0161] Specifically, it receives user input. The user puts forward specific modification requirements through natural language questions or instructions, such as "add a comparative analysis of the second-quarter sales data" or "change the color theme of the report to blue". It parses the user input, uses a large language model to parse the user's natural language input, and identifies the specific parts (such as charts, text paragraphs, etc.) that the user hopes to adjust and their adjustment methods (such as adding, deleting, modifying). It combines the context information, combines the user's context information and historical behavior data, and further refines the user's intention to ensure that the adjustment meets the user's actual needs.

[0162] This step improves the intelligence level of this method, reduces the possibility of misunderstandings and misoperations, and enhances the user experience.

[0163] Secondly, it adjusts the report content and / or mode. According to the user input and parsing results, it dynamically adjusts the report content and mode to ensure that the report always meets the user's expectations.

[0164] Specifically, it dynamically adjusts the content. According to the parsing results, it dynamically adjusts the charts, text, and other elements in the report. For example, it adds a new data analysis module, modifies the chart type, or updates the text description.

[0165] It provides a real-time preview. After each adjustment, it provides a real-time preview, allowing the user to immediately see the change effect of the report and further modify it as needed.

[0166] This step enhances the flexibility and personalized customization ability of the report, enables each report to precisely match the user's specific needs, and improves the quality and practicality of the report.

[0167] Thirdly, it outputs the report. It generates and outputs the final version of the report that meets the user's needs.

[0168] Specifically, it generates the report, integrates the adjusted report content into the selected template to generate a complete report. It supports multi-format output and supports exporting the generated report to multiple formats such as PDF, HTML, Word, etc., facilitating the user to view and share on different devices and platforms.

[0169] This step enhances the flexibility and convenience of the report, meets the diverse distribution needs, and improves work efficiency.

[0170] Thirdly, it parses the user's re-input for the output report. Through multiple iterations and optimizations, it ensures that the report fully meets the user's expectations.

[0171] Specifically, upon receiving user feedback, after the user views the preliminary report, they can submit further modification requests through natural language questions or instructions. Parse and adjust by parsing the user's feedback, dynamically adjusting the content and format of the report, and providing a real-time preview so that the user can immediately see the adjustment effect. Repeat the adjustment. The user can submit modification requests multiple times, and this method will quickly respond each time and display the updated report content until the user has no input or confirms satisfaction.

[0172] This step provides a highly interactive user experience, enabling the user to repeatedly confirm and optimize the report before generating the final version, ensuring the quality and satisfaction of the final report.

[0173] Finally, analyze the user satisfaction information to form user historical behavior data. Record the user's feedback and satisfaction information to form user historical behavior data, providing a reference for future personalized services.

[0174] Specifically, collect user feedback, recording the user's feedback during each adjustment process and the information finally confirmed. Analyze the user satisfaction information, and by analyzing the user's satisfaction information, understand the user's preferences and common patterns. Form historical behavior data, storing the analysis results as the user's historical behavior data for future personalized services.

[0175] This step improves the intelligence level of this method, making future report generation more in line with the user's habits and needs, and enhancing the user experience.

[0176] Generally speaking, the present invention is flexible and practical in different application scenarios, capable of improving the user experience and report quality, and solving the problems of relying on manual operations and lacking personalized customization in the prior art.

[0177] As a preferred embodiment of the present invention, according to the original data for generating the report, using a large language model and combining industry knowledge, obtain an interpretation of the report data. Specifically,

[0178] Combine industry knowledge and historical report generation data, use a large language model to perform context analysis on the original data, and extract key information;

[0179] Among them, the key information at least includes business driving factors and potential risk points.

[0180] The main purpose of this embodiment is to perform in-depth context analysis on the original data, combine industry knowledge and historical report generation data, and use a large language model (LLMs) to extract key information. These key information not only include business driving factors but also cover potential risk points, thereby providing users with comprehensive and in-depth data insights to support them in making more accurate decisions.

[0181] After obtaining raw data, we combine industry knowledge and historical reports to generate data, then use large language models to conduct in-depth analysis of the data to identify key business drivers and potential risk points. This process not only improves the accuracy and depth of data analysis, but also provides users with valuable business insights, helping them better understand market dynamics and business trends.

[0182] First, collect and integrate raw data to ensure that data obtained from multiple sources are carefully organized and categorized to provide accurate and contextually relevant data support for subsequent analysis.

[0183] Specifically, data collection involves collecting raw data from multiple channels, including internal enterprise databases, external API services, and the internet, and meticulously organizing and categorizing it. Data preprocessing involves cleaning, standardizing, and normalizing the collected raw data, removing duplicates, filling in missing values, and addressing outliers.

[0184] This step improves the quality and consistency of the data and ensures that the basic data for subsequent analysis are reliable and comprehensive.

[0185] Secondly, we combine industry knowledge for contextual analysis. By combining industry background and historical data, we can enhance the depth and accuracy of data analysis and provide more targeted insights.

[0186] Specifically, it loads industry knowledge bases related to specific industries, such as market dynamics, competitor information, and industry standards. This data is combined with historical report data and referenced to historical report generation data to understand past analysis results and user feedback from similar scenarios. Deep contextual analysis leverages large language models to extract deep insights from data and identify key business drivers and potential risk points. For example, this can be used to analyze seasonal fluctuations in sales data, the impact of unexpected events, or changes in long-term trends.

[0187] This step improves the comprehensiveness and depth of data analysis, helping users not only see the data but also understand the stories behind the data, thereby enhancing the effectiveness and pertinence of decision-making.

[0188] Secondly, extract key information. Extract the most valuable key information from the analysis results to provide users with clear business insights.

[0189] Specifically, identify business drivers. Based on the results of contextual analysis, identify the main drivers affecting business performance, such as changes in market demand, new product launches, and cost control measures. Identify potential risk points and analyze potential risks, such as intensified market competition, supply chain disruptions, and policy changes. Generate detailed explanations, converting the identified business drivers and potential risk points into easy-to-understand text descriptions to form detailed explanations of the report data.

[0190] This step simplifies the process of understanding complex data, enabling non-expert users to easily obtain valuable business insights and enhancing the readability and practicality of the report.

[0191] It should be noted that in this embodiment, the analysis results can be verified and optimized. By verifying and optimizing the analysis results, their accuracy and reliability are ensured.

[0192] Specifically, cross-validation is carried out by cross-validating with other data sources or independent analysis tools to ensure the accuracy of the analysis results. The user feedback mechanism collects users' feedback opinions to further optimize the analysis model and methods, improving the intelligence level of this method.

[0193] This step improves the credibility and user satisfaction of this method, making future analysis results more accurate and reliable.

[0194] Generally speaking, through the above steps, the present invention can generate data by combining industry knowledge and historical reports, use large language models to perform context analysis on the original data, and extract key business driving factors and potential risk points. This intelligent analysis method not only improves the accuracy and depth of data analysis but also provides users with valuable business insights, helping them better understand market dynamics and business trends, and solving the problems of relying on manual operations and lacking in-depth analysis capabilities in the prior art.

[0195] As a preferred embodiment of the present invention, the intelligent report generation method based on large language models further includes

[0196] collecting the original data for generating the report, and the original data includes basic data and detailed data

[0197] wherein the basic data includes at least any one of the functional positioning, title, and introduction of the report, and the detailed data includes at least multiple types of indicators and dimensions to explain the detailed data.

[0198] The main purpose of this embodiment is to ensure that the original data for generating the report can be efficiently and accurately collected from multiple data sources. These data not only include basic data (such as the functional positioning, title, and introduction of the report) but also detailed business indicator and dimension data, so as to provide comprehensive and structured input for subsequent data analysis and report generation.

[0199] Before generating a report, two main types of data need to be collected and organized: basic data and detailed data. Basic data provides the basic framework and background information of the report, while detailed data contains specific business metrics and dimensions, which will be used for in-depth analysis and interpretation. Through this meticulous data organization method, it can ensure that the finally generated report has both a clear structure and rich detailed content.

[0200] First, collect basic data to obtain the basic framework and background information of the report, providing context support for subsequent analysis.

[0201] Specifically, define the types of basic data. Basic data should include at least any one of the functional positioning, title, and introduction of the report. For example,

[0202] Functional positioning: Such as business analysis, market trend prediction, financial report, etc.;

[0203] Title: Clearly indicate the theme of the report, such as "Sales Analysis for the First Quarter of 2023";

[0204] Introduction: Provide a concise description of the main content and purpose of the report.

[0205] Data source. Basic data can be extracted from user input, preset templates, or historical reports. For example, users can directly fill in the report title and introduction in the system interface or select a preset template to automatically generate relevant content.

[0206] This step provides a clear framework and background information for the report, helping users quickly understand the purpose and content of the report.

[0207] Secondly, collect detailed data. Obtain specific business metric and dimension data, providing detailed information support for subsequent in-depth analysis.

[0208] Specifically, define the types of detailed data. Detailed data should include at least multiple types of metrics and dimensions, which can specifically include but are not limited to the following,

[0209] Component name: The names of each data component included in the report, using industry-related terms, such as "Sales Distribution Chart", "Customer Growth Trend Table", etc.;

[0210] Component type: The type of data component, such as chart, list, text, etc., which can be specifically divided into bar chart, pie chart, line chart, etc.;

[0211] Component title: The title displayed above each component in the report, providing users with an intuitive understanding of the content of the component;

[0212] Dataset name: If the component is associated with a specific dataset, the name of the dataset needs to be specified;

[0213] Parameters: Specify the parameters required for generating components, such as time range, region, product type, etc. These parameters are all expressed in common industry terms that are easy to understand.

[0214] Data display: The data content directly displayed in the component. This data is usually pre - processed and directly used for display, rather than the original data.

[0215] Data sources: Detailed data can be obtained from multiple channels, including enterprise internal databases, external API services, and Internet data. For example,

[0216] Enterprise internal database: Obtain internal data such as sales records and financial statements.

[0217] External API services: Call third - party APIs to obtain external data such as market dynamics and competitor information.

[0218] Internet data: Use web scraping technology to capture unstructured data such as the latest industry news and social media comments.

[0219] This step provides comprehensive and detailed business data, laying a solid foundation for subsequent in - depth analysis and report generation.

[0220] Again, data integration and pre - processing. Ensure that the collected raw data undergoes cleaning, standardization, and normalization to improve the quality and consistency of the data.

[0221] Specifically, performing pre - processing on the raw data includes at least any one of data cleaning, data standardization and normalization, data type conversion, and data enrichment.

[0222] Among them, data cleaning includes removing duplicate data, handling missing values, and handling outliers; data standardization and normalization include format standardization and normalization processing; data enrichment includes deriving variables and data merging.

[0223] The main purpose of this embodiment is to ensure the quality and consistency of the data by pre - processing the raw data. The pre - processing includes operations such as data cleaning, data standardization and normalization, data type conversion, and data enrichment. These steps can remove duplicate data, handle missing values and outliers, unify the data format, and generate new variables or merge data sets, thereby providing high - quality basic data for subsequent data analysis and report generation.

[0224] Among them, data cleaning removes duplicate data, handles missing values and outliers, and improves the quality of the data.

[0225] Including, removing duplicate data: Identify and delete duplicate records to avoid bias in data analysis. By comparing key fields of records (such as ID, timestamp, etc.), identify exactly the same records and delete the redundant parts;

[0226] Handling missing values: Fill in, ignore, or delete missing data. Select an appropriate handling method according to the data attributes and context. For example, use the mean, median, or estimation based on a prediction model to fill in missing values; for missing values in key fields, the record may be directly deleted.

[0227] Handling outliers: Identify and handle outliers, such as using box plots or standard deviation methods to determine the normal range of data. Identify outliers through statistical analysis methods (such as box plots, Z - scores), and decide whether to correct or delete these outliers according to the specific situation.

[0228] This step improves the integrity and accuracy of the data and reduces the impact of noise on subsequent analysis.

[0229] Data standardization and normalization: Convert all data into a unified format to reduce the impact of data with different magnitudes.

[0230] Including, format standardization: Standardize data such as dates, times, and currencies into a standard format, such as the ISO8601 date format and currency amounts with two decimal places. Define the standard format according to the data type and apply the corresponding conversion rules.

[0231] Normalization processing: Perform normalization processing on numerical data so that its range falls within a standard interval (such as 0 to 1) for easy comparison and analysis. Use the min - max scaling method or other normalization algorithms to map the data to the specified interval.

[0232] This step ensures the consistency and comparability of the data and reduces the impact of data with different magnitudes on the analysis results.

[0233] Data type conversion: Adjust the storage type of data according to the actual meaning and use of the data to ensure the correctness of subsequent operations.

[0234] Including, type adjustment: Convert text - formatted numbers to numerical types, convert string - type dates to date types, etc. Identify the data type through regular expressions or pattern matching and apply the corresponding conversion functions.

[0235] Type verification: Verify whether the converted data type meets the expectations to ensure the accuracy and consistency of the data. Confirm the correctness of the conversion result through sampling inspection and logical verification.

[0236] [[ID=

[0237] Data enrichment generates new variables based on existing data, revealing additional value in the data.

[0238] Including derived variables: Calculate new variables such as the number of days between dates, the frequency of summarizing categorical data, etc. Design reasonable calculation formulas for derived variables according to business requirements and data analysis objectives.

[0239] Data merging: Combine data from different sources into a unified dataset, providing a more comprehensive perspective for comprehensive analysis. Align the data through the primary key or common fields to ensure the integrity and consistency of the merged dataset.

[0240] This step enhances the depth and breadth of the data, provides more analysis dimensions, and improves the value of the report.

[0241] Generally speaking, through the above steps, the present invention can efficiently and accurately collect the original data for generating reports, including basic data and detailed data. This meticulous data organization method not only provides a clear structure and background information for the report, but also lays a solid foundation for subsequent in-depth analysis and report generation. These innovations solve the problems of single data source and low quality in the prior art, demonstrating its significant advantages in the field of data processing. Through flexible data source integration and intelligent report generation.

[0242] The present invention also provides an intelligent report generation system based on a large language model, including:

[0243] A data collection module for obtaining the original data for generating reports;

[0244] A data preprocessing module for performing preprocessing on the original data;

[0245] A large language model processing module for obtaining report data interpretations based on the original data for generating reports and combining industry knowledge;

[0246] A report generation module for intelligently generating reports according to the report data interpretations and combining pattern requirements, and adjusting the content and / or pattern of the report according to user input.

[0247] Therefore, any effects of the intelligent report generation method based on the large language model can be achieved, which will not be elaborated here.

[0248] The present invention further provides an electronic device, including:

[0249] A memory for storing computer instructions;

[0250] A processor for implementing the intelligent report generation method based on a large language model when executing the computer instructions.

[0251] Therefore, any effects of the intelligent report generation method based on a large language model can be achieved, which will not be elaborated here.

[0252] In the present invention, those not described can be implemented by adopting or referring to existing technologies.

[0253] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and each embodiment focuses on the differences from other embodiments.

[0254] The above are only embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, various changes and modifications can be made to the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. An intelligent report generation method based on large language models, characterized in that, Including: Based on the original data for generating a report, using a large language model and combining industry knowledge, obtain an interpretation of the report data; Based on the interpretation of the report data, and in combination with the mode requirements, intelligently generate a report; According to user input, adjust the content and / or mode of the report.

2. The intelligent report generation method based on a large language model according to claim 1, wherein Based on the interpretation of the report data, and in combination with the mode requirements, intelligently generate a report. Specifically, Based on the interpretation of the report data, and in combination with the user's historical behavior, intelligently generate a personalized report; and / or, Based on the interpretation of the report data, and in combination with the user's language questions, intelligently generate a personalized report; and / or, Based on the interpretation of the report data, and in combination with the user's preset template selection, intelligently generate a personalized report; and / or, Based on the interpretation of the report data, and in combination with the report output device, intelligently generate a personalized report.

3. The intelligent report generation method based on a large language model according to claim 2, wherein Intelligently generate a report. Specifically, Using a large language model, based on the interpretation of the report data, automatically write the content of the report, In combination with the mode requirements, adjust the report mode and intelligently generate a report.

4. The intelligent report generation method based on a large language model according to claim 1, wherein According to user input, adjust the content and / or mode of the report. Specifically, Parse the user input, adjust the report content and / or mode, and output the report; Parse the user's re-input of the output report, repeatedly adjust the report content and / or mode until the user has no input or confirms satisfaction; Analyze the user satisfaction information to form user historical behavior data.

5. The intelligent report generation method based on a large language model according to claim 1, wherein Based on the original data for generating a report, using a large language model and combining industry knowledge, obtain an interpretation of the report data. Specifically, Combining industry knowledge and historical report generation data, using a large language model, perform context analysis on the original data and extract key information; Among them, the key information at least includes business driving factors and potential risk points.

6. The intelligent report generation method based on a large language model according to claim 1, wherein It also includes, Collect the original data for generating a report, and the original data includes basic data and detailed data, Among them, the basic data at least includes any one of the functional positioning, title, and introduction of the report, and the detailed data at least includes multiple types of indicators and dimensions to interpret the detailed data.

7. The intelligent report generation method based on a large language model according to claim 6, characterized in that The acquisition of the original data at least includes any one of the data in the enterprise internal database, external programming interface service, and the Internet.

8. The intelligent report generation method based on a large language model according to claim 1, characterized in that Perform preprocessing on the original data, at least including any one of data cleaning, data standardization and normalization, data type conversion, and data enrichment, Among them, data cleaning includes removing duplicate data, handling missing values, and handling outliers; data standardization and normalization include format standardization and normalization processing; data enrichment includes derivative variables and data merging.

9. An intelligent report generation system based on a large language model, characterized in that, Including: A data collection module for obtaining the original data for generating a report; A data preprocessing module for performing preprocessing on the original data; A large language model processing module for obtaining an interpretation of the report data based on the original data for generating a report and combining industry knowledge; A report generation module, configured to intelligently generate a report based on the interpretation of the report data in combination with the pattern requirements, and adjust the content and / or pattern of the report according to user input.

10. An electronic device, characterized in that, Comprising: A memory, configured to store computer instructions; A processor, configured to implement the intelligent report generation method based on a large language model according to any one of claims 1 to 8 when executing the computer instructions.

Images