Method and system for automatically interpreting table data by using large model and medium

Abstract

The invention discloses a method and a system for automatically interpreting table data by using a large model, and a medium, mainly relates to the technical field of table data interpretation, and is used for solving the problem that most of existing data analysis systems need users to have certain data analysis knowledge. The problem that an existing data analysis method is often designed for specific types of data and is difficult to adapt to diversified data sets is solved. Comprising the following steps: determining a corresponding processing analysis algorithm and a preset prompt template, and filling an output result of the processing analysis algorithm into the preset prompt template corresponding to the processing analysis algorithm; inputting the preset general prompt template and all the preset prompt templates of the single to-be-interpreted meter data without the preset main and foreign key relationship into a large model, and inputting the preset general prompt template and all the preset prompt templates of all the to-be-interpreted meter data with the preset main and foreign key relationship into the large model as a whole, and obtaining a final data interpretation result of large model integration.

Description

Technical Field

[0001] The present application relates to the technical field of row table data interpretation, and in particular to a method, system and medium for automatically interpreting table data using a large model. Background Art

[0002] With the advent of the big data era, the amount of data has exploded, and how to extract valuable information from massive data has become an important issue. Tabular data, as one of the most common forms of data representation, is widely used in many fields such as business intelligence, social sciences, and bioinformatics. Traditional data analysis methods require data scientists to manually select analysis models and parameters, which is not only time-consuming and labor-intensive, but also difficult to adapt to the rapidly changing business environment. In addition, data analysis needs to consider multiple factors, and traditional analysis methods are difficult to quickly and accurately draw conclusions.

[0003] In order to solve the above problems, artificial intelligence technology has been increasingly used in the field of data analysis in recent years. The development of technologies such as machine learning and deep learning enables computers to automatically learn patterns from data and make predictions and classifications. However, most existing data analysis systems require users to have certain data analysis knowledge in order to choose appropriate analysis methods. This limits the popularization and application of data analysis technology. In addition, existing data analysis methods are often designed for specific types of data and are difficult to adapt to diverse data sets.

[0004] Therefore, there is an urgent need for a method, system and medium for automatically interpreting table data using a large model to solve the problem that the existing data analysis system requires users to have certain data analysis knowledge in order to choose the appropriate analysis method, and the existing data analysis methods are often designed for specific types of data and are difficult to adapt to diverse data sets. Summary of the invention

[0005] In response to the above-mentioned deficiencies in the prior art, the present application provides a method, system and medium for automatically interpreting table data using a large model, so as to solve the problem that the existing data analysis system requires users to have certain data analysis knowledge in order to select a suitable analysis method, and the existing data analysis methods are often designed for specific types of data and are difficult to adapt to diverse data sets.

[0006] In a first aspect, the present application provides a method for automatically interpreting table data using a large model, the method comprising: When a number of table data to be interpreted are obtained, the table data to be interpreted that have a preset primary and foreign key relationship are determined according to the preset primary and foreign key relationship; according to whether there is a preset primary and foreign key relationship, the corresponding preset integrity verification program is called to obtain the integrity analysis data of each table data to be interpreted, and the integrity analysis data is filled into the preset general prompt template; according to the column name and annotation data of the table data to be interpreted, the category of the current table data to be interpreted is obtained using the big model; wherein the category is at least divided into: numerical table data, categorical table data, and time series table data; according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, the corresponding processing and analysis algorithm and the preset prompt template are determined, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm; the preset general prompt template and all the preset prompt templates of a single table data to be interpreted that does not have a preset primary and foreign key relationship are input into the big model, and the preset general prompt template and all the preset prompt templates of all the table data to be interpreted that have a preset primary and foreign key relationship are input into the big model as a whole to obtain the final data interpretation result integrated by the big model.

[0007] The method for automatically interpreting table data provided by the embodiment of the present application, the existing data analysis method is often designed for specific types of data, and it is difficult to adapt to diverse data sets. However, the present application automatically classifies table data of different types (such as numerical type, classification type, time series type) through a large model, and selects the corresponding processing and analysis algorithm according to the classification results, thereby improving the adaptability and flexibility of data analysis. It solves the problem that the existing data analysis system requires users to have certain data analysis knowledge to select a suitable analysis method, and the existing data analysis methods are often designed for specific types of data, and it is difficult to adapt to diverse data sets. In addition, the present application also considers the table data to be interpreted with a preset primary and foreign key relationship, and by inputting these data into the large model as a whole, the integrated analysis of multi-table data is realized. This helps to reveal the potential connections and laws between the data and provide more comprehensive support for decision-making.

[0008] In one implementation of the present application, the preset integrity check program includes at least: a default value statistics program, a data type compliance check program, a format consistency check program, a duplicate data check program, and a data range compliance check program.

[0009] In one implementation of the present application, according to the preset primary and foreign key relationships, determining the table data to be interpreted that has the preset primary and foreign key relationships specifically includes: Read the primary keys and foreign keys in a number of table data to be read; determine that a number of table data to be read whose primary keys or foreign keys exist in a preset primary-foreign key relationship is the table data to be read having the preset primary-foreign key relationship.

[0010] In one implementation of the present application, the corresponding processing and analysis algorithm and the preset prompt template are determined according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm, specifically including: When the category of the table data to be interpreted is numerical table data, the corresponding processing and analysis algorithms are determined to be preset maximum and variance statistical programs, data distribution range statistical programs, outlier detection programs, and cluster processing programs; wherein the output results of the maximum and variance statistical programs at least include: maximum value, minimum value, and variance, the output results of the data distribution range statistical programs at least include: data distribution range, the output results of the outlier detection programs at least include: whether there are outliers and the number of outliers, and the output results of the cluster processing programs at least include cluster centers, the number of clusters, the number of samples within the clusters, and the data within the clusters; the corresponding preset prompt templates are determined to be preset statistical analysis prompt templates, outlier detection prompt templates, and cluster analysis prompt templates; wherein the preset prompt templates contain preset analysis prompt words; the output results of the maximum and variance statistical programs and the data distribution range statistical programs are filled into the statistical analysis prompt template; the output results of the outlier detection program are filled into the outlier detection prompt template; the output results of the cluster processing program are filled into the cluster analysis prompt template.

[0011] In one implementation of the present application, the corresponding processing and analysis algorithm and the preset prompt template are determined according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm, specifically including: When the category of the table data to be interpreted is categorized table data, the corresponding processing and analysis algorithm is determined to be a preset association rule mining program, a causal inference analysis program, and an interpretability analysis program; wherein the output result of the association rule mining program at least includes: the support and confidence between the condition column and the result column in the table data to be interpreted, and the output result of the causal inference analysis program at least includes: the causal effect value and the significance measurement P value between the data value in the column of the table data to be interpreted and the preset factor; the output result of the interpretability analysis program at least includes: the importance of the preset feature in the table data to be interpreted; the corresponding preset prompt template is determined to be a preset association rule mining prompt template, a causal inference analysis prompt template, an interpretability analysis prompt template, and a predictive evaluation prompt template; the output result of the association rule mining program is filled into the association rule mining prompt template; the output result of the causal inference analysis program is filled into the causal inference analysis prompt template; the output result of the interpretability analysis program is filled into the interpretability analysis prompt template.

[0012] In one implementation of the present application, the corresponding processing and analysis algorithm and the preset prompt template are determined according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm, specifically including: When the category of the table data to be interpreted is time series table data, the corresponding processing and analysis algorithm is determined to be a single-table time series prediction program or a multi-table time series prediction program; wherein, the output results of the single-table time series prediction program and the multi-table time series prediction program are both prediction data for a preset future time period; the corresponding preset prompt template is determined to be a multi-table time series analysis prompt template and a single-table time series analysis prompt template; whether the current table data to be interpreted has a preset primary and foreign key relationship is obtained, and according to whether it has a preset primary and foreign key relationship, a specific processing and analysis algorithm is determined, and the output result is imported into the preset prompt template corresponding to the time series table data; wherein, the previous table data to be interpreted with a preset primary and foreign key relationship corresponds to the multi-table time series prediction program and the multi-table time series analysis prompt template.

[0013] In a second aspect, the present application provides a system for automatically interpreting table data using a large model, the system comprising: The relationship interpretation module is used to determine the table data to be interpreted that has a preset primary and foreign key relationship when obtaining a number of table data to be interpreted according to the preset primary and foreign key relationship; the prompt filling module is used to call the corresponding preset integrity verification program according to whether there is a preset primary and foreign key relationship, obtain the integrity analysis data of each table data to be interpreted, and fill the integrity analysis data into the preset general prompt template; according to the column name and annotation data of the table data to be interpreted, use the big model to obtain the category of the current table data to be interpreted; wherein the category is at least divided into: numerical table data, categorical table data, and time series table data; according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, determine the corresponding processing and analysis algorithm and the preset prompt template, and fill the output result of the processing and analysis algorithm into the preset prompt template corresponding to the processing and analysis algorithm; the model interpretation module is used to input the preset general prompt template and all the preset prompt templates of a single table data to be interpreted that does not have a preset primary and foreign key relationship into the big model, and input the preset general prompt template and all the preset prompt templates of all the table data to be interpreted that have a preset primary and foreign key relationship into the big model as a whole, so as to obtain the final data interpretation result integrated by the big model.

[0014] In one implementation of the present application, the relationship interpretation module includes a relationship determination unit, which is used to read the primary keys and foreign keys in a number of table data to be interpreted; and determine that a number of table data to be interpreted whose primary keys or foreign keys exist in a preset primary-foreign key relationship are table data to be interpreted that have a preset primary-foreign key relationship.

[0015] In one implementation of the present application, the model interpretation module includes a first interpretation unit, which is used to determine, when the category of the table data to be interpreted is time series table data, the corresponding processing and analysis algorithm is a single-table time series prediction program or a multi-table time series prediction program; wherein the output results of the single-table time series prediction program and the multi-table time series prediction program are both prediction data for a preset future time period; determine the corresponding preset prompt template to be a multi-table time series analysis prompt template and a single-table time series analysis prompt template; obtain whether the current table data to be interpreted has a preset primary and foreign key relationship, determine a specific processing and analysis algorithm based on whether it has a preset primary and foreign key relationship, and import the output result into the preset prompt template corresponding to the time series table data; wherein the previous table data to be interpreted with a preset primary and foreign key relationship corresponds to the multi-table time series prediction program and the multi-table time series analysis prompt template.

[0016] In a third aspect, the present application provides a non-volatile computer storage medium having computer instructions stored thereon, which, when executed, implement a method for automatically interpreting table data using a large model as described above.

[0017] Those skilled in the art can understand that the present application has at least the following beneficial effects: The present application provides a method, system and medium for automatic interpretation of table data using a large model, which is different from the existing data analysis methods that are often designed for specific types of data and are difficult to adapt to diverse data sets. The present application automatically classifies table data of different types (such as numerical, categorical, and time series) through a large model, and selects corresponding processing and analysis algorithms based on the classification results, thereby improving the adaptability and flexibility of data analysis. And selecting the corresponding processing and analysis algorithm based on the classification results solves the problem that the existing data analysis system requires users to have certain data analysis knowledge before they can choose a suitable analysis method. In addition, the present application also considers the table data to be interpreted with a preset primary and foreign key relationship, and by inputting these data as a whole into the large model, the integrated analysis of multi-table data is realized. This helps to reveal the potential connections and laws between data and provide more comprehensive support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Some embodiments of the present disclosure are described below with reference to the accompanying drawings, in which: Figure 1 It is a flow chart of a method for automatically interpreting table data using a large model provided in an embodiment of the present application.

[0019] Figure 2 It is a schematic diagram of the internal structure of a system for automatically interpreting table data using a large model, provided in an embodiment of the present application. DETAILED DESCRIPTION

[0020] It should be understood by those skilled in the art that the embodiments described below are only preferred embodiments of the present disclosure, and do not mean that the present disclosure can only be implemented through the preferred embodiments. The preferred embodiments are only used to explain the technical principles of the present disclosure, and are not used to limit the protection scope of the present disclosure. Based on the preferred embodiments provided by the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work should still fall within the protection scope of the present disclosure.

[0021] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0022] In the business field, for example, the annual consumption data of Beijing stores, which usually contains multiple dimensions such as time, amount, customer information, etc., traditional data analysis methods require data scientists to manually select analysis models and parameters, which is not only time-consuming and labor-intensive, but also difficult to adapt to the rapidly changing business environment. In the agricultural field, such as the watermelon dataset, the dataset contains various attributes of watermelons, such as color, size, weight, etc., as well as labels of whether the watermelon is good or not. The analysis of these data requires consideration of multiple factors, and traditional analysis methods are difficult to quickly and accurately draw conclusions.

[0023] In order to solve the above problems, artificial intelligence technology has been increasingly used in the field of data analysis in recent years. The development of machine learning, deep learning and other technologies has enabled computers to automatically learn patterns from data and make predictions and classifications. However, most existing data analysis systems require users to have certain data analysis knowledge in order to select appropriate analysis methods. This limits the popularization and application of data analysis technology.

[0024] In addition, existing data analysis methods are often designed for specific types of data and are difficult to adapt to diverse data sets. For example, for tabular data, it is necessary to consider multiple factors such as the type, distribution, and relevance of the data to select appropriate statistical methods or machine learning models. For text data in tables, such as comments on social media, sentiment analysis, topic modeling, and other operations are required. The selection and application of these analysis methods require users to have rich data analysis experience and a deep understanding of the data.

[0025] Therefore, developing a solution that can automatically identify data features and automatically select appropriate analysis methods based on data features is of great significance for improving the efficiency and accuracy of data analysis. This can not only lower the threshold of data analysis, making it easy for non-professionals to perform data analysis, but also improve the degree of automation of data analysis and adapt to the rapidly changing data environment.

[0026] Based on this background, this application proposes a method, system and medium for automatically interpreting table data using a large model. This application first performs integrity verification on the table data (table data to be interpreted), then establishes the association relationship between the table data to be interpreted based on the primary and foreign key relationships, and then performs corresponding analysis based on the specific classification of the table data to be interpreted, such as outlier detection and correlation analysis. The analysis results are converted into prompts, input into the large model for integration and suggestion generation, and finally the interpretation results are output.

[0027] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.

[0028] The embodiment provides a method for automatically interpreting table data using a large model, such as Figure 1 As shown, the method provided in the embodiment of the present application mainly includes the following steps: Step 110 : when a plurality of table data to be read are obtained, based on the preset primary and foreign key relationships, the table data to be read having the preset primary and foreign key relationships are determined.

[0029] This step can be specifically as follows: Read the primary keys and foreign keys in a number of table data to be read; determine that a number of table data to be read whose primary keys or foreign keys exist in a preset primary-foreign key relationship is the table data to be read having the preset primary-foreign key relationship.

[0030] Step 120: Based on whether there is a preset primary and foreign key relationship, call the corresponding preset integrity check program to obtain the integrity analysis data of each table data to be interpreted, and fill the integrity analysis data into the preset general prompt template.

[0031] It should be noted that the preset integrity check program can be: a default value statistics program, a data type conforming to a preset type check program, a format consistency check program, a duplicate data check program, and a data range conforming to a preset range check program. Specifically, the default value statistics program is a program that can count the default values ​​of data in the table, the data type conforming to the preset type check program is an existing program that can detect whether the data type conforms to the preset type, the format consistency check program is a program that can detect whether the current data format is completely consistent, the duplicate data check program is a program that can detect whether there is duplicate data in the current data, and the data range conforming to the preset range check program is a program that checks whether the current data range conforms to the preset range.

[0032] The preset general prompt template without preset primary and foreign key relationships (single table) can be: For the table {enter table name here}, which has {enter number of columns here} columns, the column names are: {enter column name here}, and the column comments are: {enter column comment here}, the integrity check results are as follows: Default value statistics: {Enter default value statistics here / no default value}; Data type check: {Enter the data type check result here / the data type of each column is correct}; Format consistency check: {Enter the format consistency check result here / the data format is completely consistent}; Repeatability check: {Enter the repeatability check result / no duplicate data here}; Range check: {Enter the range check results here / the data in the table are all within a reasonable range}.

[0033] It should be noted that the table name, number of columns, column name, and column comment are data that the table data to be interpreted has.

[0034] The preset common prompt template with preset primary and foreign key relationships (multiple tables) can be: "The primary and foreign key relationships between the tables are: {enter the known primary and foreign key relationships here}. Among them, for the table {enter the table name here}, it has {enter the number of columns here} columns, the column names are: {enter the column name here}, and the column comments are: {enter the column comment here}. The integrity check results are as follows: Default value statistics: {Enter default value statistics here / no default value}; Data type check: {Enter the data type check result here / the data type of each column is correct}; Format consistency check: {Enter the format consistency check result here / the data format is completely consistent}; Repeatability check: {Enter the repeatability check result / no duplicate data here}; Range check: {Enter the range check results here / the data in the table are all within a reasonable range}".

[0035] Step 130: According to the column names and annotation data of the table data to be interpreted, the category of the current table data to be interpreted is obtained using the large model.

[0036] It should be noted that the categories are at least divided into: numerical table data, categorical table data, and time series table data.

[0037] Step 140: determine the corresponding processing and analysis algorithm and preset prompt template according to the category of the table data to be interpreted and whether it has a preset primary and foreign key relationship, and fill the output result of the processing and analysis algorithm into the preset prompt template corresponding to the processing and analysis algorithm.

[0038] In some implementations, this step may specifically include: When the category of the table data to be interpreted is numerical table data, the corresponding processing and analysis algorithms are determined to be preset maximum and variance statistical programs, data distribution range statistical programs, outlier detection programs, and cluster processing programs; wherein the output results of the maximum and variance statistical programs at least include: maximum value, minimum value, and variance, the output results of the data distribution range statistical programs at least include: data distribution range, the output results of the outlier detection programs at least include: whether there are outliers and the number of outliers, and the output results of the cluster processing programs at least include cluster centers, the number of clusters, the number of samples within the clusters, and the data within the clusters; the corresponding preset prompt templates are determined to be preset statistical analysis prompt templates, outlier detection prompt templates, and cluster analysis prompt templates; wherein the preset prompt templates contain preset analysis prompt words; the output results of the maximum and variance statistical programs and the data distribution range statistical programs are filled into the statistical analysis prompt template; the output results of the outlier detection program are filled into the outlier detection prompt template; the output results of the cluster processing program are filled into the cluster analysis prompt template.

[0039] As an example, the statistical analysis prompt template may be: "Now you need to summarize the statistical results of a set of data. The input parts are the maximum value and variance statistics and data distribution statistics: Maximum value and variance statistics input: {enter the maximum value and variance statistics results here}; Data distribution statistics input: {enter the data distribution results here}. If the input is empty, no explanation will be given. A summary analysis must be conducted from four aspects: maximum value and variance statistics and data distribution statistics.

[0040] In the maximum value analysis, it is necessary to combine the given other column data, analyze under what conditions the highest or lowest value is reached, and give a simple explanation for it.

[0041] Please enter your answer in the following format: Data statistical analysis summary: 1. Maximum value and variance statistical analysis, maximum value and variance statistical content output; 2. Data distribution statistical analysis, data distribution statistical content output; 3. Summary: summary content output.

[0042] As an example, the outlier detection prompt template may be: "Now you need to summarize the outlier detection results for a set of data. Input: {Enter the outlier detection results here}; If the input is empty, it means that there are no outliers in the data, and directly answer that there are no outliers in the data.

[0043] Please enter your answer in the following format: Data outlier detection result summary; data outlier detection result content output".

[0044] As an example, the cluster analysis prompt template may be: "Now you need to summarize the results of cluster analysis of a set of data. Input: {Enter the results of cluster analysis here (including cluster center, number of clusters, number of samples within clusters, statistical analysis results of data within clusters, etc.)}; If the input is empty, it means that the data cannot be clustered. The direct answer is that the data is not suitable for cluster analysis.

[0045] Please output your answer in the following format: Summary of cluster analysis results, Output of cluster analysis results content".

[0046] In some implementations, this step may specifically include: When the category of the table data to be interpreted is categorized table data, the corresponding processing and analysis algorithm is determined to be a preset association rule mining program, a causal inference analysis program, and an interpretability analysis program; wherein the output result of the association rule mining program at least includes: the support and confidence between the condition column and the result column in the table data to be interpreted, and the output result of the causal inference analysis program at least includes: the causal effect value and the significance measurement P value between the data value in the column of the table data to be interpreted and the preset factor; the output result of the interpretability analysis program at least includes: the importance of the preset feature in the table data to be interpreted; the corresponding preset prompt template is determined to be a preset association rule mining prompt template, a causal inference analysis prompt template, an interpretability analysis prompt template, and a predictive evaluation prompt template; the output result of the association rule mining program is filled into the association rule mining prompt template; the output result of the causal inference analysis program is filled into the causal inference analysis prompt template; the output result of the interpretability analysis program is filled into the interpretability analysis prompt template.

[0047] As an example, the association rule mining prompt template may be specifically: “Through association rule mining, we found the following rules to be significantly associated with the category column [enter category column name here]: - Rule 1: If [condition 1], then [result 1], support is [support 1], confidence is [confidence 1].

[0048] - Rule 2: If [condition 2], then [result 2], support is [support 2], confidence is [confidence 2]. ... The above rules describe the relationship between different features and classification results. Please summarize according to the above rules and put forward your reasonable suggestions.

[0050] As an example, the causal inference analysis prompt template may be specifically: “Causal inference analysis shows that the following factors have a significant effect on the category column [enter category column name here]: - Factor 1: [Factor 1 name], estimated causal effect is [effect size 1], p-value is [p-value 1].

[0051] - Factor 2: [Factor 2 name], estimated causal effect is [effect size 2], p-value is [p-value 2]. ... The causal relationship of the above factors provides the basis for classification decisions. Please summarize based on the above causal relationship and put forward your reasonable suggestions.

[0053] As an example, the explainability analysis prompt template can be specifically: "Using the decision tree model for interpretability analysis, we know that the following features have an important impact on the prediction of the category column [enter the category column name here]: - Feature 1: [Feature 1 name], with importance score of [Importance score 1].

[0054] - Feature 2: [Feature 2 name], with importance score of [Importance score 2]. ... The relative importance of these features is important to understanding how classification decisions are made. Please summarize based on the above characteristics and give reasonable suggestions."

[0056] In addition, for further analysis, the output of the interpretability analysis program can be input into the prediction evaluation program to obtain the output result of the prediction evaluation. The output result is input into the prediction evaluation prompt template. The prediction evaluation prompt template can be specifically: “Evaluation of the predictions of the decision tree model showed that: - Cross-validation accuracy: [enter accuracy here]; - AUC value: [Enter the AUC value here]; - Recall: [enter recall here]; - F1 score: [enter F1 score here]".

[0057] It can be understood by those skilled in the art that for categorized data, when the amount of data is judged to be appropriate, additional data interpretation functions such as association rule mining, causal analysis inference, interpretable analysis, and predictive evaluation are added. The classification column is used as a label to analyze the potential association rules and causal relationships between other columns and the classification column. The decision tree model is used to fit the data in the table to obtain the basis for classification (interpretability analysis), and the fitting situation is explained (predictive evaluation), such as using cross-validation, AUC, accuracy, recall and other indicators. Solvability analysis and predictive evaluation must be performed simultaneously. A new prompt is formed based on the obtained results, and the final prompt is obtained by splicing it with the general template before and after, and the large model is input to obtain the interpretation result.

[0058] In some implementations, this step may specifically include: When the category of the table data to be interpreted is time series table data, the corresponding processing and analysis algorithm is determined to be a single-table time series prediction program or a multi-table time series prediction program; wherein, the output results of the single-table time series prediction program and the multi-table time series prediction program are both prediction data for a preset future time period; the corresponding preset prompt template is determined to be a multi-table time series analysis prompt template and a single-table time series analysis prompt template; whether the current table data to be interpreted has a preset primary and foreign key relationship is obtained, and according to whether it has a preset primary and foreign key relationship, a specific processing and analysis algorithm is determined, and the output result is imported into the preset prompt template corresponding to the time series table data; wherein, the previous table data to be interpreted with a preset primary and foreign key relationship corresponds to the multi-table time series prediction program and the multi-table time series analysis prompt template.

[0059] As an example, the single-table time series analysis prompt template may specifically be: “Now you need to perform trend analysis on the future forecast values ​​of the time series related data columns in the input table data. Input: {Enter the time series analysis results here (including development trends, future forecast values, etc.)}; Please output your answer in the following format: [Time Series Trend Analysis Summary] Time Series Trend Analysis Content Output".

[0060] As an example, the multi-table time series analysis prompt template may specifically be: "Now you need to perform trend analysis on the future forecast values ​​of the time series related data columns in the input table data. Input: The primary and foreign key relationships between the tables are: {enter the known primary and foreign key relationships here}, {enter the time series analysis results here (including development trends, forecast values ​​for future time, etc.)}; Please output your answer in the following format: [Time Series Trend Analysis Summary] Time Series Trend Analysis Content Output".

[0061] Step 150: input the preset general prompt template and all preset prompt templates of a single table data to be interpreted that does not have a preset primary-foreign key relationship into the big model, and input the preset general prompt template and all preset prompt templates of all table data to be interpreted that have a preset primary-foreign key relationship into the big model as a whole, to obtain the final data interpretation result integrated by the big model.

[0062] In addition, this application Figure 2 A system for automatically interpreting table data using a large model is provided in an embodiment of the present application. Figure 2 As shown, the system provided in the embodiment of the present application mainly includes: The relationship interpretation module 210 is used to determine the table data to be interpreted that has a preset primary and foreign key relationship according to the preset primary and foreign key relationship when a plurality of table data to be interpreted are obtained.

[0063] The relationship interpretation module 210 includes a relationship determination unit for reading primary keys and foreign keys in a number of table data to be interpreted; and determining that a number of table data to be interpreted whose primary keys or foreign keys exist in a preset primary-foreign key relationship are table data to be interpreted with a preset primary-foreign key relationship.

[0064] The prompt filling module 220 is used to call the corresponding preset integrity verification program according to whether there is a preset primary-foreign key relationship, obtain the integrity analysis data of each table data to be interpreted, and fill the integrity analysis data into the preset general prompt template; according to the column name and annotation data of the table data to be interpreted, use the big model to obtain the category of the current table data to be interpreted; wherein the category is at least divided into: numerical table data, categorical table data, and time series table data; according to the category of the table data to be interpreted and whether there is a preset primary-foreign key relationship, determine the corresponding processing and analysis algorithm and the preset prompt template, and fill the output result of the processing and analysis algorithm into the preset prompt template corresponding to the processing and analysis algorithm.

[0065] The model interpretation module 230 is used to input the preset general prompt template and all preset prompt templates of a single table data to be interpreted that does not have a preset primary-foreign key relationship into the big model, and input the preset general prompt template and all preset prompt templates of all table data to be interpreted that have a preset primary-foreign key relationship into the big model as a whole, to obtain the final data interpretation result integrated by the big model.

[0066] The model interpretation module 230 includes a first interpretation unit, which is used to determine the corresponding processing and analysis algorithm as a single-table time series prediction program or a multi-table time series prediction program when the category of the table data to be interpreted is time series table data; wherein the output results of the single-table time series prediction program and the multi-table time series prediction program are both prediction data for a preset future time period; determine the corresponding preset prompt template as a multi-table time series analysis prompt template and a single-table time series analysis prompt template; obtain whether the current table data to be interpreted has a preset primary and foreign key relationship, determine a specific processing and analysis algorithm based on whether it has a preset primary and foreign key relationship, and import the output result into the preset prompt template corresponding to the time series table data; wherein the previous table data to be interpreted with a preset primary and foreign key relationship corresponds to the multi-table time series prediction program and the multi-table time series analysis prompt template.

[0067] In addition, an embodiment of the present application further provides a non-volatile computer storage medium on which executable instructions are stored. When the executable instructions are executed, a method for automatically interpreting table data using a large model as described above is implemented.

[0068] So far, the technical solutions of the present disclosure have been described in combination with the above multiple embodiments, but it is easy for those skilled in the art to understand that the protection scope of the present disclosure is not limited to these specific embodiments. Without departing from the technical principles of the present disclosure, those skilled in the art can split and combine the technical solutions in the above-mentioned various embodiments, and can also make equivalent changes or replacements to the relevant technical features. Any changes, equivalent replacements, improvements, etc. made within the technical concept and / or technical principle of the present disclosure will fall within the protection scope of the present disclosure.

Claims

1. A method for automatically interpreting table data using a large model, characterized in that: The method comprises: When a number of table data to be read are obtained, the table data to be read having the preset primary and foreign key relationship is determined according to the preset primary and foreign key relationship; According to whether there is a preset primary and foreign key relationship, the corresponding preset integrity check program is called to obtain the integrity analysis data of each table data to be interpreted, and the integrity analysis data is filled into the preset general prompt template; According to the column names and annotation data of the table data to be interpreted, the category of the current table data to be interpreted is obtained by using the large model; wherein the categories are at least divided into: numerical table data, categorical table data, and time series table data; Determine the corresponding processing and analysis algorithm and preset prompt template according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, and fill the output result of the processing and analysis algorithm into the preset prompt template corresponding to the processing and analysis algorithm; The preset general prompt template and all preset prompt templates of a single table data to be interpreted that does not have a preset primary-foreign key relationship are input into the big model, and the preset general prompt template and all preset prompt templates of all table data to be interpreted that have a preset primary-foreign key relationship are input into the big model as a whole to obtain the final data interpretation result integrated by the big model.

2. The method for automatically interpreting table data using a large model according to claim 1, characterized in that: The preset integrity check program at least includes: a default value statistics program, a data type compliance check program, a format consistency check program, a duplicate data check program, and a data range compliance check program.

3. The method for automatically interpreting table data using a large model according to claim 1, characterized in that: According to the preset primary and foreign key relationships, the table data to be read with the preset primary and foreign key relationships is determined, specifically including: Read the primary keys and foreign keys in several tables to be interpreted; It is determined that a plurality of to-be-read table data whose primary keys or foreign keys exist in a preset primary-foreign key relationship are to-be-read table data having a preset primary-foreign key relationship.

4. The method for automatically interpreting table data using a large model according to claim 1, characterized in that: According to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, the corresponding processing and analysis algorithm and the preset prompt template are determined, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm, which specifically includes: When the category of the table data to be interpreted is numerical table data, the corresponding processing and analysis algorithm is determined to be a preset maximum value and variance statistical program, a data distribution range statistical program, an outlier detection program, and a cluster processing program; wherein the output result of the maximum value and variance statistical program at least includes: the maximum value, the minimum value, and the variance, the output result of the data distribution range statistical program at least includes: the data distribution range, the output result of the outlier detection program at least includes: whether there are outliers and the number of outliers, and the output result of the cluster processing program at least includes the cluster center, the number of clusters, the number of samples in the cluster, and the data in the cluster; Determine that the corresponding preset prompt template is a preset statistical analysis prompt template, an outlier detection prompt template, and a cluster analysis prompt template; wherein the preset prompt template includes a preset analysis prompt word; Fill the output results of the maximum value and variance statistics program and the data distribution range statistics program into the statistical analysis prompt template; fill the output results of the outlier detection program into the outlier detection prompt template; fill the output results of the cluster processing program into the cluster analysis prompt template.

5. The method for automatically interpreting table data using a large model according to claim 1, characterized in that: According to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, the corresponding processing and analysis algorithm and the preset prompt template are determined, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm, which specifically includes: When the category of the table data to be interpreted is categorized table data, the corresponding processing and analysis algorithm is determined to be a preset association rule mining program, a causal inference analysis program, and an interpretability analysis program; wherein the output result of the association rule mining program at least includes: the support and confidence between the condition column and the result column in the table data to be interpreted, and the output result of the causal inference analysis program at least includes: the causal effect value and the significance measurement P value between the data value in the column of the table data to be interpreted and the preset factor; the output result of the interpretability analysis program at least includes: the importance of the preset feature in the table data to be interpreted; Determine that the corresponding preset prompt template is a preset association rule mining prompt template, a causal inference analysis prompt template, and an explainability analysis prompt template; Fill the output results of the association rule mining program into the association rule mining prompt template; fill the output results of the causal inference analysis program into the causal inference analysis prompt template; fill the output results of the explainability analysis program into the explainability analysis prompt template.

6. The method for automatically interpreting table data using a large model according to claim 1, characterized in that: According to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, the corresponding processing and analysis algorithm and the preset prompt template are determined, and the output result of the processing and analysis algorithm is filled into the preset prompt template corresponding to the processing and analysis algorithm, which specifically includes: When the category of the table data to be interpreted is time series table data, the corresponding processing and analysis algorithm is determined to be a single-table time series prediction program and a multi-table time series prediction program; wherein the output results of the single-table time series prediction program and the multi-table time series prediction program are both prediction data for a preset future time period; Determine that the corresponding preset prompt template is a multi-table time series analysis prompt template and a single-table time series analysis prompt template; Obtain whether the current table data to be interpreted has a preset primary-foreign key relationship, determine the specific processing and analysis algorithm based on whether it has a preset primary-foreign key relationship, and import the output result into the preset prompt template corresponding to the time series table data; among them, the previous table data to be interpreted with a preset primary-foreign key relationship corresponds to the multi-table time series prediction program and the multi-table time series analysis prompt template.

7. A system for automatically interpreting table data using a large model, characterized in that: The system comprises: The relationship interpretation module is used to determine the table data to be interpreted that has a preset primary and foreign key relationship when obtaining a number of table data to be interpreted according to the preset primary and foreign key relationship; The prompt filling module is used to call the corresponding preset integrity check program according to whether there is a preset primary and foreign key relationship, obtain the integrity analysis data of each table data to be interpreted, and fill the integrity analysis data into the preset general prompt template; according to the column name and annotation data of the table data to be interpreted, use the big model to obtain the category of the current table data to be interpreted; wherein the category is at least divided into: numerical table data, categorical table data, and time series table data; according to the category of the table data to be interpreted and whether there is a preset primary and foreign key relationship, determine the corresponding processing and analysis algorithm and the preset prompt template, and fill the output result of the processing and analysis algorithm into the preset prompt template corresponding to the processing and analysis algorithm; The model interpretation module is used to input the preset general prompt template and all preset prompt templates of a single table data to be interpreted that does not have a preset primary-foreign key relationship into the big model, and to input the preset general prompt template and all preset prompt templates of all table data to be interpreted that have a preset primary-foreign key relationship into the big model as a whole, so as to obtain the final data interpretation result integrated by the big model.

8. The system for automatically interpreting table data using a large model according to claim 7, characterized in that: The relationship interpretation module includes a relationship determination unit, Used to read the primary keys and foreign keys in several tables to be interpreted; It is determined that a plurality of to-be-read table data whose primary keys or foreign keys exist in a preset primary-foreign key relationship are to-be-read table data having a preset primary-foreign key relationship.

9. The system for automatically interpreting table data using a large model according to claim 7, characterized in that: The model interpretation module includes a first interpretation unit, When the category of the table data to be interpreted is time series table data, determine the corresponding processing and analysis algorithm as a single-table time series prediction program or a multi-table time series prediction program; wherein the output results of the single-table time series prediction program and the multi-table time series prediction program are both prediction data for a preset future time period; determine the corresponding preset prompt template as a multi-table time series analysis prompt template and a single-table time series analysis prompt template; obtain whether the current table data to be interpreted has a preset primary and foreign key relationship, determine the specific processing and analysis algorithm according to whether it has the preset primary and foreign key relationship, and import the output result into the preset prompt template corresponding to the time series table data; wherein the previous table data to be interpreted with the preset primary and foreign key relationship corresponds to the multi-table time series prediction program and the multi-table time series analysis prompt template.

10. A non-volatile computer storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the method for automatically interpreting table data using a large model as described in any one of claims 1 to 6 is implemented.

Images