Enterprise data visual analysis method and system based on natural language

Abstract

The invention relates to the technical field of enterprise data analysis, and discloses an enterprise data visual analysis method and system based on a natural language, and the method comprises the steps: collecting enterprise business data and a natural language query text, and determining a target analysis dimension; dividing core, auxiliary and associated data subtypes according to the analysis intention, and constructing a to-be-analyzed dimension set; calculating a dimension correlation degree between the subtypes, generating an analysis correlation graph and determining an optimal generation tree; and determining a visualization priority order based on the optimal generation tree, and sequentially generating visualization results. The system comprises a data acquisition unit, a dimension division unit, an association graph construction unit, a priority determination unit and a visualization generation unit. According to the invention, through natural language interaction and intelligent association analysis, the accuracy and efficiency of enterprise data visualization are improved, and the method is suitable for data analysis requirements in complex business scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise data analysis, and in particular to a natural language-based enterprise data visualization analysis method and system. Background Art

[0002] With the rapid growth of enterprise data, efficiently extracting valuable information from massive amounts of business data has become a key challenge for business decision-making. Traditional data analysis methods often rely on manually written queries or fixed reporting tools. This approach is not only inefficient but also requires high technical skills. For non-technical business personnel, complex query syntax and data analysis processes often become bottlenecks that hinder rapid insight.

[0003] Currently, there are several data analysis tools on the market that use natural language processing (NLP) to input queries using natural language. However, these tools still have significant limitations in practical applications. Most tools can only handle simple query intent and cannot deeply understand complex analytical needs, such as multi-dimensional correlation analysis or dynamic data sub-typing. Existing methods lack intelligent prioritization during visualization generation, resulting in redundant or unfocused charts, reducing user experience and analytical efficiency. Traditional systems often ignore the correlations between data sub-types, making it difficult to build a globally optimized visualization solution.

[0004] To address these issues, a method for enterprise data visualization analysis that combines natural language processing with graph optimization techniques is needed. This method aims to achieve the following: accurately identify the core analytical intent in natural language queries and dynamically divide data into subtypes; quantify the correlations between data subtypes, construct an analytical correlation graph, and determine the optimal spanning tree; and generate visualization results based on priority order, ensuring clear presentation logic and a clear focus. By introducing multidimensional correlation calculation and a spanning tree optimization algorithm, this invention effectively addresses the existing problems of single analysis dimensions and fragmented visualization results, providing a more efficient and intelligent solution for enterprise data visualization analysis. Summary of the Invention

[0005] The purpose of the present invention is to provide a natural language-based enterprise data visualization analysis method and system to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a natural language-based enterprise data visualization analysis method, the method comprising: Collect the enterprise business data to be analyzed and the corresponding natural language query text, and obtain the target analysis dimension based on the enterprise business data and the natural language query text; Based on the analysis intent of the natural language query text, the data subtypes covered by the target analysis dimension are divided into core data subtypes and auxiliary data subtypes, and associated data subtypes are determined among the data subtypes not covered by the target analysis dimension, wherein the core data subtypes, auxiliary data subtypes, and associated data subtypes constitute a set of dimensions to be analyzed; Determine the dimensional correlation between any two adjacent subtypes in the set of dimensions to be analyzed, and construct an analysis correlation graph of the set of dimensions to be analyzed with the subtypes in the set of dimensions to be analyzed as vertices and the dimensional correlation as edges connecting the vertices; Determining an optimal spanning tree of the analysis association graph, and determining a visualization priority order of each subtype in the set of dimensions to be analyzed based on the optimal spanning tree; The visualization results of each subtype in the set of dimensions to be analyzed are generated in sequence according to the visualization priority order.

[0007] Preferably, obtaining target analysis dimensions based on the enterprise business data and natural language query text includes: According to the natural language query text, counting analysis dimensions in the historical analysis task that match the enterprise business data; According to the semantic features of the natural language query text in each analysis dimension, the statistically obtained analysis dimensions are classified to form one or more analysis dimension sets; A target analysis dimension set that matches the current analysis intent of the natural language query text is determined in the analysis dimension set, and the dimensions covered by the target analysis dimension set are used as the target analysis dimensions.

[0008] Preferably, classifying the statistically obtained analysis dimensions according to the semantic features of the natural language query text in each analysis dimension includes: Obtaining a predefined semantic feature standard interval in the enterprise data management system; For any analysis dimension obtained through statistics, the semantic feature standard interval to which the semantic features of the natural language query text in the analysis dimension belong is identified, and the analysis dimension is divided into an analysis dimension set corresponding to the identified semantic feature standard interval.

[0009] Preferably, based on the analysis intent of the natural language query text, dividing the data subtypes covered by the target analysis dimension into core data subtypes and auxiliary data subtypes includes: Determining the core analysis object and auxiliary analysis objects represented by the analysis intent of the natural language query text; Determining a first data coverage range of the core analysis object in the target analysis dimension, and determining a second data coverage range of the auxiliary analysis object in the target analysis dimension; The data subtype covered by the first data coverage range within the enterprise business data range is determined as a core data subtype, and the data subtype covered by the second data coverage range within the enterprise business data range is determined as an auxiliary data subtype.

[0010] Preferably, determining the associated data subtypes in the data subtypes not covered by the target analysis dimension includes: Setting a first extension range for a core data subtype in the target analysis dimension, and setting a second extension range for an auxiliary data subtype in the target analysis dimension; For the data subtype not covered by the target analysis dimension, if it is adjacent to the core data subtype, the data subtype of the first extended range is selected as the associated data subtype; if it is adjacent to the auxiliary data subtype, the data subtype of the second extended range is selected as the associated data subtype.

[0011] Preferably, determining the dimensional correlation between any two adjacent subtypes in the set of dimensions to be analyzed includes: For adjacent first and second subtypes in the set of dimensions to be analyzed, identifying a type category of each of the first and second subtypes, wherein the type category includes one of a core data subtype, an auxiliary data subtype, and an associated data subtype; determining, based on the identified type category, a first dimension associated parameter of the first subtype, and determining a second dimension associated parameter of the second subtype; A difference value between the first dimensional association parameter and the second dimensional association parameter is calculated, and a dimensional association degree between the first subtype and the second subtype is determined according to the difference value.

[0012] Preferably, determining the dimensional association between the first subtype and the second subtype according to the difference value includes: Identifying dimension-related parameters of each subtype in the set of dimensions to be analyzed, and determining an average value of the identified dimension-related parameters; determining a plurality of difference intervals according to the average value, and determining a target difference interval in which the calculated difference value between the first dimension associated parameter and the second dimension associated parameter lies; The preset correlation degree corresponding to the target difference interval is determined as the dimensional correlation degree between the first subtype and the second subtype.

[0013] Preferably, generating visualization results of each subtype in the set of dimensions to be analyzed in sequence according to the visualization priority order includes: Identifying generation parameters of each generation node in the visualization priority order, and constructing an error feature sequence based on the identified generation parameters, wherein the arrangement order of each parameter in the error feature sequence is consistent with the visualization priority order; Inputting the error feature sequence into an error correction model to generate a visualization action output for each of the generation nodes through the error correction model; According to the generated visualization action outputs, corresponding visualization results of each subtype are generated.

[0014] Preferably, the present invention further includes a natural language-based enterprise data visualization analysis system, the system comprising: A data acquisition unit is used to collect the enterprise business data to be analyzed and the corresponding natural language query text, and obtain the target analysis dimension based on the enterprise business data and the natural language query text; a dimension division unit, configured to divide the data subtypes covered by the target analysis dimension into core data subtypes and auxiliary data subtypes based on the analysis intent of the natural language query text, and determine associated data subtypes among the data subtypes not covered by the target analysis dimension, wherein the core data subtypes, auxiliary data subtypes, and associated data subtypes constitute a set of dimensions to be analyzed; an association graph construction unit, configured to determine the dimensional association degree between any two adjacent subtypes in the set of dimensions to be analyzed, and to construct an analysis association graph of the set of dimensions to be analyzed, using the subtypes in the set of dimensions to be analyzed as vertices and the dimensional association degrees as edges connecting the vertices; a priority determination unit, configured to determine an optimal spanning tree of the analysis association graph, and determine a visualization priority order of each subtype in the set of dimensions to be analyzed based on the optimal spanning tree; The visualization generating unit is configured to generate visualization results of each subtype in the set of dimensions to be analyzed in sequence according to the visualization priority order.

[0015] Preferably, the visualization generation unit is specifically used to identify the generation parameters of each generation node in the visualization priority order, and construct an error feature sequence based on the identified generation parameters, and the arrangement order of each parameter in the error feature sequence is consistent with the visualization priority order; input the error feature sequence into the error correction model to generate a visualization action output for each of the generation nodes through the error correction model; and generate corresponding visualization results of each subtype according to each generated visualization action output.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention improves the intelligence and efficiency of enterprise data visualization analysis by introducing natural language processing technology and graph theory optimization algorithms. Based on the analytical intent of natural language query text, it dynamically divides core data into subtypes, auxiliary data subtypes, and associated data subtypes, accurately capturing user needs and avoiding the information omissions caused by the single analysis dimension in traditional methods. By quantifying the correlation between data subtypes and constructing an analytical correlation graph, the system can optimize data display logic from a global perspective, ensuring that the visualization results are both comprehensive and focused.

[0017] Using an optimal spanning tree algorithm to prioritize visualizations effectively addresses the issues of redundant charts or confusing display logic in existing technologies. The system automatically generates clear, prioritized visualizations without requiring manual adjustments or filtering. This feature is particularly well-suited for multi-dimensional analysis in complex business scenarios, such as a comprehensive display of sales trends and regional correlations.

[0018] This paper dynamically optimizes the visualization generation process through an error correction model, further improving the accuracy and stability of the results. The introduction of error signature sequences enables the system to automatically correct for deviations caused by data noise or query ambiguity, ensuring that the final output chart is highly consistent with the user's intent.

[0019] In practical applications, this invention significantly lowers the technical barriers to enterprise data analysis. Business personnel no longer need to master specialized query languages ​​or data analysis tools; they can simply input natural language descriptions to obtain high-quality visualization results. Furthermore, the system is highly scalable and adaptable to the enterprise data management needs of various industries, such as multi-scenario analysis tasks in finance, retail, and manufacturing. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a diagram showing the working principle of the natural language-based enterprise data visualization analysis method of the present invention; Figure 2 Design diagram for data subtyping method; Figure 3 A design diagram for determining methods for Linked Data subtyping; Figure 4 A design diagram for the method of determining dimension correlation. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] See also Figure 1-Figure 4 The present invention relates to a natural language-based enterprise data visualization analysis method, and the specific implementation steps are as follows: Collect the business data to be analyzed and the corresponding natural language query text, and then obtain the target analysis dimension based on the business data and natural language query text. Business data includes all types of data generated during the enterprise's operations, such as financial data, sales data, and production data. Natural language query text is the query request submitted by the user in natural language.

[0023] Based on the analytical intent of the natural language query text, the data subtypes covered by the target analysis dimension are divided into core data subtypes and auxiliary data subtypes. Related data subtypes are then identified within the data subtypes not covered by the target analysis dimension. The core data subtypes, auxiliary data subtypes, and related data subtypes constitute the set of dimensions to be analyzed. The analytical intent reflects the core purpose and focus of the user's query.

[0024] Determine the dimensional correlation between any two adjacent subtypes in the set of dimensions to be analyzed. Construct an analysis correlation graph for the set of dimensions to be analyzed, using the subtypes in the set of dimensions as vertices and the dimensional correlation as edges connecting the vertices. Dimension correlation measures the degree of association between subtypes.

[0025] Determine the optimal spanning tree for the analysis association graph and, based on the optimal spanning tree, determine the visualization priority order for each subtype in the set of dimensions to be analyzed. The optimal spanning tree can optimize the connection relationships between subtypes, thereby determining a reasonable visualization order.

[0026] Generate visualization results for each subtype of the dimension set to be analyzed in order of visualization priority. The visualization results present data information in the form of intuitive charts, graphs, etc.

[0027] Example 1:

[0028] When acquiring the target analysis dimension, the company's business data to be analyzed and the corresponding natural language query text are collected. Business data includes various data generated by the company throughout its operations, such as income and expenditure details and assets and liabilities in financial data, order information and customer purchase records in sales data, and output and raw material consumption in production data. Natural language query text is the user's query request expressed in everyday language, such as "Analyze the changing sales performance trends of each region this quarter" or "View the company's cost structure over the past year."

[0029] Based on the natural language query text, the analysis dimensions in historical analysis tasks that match the enterprise business data are counted. The historical analysis task database stores similar queries that have been processed in the past and their corresponding analysis dimensions. The system will perform semantic analysis on the natural language query text, extract the keywords, key phrases, and core intent of the query, and then compare these features with the queries in the historical analysis tasks to find historical analysis tasks that are similar in semantics and business logic, and then count the matching analysis dimensions. For example, when the natural language query text is "Analyze the sales in the second quarter of the East China region", the system will retrieve the analysis tasks in the historical analysis tasks involving keywords such as "East China region", "second quarter", and "sales", and count the analysis dimensions used by these tasks, such as time dimension, regional dimension, sales indicator, etc.

[0030] Based on the semantic features of the natural language query text within each analysis dimension, the statistically derived analysis dimensions are categorized to form one or more analysis dimension sets. Semantic feature analysis encompasses multiple aspects, including query keywords and their weights, sentence grammatical structure, semantic relationships (such as causal relationships, parallel relationships, and subordinate relationships), and context. To achieve accurate categorization, it is necessary to obtain predefined semantic feature standard intervals within the enterprise data management system. These standard intervals are formed by classifying and defining common semantic features based on the enterprise's business characteristics and data management requirements. For example, semantic features might be categorized into different standard intervals, such as "time-related," "region-related," "product-related," and "financial indicator-related," each with clear feature descriptions and scope definitions.

[0031] For any statistically derived analysis dimension, the system identifies the semantic features of the natural language query text within that analysis dimension and determines the specific semantic feature standard range to which it belongs. During this identification process, the system applies natural language processing techniques, such as lexical analysis, syntactic analysis, and semantic understanding, to conduct an in-depth analysis of the query text. For example, for the query text corresponding to an analysis dimension, "Analyze the sales of product A in North China in 2024," the system extracts semantic features such as "North China" (region-related), "2024" (time-related), "Product A" (product-related), and "sales" (indicator-related). It then matches these features with the semantic feature standard range to determine that the analysis dimension should be classified into the analysis dimension set corresponding to the semantic feature standard range related to "region-time-product-indicator."

[0032] After categorizing each analysis dimension, the system identifies a target analysis dimension set within the analysis dimension set that matches the current analysis intent of the natural language query. Analysis intent represents the core purpose and focus of the user's query, and requires a deep understanding of the natural language query text to grasp it. The system then performs a more detailed analysis of the current natural language query text to clarify the user's desired content, focus, and desired analysis perspective. For example, if the user's query text is "Compare the profit growth rates of different regions in the third quarter," the analysis intent is to conduct a comparative analysis of the profit growth rate metric based on region and time. The system then compares the characteristics of each analysis dimension set with the current analysis intent, selecting the analysis dimension set that best meets the requirements in terms of semantics, business logic, and analysis perspective, and identifies it as the target analysis dimension set.

[0033] The dimensions covered by the target analysis dimension set are considered target analysis dimensions. The target analysis dimension set contains multiple analysis dimensions that match the current analysis intent. Together, these dimensions form the framework for analyzing user queries. For example, the target analysis dimension set might include region, time, and profit indicator dimensions. The target analysis dimensions are these specific dimensions, and the system will use these dimensions to further analyze and process enterprise business data.

[0034] Throughout the process of acquiring target analysis dimensions, the system continuously processes and analyzes natural language query text and enterprise business data to ensure that the acquired target analysis dimensions accurately reflect the user's query requirements, laying a solid foundation for subsequent data visualization analysis. This process also fully leverages data from historical analysis tasks and predefined semantic feature standard intervals, improving the efficiency and accuracy of acquiring target analysis dimensions and enabling the system to respond to user queries more quickly and accurately.

[0035] Example 2:

[0036] When dividing core data subtypes and auxiliary data subtypes based on the analysis intent of natural language query text, it is necessary to clarify the core analysis objects and auxiliary analysis objects represented by the analysis intent of the natural language query text. The process of extracting analysis intent relies on natural language processing technology. The system will perform deep semantic analysis on the query text to identify the entities, relationships, and core demands of the user. For example, when the query text is "Analyze the proportion of R&D costs of each department in the first half of 2024 and its impact on product iteration efficiency", semantic analysis can determine that "R&D costs" is one of the core analysis objects, and "product iteration efficiency" is an auxiliary analysis object, because the user's core purpose is to understand the proportion of R&D costs, and at the same time, it is necessary to conduct a comprehensive analysis based on its impact on product iteration efficiency.

[0037] After determining the core and auxiliary analysis objects, you need to define their data coverage within the target analysis dimensions. Target analysis dimensions are specific analysis dimensions previously derived from enterprise business data and natural language query text, such as time, department, cost indicator, and efficiency indicator. For the core analysis object "R&D cost," when determining its primary data coverage within the target analysis dimensions, it's important to consider the specific scope of the core analysis object across each dimension. In terms of the time dimension, coverage is for the first half of 2024; in terms of the department dimension, coverage is for all R&D-related departments within the enterprise; and in terms of the cost indicator dimension, coverage is for all R&D cost components, such as employee salaries, equipment procurement costs, and material consumption. For the auxiliary analysis object "product iteration efficiency," when determining its secondary data coverage within the target analysis dimensions, similar definitions are required within the relevant dimensions. In terms of the time dimension, consistent with the core analysis object, coverage is for the first half of 2024; in terms of the department dimension, coverage is for departments with product iteration tasks; and in terms of the efficiency indicator dimension, coverage is for specific indicators such as product iteration cycle, new version release frequency, and demand response time.

[0038] The data subtypes covered by the first data coverage within the scope of enterprise business data are identified as core data subtypes. Enterprise business data includes data of multiple dimensions and types. The first data coverage matches and filters this data to determine specific data subtypes. For example, in an enterprise's financial data, data subtypes related to R&D costs may include personnel salary data, R&D equipment procurement data, and R&D material consumption data; while human resources data may include data subtypes such as the number of R&D personnel, job distribution, and years of service. These data subtypes are all covered by the first data coverage within the scope of enterprise business data and are therefore identified as core data subtypes.

[0039] Similarly, the data subtypes covered by the second data coverage within the scope of enterprise business data are determined to be auxiliary data subtypes. For the auxiliary analysis object "product iteration efficiency," its second data coverage in enterprise business data may include product R&D project management data, such as records of the iteration cycles of various projects; product version release data, such as the time and content of new version releases; and demand management data, such as the time it takes to receive, process, and respond to user requirements. These data subtypes constitute auxiliary data subtypes. Although they are not the core of user queries, they play an important auxiliary role in comprehensively analyzing the core analysis object.

[0040] During this process, it is important to ensure that the division of data subtypes accurately reflects the relationship between the analysis intent and the data. The system needs to have a clear understanding of the structure and content of the enterprise's business data and be able to accurately identify the corresponding subtypes based on the data coverage. For example, when determining the core data subtypes of R&D costs, it is necessary to distinguish between R&D costs and other cost types, and avoid including data subtypes that are not R&D costs; when determining the auxiliary data subtypes of product iteration efficiency, it is necessary to ensure that the selected data subtypes can indeed reflect all aspects of product iteration efficiency, rather than irrelevant or unrelated data.

[0041] Analysis intent can be complex and multi-layered, requiring the system to handle it flexibly. For example, some query texts may involve multiple core analysis objects and auxiliary analysis objects. In this case, the data coverage and corresponding subtypes need to be determined for each object separately. In another example, analysis intent may change with the context or further user interaction, which requires a certain degree of dynamic adjustment capability when dividing data subtypes. However, in this embodiment, the analysis intent and data subtype division are mainly determined based on the current natural language query text.

[0042] Example 3:

[0043] When determining associated data subtypes in data subtypes not covered by the target analysis dimension, you first need to set the first extended range for the core data subtype in the target analysis dimension, and set the second extended range for the auxiliary data subtype in the target analysis dimension. The setting of the extended range needs to be determined in combination with the inherent logic of the enterprise business data and the correlation between the data. For core data subtypes, the first extended range is usually a data range that is closely related to the core data subtype in business logic and can further supplement the core data analysis. For example, if the core data subtype is "quarterly sales of product A", its first extended range may include data ranges such as product A's sales cost, inventory changes, and marketing expenses. These data are directly related to sales in business and can help to analyze the core data more comprehensively.

[0044] The second extension of the auxiliary data subtype is to select data related to the auxiliary data subtype. For example, if the auxiliary data subtype is "Customer Satisfaction Survey Results," its second extension might include data such as customer purchase frequency, complaint records, and willingness to recommend. This data can supplement the information in the auxiliary data subtype from different perspectives and provide more support for the overall analysis.

[0045] After setting the expansion scope, you need to determine the proximity of data subtypes not covered by the target analysis dimension to core or auxiliary data subtypes. Based on this, you can select data subtypes within the expansion scope as associated data subtypes. "Adjacent" here doesn't just refer to physical storage location; more importantly, it refers to close connections in business logic, data structure, or semantic relationships.

[0046] Specifically, for a data subtype that is not covered by a target analysis dimension, the system will first analyze its association with the core data subtype. For example, suppose the data subtype that is not covered by the target analysis dimension is "raw material price fluctuations of product A", and the core data subtype is "production cost of product A". Since raw material price fluctuations directly affect production costs, the two have a close causal relationship in business logic, so this data subtype can be regarded as adjacent to the core data subtype. At this time, the system will select the corresponding data subtype from the first extended range set for the core data subtype as the associated data subtype. If the first extended range contains a data subtype related to "raw material price fluctuations", it will be determined as an associated data subtype.

[0047] If the data subtype not covered by the target analysis dimension is adjacent to the auxiliary data subtype, the related data subtype is selected from the second extended scope. For example, the auxiliary data subtype is "user evaluation of the function of product B", and the data subtype not covered by the target analysis dimension is "function iteration record of product B". Function evaluation and function iteration record are related in business logic. User evaluation may affect function iteration, and function iteration may change subsequent user evaluation. Therefore, the two are in an adjacent relationship. At this time, the system will search the second extended scope set for the auxiliary data subtype to see whether it contains data subtypes related to "function iteration records". If so, it will be determined as a related data subtype.

[0048] When determining the adjacent relationships between data subtypes, the system uses data association analysis technology to conduct a comprehensive analysis of the data subtype's field attributes, semantic features, and data flow within the business process. For example, this can be done by analyzing whether key fields within the data subtypes are associated, such as shared fields like product ID and timestamp; or by using semantic analysis to determine whether the business concepts represented by the data subtypes have a logical connection, causal relationship, or complementary relationship.

[0049] Furthermore, the extension scope is not set in stone and can be flexibly adjusted based on the business characteristics and data structure of an enterprise. The same data subtype may have different degrees of business relevance across different industries or enterprises, so the first and second extension scopes need to be appropriately defined based on actual circumstances. For example, in a manufacturing enterprise, product production hour data may have a strong correlation with the production cost core data subtype and should be included in the first extension scope. In contrast, in a service enterprise, similar data may have a lower degree of relevance and need not be included in the extension scope.

[0050] The process of identifying associated data subtypes is crucial for building a complete set of dimensions to be analyzed. Although these associated data subtypes aren't directly covered by the target analysis dimensions, their association with core or auxiliary data subtypes can enrich the dimensionality and depth of analysis, making the visualization results more comprehensive. For example, when analyzing sales performance, in addition to core sales data and auxiliary customer satisfaction data, associated data subtypes such as market competition data and promotional activity data can help companies better understand the causes and trends of sales performance fluctuations.

[0051] Example 4:

[0052] When determining the dimensional correlation between any two adjacent subtypes in the dimension set to be analyzed, it is necessary to identify the type categories of the first and second adjacent subtypes in the dimension set to be analyzed. The type category includes one of the core data subtype, auxiliary data subtype, and associated data subtype. For example, assuming that there are adjacent "R&D costs of each department in the third quarter of 2024" (core data subtype) and "Educational education distribution of R&D personnel" (auxiliary data subtype) in the dimension set to be analyzed, the system must first clarify the type of the two. The identification process is based on the previous dimension division logic. The core data subtype corresponds to the core object coverage data of the analysis intention, the auxiliary data subtype corresponds to the auxiliary object coverage data, and the associated data subtype is the supplementary data determined by the extended scope.

[0053] After identifying the type category, the first-dimensional association parameters for the first subtype and the second-dimensional association parameters for the second subtype need to be determined based on the type category. Dimensional association parameters are characteristic values ​​used to measure the degree of association between subtypes, and their values ​​are related to the type category. For example, core data subtypes may correspond to higher basic association weights, auxiliary data subtypes to a lower value, and associated data subtypes to a relatively lower weight. Specific parameters may include the degree of overlap in data fields, the strength of association within business processes, and semantic relevance. Taking "R&D Cost" (core) and "Educational Distribution of R&D Personnel" (auxiliary) as examples, the first-dimensional association parameters can be set to reflect the strength of the association between cost data and personnel structure data in the business logic, such as an indicator of the historical correlation between R&D personnel's education level and R&D investment. The second-dimensional association parameters correspond to the associated characteristic parameters of the education distribution data itself, such as an indicator of the potential correlation between different education levels and R&D efficiency.

[0054] Calculate the difference between the first dimension's correlation parameter and the second dimension's correlation parameter. The calculation method for this difference depends on the specific form of the parameter. If the parameter is numeric, the absolute or relative difference can be directly calculated. If the parameter is a semantic feature parameter, the degree of difference can be calculated using a semantic similarity algorithm. For example, if the first dimension's correlation parameter is 0.8 (indicating a strong correlation between R&D costs and personnel structure) and the second dimension's correlation parameter is 0.5 (indicating a moderate correlation between educational background distribution and R&D efficiency), the absolute difference between the two is 0.3.

[0055] To determine dimension correlation based on variance, we first need to identify the dimension correlation parameters for each subtype in the dimension set to be analyzed and calculate the average of these parameters. For example, if the dimension set to be analyzed contains five subtypes, and their dimension correlation parameters are 0.6, 0.8, 0.5, 0.7, and 0.9, respectively, the average value is (0.6 + 0.8 + 0.5 + 0.7 + 0.9) / 5 = 0.7.

[0056] Multiple variance intervals are then determined based on the average value. These intervals should cover the range of possible variance values ​​and reflect different levels of association. For example, using an average value of 0.7 as a benchmark, the variance intervals could be: [0, 0.2) for "high association," [0.2, 0.4) for "medium-high association," [0.4, 0.6) for "medium association," and [0.6, 1.0] for "low association." The specific intervals can be adjusted based on business needs and data characteristics to ensure that the intervals reasonably reflect the differences in association strength between subtypes.

[0057] Determine the target difference range that the calculated difference value falls in. Taking the aforementioned difference value of 0.3 as an example, it falls in the range [0.2, 0.4), which corresponds to the "medium-high correlation" range.

[0058] The preset correlation degree corresponding to the target difference interval is determined as the dimensional correlation degree between the first subtype and the second subtype. For example, the preset correlation degree corresponding to the "medium-high correlation" interval can be set to 0.75 (this value is only an example; the actual preset value should be determined based on business logic and historical data), indicating that the degree of correlation between the two adjacent subtypes is medium-high.

[0059] For another example, if the adjacent subtypes in the dimension set to be analyzed are "Product A Sales" (core data subtype, first dimension association parameter 0.9) and "Market Competitive Product Price" (association data subtype, second dimension association parameter 0.3), the difference between the two is 0.6, which is in the range of [0.6, 1.0], corresponding to "low correlation", and the preset correlation degree is 0.3, indicating that the correlation between sales and competitive product prices is low.

[0060] In practice, determining dimension association parameters requires comprehensive consideration of multiple factors. For example, for "Customer Complaint Rate" (an auxiliary data subtype) and "Product Return Rate" (an associated data subtype), dimension association parameters can be set based on indicators such as the co-occurrence frequency of complaint and return rates in historical data and the lagged correlation in time series. When calculating the difference value, if the association parameter for the complaint rate is 0.6 and the association parameter for the return rate is 0.4, the difference value is 0.2, which is in the interval [0, 0.2) and corresponds to "high correlation." The preset correlation degree is 0.9, indicating that the two are closely related in business.

[0061] The system needs to standardize the value logic of dimension association parameters to ensure that parameters of different subtypes are comparable. For example, the parameter value range of the core data subtype can be set to [0.7, 1.0], the auxiliary data subtype to [0.4, 0.7), and the associated data subtype to [0.1, 0.4]. This reflects the difference in importance of different types in the analysis, which in turn affects the calculation results of the correlation degree.

[0062] The process of determining dimension correlation is crucial for constructing the analytical correlation graph. It directly determines the weights of the edges between vertices (subtypes) in the graph, which in turn influences the subsequent construction of the optimal spanning tree and the determination of visualization priorities. By scientifically and rationally calculating dimension correlation, we can ensure that the analytical correlation graph accurately reflects the inherent connections between data subtypes, enabling the visualization process to be presented in an orderly manner based on the closeness of data correlations, thereby providing users with logically clear and focused data analysis results. The entire process requires the system to have the ability to accurately identify data types, a deep understanding of business logic, and the rational design of parameter calculations to ensure the accuracy and effectiveness of dimension correlations.

[0063] Example 5:

[0064] When generating visualization results for each subtype within the set of dimensions to be analyzed, following the visualization priority order, it is necessary to identify the generation parameters for each generation node in the visualization priority order and construct an error signature sequence based on these parameters. Generation parameters encompass various attributes of the data subtype, such as data format (e.g., numeric, text, time series), data precision (e.g., retaining two decimal places), data dimension (e.g., time dimension, region dimension), and data relationships (e.g., the strength of association with other subtypes). For example, for "R&D costs by department in the third quarter of 2024" (a core data subtype) and "Educational level distribution of R&D personnel" (an auxiliary data subtype), generation parameters for the former might include data time range (July-September 2024), department classification (e.g., technology department, product department), and cost components (e.g., salaries, equipment purchases). Generation parameters for the latter might include educational level (bachelor's, master's, doctoral), and personnel counting method.

[0065] After identifying the generation parameters, the system arranges them according to the visualization priority order to construct an error signature sequence. The order of the parameters in this sequence strictly aligns with the visualization priority order to ensure logical coherence in subsequent processing. For example, if the visualization priority order is "R&D cost → educational background distribution → competitor price," the parameters in the error signature sequence will be generated parameters for R&D cost, then educational background distribution, and finally competitor price. Each parameter is assigned a position in the sequence, reflecting its priority and importance in the visualization process.

[0066] The error feature sequence is input into the error correction model, which generates visual action outputs for each generated node. The error correction model is trained using historical data and machine learning algorithms. It can identify potential errors in the feature sequence and generate corresponding corrective actions. For example, when the error feature sequence contains inconsistent data formats (e.g., R&D cost amounts are expressed in both 10,000 yuan and yuan) or missing dimensions (e.g., educational background distribution without statistical time), the model generates visual action outputs such as standardized format conversion and supplementary time dimensions based on preset rules and training experience.

[0067] Taking the "R&D cost" generation node as an example, if there is an error in its generation parameters that "cost data is not classified and summarized by department", the error correction model will identify the deviation between the parameter and the preset standard (need to be classified by department), and then generate a visual action output of "reaggregating data by department dimension"; for the "education distribution" generation node, if the "number of master's degree personnel statistics omits a department" in the parameters, the model will generate an action output of "supplementing the master's degree personnel data of XX department".

[0068] After generating visualization action outputs, the system sequentially generates visualization results for each subtype based on these actions. The presentation format of visualization results is determined by the data characteristics and analysis requirements, and can include bar charts, line charts, pie charts, heat maps, and more. For example, "R&D cost" data can be generated into a department-by-department bar chart, visually displaying the cost contribution of each department; "education level distribution" data can be generated into a pie chart, showing the proportion of employees at different educational levels; and "competing product prices" data can be generated into a time series line chart, reflecting price fluctuation trends.

[0069] When generating visualization results, the system strictly adheres to the instructions for visualization action output to ensure that errors are corrected. For example, when processing R&D cost data, the data is first processed according to the "aggregate by department" action before generating a bar chart. When processing educational background distribution data, missing department data is first supplemented before generating a pie chart, thus ensuring the accuracy and completeness of the visualization results.

[0070] For example, suppose the subtypes in the dimension set to be analyzed are "Product A Sales" (core, first priority), "Customer Satisfaction" (auxiliary, second priority), and "Promotional Investment" (association, third priority). The generation parameters for "Sales" include the time period (January-December 2024), regional distribution (East China, North China, etc.), and sales value, but there is an error: "The regional dimension lacks data for South China." The generation parameters for "Customer Satisfaction" include the satisfaction score range and customer group classification, but there is an error: "The rating data format is inconsistent (a mixture of a 100-point and a 5-point scale)." The generation parameters for "Promotional Investment" include the activity time and investment amount, but there is an error: "The corresponding product model is not labeled."

[0071] When constructing the error feature sequence, the generation parameters of these three subtypes were prioritized, resulting in a sequence containing error information such as missing regions, inconsistent formats, and unlabeled models. This sequence was then fed into the error correction model, generating actions for "sales volume": "Supplement sales data for the South China region," "convert satisfaction scores to a uniform 100-point rating," and "Supplement the active model labeling for Product A" for "promotional investment."

[0072] Based on these actions, the system processes the data: it adds statistics for the South China region to the sales data, converts the satisfaction data to a unified percentage system, and annotates the promotional investment data with the activity model of Product A. Then, the system generates visualizations for each of these. For example, a bar chart showing regional sales in South China is generated for sales data, a line chart showing the score distribution in a unified format for satisfaction data, and a heat map showing the correlation between product A's investment and sales is generated for promotional investment data. These visualizations accurately reflect the business situation behind the data, providing users with a clear and reliable analytical view.

[0073] In practical applications, error correction model training requires extensive historical visualization task data. By learning from past error types and correction methods, the model's judgment and output capabilities are continuously optimized. For example, when errors with a "missing time dimension" occur repeatedly, the model strengthens its detection logic for the time parameter, improving the accuracy of identifying similar errors.

[0074] Action visualization output also needs to consider user needs and industry standards. For example, the financial industry may require that visualization results include data source annotations, while the retail industry may be more concerned with the frequency of real-time sales data updates. When generating action output, the system combines industry characteristics and user-defined visualization rules to ensure that the results meet the requirements of the actual application scenario.

[0075] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0076] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A natural language-based enterprise data visualization analysis method, characterized in that: The method comprises: Collect the enterprise business data to be analyzed and the corresponding natural language query text, and obtain the target analysis dimension based on the enterprise business data and the natural language query text; Based on the analysis intent of the natural language query text, the data subtypes covered by the target analysis dimension are divided into core data subtypes and auxiliary data subtypes, and associated data subtypes are determined among the data subtypes not covered by the target analysis dimension, wherein the core data subtypes, auxiliary data subtypes, and associated data subtypes constitute a set of dimensions to be analyzed; Determine the dimensional association between any two adjacent subtypes in the set of dimensions to be analyzed, and construct an analysis association graph of the set of dimensions to be analyzed with the subtypes in the set of dimensions to be analyzed as vertices and the dimensional association as edges connecting the vertices; Determining an optimal spanning tree of the analysis association graph, and determining a visualization priority order of each subtype in the set of dimensions to be analyzed based on the optimal spanning tree; The visualization results of each subtype in the set of dimensions to be analyzed are generated in sequence according to the visualization priority order.

2. The method for enterprise data visualization analysis based on natural language according to claim 1, characterized in that: The target analysis dimensions obtained based on the enterprise business data and natural language query text include: According to the natural language query text, counting analysis dimensions in the historical analysis task that match the enterprise business data; According to the semantic features of the natural language query text in each analysis dimension, the statistically obtained analysis dimensions are classified to form one or more analysis dimension sets; A target analysis dimension set that matches the current analysis intent of the natural language query text is determined in the analysis dimension set, and the dimensions covered by the target analysis dimension set are used as the target analysis dimensions.

3. The method for enterprise data visualization analysis based on natural language according to claim 2, characterized in that: Based on the semantic features of the natural language query text in each analysis dimension, the statistically obtained analysis dimensions are classified into the following categories: Obtaining a predefined semantic feature standard interval in the enterprise data management system; For any analysis dimension obtained through statistics, the semantic feature standard interval to which the semantic features of the natural language query text in the analysis dimension belong is identified, and the analysis dimension is divided into an analysis dimension set corresponding to the identified semantic feature standard interval.

4. The method for enterprise data visualization analysis based on natural language according to claim 1, characterized in that: Based on the analysis intent of the natural language query text, the data subtypes covered by the target analysis dimension are divided into core data subtypes and auxiliary data subtypes, including: Determining the core analysis object and auxiliary analysis objects represented by the analysis intent of the natural language query text; Determining a first data coverage range of the core analysis object in the target analysis dimension, and determining a second data coverage range of the auxiliary analysis object in the target analysis dimension; The data subtype covered by the first data coverage range within the enterprise business data range is determined as a core data subtype, and the data subtype covered by the second data coverage range within the enterprise business data range is determined as an auxiliary data subtype.

5. The method for enterprise data visualization analysis based on natural language according to claim 1, characterized in that: Determining the associated data subtypes from the data subtypes not covered by the target analysis dimension includes: Setting a first extension range for a core data subtype in the target analysis dimension, and setting a second extension range for an auxiliary data subtype in the target analysis dimension; For the data subtype not covered by the target analysis dimension, if it is adjacent to the core data subtype, the data subtype of the first extended range is selected as the associated data subtype; if it is adjacent to the auxiliary data subtype, the data subtype of the second extended range is selected as the associated data subtype.

6. The method for enterprise data visualization analysis based on natural language according to claim 1, characterized in that: Determining the dimensional correlation between any two adjacent subtypes in the set of dimensions to be analyzed includes: For adjacent first and second subtypes in the set of dimensions to be analyzed, identifying a type category of each of the first and second subtypes, wherein the type category includes one of a core data subtype, an auxiliary data subtype, and an associated data subtype; determining, based on the identified type category, a first dimension associated parameter of the first subtype, and determining a second dimension associated parameter of the second subtype; A difference value between the first dimensional association parameter and the second dimensional association parameter is calculated, and a dimensional association degree between the first subtype and the second subtype is determined according to the difference value.

7. The method for enterprise data visualization analysis based on natural language according to claim 6, characterized in that: Determining the dimensional association between the first subtype and the second subtype according to the difference value includes: Identifying dimension-related parameters of each subtype in the set of dimensions to be analyzed, and determining an average value of the identified dimension-related parameters; determining a plurality of difference intervals according to the average value, and determining a target difference interval in which the calculated difference value between the first dimension associated parameter and the second dimension associated parameter lies; The preset correlation degree corresponding to the target difference interval is determined as the dimensional correlation degree between the first subtype and the second subtype.

8. The method for enterprise data visualization analysis based on natural language according to claim 1, characterized in that: Generating visualization results of each subtype in the set of dimensions to be analyzed in sequence according to the visualization priority order includes: Identifying generation parameters of each generation node in the visualization priority order, and constructing an error feature sequence based on the identified generation parameters, wherein the arrangement order of each parameter in the error feature sequence is consistent with the visualization priority order; Inputting the error feature sequence into an error correction model to generate a visualization action output for each of the generation nodes through the error correction model; According to the generated visualization action outputs, corresponding visualization results of each subtype are generated.

9. A natural language-based enterprise data visualization analysis system, characterized in that: The system comprises: A data acquisition unit is used to collect the enterprise business data to be analyzed and the corresponding natural language query text, and obtain the target analysis dimension based on the enterprise business data and the natural language query text; a dimension division unit, configured to divide the data subtypes covered by the target analysis dimension into core data subtypes and auxiliary data subtypes based on the analysis intent of the natural language query text, and determine associated data subtypes among the data subtypes not covered by the target analysis dimension, wherein the core data subtypes, auxiliary data subtypes, and associated data subtypes constitute a set of dimensions to be analyzed; an association graph construction unit, configured to determine the dimensional association degree between any two adjacent subtypes in the set of dimensions to be analyzed, and to construct an analysis association graph of the set of dimensions to be analyzed, using the subtypes in the set of dimensions to be analyzed as vertices and the dimensional association degrees as edges connecting the vertices; a priority determination unit, configured to determine an optimal spanning tree of the analysis association graph, and determine a visualization priority order of each subtype in the set of dimensions to be analyzed based on the optimal spanning tree; The visualization generating unit is configured to generate visualization results of each subtype in the set of dimensions to be analyzed in sequence according to the visualization priority order.

10. The system according to claim 9, characterized in that The visualization generation unit is specifically configured to identify generation parameters of each generation node in the visualization priority order, and construct an error feature sequence based on the identified generation parameters, wherein the arrangement order of each parameter in the error feature sequence is consistent with the visualization priority order; The error feature sequence is input into an error correction model to generate a visualization action output for each of the generation nodes through the error correction model; and corresponding visualization results of each subtype are generated according to each of the generated visualization action outputs.

Images