Display system for realizing graphical data through natural language interaction
By generating a graphical data display system through natural language interaction, the problems of data integration and high professional knowledge requirements are solved, the data visualization effect and analysis accuracy are improved, and the enterprise management efficiency is increased.
Patent Information
- Application Number
- CN202511035869.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-25
- Publication Date
- 2025-11-11
AI Technical Summary
Existing data analysis systems suffer from limitations in data integration, high requirements for professional knowledge, and poor data visualization, which negatively impacts the user experience and decision-making accuracy for enterprise managers.
A graphical data display system based on natural language interaction is implemented, including a user interaction module, a graphics generation module, and a business application module. It utilizes natural language processing models and database mapping rules to generate graphical data displays that conform to user intent.
It lowers the professional knowledge requirements for data analysis, improves data visualization and enterprise management efficiency, and enhances the accuracy of data analysis and decision-making.
Smart Images
Figure CN120929653A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of data transformation technology, specifically, it relates to a system for displaying graphical data through natural language interaction. Background Technology
[0002] With the development of information technology, the analysis and processing of data for every enterprise has become increasingly important. However, while existing data analysis systems can perform data analysis to a certain extent, they have some problems and shortcomings.
[0003] The diverse and complex sources of data, with enterprise data potentially scattered across different business units and varied database structures, make data integration an exceptionally arduous task. Enterprise managers struggle to manage this disorganized data in a unified manner. Furthermore, the data analysis methods required are relatively traditional, typically employing statistical or machine learning approaches for complex preprocessing and cleaning. This demands a high level of professional knowledge and programming skills from users, posing a significant obstacle, especially for most enterprise managers. Simultaneously, the data visualization is inadequate, failing to effectively and intuitively present the analysis results tailored to the characteristics of the enterprise data, impacting the user experience for enterprise managers and hindering their ability to make accurate business decisions. Summary of the Invention
[0004] To address the aforementioned problems and technical deficiencies, this application adopts the following technical solution: a system for displaying graphical data through natural language interaction, comprising:
[0005] The user interaction module is used to interact with users in natural language. It parses the natural language input by users and extracts keywords to obtain user intent and keywords, and generates database language based on user intent and keywords.
[0006] The graph generation module is used to read target data using a database language, perform graphical transformation on the target data based on mapping rules using a data graphing model, and generate and display graphical data.
[0007] The data prediction module is used to read historical data of the target data, generate historical data change curves based on the historical data, calculate the change function of the target data based on the historical change curves, use the change function to predict and calculate the subsequent data of the target data within a preset period, and then use a data graphics model based on mapping rules to perform graphical transformation on the subsequent data, generating and displaying graphical prediction data.
[0008] The business application module is used to preset multiple business scenario templates, select the corresponding business scenario template according to the user's intent, and perform style conversion on the generated graphical data so that the style of the generated and displayed graphical data matches the user's intent and application scenario.
[0009] Preferably, the process of parsing natural language and extracting keywords is as follows:
[0010] The natural language processing model is used to perform word segmentation calculation on the natural language and obtain the word segmentation vector for each word.
[0011] The cross-entropy loss algorithm is used to calculate the word segmentation vectors to obtain the intent classification of each word segmentation;
[0012] The sequence labeling loss algorithm is used to calculate the word segmentation vectors and obtain the part-of-speech tag for each word segmentation.
[0013] By combining intent classification and part-of-speech tagging, each word segment is identified to obtain keywords;
[0014] The key intent vector of keywords is calculated using an attention mechanism based on intent classification.
[0015] Combine the key intent vectors of all keywords to generate a structured intent, which is then used as the user intent.
[0016] Furthermore, the generation of structured intent includes:
[0017] A domain-specific dictionary-enhanced word segmentation engine based on the BERT model is used to generate word segmentation vectors. It takes key intent vectors as input and outputs key intent sequences with contextual semantics.
[0018] The key intent sequence is input into a multi-layer intent classifier, which outputs a probability distribution. Based on the probability distribution, a BiLSTM-CRF model is used to label parts of speech and filter out non-key function words.
[0019] For the selected keywords, a multi-head attention mechanism is used to calculate the weights. Based on the time-related keywords and the historical data range, the logic is calculated based on the trigger ratio of the metric keywords, and the combined vectors are used to generate structured intents.
[0020] Furthermore, the user interaction module also includes:
[0021] The word segmentation vector optimization module introduces dynamic dimension adaptation technology in the word segmentation vector calculation process. It automatically adjusts the vector dimension according to the semantic domain to which the word belongs. The vector dimension is dynamically expanded through the domain word vector fine-tuning model. While maintaining the basic dimension of ordinary words, it dynamically expands the dimension of professional words.
[0022] The multi-granularity intent classification module is used to superimpose a hierarchical classification strategy on the basis of the cross-entropy loss algorithm. First, the first-level intent is determined by coarse-grained classification, and then the second-level intent is identified by fine-grained classification. The intent probability distribution matrix is generated by combining the two-level classification results. When the intent graph probability is lower than the threshold, multi-intent fusion processing is automatically triggered.
[0023] The keyword ambiguity resolution module is used to establish a polysemous semantic disambiguation knowledge base, store the context association rules of common ambiguous words in the industry, and perform secondary verification by calculating the co-occurrence probability of word segmentation and context, combined with the entity type library of the database.
[0024] Preferably, the database language generation process includes:
[0025] Perform attribute analysis on the data types in the database to obtain the attribute dimension characteristics of each data type;
[0026] Analyze the attribute dimension features, generate lightweight mapping values for data types, compress the original values into 16-bit hash codes, establish a mapping relationship with the coordinate axes, and generate mapping rules based on the mapping relationship;
[0027] Establish a lightweight mapping value conversion relationship between keywords and data types in the database, and construct a key entity topology graph;
[0028] The extracted keywords are semantically identified, and the identified semantics are matched with the data types in the database to identify the keyword types;
[0029] Based on the key entity topology graph, the matched keyword types are used as key entities that correspond one-to-one with the keywords.
[0030] Generate data connection relationships between key entities based on user intent, and translate user intent into database language based on data connection relationships and key entities.
[0031] Furthermore, before graphically converting the database data, it is necessary to establish a composite mapping formula based on the mapping rules, use the composite mapping formula to generate a lightweight mapping index between mapping values, and read other database data that references non-target data.
[0032] Furthermore, the graphical transformation involves cleaning, desensitizing, and formatting the read data, then using a mapping engine to map the data type and value to corresponding mapping coordinates to generate a mapping image. Finally, a vector rendering algorithm is used to render the mapping image to generate graphical data.
[0033] Preferably, the generation of the graphical prediction data includes:
[0034] Data is anonymized and made secure according to user permission scope, and historical data of the target data is converted into a standardized JSON matrix using a format converter;
[0035] Based on mapping rules, data points in historical data are converted into polar coordinates, and control points are adaptively added for abrupt change periods to generate historical data change curves.
[0036] Read historical data from the historical change curve, use the change function to perform fitting calculations, and obtain subsequent data within the preset period;
[0037] Based on mapping rules, a predicted change curve is generated for subsequent data. This predicted change curve is superimposed on the historical change curve as a semi-transparent gradient curve, and confidence intervals are marked with color bands.
[0038] The piecewise Bézier curve algorithm is used to render the changing curves, generate vector graphics, and obtain graphical prediction data.
[0039] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the program, it implements the content of a graphical data display system that achieves interactive natural language communication as described above.
[0040] A computer-readable storage medium having a computer program stored thereon, characterized in that, when executed by a processor, the computer program implements the content of a graphical data display system that achieves interactive natural language communication as described above.
[0041] Compared to existing technologies, the beneficial effects of this application are as follows:
[0042] (1) This application obtains user intent and keywords by parsing the natural language input by the user and extracting keywords, and generates database language based on user intent and keywords, so that even non-professionals can generate database language that meets their own intent, reducing the threshold of use, facilitating the integration of complex business data of enterprises, and simplifying the data analysis process.
[0043] (2) This application improves the visualization effect by converting the user's natural language into database language and then converting the target data read by the database language into graphical data, so as to provide enterprise managers with intuitive and effective visualization graphics, improve the accuracy of enterprise data analysis and the efficiency of enterprise management. Attached Figure Description
[0044] In the attached diagram:
[0045] Figure 1 This is a schematic diagram of the system structure according to an embodiment of this application;
[0046] Figure 2 This is a schematic diagram of the device structure according to an embodiment of this application. Detailed Implementation
[0047] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of this application, but not all embodiments. Generally, the components of the embodiments of this application described and shown in the accompanying drawings can be arranged and designed in various different configurations.
[0048] Example 1
[0049] like Figure 1 As shown, a system for displaying graphical data through natural language interaction includes:
[0050] The user interaction module interacts with users in natural language, parses the natural language input by users and extracts keywords to obtain user intent and keywords, and generates database language based on user intent and keywords.
[0051] The process of parsing natural language and extracting keywords is as follows:
[0052] The natural language processing model is used to perform word segmentation calculation on the natural language and obtain the word segmentation vector for each word.
[0053] The cross-entropy loss algorithm is used to calculate the word segmentation vectors to obtain the intent classification of each word segmentation;
[0054] The sequence labeling loss algorithm is used to calculate the word segmentation vectors and obtain the part-of-speech tag for each word segmentation.
[0055] By combining intent classification and part-of-speech tagging, each word segment is identified to obtain keywords;
[0056] The key intent vector of keywords is calculated using an attention mechanism based on intent classification.
[0057] Combine the key intent vectors of all keywords to generate a structured intent, which is then used as the user intent.
[0058] Generating structured intents includes:
[0059] A domain-specific dictionary-enhanced word segmentation engine based on the BERT model is used to generate word segmentation vectors. It takes key intent vectors as input and outputs key intent sequences with contextual semantics.
[0060] The key intent sequence is input into a multi-layer intent classifier, which outputs a probability distribution. Based on the probability distribution, a BiLSTM-CRF model is used to label parts of speech and filter out non-key function words.
[0061] For the selected keywords, a multi-head attention mechanism is used to calculate the weights. Based on the time-related keywords and the historical data range, the logic is calculated based on the trigger ratio of the metric keywords, and the combined vectors are used to generate structured intents.
[0062] The user interaction module also includes:
[0063] The word segmentation vector optimization module introduces dynamic dimension adaptation technology in the word segmentation vector calculation process. It automatically adjusts the vector dimension according to the semantic domain to which the word belongs. The vector dimension is dynamically expanded through the domain word vector fine-tuning model. While maintaining the basic dimension of ordinary words, it dynamically expands the dimension of professional words.
[0064] The multi-granularity intent classification module is used to superimpose a hierarchical classification strategy on the basis of the cross-entropy loss algorithm. First, the first-level intent is determined by coarse-grained classification, and then the second-level intent is identified by fine-grained classification. The intent probability distribution matrix is generated by combining the two-level classification results. When the intent graph probability is lower than the threshold, multi-intent fusion processing is automatically triggered.
[0065] The keyword ambiguity resolution module is used to establish a polysemous semantic disambiguation knowledge base, store the context association rules of common ambiguous words in the industry, and perform secondary verification by calculating the co-occurrence probability of word segmentation and context, combined with the entity type library of the database.
[0066] The process of generating a database language includes:
[0067] Perform attribute analysis on the data types in the database to obtain the attribute dimension characteristics of each data type;
[0068] Analyze the attribute dimension features, generate lightweight mapping values for data types, compress the original values into 16-bit hash codes, establish a mapping relationship with the coordinate axes, and generate mapping rules based on the mapping relationship;
[0069] Establish a lightweight mapping value conversion relationship between keywords and data types in the database, and construct a key entity topology graph;
[0070] The extracted keywords are semantically identified, and the identified semantics are matched with the data types in the database to identify the keyword types;
[0071] Based on the key entity topology graph, the matched keyword types are used as key entities that correspond one-to-one with the keywords.
[0072] Generate data connection relationships between key entities based on user intent, and translate user intent into database language based on data connection relationships and key entities.
[0073] The graph generation module uses a database language to read the target data, and uses a data graphing model based on mapping rules to perform graphical transformation on the target data, generating and displaying graphical data.
[0074] Before graphically transforming database data, each data type needs to be converted into a lightweight mapping value. A composite mapping formula is then established based on the mapping rules. The composite mapping formula is used to generate a mapping index between lightweight mapping values and to read other database data that references non-target data.
[0075] The graphical transformation involves cleaning, desensitizing, and formatting the read data, then using a mapping engine to map the data type and value to corresponding coordinates to generate a mapped image. Finally, a vector rendering algorithm is used to render the mapped image to generate graphical data.
[0076] The data prediction model is used to read historical data of the target data, generate historical data change curves based on the historical data, calculate the change function of the target data based on the historical change curves, use the change function to predict and calculate the subsequent data of the target data within a preset period, and then use a data graphics model based on mapping rules to graphically transform the subsequent data, generating and displaying graphical prediction data.
[0077] The generation of graphical prediction data includes:
[0078] Data is anonymized and made secure according to user permission scope, and historical data of the target data is converted into a standardized JSON matrix using a format converter;
[0079] Based on mapping rules, data points in historical data are converted into polar coordinates, and control points are adaptively added for abrupt change periods to generate historical data change curves.
[0080] Read historical data from the historical change curve, use the change function to perform fitting calculations, and obtain subsequent data within the preset period;
[0081] Based on mapping rules, a predicted change curve is generated for subsequent data. This predicted change curve is superimposed on the historical change curve as a semi-transparent gradient curve, and confidence intervals are marked with color bands.
[0082] The piecewise Bézier curve algorithm is used to render the changing curves, generate vector graphics, and obtain graphical prediction data.
[0083] The business application module is used to preset multiple business scenario templates, select the corresponding business scenario template according to the user's intent, and perform style conversion on the generated graphical data so that the style of the generated and displayed graphical data matches the user's intent and application scenario.
[0084] Example 2
[0085] like Figure 2As shown, from a hardware perspective, this application provides an embodiment of an electronic device that implements all or part of a graphical data display system through natural language interaction. The electronic device includes a service processor and a distributed memory. The service processor is connected to the memory. The distributed memory stores a service self-management program configured to store machine-readable instructions. The service processor executes the service self-management program. When the instructions are executed by the processor, they implement the graphical data display system that implements graphical data display through natural language interaction as described above.
[0086] Example 3
[0087] Embodiments of this application also provide a computer-readable storage medium capable of implementing a graphical data display system that uses natural language interaction and is executed by a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all the contents of the graphical data display system that uses natural language interaction and is executed by a server or client as the execution subject in the above embodiments.
[0088] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications, improvements, and substitutions without departing from the concept of this application, and these all fall within the protection scope of this application.
Claims
1. A system for displaying graphical data through natural language interaction, characterized in that, include: The user interaction module is used to interact with users in natural language. It parses the natural language input by users and extracts keywords to obtain user intent and keywords, and generates database language based on user intent and keywords. The graph generation module is used to read target data using a database language, perform graphical transformation on the target data based on mapping rules using a data graphing model, and generate and display graphical data. The data prediction module is used to read historical data of the target data, generate historical data change curves based on the historical data, calculate the change function of the target data based on the historical change curves, use the change function to predict and calculate the subsequent data of the target data within a preset period, and then use a data graphics model based on mapping rules to perform graphical transformation on the subsequent data, generating and displaying graphical prediction data. The business application module is used to preset multiple business scenario templates, select the corresponding business scenario template according to the user's intent, and perform style conversion on the generated graphical data so that the style of the generated and displayed graphical data matches the user's intent and application scenario.
2. The system for displaying graphical data through natural language interaction according to claim 1, characterized in that, The process of parsing natural language and extracting keywords is as follows: The natural language processing model is used to perform word segmentation calculation on the natural language and obtain the word segmentation vector for each word. The cross-entropy loss algorithm is used to calculate the word segmentation vectors to obtain the intent classification of each word segmentation; The sequence labeling loss algorithm is used to calculate the word segmentation vectors and obtain the part-of-speech tag for each word segmentation. By combining intent classification and part-of-speech tagging, each word segment is identified to obtain keywords; The key intent vector of keywords is calculated using an attention mechanism based on intent classification. Combine the key intent vectors of all keywords to generate a structured intent, which is then used as the user intent.
3. The graphical data display system based on natural language interaction according to claim 2, characterized in that, The generated structured intent includes: A domain-specific dictionary-enhanced word segmentation engine based on the BERT model is used to generate word segmentation vectors. It takes key intent vectors as input and outputs key intent sequences with contextual semantics. The key intent sequence is input into a multi-layer intent classifier, which outputs a probability distribution. Based on the probability distribution, a BiLSTM-CRF model is used to label parts of speech and filter out non-key function words. For the selected keywords, a multi-head attention mechanism is used to calculate the weights. Based on the time-related keywords and the historical data range, the logic is calculated based on the trigger ratio of the metric keywords, and the combined vectors are used to generate structured intents.
4. The graphical data display system based on natural language interaction according to claim 3, characterized in that, The user interaction module also includes: The word segmentation vector optimization module introduces dynamic dimension adaptation technology in the word segmentation vector calculation process. It automatically adjusts the vector dimension according to the semantic domain to which the word belongs. The vector dimension is dynamically expanded through the domain word vector fine-tuning model. While maintaining the basic dimension of ordinary words, it dynamically expands the dimension of professional words. The multi-granularity intent classification module is used to superimpose a hierarchical classification strategy on the basis of the cross-entropy loss algorithm. First, the first-level intent is determined by coarse-grained classification, and then the second-level intent is identified by fine-grained classification. The intent probability distribution matrix is generated by combining the two-level classification results. When the intent graph probability is lower than the threshold, multi-intent fusion processing is automatically triggered. The keyword ambiguity resolution module is used to establish a polysemous semantic disambiguation knowledge base, store the context association rules of common ambiguous words in the industry, and perform secondary verification by calculating the co-occurrence probability of word segmentation and context, combined with the entity type library of the database.
5. A system for displaying graphical data through natural language interaction according to claim 1, characterized in that, The generation process of the database language includes: Perform attribute analysis on the data types in the database to obtain the attribute dimension characteristics of each data type; Analyze the attribute dimension features, generate lightweight mapping values for data types, compress the original values into 16-bit hash codes, establish a mapping relationship with the coordinate axes, and generate mapping rules based on the mapping relationship; Establish a lightweight mapping value conversion relationship between keywords and data types in the database, and construct a key entity topology graph; The extracted keywords are semantically identified, and the identified semantics are matched with the data types in the database to identify the keyword types; Based on the key entity topology graph, the matched keyword types are used as key entities that correspond one-to-one with the keywords. Generate data connection relationships between key entities based on user intent, and translate user intent into database language based on data connection relationships and key entities.
6. A system for displaying graphical data through natural language interaction according to claim 5, characterized in that, Before graphically converting the database data, it is necessary to establish a composite mapping formula based on the mapping rules, use the composite mapping formula to generate a lightweight mapping index between mapping values, and read other database data that references non-target data.
7. A system for displaying graphical data through natural language interaction according to claim 6, characterized in that, The graphical transformation involves cleaning, desensitizing, and formatting the read data, then using a mapping engine to map the data type and value to corresponding mapping coordinates to generate a mapping image. Finally, a vector rendering algorithm is used to render the mapping image to generate graphical data.
8. A system for displaying graphical data through natural language interaction according to claim 1, characterized in that, The generation of the graphical prediction data includes: Data is anonymized and made secure according to user permission scope, and the historical data of the target data is converted into a standardized JSON matrix using a format converter; Based on mapping rules, data points in historical data are converted into polar coordinates, and control points are adaptively added for abrupt change periods to generate historical data change curves. Read historical data from the historical change curve, use the change function to perform fitting calculations, and obtain subsequent data within the preset period; Based on mapping rules, a predicted change curve is generated for subsequent data. This predicted change curve is superimposed on the historical change curve as a semi-transparent gradient curve, and confidence intervals are marked with color bands. The piecewise Bézier curve algorithm is used to render the changing curves, generate vector graphics, and obtain graphical prediction data.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the content of the graphical data display system that achieves graphical data display through natural language interaction as described in claim 1.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the content of the graphical data display system described in claim 1, which uses natural language interaction.