Dynamic slide updating method, system, terminal and storage medium based on user instruction and custom template

Abstract

This invention relates to the field of data processing technology, and discloses a method, system, terminal, and storage medium for dynamic slideshow updates based on user instructions and custom templates. The method includes: analyzing the slideshow page to obtain visual prediction boxes and underlying object boxes, and matching them to obtain geometric objects; determining the statistical functions of the geometric objects in a finite function library, and extracting specific parameters from the statistical functions; decomposing and reorganizing the geometric objects to obtain corresponding computational pipelines; extracting keywords from language instructions to determine specific parameters and computational pipelines; and constructing a target slideshow file based on the specific parameters, computational pipelines, and updated data. This invention, by reconstructing the computational logic, achieves a fully automated closed loop of instruction parsing, data recalculation, and rendering generation, enabling end-to-end linked updates of data, charts, and textual conclusions. This ensures that the updated textual summary accurately reflects the trend of the new data and significantly reduces the labor costs of maintaining periodic reports.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, system, terminal, and computer-readable storage medium for dynamic slideshow updates based on user instructions and custom templates. Background Technology

[0002] Slides (presentations) are the core medium for data reporting. They are very common in data analysis scenarios and in the production of periodic reports (such as weekly and monthly reports).

[0003] At the regional development level, current research focuses on analyzing single factors such as industry, economy, and transportation, rarely taking a systemic perspective to conduct quantitative analysis. As a result, the analytical results are limited and inaccurate. Such reports are usually based on historical slides as templates. As new data is generated, it is necessary to update the chart data, table values, and corresponding textual conclusions while maintaining the original professional layout and visual style. This process relies excessively on manual work, which is time-consuming and labor-intensive, and the accuracy of the data points cannot be guaranteed.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a dynamic slideshow update method, system, terminal, and computer-readable storage medium based on user instructions and custom templates. This aims to solve the problems in the prior art where slideshow updates mostly use fixed templates, which cannot adapt to user-defined personalized templates, and rely too heavily on manual processing, which easily leads to inaccurate data statistics results.

[0006] To achieve the above objectives, the present invention provides a dynamic slideshow update method based on user commands and custom templates, the method comprising the following steps: Each slide in the PowerPoint file is analyzed to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide. After matching each visual prediction box and each underlying object box, the corresponding geometric object is obtained. In the existing finite function library, determine the statistical functions corresponding to multiple geometric objects, and extract the corresponding specific parameters from the corresponding statistical functions for each geometric object; The multiple geometric objects are decomposed and recombined to obtain the computational pipeline corresponding to each geometric object; The system acquires user-input language commands and update data, extracts multiple keywords from the language commands, determines multiple specific parameters and multiple computational pipelines based on each keyword, and constructs a target slide file based on each specific parameter, each computational pipeline, and the update data.

[0007] Optionally, the dynamic slide update method based on user instructions and custom templates, wherein analyzing each slide page in the slide file to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page, and matching each visual prediction box and each underlying object box to obtain the corresponding geometric object, specifically includes: Obtain the PowerPoint file input by the user, construct a visual language model and an underlying document parser, and input each slide page of the PowerPoint file into the visual language model and the underlying document parser for analysis; For each slide page, the visual language model analyzes each page element within the slide page to obtain the semantic category and corresponding visual prediction box for each page element; For each slide page, the underlying document parser extracts each style metadata within the slide page to obtain the underlying object frame of each style metadata. Calculate the intersection-union ratio (IUGR) between each visual prediction box and each underlying object box. When the IUGR exceeds a preset threshold, add the semantic category corresponding to the visual prediction box to the style metadata corresponding to the underlying object box, and finally obtain all geometric objects.

[0008] Optionally, the dynamic slide update method based on user instructions and custom templates, wherein the step of determining statistical functions corresponding to multiple geometric objects in a pre-built finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object, further includes: Define multiple semantic slots in the underlying database, wherein the semantic slots are used to match query conditions in the underlying database; The scope description of each table title is analyzed using a large language model. The unstructured text of each table title is converted into a structured slot and filled into the corresponding semantic slot. The table header corresponding to each table title is analyzed using a large language model. The dimension information of each table header is matched with multiple field names in the underlying database to obtain the corresponding matching score. Based on all the matching scores, multiple target table titles that meet the preset conditions are filtered out, and all the semantic slots of each target table title are filtered out to obtain the semantically matched target table name.

[0009] Optionally, in the dynamic slide update method based on user instructions and custom templates, the geometric objects include: line charts, bar charts, scatter plots, and pie charts; The step of determining statistical functions corresponding to multiple geometric objects in a pre-constructed finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object, specifically includes: In the existing finite function library, determine the trend analysis function corresponding to the line chart, the summary statistics function corresponding to the bar chart, the regression function corresponding to the scatter plot, and the proportion analysis function corresponding to the pie chart. Multiple data distribution patterns and multiple target correlation information are extracted from the trend analysis function, the summary statistics function, the regression function, and the pie chart; Based on the target table names of the line chart, the bar chart, the scatter plot, and the pie chart, each data distribution pattern and each target association information is added to the corresponding semantic slot.

[0010] Optionally, in the dynamic slide update method based on user instructions and custom templates, the geometric object includes: a user-defined complex chart; The step of decomposing and recombining multiple geometric objects to obtain a computational pipeline corresponding to each geometric object specifically includes: For each of the complex charts, the computational logic in the complex chart is structured using the large language model to obtain the table structure type, dimension header, constraint specification, source field mapping, and aggregation operator in the complex chart; Based on the table structure type, the dimension header, the constraint specification, and the source field mapping are defined as multiple query statements in the database, and the aggregation operator is embedded into each query statement to obtain multiple aggregation expressions; All the aggregate expressions are encapsulated to obtain the calculation pipeline corresponding to each of the table structure types.

[0011] Optionally, the dynamic slide update method based on user instructions and custom templates, wherein obtaining user-input language instructions and update data, extracting multiple keywords from the language instructions, determining multiple specific parameters and multiple computational pipelines based on each keyword, and constructing a target slide file based on each specific parameter, each computational pipeline, and the update data, specifically includes: Obtain the user's input language commands and the updated data of the current task, analyze the language commands, and obtain multiple keywords of the current task; Based on each keyword, determine all the geometric objects required for the current task; Extract the computation pipelines corresponding to all the geometric objects and multiple target table names containing specific parameters; The frame of the slide file is updated based on all the computation pipelines and all the target table names, and the updated data is populated into the frame to obtain the target slide file for the current task; The target slide file includes several new statistical data points.

[0012] Optionally, the dynamic slide update method based on user instructions and custom templates, wherein the steps of obtaining user-input language instructions and update data, extracting multiple keywords from the language instructions, determining multiple specific parameters and multiple computational pipelines based on each keyword, constructing a target slide file based on each specific parameter, each computational pipeline, and the update data, further include: By comparing all stored historical data with all the new statistical data, the corresponding data difference information, trend change data, dimension change data, and abnormal fluctuation data are obtained. After structuring the data difference information, the trend change data, the dimension change data, and the abnormal fluctuation data respectively, we obtain the difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector. The difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector are input into the constructed analysis model for analysis, and the output includes a description of trend changes, logical causal relationships, and trend consistency information. The accuracy of the target slide file is obtained by performing contextual consistency verification and multi-dimensional cross-validation based on the trend change description, the logical causal relationship, and the trend consistency information.

[0013] Furthermore, to achieve the above objectives, the present invention also provides a dynamic slideshow update system based on user commands and custom templates, wherein the dynamic slideshow update system based on user commands and custom templates includes: The multimodal layout module is used to analyze each slide page in the slide file, obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page, and obtain the corresponding geometric object by matching each visual prediction box and each underlying object box. A logic extraction module, configured to determine statistical functions corresponding to multiple geometric objects in a pre-constructed finite function library, and extract corresponding specific parameters from the corresponding statistical functions for each geometric object; A logic reconstruction module, configured to decompose and recombine multiple geometric objects to obtain a calculation pipeline corresponding to each geometric object; An instruction parsing module, configured to obtain a language instruction and updated data input by a user, extract multiple keywords from the language instruction, determine multiple specific parameters and multiple calculation pipelines according to each keyword, and construct a target slide file according to each specific parameter, each calculation pipeline, and the updated data.

[0014] In addition, to achieve the above object, the present invention further provides a terminal, wherein the terminal includes: a memory, a processor, and a dynamic slide update program based on user instructions and a custom template stored on the memory and executable on the processor. When the dynamic slide update program based on user instructions and a custom template is executed by the processor, the steps of the above-mentioned dynamic slide update method based on user instructions and a custom template are implemented.

[0015] In addition, to achieve the above object, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a dynamic slide update program based on user instructions and a custom template. When the dynamic slide update program based on user instructions and a custom template is executed by a processor, the steps of the above-mentioned dynamic slide update method based on user instructions and a custom template are implemented.

[0016] In the present invention, each slide page in a slide file is analyzed to obtain multiple visual prediction frames and multiple underlying object frames for each slide page. After matching each visual prediction frame and each underlying object frame, corresponding geometric objects are obtained; statistical functions corresponding to multiple geometric objects are determined in a pre-constructed finite function library, and corresponding specific parameters are extracted from the corresponding statistical functions for each geometric object; multiple geometric objects are decomposed and recombined to obtain a calculation pipeline corresponding to each geometric object; a language instruction and updated data input by a user are obtained, multiple keywords are extracted from the language instruction, multiple specific parameters and multiple calculation pipelines are determined according to each keyword, and a target slide file is constructed according to each specific parameter, each calculation pipeline, and the updated data. By reconstructing the calculation logic, the present invention realizes a full-automatic closed loop of instruction parsing - data recalculation - rendering generation, realizes end-to-end linkage update of data, charts, and text conclusions, ensures that the updated text summary accurately reflects the trend of new data, and greatly reduces the labor cost of periodic report maintenance. Attached Figure Description

[0017] Figure 1 This is a flowchart of a preferred embodiment of the dynamic slideshow update method based on user instructions and custom templates of the present invention; Figure 2 This is a schematic diagram illustrating the multimodal slide layout parsing of a preferred embodiment of the dynamic slide update method based on user instructions and custom templates of the present invention. Figure 3 This is a schematic diagram illustrating the table and chart logic extraction of a preferred embodiment of the dynamic slide update method based on user instructions and custom templates of the present invention. Figure 4 This is a schematic diagram of a preferred embodiment of the dynamic slide update method based on user commands and custom templates of the present invention, showing command-driven update. Figure 5 This is a structural diagram of a preferred embodiment of the dynamic slide update system based on user instructions and custom templates of the present invention; Figure 6 This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] The preferred embodiment of the present invention describes a dynamic slideshow update method based on user commands and custom templates, such as... Figure 1 As shown, the dynamic slide update method based on user commands and custom templates includes the following steps: Step S10: Analyze each slide page in the slide file to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page. After matching each visual prediction box and each underlying object box, obtain the corresponding geometric object.

[0020] Most existing technologies require users to adapt to fixed, pre-set templates, and cannot handle user-provided, diverse, and highly personalized slides (i.e., "user-defined templates"). Once the user adjusts the layout or uses a non-standard design, the system cannot recognize it. To solve this problem, this invention combines a Vision-Language Model (VLM) with underlying code analysis (e.g., Python-pptx, a Python library for reading PPT files) to perform multimodal layout analysis on any user template and reverse engineer to extract the "data source" and "computational logic" behind tables and charts.

[0021] Specifically, the user-input slide file is obtained, a visual language model and an underlying document parser are constructed, and each slide page in the slide file is input into the visual language model and the underlying document parser for analysis. For each slide page, the visual language model analyzes each page element within the slide page to obtain the semantic category and corresponding visual prediction box for each page element; For each slide page, the underlying document parser extracts each style metadata within the slide page to obtain the underlying object frame of each style metadata. Calculate the intersection-union ratio (IUGR) between each visual prediction box and each underlying object box. When the IUGR exceeds a preset threshold, add the semantic category corresponding to the visual prediction box to the style metadata corresponding to the underlying object box, and finally obtain all geometric objects.

[0022] Among them, such as Figure 2 As shown, in this stage, the slides are parsed and the logic is restructured in order to solve the mapping problem from unstructured documents to structured logic.

[0023] Specifically, in terms of perceiving visual semantics, a visual language model is used to take a slide page as an image input to predict the semantic category of all page elements within the slide page and predict the bounding box of each page element; the visual language model is then used to parse the semantics of the slide page content to obtain the underlying logic.

[0024] Furthermore, the underlying document parser is used to scan the structure of the slide page, extract the precise geometric coordinates of all style metadata, and finally, by calculating the intersection-union ratio of the visual prediction box and the underlying object box, it is determined whether the matching degree between the two exceeds a threshold (in the embodiments disclosed in this invention, the threshold can be defined as 0.5). If it exceeds the threshold, the semantic category corresponding to the visual prediction box can be added to the style metadata corresponding to the underlying object box. In this way, a structured page representation result with both semantic role and precise location information is constructed.

[0025] Furthermore, multiple semantic slots are defined in the underlying database, wherein the semantic slots are used to match query conditions in the underlying database; The scope description of each table title is analyzed using a large language model. The unstructured text of each table title is converted into a structured slot and filled into the corresponding semantic slot. The table header corresponding to each table title is analyzed using a large language model. The dimension information of each table header is matched with multiple field names in the underlying database to obtain the corresponding matching score. Based on all the matching scores, multiple target table titles that meet the preset conditions are filtered out, and all the semantic slots of each target table title are filtered out to obtain the semantically matched target table name.

[0026] In order to "anchor" static slide elements to a specific database context, this invention uses a slot-filling mechanism to infer implicit data sources, thereby semantically aligning and dynamically associating unstructured or semi-structured table content (such as tables in slides) with the underlying structured database.

[0027] Specifically, firstly, a set of standard semantic slots aligned with the underlying database is defined, such as time span, geographical region, and business type. These slots correspond to fields or query conditions in the database tables, forming the target framework for subsequent information extraction. Then, semantic clues are extracted by analyzing the table titles and headers. Analyzing the scope descriptions in the table titles yields high-level semantic information used to determine the analysis dimensions and scope. Analyzing the dimensional information in the headers reveals the column dimensions reflecting the data, allowing inference of the database table structure. Finally, a large language model combines header vocabulary with field names in the underlying database for similarity matching, improving the accuracy of table name inference. Ultimately, the target table name with the highest probability is output based on the semantic matching degree, effectively "anchoring" static slide elements to the dynamic data source, providing foundational support for subsequent automatic refresh, drill-down analysis, and interactive queries.

[0028] Step S20: Determine the statistical functions corresponding to multiple geometric objects in the constructed finite function library, and extract the corresponding specific parameters from the corresponding statistical functions for each geometric object.

[0029] The geometric objects include: line charts, bar charts, scatter plots, and pie charts. To handle geometric objects of varying complexity, this invention designs two complementary logical extraction mechanisms in the calculation process from raw data to the final chart display, such as... Figure 3 As shown, they are used for analyzing and processing common charts and for analyzing and processing complex charts defined by users, respectively.

[0030] Specifically, in the finite function library that has been constructed, the trend analysis function corresponding to the line chart, the summary statistics function corresponding to the bar chart, the regression function corresponding to the scatter plot, and the proportion analysis function corresponding to the pie chart are determined respectively; Multiple data distribution patterns and multiple target correlation information are extracted from the trend analysis function, the summary statistics function, the regression function, and the pie chart; Based on the target table names of the line chart, the bar chart, the scatter plot, and the pie chart, each data distribution pattern and each target association information is added to the corresponding semantic slot.

[0031] For common chart types, the characteristics of the target chart are first identified to determine its type (standard types such as bar charts, line charts, scatter plots, and pie charts are disclosed in the embodiments of this invention, but are not limited to these). The specific statistical function used to generate the target chart is then matched from the database. For example, for bar charts, trend analysis functions (such as linear regression and exponential fitting) are extracted. Then, the number of data series, grouping dimensions, axis scale range, legend labels, and other features in the chart are analyzed to construct a preliminary data domain and model (for example, if there are multiple colored lines in a line chart, the system can identify them as "multiple time series", providing a grouping basis for subsequent function selection).

[0032] Furthermore, based on a standardized statistical function library, each function performs a strong correlation analysis with a specific chart type and data distribution model, thereby adding the data distribution pattern and target correlation information to the corresponding semantic slots. This method can effectively handle the requirements of common charts, is suitable for standard report scenarios, and has a high accuracy rate in the final data update.

[0033] Step S30: Decompose and reorganize the multiple geometric objects to obtain the computational pipeline corresponding to each geometric object.

[0034] The geometric objects include user-defined complex graphs. For user-defined complex graphs, this invention discloses a general analysis table synthesis interface. This interface does not rely on specific predefined functions, but deconstructs the unknown computational logic into five atomic reasoning components through a large language model and reassembles them into an executable computational pipeline.

[0035] Specifically, for each of the complex charts, the computational logic in the complex chart is structured using the large language model to obtain the table structure type, dimension header, constraint specification, source field mapping, and aggregation operator in the complex chart; Based on the table structure type, the dimension header, the constraint specification, and the source field mapping are defined as multiple query statements in the database, and the aggregation operator is embedded into each query statement to obtain multiple aggregation expressions; All the aggregate expressions are encapsulated to obtain the calculation pipeline corresponding to each of the table structure types.

[0036] When the system receives an undefined computation request, the large language model first performs semantic parsing, deconstructing it into five core atomic reasoning components: Table structure type refers to identifying the two-dimensional layout paradigm of the data, including fact tables (containing measures, such as sales data), dimension tables (containing descriptive information, such as time and location), and wide tables (including merged tables of multiple dimensions and measures); Dimension headers refer to extracting dimensional information used for grouping or filtering (such as region, time, and product category); Source field mapping maps field names in natural language to actual fields in the database, such as mapping "region" to "region_name"; Aggregation operators are the aggregation operations that the system identifies that need to be performed, including mathematical operations such as average, maximum, and sum.

[0037] Furthermore, after decomposing the five atomic inference components, they can be combined to obtain an executable computation pipeline. Through this process, there is no need to write computation rules in advance, which can cope with complex and ever-changing business needs. Each atomic component can be optimized independently, which is convenient for maintenance and upgrades. The conversion from natural language to structured logic is achieved through a large language model, which reduces the cost of manual configuration.

[0038] Step S40: Obtain the user-input language instructions and update data, extract multiple keywords from the language instructions, determine multiple specific parameters and multiple calculation pipelines based on each keyword, and construct a target slide file based on each specific parameter, each calculation pipeline, and the update data.

[0039] Specifically, the system acquires the user's input language commands and the updated data of the current task, analyzes the language commands, and obtains multiple keywords for the current task. Based on each keyword, determine all the geometric objects required for the current task; Extract the computation pipelines corresponding to all the geometric objects and multiple target table names containing specific parameters; The frame of the slide file is updated based on all the computation pipelines and all the target table names, and the updated data is populated into the frame to obtain the target slide file for the current task; The target slide file includes several new statistical data points.

[0040] Among them, such as Figure 4As shown, based on the user's current input language command and the data to be updated (where the slide file is the slide file processed in the previous steps), the user's natural language modification command (such as "change the statistical range to 2024") is parsed into a change operation on the extracted logical parameters (slots or specific parameters, selected according to the chart type chosen by the user (common charts or custom complex charts)), generating a new execution specification; then, based on the updated parameter status, a structured query statement is automatically generated to retrieve the latest original data from the database; finally, using the updated original data, the extracted closed-domain function or open-domain synthesis interface is called again to calculate and generate new statistical results.

[0041] Furthermore, all stored historical data and all the aforementioned new statistical data are compared to obtain corresponding data difference information, trend change data, dimension change data, and abnormal fluctuation data; After structuring the data difference information, the trend change data, the dimension change data, and the abnormal fluctuation data respectively, we obtain the difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector. The difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector are input into the constructed analysis model for analysis, and the output includes a description of trend changes, logical causal relationships, and trend consistency information. The accuracy of the target slide file is obtained by performing contextual consistency verification and multi-dimensional cross-validation based on the trend change description, the logical causal relationship, and the trend consistency information.

[0042] To further address the issue of delayed textual conclusions after chart updates, this invention discloses a factual consistency verification mechanism in its embodiments. When generating new summary text, difference information (such as numerical differences, trend differences, dimensional differences, and abnormal fluctuations) is extracted from historical and current data. This difference information is structured into difference vectors, which are then input into the model as the display context. This mechanism ensures that the model explicitly knows which data has changed when generating the summary, rather than relying solely on the current data.

[0043] Furthermore, when generating summaries, the model is required to perform trend comparison analysis, including trend identification (the model needs to identify the consistency of trends between current data and historical data), causal reasoning (the model needs to determine whether the differences have a logical causal relationship), and trend change description (clearly outputting a specific description of trend changes). In addition, to ensure the logical consistency between the text and graphics, the summary text must follow a fixed structure and be output according to a structured output template. The output must be consistent with the key change points in the difference information input to avoid omissions or misjudgments. Finally, the accuracy of trend changes is verified from multiple perspectives, such as time series, dimensional comparison, and statistical significance.

[0044] This invention achieves a fully automated closed loop of instruction parsing, data recalculation, and rendering generation by reconstructing the calculation logic. It enables end-to-end linkage updates of data, charts, and textual conclusions, ensuring that the updated textual summary accurately reflects the trend of the new data and significantly reducing the manpower cost of maintaining periodic reports.

[0045] Furthermore, such as Figure 5 As shown, based on the above-described dynamic slide update method based on user commands and custom templates, the present invention also provides a dynamic slide update system based on user commands and custom templates, wherein the dynamic slide update system based on user commands and custom templates includes: The multimodal layout module 51 is used to analyze each slide page in the slide file, obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page, and obtain the corresponding geometric object after matching each visual prediction box and each underlying object box. The logic extraction module 52 is used to determine the statistical functions corresponding to multiple geometric objects in the finite function library that has been built, and to extract the corresponding specific parameters from the corresponding statistical functions for each geometric object; The logic reconstruction module 53 is used to decompose and reorganize multiple geometric objects to obtain the computational pipeline corresponding to each geometric object; The instruction parsing module 54 is used to obtain the language instructions and update data input by the user, extract multiple keywords from the language instructions, determine multiple specific parameters and multiple calculation pipelines based on each keyword, and construct a target slide file based on each specific parameter, each calculation pipeline and the update data.

[0046] Furthermore, such as Figure 6 As shown, based on the above-described dynamic slideshow update method and system based on user instructions and custom templates, the present invention also provides a terminal, which includes a processor 10, a memory 20, and a display 30. Figure 6Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0047] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a dynamic slideshow update program 40 based on user instructions and custom templates. This dynamic slideshow update program 40 can be executed by the processor 10 to implement the dynamic slideshow update method based on user instructions and custom templates in this application.

[0048] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the dynamic slide update method based on user instructions and custom templates.

[0049] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components of the terminal communicate with each other via a system bus.

[0050] In one embodiment, when the processor 10 executes the dynamic slide update program 40 based on user instructions and a custom template stored in the memory 20, the following steps are performed: Each slide in the PowerPoint file is analyzed to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide. After matching each visual prediction box and each underlying object box, the corresponding geometric object is obtained. In the existing finite function library, determine the statistical functions corresponding to multiple geometric objects, and extract the corresponding specific parameters from the corresponding statistical functions for each geometric object; The multiple geometric objects are decomposed and recombined to obtain the computational pipeline corresponding to each geometric object; The system acquires user-input language commands and update data, extracts multiple keywords from the language commands, determines multiple specific parameters and multiple computational pipelines based on each keyword, and constructs a target slide file based on each specific parameter, each computational pipeline, and the update data.

[0051] Specifically, the process of analyzing each slide page in the PowerPoint file to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page, and matching each visual prediction box and each underlying object box to obtain the corresponding geometric object, includes: Obtain the PowerPoint file input by the user, construct a visual language model and an underlying document parser, and input each slide page of the PowerPoint file into the visual language model and the underlying document parser for analysis; For each slide page, the visual language model analyzes each page element within the slide page to obtain the semantic category and corresponding visual prediction box for each page element; For each slide page, the underlying document parser extracts each style metadata within the slide page to obtain the underlying object frame of each style metadata. Calculate the intersection-union ratio (IUGR) between each visual prediction box and each underlying object box. When the IUGR exceeds a preset threshold, add the semantic category corresponding to the visual prediction box to the style metadata corresponding to the underlying object box, and finally obtain all geometric objects.

[0052] The step of determining statistical functions corresponding to multiple geometric objects in a pre-constructed finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object, further includes: Define multiple semantic slots in the underlying database, wherein the semantic slots are used to match query conditions in the underlying database; The scope description of each table title is analyzed using a large language model. The unstructured text of each table title is converted into a structured slot and filled into the corresponding semantic slot. The table header corresponding to each table title is analyzed using a large language model. The dimension information of each table header is matched with multiple field names in the underlying database to obtain the corresponding matching score. Based on all the matching scores, multiple target table titles that meet the preset conditions are filtered out, and all the semantic slots of each target table title are filtered out to obtain the semantically matched target table name.

[0053] The geometric objects include: line charts, bar charts, scatter plots, and pie charts; The step of determining statistical functions corresponding to multiple geometric objects in a pre-constructed finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object, specifically includes: In the existing finite function library, determine the trend analysis function corresponding to the line chart, the summary statistics function corresponding to the bar chart, the regression function corresponding to the scatter plot, and the proportion analysis function corresponding to the pie chart. Multiple data distribution patterns and multiple target correlation information are extracted from the trend analysis function, the summary statistics function, the regression function, and the pie chart; Based on the target table names of the line chart, the bar chart, the scatter plot, and the pie chart, each data distribution pattern and each target association information is added to the corresponding semantic slot.

[0054] The geometric objects include: user-defined complex diagrams; The step of decomposing and recombining multiple geometric objects to obtain a computational pipeline corresponding to each geometric object specifically includes: For each of the complex charts, the computational logic in the complex chart is structured using the large language model to obtain the table structure type, dimension header, constraint specification, source field mapping, and aggregation operator in the complex chart; Based on the table structure type, the dimension header, the constraint specification, and the source field mapping are defined as multiple query statements in the database, and the aggregation operator is embedded into each query statement to obtain multiple aggregation expressions; All the aggregate expressions are encapsulated to obtain the calculation pipeline corresponding to each of the table structure types.

[0055] The steps of acquiring user-input language commands and update data, extracting multiple keywords from the language commands, determining multiple specific parameters and multiple computational pipelines based on each keyword, and constructing a target slide file based on each specific parameter, each computational pipeline, and the update data specifically include: Obtain the user's input language commands and the updated data of the current task, analyze the language commands, and obtain multiple keywords of the current task; Based on each keyword, determine all the geometric objects required for the current task; Extract the computation pipelines corresponding to all the geometric objects and multiple target table names containing specific parameters; The frame of the slide file is updated based on all the computation pipelines and all the target table names, and the updated data is populated into the frame to obtain the target slide file for the current task; The target slide file includes several new statistical data points.

[0056] The process includes obtaining user-input language commands and update data, extracting multiple keywords from the language commands, determining multiple specific parameters and multiple computational pipelines based on each keyword, constructing a target slide file based on each specific parameter, each computational pipeline, and the update data, and then further including: By comparing all stored historical data with all the new statistical data, the corresponding data difference information, trend change data, dimension change data, and abnormal fluctuation data are obtained. After structuring the data difference information, the trend change data, the dimension change data, and the abnormal fluctuation data respectively, we obtain the difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector. The difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector are input into the constructed analysis model for analysis, and the output includes a description of trend changes, logical causal relationships, and trend consistency information. The accuracy of the target slide file is obtained by performing contextual consistency verification and multi-dimensional cross-validation based on the trend change description, the logical causal relationship, and the trend consistency information.

[0057] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a dynamic slide update program based on user instructions and a custom template, and the dynamic slide update program based on user instructions and a custom template, when executed by a processor, implements the steps of the dynamic slide update method based on user instructions and a custom template as described above.

[0058] In summary, this invention provides a dynamic slideshow update method and related equipment based on user commands and custom templates. The method includes: analyzing each slide page in a slideshow file to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page; matching each visual prediction box and each underlying object box to obtain corresponding geometric objects; determining statistical functions corresponding to multiple geometric objects in a pre-built finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object; decomposing and recombining the multiple geometric objects to obtain computational pipelines corresponding to each geometric object; acquiring user-input language commands and update data; extracting multiple keywords from the language commands; determining multiple specific parameters and multiple computational pipelines based on each keyword; and constructing a target slideshow file based on each specific parameter, each computational pipeline, and the update data. This invention, by reconstructing the computational logic, achieves a fully automated closed loop of command parsing, data recalculation, and rendering generation, realizing end-to-end linkage updates of data, charts, and textual conclusions, ensuring that the updated textual summary accurately reflects the trend of new data, and significantly reducing the labor costs of periodic report maintenance.

[0059] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0060] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0061] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A dynamic slideshow update method based on user commands and custom templates, characterized in that, The dynamic slide update method based on user commands and custom templates includes: Each slide in the PowerPoint file is analyzed to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide. After matching each visual prediction box and each underlying object box, the corresponding geometric object is obtained. In the existing finite function library, determine the statistical functions corresponding to multiple geometric objects, and extract the corresponding specific parameters from the corresponding statistical functions for each geometric object; The multiple geometric objects are decomposed and recombined to obtain the computational pipeline corresponding to each geometric object; The system acquires user-input language commands and update data, extracts multiple keywords from the language commands, determines multiple specific parameters and multiple computational pipelines based on each keyword, and constructs a target slide file based on each specific parameter, each computational pipeline, and the update data.

2. The dynamic slideshow update method based on user instructions and custom templates according to claim 1, characterized in that, The process involves analyzing each slide page in the PowerPoint file to obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page. Matching each visual prediction box and each underlying object box yields the corresponding geometric object. Specifically, this includes: Obtain the PowerPoint file input by the user, construct a visual language model and an underlying document parser, and input each slide page of the PowerPoint file into the visual language model and the underlying document parser for analysis; For each slide page, the visual language model analyzes each page element within the slide page to obtain the semantic category and corresponding visual prediction box for each page element; For each slide page, the underlying document parser extracts each style metadata within the slide page to obtain the underlying object frame of each style metadata. Calculate the intersection-union ratio (IUGR) between each visual prediction box and each underlying object box. When the IUGR exceeds a preset threshold, add the semantic category corresponding to the visual prediction box to the style metadata corresponding to the underlying object box, and finally obtain all geometric objects.

3. The dynamic slideshow update method based on user instructions and custom templates according to claim 1, characterized in that, The step of determining statistical functions corresponding to multiple geometric objects in a pre-constructed finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object, further includes: Define multiple semantic slots in the underlying database, wherein the semantic slots are used to match query conditions in the underlying database; The scope description of each table title is analyzed using a large language model. The unstructured text of each table title is converted into a structured slot and filled into the corresponding semantic slot. The table header corresponding to each table title is analyzed using a large language model. The dimension information of each table header is matched with multiple field names in the underlying database to obtain the corresponding matching score. Based on all the matching scores, multiple target table titles that meet the preset conditions are filtered out, and all the semantic slots of each target table title are filtered out to obtain the semantically matched target table name.

4. The dynamic slideshow update method based on user instructions and custom templates according to claim 3, characterized in that, The geometric objects include: line charts, bar charts, scatter plots, and pie charts; The step of determining statistical functions corresponding to multiple geometric objects in a pre-constructed finite function library, and extracting specific parameters from the corresponding statistical functions for each geometric object, specifically includes: In the existing finite function library, determine the trend analysis function corresponding to the line chart, the summary statistics function corresponding to the bar chart, the regression function corresponding to the scatter plot, and the proportion analysis function corresponding to the pie chart. Multiple data distribution patterns and multiple target correlation information are extracted from the trend analysis function, the summary statistics function, the regression function, and the pie chart; Based on the target table names of the line chart, the bar chart, the scatter plot, and the pie chart, each data distribution pattern and each target association information is added to the corresponding semantic slot.

5. The dynamic slideshow update method based on user instructions and custom templates according to claim 3, characterized in that, The geometric objects include: user-defined complex diagrams; The step of decomposing and recombining multiple geometric objects to obtain a computational pipeline corresponding to each geometric object specifically includes: For each of the complex charts, the computational logic in the complex chart is structured using the large language model to obtain the table structure type, dimension header, constraint specification, source field mapping, and aggregation operator in the complex chart; Based on the table structure type, the dimension header, the constraint specification, and the source field mapping are defined as multiple query statements in the database, and the aggregation operator is embedded into each query statement to obtain multiple aggregation expressions; All the aggregate expressions are encapsulated to obtain the calculation pipeline corresponding to each of the table structure types.

6. The dynamic slideshow update method based on user instructions and custom templates according to claim 3, characterized in that, The process of acquiring user-input language commands and update data, extracting multiple keywords from the language commands, determining multiple specific parameters and multiple computational pipelines based on each keyword, and constructing a target slide file based on each specific parameter, each computational pipeline, and the update data specifically includes: Obtain the user's input language commands and the updated data of the current task, analyze the language commands, and obtain multiple keywords of the current task; Based on each keyword, determine all the geometric objects required for the current task; Extract the computation pipelines corresponding to all the geometric objects and multiple target table names containing specific parameters; The frame of the slide file is updated based on all the computation pipelines and all the target table names, and the updated data is populated into the frame to obtain the target slide file for the current task; The target slide file includes several new statistical data points.

7. The dynamic slideshow update method based on user instructions and custom templates according to claim 6, characterized in that, The process of acquiring user-input language commands and update data, extracting multiple keywords from the language commands, determining multiple specific parameters and multiple computational pipelines based on each keyword, constructing a target slide file based on each specific parameter, each computational pipeline, and the update data, further includes: By comparing all stored historical data with all the new statistical data, the corresponding data difference information, trend change data, dimension change data, and abnormal fluctuation data are obtained. After structuring the data difference information, the trend change data, the dimension change data, and the abnormal fluctuation data respectively, we obtain the difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector. The difference magnitude vector, the difference direction vector, the difference significance vector, and the difference dimension vector are input into the constructed analysis model for analysis, and the output includes a description of trend changes, logical causal relationships, and trend consistency information. The accuracy of the target slide file is obtained by performing contextual consistency verification and multi-dimensional cross-validation based on the trend change description, the logical causal relationship, and the trend consistency information.

8. A dynamic slideshow update system based on user commands and custom templates, characterized in that, The dynamic slide update system based on user commands and custom templates is used to implement the dynamic slide update method based on user commands and custom templates as described in any one of claims 1-7, wherein the dynamic slide update system based on user commands and custom templates includes: The multimodal layout module is used to analyze each slide page in the slide file, obtain multiple visual prediction boxes and multiple underlying object boxes for each slide page, and obtain the corresponding geometric object after matching each visual prediction box and each underlying object box. The logic extraction module is used to determine the statistical functions corresponding to multiple geometric objects in the finite function library that has been built, and to extract the corresponding specific parameters from the corresponding statistical functions for each geometric object; The logic reconstruction module is used to decompose and reorganize multiple geometric objects to obtain the computational pipeline corresponding to each geometric object; The instruction parsing module is used to obtain the language instructions and update data input by the user, extract multiple keywords from the language instructions, determine multiple specific parameters and multiple calculation pipelines based on each keyword, and construct a target slide file based on each specific parameter, each calculation pipeline and the update data.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and a dynamic slide update program based on user instructions and custom templates stored in the memory and executable on the processor. When the dynamic slide update program based on user instructions and custom templates is executed by the processor, it implements the steps of the dynamic slide update method based on user instructions and custom templates as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a dynamic slide update program based on user instructions and a custom template, which, when executed by a processor, implements the steps of the dynamic slide update method based on user instructions and a custom template as described in any one of claims 1-7.

Images