Data processing method, device, equipment and storage medium

By performing field recognition of statistical data and using large language models to generate recommendation problems, the problem of insufficient efficiency and comprehensiveness of existing data analysis tools is solved, and efficient and accurate data analysis is achieved.

CN118673900BActive Publication Date: 2024-11-26TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411165725.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-26
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

Existing data analysis tools and systems have limitations in efficiency and comprehensiveness, making it difficult to effectively extract and analyze large amounts of complex statistical data.

Method used

By performing field identification processing on the input statistical data, data description information is generated, and recommendation problems are generated using large language models to improve the efficiency and accuracy of data analysis and reduce dependence on artificial intelligence.

Benefits of technology

提升了数据分析的效率和准确性,提供了更广泛的数据分析方向,降低了数据分析的成本和难度,提高了数据分析的可行性和普适性。

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118673900B_ABST
    Figure CN118673900B_ABST
Patent Text Reader

Abstract

The embodiment of the present application discloses a data processing method, device, equipment and storage medium, which can be applied to the field of computer technology. The method includes: in response to obtaining statistical data input by the target object, performing field recognition processing on the statistical data to obtain a field recognition result, the field recognition result includes the recognized statistical indicators and statistical dimensions; determining data description information based on the statistical data and the field recognition result, the data description information is used to describe the data characteristics of the statistical data; inputting the field recognition result into a large language model to obtain a recommended question, the recommended question is used to guide the target object to query and ask questions about the specified data characteristics of the statistical data; displaying the data description information and the recommended question. By adopting the embodiment of the present application, the efficiency of data interaction can be improved, the accuracy of data analysis can be improved, and the applicability is high.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data processing method, device, equipment and storage medium. Background Art

[0002] In this digital world, statistics are an indispensable resource in our work and life. These statistics come from various sources, including social media, e-commerce, and enterprise systems. The scale and complexity of these statistics are growing every day, and it is not easy to extract valuable information from them. With the growing demand for data analysis, more and more data analysis tools and systems have become indispensable means of data analysis in order to simplify the data analysis process and improve data analysis efficiency.

[0003] However, the data analysis capabilities of existing tools and systems are limited, the efficiency and comprehensiveness of data analysis results are poor, and there are significant limitations. Summary of the invention

[0004] The embodiments of the present application provide a data processing method, apparatus, device, and storage medium, which can improve data analysis efficiency and accuracy and have high applicability.

[0005] On the one hand, an embodiment of the present application provides a data processing method, the method comprising:

[0006] In response to obtaining the statistical data input by the target object, performing field recognition processing on the statistical data to obtain a field recognition result, wherein the field recognition result includes the recognized statistical indicators and statistical dimensions;

[0007] Determine data description information according to the statistical data and the field identification result, where the data description information is used to describe data characteristics of the statistical data;

[0008] Inputting the field recognition results into a large language model to obtain a recommendation question, wherein the recommendation question is used to guide the target object to ask a query about the specified data features of the statistical data;

[0009] Display the above data description information and the above recommended questions.

[0010] On the other hand, an embodiment of the present application provides a data processing device, the device comprising:

[0011] A field recognition module is used to, in response to obtaining statistical data input by a target object, perform field recognition processing on the statistical data to obtain a field recognition result, wherein the field recognition result includes the recognized statistical indicators and statistical dimensions;

[0012] A data analysis module, used to determine data description information according to the statistical data and the field identification result, wherein the data description information is used to describe data characteristics of the statistical data;

[0013] The data analysis module is used to input the field recognition results into a large language model to obtain recommended questions, and the recommended questions are used to guide the target object to query and ask questions about the specified data features of the statistical data;

[0014] The information display module is used to display the above data description information and the above recommended questions.

[0015] On the other hand, an embodiment of the present application provides an electronic device, including a processor and a memory, wherein the processor and the memory are connected to each other;

[0016] The memory is used to store computer programs;

[0017] The above-mentioned processor is used to execute the data processing method provided in the embodiment of the present application when calling the above-mentioned computer program.

[0018] On the other hand, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the data processing method provided by the embodiment of the present application.

[0019] On the other hand, an embodiment of the present application provides a computer program product, which includes a computer program. When the above computer program is executed by a processor, it implements the data processing method provided by the embodiment of the present application.

[0020] In an embodiment of the present application, after obtaining the statistical data input by the target object, the statistical data can be firstly identified by field recognition to obtain information such as statistical indicators, statistical dimensions, and field types in the statistical data, so that when performing data analysis on the statistical data, the structure and key information points of the statistical data can be quickly understood, avoiding frequent queries or confirmations of relevant field features during the data analysis process, thereby improving data analysis efficiency and improving the accuracy of data analysis results. By determining and displaying the data description information of the statistical data, the basic data features of the statistical data at all levels can be quickly extracted in an all-round manner, which helps the target object to quickly understand the basic features of the statistical data. By generating and displaying recommendation questions through a large language model, it can not only provide the target object with a wider range of data analysis directions, improve the efficiency of human-computer interaction and data analysis experience, but also reduce the reliance on artificial intelligence when the data description information does not rely on a large language model, thereby reducing the cost and difficulty of data analysis and improving the feasibility and universality of data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 It is a scenario schematic diagram of the data processing method provided in the embodiment of the present application;

[0023] Figure 2 It is a flowchart of a data processing method provided in an embodiment of the present application;

[0024] Figure 3 This is a schematic diagram of a scenario for obtaining statistical data provided by an embodiment of the present application;

[0025] Figure 4 (a) is another schematic diagram of a scenario for obtaining statistical data provided in an embodiment of the present application;

[0026] Figure 4 (b) is another schematic diagram of a scenario for obtaining statistical data provided in an embodiment of the present application;

[0027] Figure 5 It is a schematic diagram of the data analysis results provided in the embodiment of the present application;

[0028] Figure 6 It is a schematic diagram of the process framework for processing query problems provided by an embodiment of the present application;

[0029] Figure 7 It is a timing diagram of the data processing method adopted in the embodiment of the present application;

[0030] Figure 8 It is a schematic diagram of the process framework of the data processing method provided in the embodiment of the present application;

[0031] Fig. 9 is a structural schematic diagram of a data processing device provided in an embodiment of the present application;

[0032] Fig.10 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0034] The data processing method provided in the embodiment of the present application can be applied to the field of human-computer interaction, and is used to perform data analysis on statistical data input by a target object, provide data description information of the statistical data to the target object, and reference more data features of the target object to query and ask questions about the statistical data, which is beneficial to improving data analysis efficiency and human-computer interaction efficiency.

[0035] See also Figure 1 , Figure 1 Schematic diagram of a data processing method provided in an embodiment of the present application. Figure 1 As shown, the target object 100 can input the statistical data that needs to be analyzed through the interactive interface. After the terminal 200 obtains the statistical data, it can send the statistical data to the server 300, so that the server 300 performs field recognition on the statistical data to obtain field recognition results, and then determines the data description information according to the field recognition results and the statistical data, and inputs the field recognition results and the statistical data into the large language model to obtain the recommendation question.

[0036] Among them, the above-mentioned field recognition results include the statistical indicators and statistical dimensions obtained by recognition.

[0037] Among them, the data description information is used to describe the data characteristics of the statistical data, such as the number of fields in the statistical data, the number of statistical indicators, and the number of statistics and dimensions, etc., which are not limited here.

[0038] The above-mentioned recommendation questions are used to guide the target object to ask questions about the specified data features of the statistical data.

[0039] In this case, the terminal 200 obtains the data description information and recommended questions, and displays the data description information and recommended questions to the target object 100, so as to feedback the data analysis results to the target object 100 through the data description information, and guide the target object 100 to continue to query other data features of the question statistical data through the recommended questions.

[0040] It should be noted that when the terminal 200 has data processing and analysis capabilities, the terminal 200 can independently perform a number of data analysis steps such as field recognition processing on statistical data, determining data description information based on statistical data and field recognition results, and determining recommendation questions based on a large language model.

[0041] Among them, the terminal 200 can be a smart phone with display function, a tablet computer, a laptop computer, a desktop computer (including a display), a smart watch, a vehicle terminal, an aircraft, a smart home appliance (such as a smart TV) or a wearable device, etc.

[0042] Among them, server 300 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud database, cloud computing, big data and artificial intelligence platform.

[0043] It should be noted that the data processing method provided in the embodiment of the present application can be executed by a server or a terminal alone, or by a server and a terminal together, and is not limited here. Figure 2 , Figure 2 Schematic diagram of the data processing method provided in the embodiment of the present application. Figure 2 As shown, the data processing method provided in the embodiment of the present application can be described by taking terminal execution as an example, and the method may include the following steps:

[0044] S21, in response to obtaining statistical data input by the target object, performing field recognition processing on the statistical data to obtain a field recognition result.

[0045] The statistical data in the embodiments of the present application are collected through observation, investigation, experiment, etc., and are obtained through processing such as sorting, classification, and aggregation. These data are used to describe, analyze, infer or predict the characteristics, laws, trends or states of a phenomenon or object.

[0046] In some feasible implementations, a data interaction interface may be displayed to the target object, so as to obtain statistical data of a table type input by the target object through the data interaction interface.

[0047] Optionally, when obtaining statistical data, text data input by the target object in the input box of the data interaction interface may be obtained. After the text data is obtained, it may be converted into statistical data of a structured table type.

[0048] Optionally, when obtaining statistical data, a data file uploaded by the target object through the data interaction interface may be obtained, and then the data content of the data file may be extracted and converted into statistical data of a table type.

[0049] Optionally, when obtaining statistical data, it is also possible to obtain chart contents uploaded by the target object through the data interaction interface, and then extract and convert the chart contents to obtain statistical data in a table type.

[0050] Optionally, when obtaining statistical data, voice data input by the target object through the data interaction interface may also be obtained, and the voice data may be converted into text data, and the text data may be converted into statistical data of a table type.

[0051] Optionally, the statistical data input by the target object is the statistical data queried through the data interaction interface before the target object triggers the data analysis operation through the data interaction interface. When obtaining statistical data through the data interaction interface, the statistical data previously queried by the target object can also be determined as the statistical data input by the target object in response to the data analysis operation triggered by the target object on the data interaction interface.

[0052] Among them, the target object can trigger the data analysis operation by inputting text or voice commands, or by related touch buttons in the data interaction interface, and there is no limitation here.

[0053] Among them, the statistical data of the previous query of the target object can be directly obtained from the data interaction interface, or obtained from the cache, and the statistical data of the previous query of the target object can also be obtained from the database by calling the interface. The specific acquisition method can be determined based on the actual application scenario requirements and is not limited here.

[0054] As an example, Figure 3 As shown, Figure 3 Schematic diagram of a scenario for obtaining statistical data provided by an embodiment of the present application. Figure 3 As shown, after querying and displaying the queried statistical data for the target object through the query instruction of the target object data or the query SQL statement, the queried statistical data can be displayed to the target object through the data interaction interface. In this case, when the target object enters the text instruction "please interpret the data", it is determined that the target object triggers the data analysis operation through the data interaction interface, and then the statistical data can be directly obtained from the data interaction interface, or obtained from the cache.

[0055] Alternatively, when a touch operation of the target object on a touch button in the interactive interface for data interpretation is detected, it can be determined that the target object triggers a data analysis operation through the data interaction interface, and statistical data can be obtained directly from the data interaction interface or from the cache.

[0056] Optionally, when acquiring statistical data, in response to a data query operation of the target object, corresponding statistical data may be queried for the target object and used as statistical data input for the target object.

[0057] As an example, Figure 4 As shown in (a) and (b), Figure 4 (a) is another schematic diagram of a scenario for obtaining statistical data provided in an embodiment of the present application. Figure 4 (b) is another schematic diagram of a scenario for obtaining statistical data provided by an embodiment of the present application. Figure 4As shown in (a) in the figure, when the target object selects the data result interpretation function, it can enter Figure 4 The data interaction interface shown in (b) in FIG. 1 shows a data interaction interface. After detecting the query statement input by the target object in the data interaction interface, the statistical data required by the target object can be queried according to the query statement, and the statistical data can be obtained synchronously as the statistical data input by the target object.

[0058] It should be noted that the aforementioned data interaction interface can be an interactive application based on artificial intelligence, or it can be a related application interface used by a terminal device to interact with a target object for data, and there is no limitation here.

[0059] In some feasible implementations, after obtaining the statistical data of the target object input, field recognition processing may be performed on the statistical data to obtain a field recognition result.

[0060] The field recognition results of the statistical data include the recognized statistical indicators, statistical dimensions, and field types of the statistical dimensions.

[0061] Among them, statistical indicators can be understood as the objects being counted, and statistical dimensions can be understood as the various aspects or angles used to measure or evaluate statistical indicators. For example, in a class transcript, each student is a statistical indicator, each subject is a statistical dimension, and the statistical value of a statistical indicator under a statistical dimension is the test score of a student in a subject. In other words, the statistical value of a statistical indicator under each statistical dimension represents the test score of a student in each subject, and the statistical value of each statistical indicator under a statistical dimension is the test score of each student in a subject.

[0062] When identifying fields in statistical data, Python methods (including regular function matching, type judgment, etc.) can be used to traverse each row and column of the tabular data to identify the statistical indicators, statistical dimensions, and field types of each statistical dimension.

[0063] Among them, the field type of each statistical dimension is used to represent the data type of the corresponding statistical dimension in the tabular data, such as int type, string type, etc., and there is no restriction here.

[0064] In order to improve the efficiency of field recognition, statistical data can be sampled and recognized to reduce resource consumption.

[0065] In an embodiment of the present application, after obtaining the statistical data input by the target object, the statistical data can be firstly identified by field recognition to obtain information such as statistical indicators, statistical dimensions, and field types in the statistical data, so that when performing data analysis on the statistical data, the structure and key information points of the statistical data can be quickly understood, avoiding frequent queries or confirmations of relevant field features during the data analysis process, thereby improving data analysis efficiency. At the same time, during the data processing process, the field recognition results can help to discover and correct potential data errors (such as data type mismatch, data missing, etc.) during the data analysis stage, thereby improving the accuracy and reliability of the data. In addition, clear field recognition results can provide a clear statistical dimension and statistical indicator basis for subsequent data analysis, which is conducive to targeted combination of specific data analysis methods to achieve accurate analysis of statistical data and improve the accuracy of data analysis results.

[0066] S22, determining data description information based on the statistical data and field identification results.

[0067] In some feasible implementations, the data description information is used to describe the data characteristics of the statistical data. The data description information includes but is not limited to at least one of the following:

[0068] Data summary information, which is used to describe the field characteristics of the field recognition result;

[0069] Dimensional statistical features, which include data features of each statistical value under each statistical dimension;

[0070] Indicator statistical characteristics, including data characteristics of each statistical value corresponding to each statistical indicator;

[0071] Supplementary feature information: Supplementary feature information is used to describe other data features besides data summary information, dimension statistical features and indicator statistical features.

[0072] In some feasible implementations, the data summary information can be specifically used to describe the number of statistical indicators in the field recognition results, the number of statistical dimensions, the specific statistical indicators included, the specific statistical dimensions included, the field type of each statistical dimension, the number of statistical dimensions of each field type, the start and end time when the statistical dimension is a time dimension, etc., which can be specifically determined based on the actual application scenario requirements and are not limited here. For example, it can also include repeated statistical indicators, repeated statistical dimensions, etc.

[0073] For example, the expression format of data summary information can be: "The result set contains xx statistical dimensions and xx statistical indicators; the current result set may contain a time dimension ds, the start time is xxxx, and the end time is xxxx".

[0074] As an example, assume that the statistical data is a class report card, each student is a statistical indicator, each subject is a statistical dimension, and the statistical value of each statistical indicator under a statistical dimension is the test score of a subject of a student. In this case, the data summary information can describe the number of students, the number of subjects, etc. included in the statistical data.

[0075] In some feasible implementations, dimensional statistical features include data features of all statistical values ​​under each statistical dimension. The data features of all statistical values ​​under each statistical dimension may include the number of statistical values, maximum value, minimum value, standard deviation, median, mode, mean, variance, number of unique values, number of common values ​​(mode), etc. The specific features can be determined based on the actual application scenario requirements and are not limited here.

[0076] The statistics of the above-mentioned data features can be determined based on different feature statistical functions.

[0077] When the field identification result includes the field type of the statistical dimension, for each statistical dimension, when determining the data characteristics of the statistical values ​​of each statistical indicator under the statistical dimension based on the statistical data and the field identification result, a characteristic statistical function matching the field type corresponding to the statistical dimension can be determined.

[0078] Furthermore, according to the matched characteristic statistical function, characteristic statistical processing is performed on the statistical value of each statistical indicator under the statistical dimension to obtain data characteristics of the statistical value of each statistical indicator under the statistical dimension.

[0079] The determined data features are determined by the function of the adopted feature statistical function. The above feature statistical function can be a describe function based on Python or other database functions, which is not limited here.

[0080] For each statistical indicator, a unique value corresponding to the statistical indicator is the statistical value of the statistical indicator in a statistical dimension, and the unique value is different from the statistical values ​​of the statistical indicator in all other statistical dimensions.

[0081] As an example, taking the class transcript as an example, the dimension statistical features include the data features of the test scores of each student in each subject. For each student, the data features of the test scores of each subject of the student may include the highest score of each subject, the lowest score of each subject, the average score of each subject, the standard deviation of the scores of each subject, the median of the test scores of each subject, the mode of the test scores of each subject, the variance of the test scores of each subject, the number of test scores that appear only once in the test scores of each subject (the number of unique values), and the number of test scores that appear most frequently in the test scores of each subject (the number of common values). The specific features can be determined based on the actual application scenario requirements and are not limited here.

[0082] In some feasible implementations, the indicator statistical characteristics include data characteristics of all statistical values ​​corresponding to each statistical indicator. The data characteristics of all statistical values ​​corresponding to each statistical indicator may include the number of statistical values, maximum value, minimum value, standard deviation, median, mode, mean, variance, number of unique values, number of common values ​​(mode), etc. The specific characteristics can be determined based on the actual application scenario requirements and are not limited here.

[0083] For each statistical dimension, a unique value corresponding to the statistical dimension is a statistical value of a statistical indicator under the statistical dimension, and the unique value is different from the numerical values ​​of the statistical values ​​of other statistical indicators under the statistical dimension.

[0084] Among them, the indicator statistical characteristics can be determined according to the characteristic statistical function, and the determined data characteristics are determined by the function function of the characteristic statistical function adopted. The characteristic statistical function can be a describe function based on Python or other database functions, which is not limited here.

[0085] As an example, taking the statistical data as a class transcript as an example, the indicator statistical features include the data features of the test scores of each student in each subject. For each subject, the data features of the test scores of each student in the subject may include the highest score of each student in the subject, the lowest score of each student in the subject, the average score of each student in the subject, the standard deviation of the test scores of each student in the subject, the median of the test scores of each student in the subject, the mode of the test scores of each student in the subject, the variance of the test scores of each student in the subject, the number of test scores that appear only once in the test scores of each student in the subject (the number of unique values), and the number of test scores that appear most frequently in the test scores of each student in the subject (the number of common values). The specific features can be determined based on the actual application scenario requirements and are not limited here.

[0086] In some feasible implementations, the supplementary feature information is used to supplement the description of other data features except for data summary information, dimension statistical features and indicator statistical features. The supplementary feature information includes but is not limited to the change trend of the statistical value of each statistical indicator under each statistical dimension, the statistical value with the most occurrences, etc., which can be determined based on the actual application scenario requirements and is not limited here.

[0087] When determining the supplementary feature information of the statistical data, the data summary information, the dimension statistical features, the indicator statistical features, the statistical data, the field recognition result and the first prompt information may be input into the large language model, and the supplementary feature information may be generated by the large language model.

[0088] Among them, the first prompt information is used to prompt the large language model to generate supplementary feature information according to the input information, that is, to prompt the large language model to generate other data features except data summary information, dimension statistical features, and indicator statistical features according to the input information.

[0089] The large language model can be used to thoroughly determine the data features that have been determined, and then the statistical data can be comprehensively analyzed to obtain other data features. This not only makes the supplementary feature information more comprehensive and sufficient, but also reduces the data processing volume of the large language model, improving and optimizing the output quality and efficiency of the large language model.

[0090] S23, input the field recognition results and statistical data into the large language model to obtain the recommendation question.

[0091] In some feasible implementations, recommended questions are used to guide the target object to query and ask questions about specified data features of statistical data, such as guiding the target object to inquire about abnormal situations in the statistical values ​​of a certain statistical indicator under various statistical dimensions, such as guiding the target object to query abnormal values ​​in statistical data, etc. The specific requirements can be determined based on the actual application scenario and are not limited here.

[0092] In some feasible implementations, when determining the recommendation question, the field recognition result may be directly input into the large language model, so that the large language model generates at least one recommendation question according to the field recognition result.

[0093] In some feasible implementations, when determining the recommendation question, the second prompt information may be determined according to the field recognition result.

[0094] The second prompt information includes a field recognition result, at least one data analysis method, and an execution condition of each data analysis method.

[0095] Among them, different data analysis methods have different data analysis functions, including attribution analysis function, anomaly detection function and trend prediction function.

[0096] Among them, the second prompt information is used to prompt the large language model to determine at least one executable data analysis method that can perform data analysis on the statistical data based on the execution conditions of each data analysis method, and generate recommended questions corresponding to each executable data analysis method based on the field recognition results.

[0097] Among them, the recommended questions corresponding to each executable data analysis method meet the execution conditions of the corresponding executable data analysis method, and are used to guide the target object to query and ask questions about data features of statistical data associated with the corresponding data analysis function.

[0098] In other words, the recommended questions corresponding to each executable data analysis method are used to guide the target object to ask questions about the data features that match the data analysis function corresponding to the executable data analysis method. In other words, the recommended questions corresponding to each executable data analysis method are the query questions that can be implemented by the executable data analysis method.

[0099] Optionally, the second prompt information may also instruct the large language model to determine a preset number of executable data analysis methods that can perform data analysis on the statistical data based on the execution conditions of each data analysis method, and then field recognition results to generate a recommendation question corresponding to each executable data analysis method.

[0100] When the number of executable data analysis methods capable of performing data analysis on statistical data is less than a preset number, a recommended question is generated for each executable data analysis method based on the field recognition result, and then a first number of recommended questions is generated based on one or more of the executable data analysis methods, where the first number is the number of missing recommended questions.

[0101] Optionally, the second prompt information may also include a data analysis function description of each data analysis method, recommended question generation requirements, and one or more of the recommended question examples corresponding to each data analysis method, so that the large language model can accurately generate recommended questions.

[0102] The above-mentioned recommended question generation requirements may restrict the generation method and the number of generated questions, etc. For example, the recommended questions of each data analysis method must strictly meet the execution conditions, each data analysis method only generates one recommended question, and if the last data analysis method has not generated a preset number of recommended questions, the corresponding number of recommended questions will be supplemented according to the last data analysis method.

[0103] As an example, the second prompt information includes the following parts:

[0104] 1. Recommended Question Generation Requirements

[0105] This part can be expressed as:

[0106] There are a series of data analysis methods that can assist in completing data analysis tasks. Please consider the execution conditions of the data analysis methods according to the order of the data analysis methods in the list and the data table provided, and generate three recommended questions that the target object can query and ask.

[0107] Among them, the recommended questions of each data analysis method must strictly meet the execution conditions. In principle, each data analysis method only generates one recommended question. If three recommended questions have not been generated by the last data analysis method, the recommended questions will be supplemented by the last data analysis method.

[0108] 2. Field recognition results

[0109] This part can be shown in Table 1.

[0110] Table 1:

[0111]

[0112] 3. Method Description

[0113] This section can include the name of the data analysis method, a description of its functionality, execution conditions, and examples of recommended questions.

[0114] Among them, this part can describe multiple data analysis methods in the format of "method name-function description-execution conditions-recommended question examples".

[0115] For example, the method description section could be arranged as follows:

[0116] Data analysis method 1—functional description—execution conditions—recommended question examples;

[0117] Data Analysis Method 2—Functional Description—Execution Conditions—Recommended Question Examples;

[0118] Data Analysis Method 3—Functional Description—Execution Conditions—Recommended Question Examples;

[0119] Among them, data analysis method 1 can be expressed as: IData Anonely Detection—data anomaly detection, which can detect whether there are outliers—requires a statistical indicator and a time dimension (statistical indicator: fieldType='index', statistical time dimension: fieldTrpe='index'&&dateType=true)—check the data anomaly of the statistical indicator cnt under the time ts condition.

[0120] Data analysis method 2 can be expressed as: Data_Attribute_Analysis—data fluctuation analysis, check the fluctuation of statistical indicators—requires a statistical indicator, a statistical dimension and a time dimension—the statistical indicator is Figure 2 When if_select_dataresource is used as the analysis dimension, the fluctuation of time ds is used.

[0121] Data analysis method three can be expressed as: Data_Describe_Simplify—simplified version of descriptive statistics—requires a statistical indicator or a statistical dimension (statistical indicator: fieldType='index', statistics and dimension: fieldType='index')—view the statistical information of the indicator cnt.

[0122] For the large language model, the large language model can generate recommendation questions according to the second prompt information through the Chain of Thoughts prompt technology.

[0123] Specifically, the large language model first determines whether Data Analysis Method 1 (Data_Anomaly_Detection) can be executed based on the field recognition results and the execution conditions of Data Analysis Method 1. When Data Analysis Method 1 can be executed, a recommendation question is generated: View statistical indicators Figure 2 The data is abnormal under the condition of time ds. Further, the large language model determines whether the data analysis method 2 (Data_Attribute_Analysis) can be executed based on the field recognition results and the execution conditions of the data analysis method 2. When the data analysis method 2 can be executed, a recommendation question is generated: the statistical indicators are Figure 2 When if_seect_dataresource is used as the analysis dimension, the fluctuation of time ds is analyzed. Further, the large language model determines whether data analysis method 2 (Data_Describe_simplify) can be executed based on the field recognition results and the execution conditions of data analysis method 3. When data analysis method 3 can be executed, a recommendation question is generated: Check the distribution of the statistical dimension if_select_dataresource.

[0124] S24, displaying data description information and recommended questions.

[0125] In some feasible implementations, after the data description information of the statistical data is determined and the recommended questions are obtained through the large language model, the data description information and the recommended questions may be displayed to the target object.

[0126] Among them, the data description information displayed to the target object includes any one or more of data summary information, dimension statistical characteristics, indicator statistical characteristics and supplementary characteristic information, which can be determined based on the actual application scenario requirements and is not limited here.

[0127] As an example, see Figure 5 , Figure 5 Schematic diagram of data analysis results provided in the embodiment of the present application. Figure 5 As shown, after obtaining the statistical data input by the target object, data description information and recommended questions can be obtained based on the method shown in steps S21-S23, and then the data analysis results including the data description information and recommended questions can be displayed to the target object through the data interaction interface.

[0128] For example, Figure 5 The data overview displayed through the data interaction interface is the data summary information in the data description information, which is used to describe the field characteristics of the field recognition results. Figure 5 The indicator statistics displayed through the data interaction interface are the indicator statistical characteristics in the data description information, which are used to describe the data characteristics of each statistical value corresponding to each statistical indicator. Figure 5 The dimensional distribution part displayed through the data interaction interface is the dimensional statistical characteristics in the data description information, which is used to describe the data characteristics of all statistical values ​​under each statistical dimension. Figure 5 The other insights displayed through the data interaction interface are the supplementary feature information in the data description information, which are used to supplement the description of other data features not mentioned in the data summary information, dimension statistical features, and indicator statistical features.

[0129] in, Figure 5 The data interaction interface can also display the recommendation questions generated by the large language model, such as the dimensional distribution of product_name, the number of is_active active customers, the number of different values ​​of bu_product, and the most common value, to guide the target object to further query and ask questions about the data characteristics of the statistical data.

[0130] Optionally, after displaying the data description information and recommended questions to the target object, feedback information from the target object regarding the data description information may also be obtained, and then the adopted data analysis method may be optimized or replaced based on the feedback information from the target object, thereby improving the quality of the generated data description information.

[0131] Among them, the feedback information of the target object can be the evaluation information input by the target object through the data interaction interface, or it can be triggered by the target object through the evaluation touch buttons (such as the like button and the dislike button) in the data interaction interface, and there is no limitation here.

[0132] By determining and displaying the data description information of statistical data (data summary information, indicator statistical information, dimension statistical characteristics and supplementary characteristic information, etc.), the basic data characteristics of statistical data at all levels can be quickly extracted in an all-round way, which helps the target object to quickly understand the basic characteristics of statistical data, such as mean, median, standard deviation, etc., which can not only provide a data basis for subsequent possible data analysis processes, but also enhance the richness and diversity of data description information, and provide subsequent data analysis direction for the target object.

[0133] Moreover, the large language model can identify the results according to the execution conditions of different data analysis methods and the fields of statistical data. It can automatically generate recommendation questions that can be used to query different data features according to the data characteristics of the statistical data and the data analysis methods that can be adopted, providing a wider range of data analysis directions for the target object. Moreover, generating recommendation questions through the large language model can not only improve the generation efficiency of recommendation questions, but also help to provide the depth and breadth of analysis of recommendation questions, thereby improving the data analysis experience of the target object in the human-computer interaction process.

[0134] In some feasible implementations, after the data description information and the recommended questions are displayed to the target object, a first query question input by the target object may be further obtained.

[0135] The first query question is used to query the specified data features of the statistical data. For example, if the statistical data is a sales report, the first query question can be a query question related to the sales situation and other sales indicators.

[0136] Since a certain number of recommended questions have been displayed to the target object, after obtaining the first query question input by the target object, it may be determined whether the first query question belongs to the recommended questions displayed to the target object.

[0137] When the first query question input by the target object is a recommended question displayed to the target object, the executable data analysis method corresponding to the first query question can be determined as a target data analysis method that matches the first query question, thereby analyzing the statistical data according to the first query question and the determined target data analysis method to obtain and display the data analysis result corresponding to the first query question.

[0138] For the convenience of description, when the first query question belongs to a recommendation question displayed to the target object, the first query question is referred to as a target recommendation question.

[0139] The executable data analysis method corresponding to the target recommendation question is the data analysis method associated with the generation of the target recommendation question, that is, the data analysis method that meets the corresponding execution conditions of the target recommendation question. In this case, the executable data analysis method corresponding to the target recommendation question is the target data analysis method that matches the first query question.

[0140] Based on this, the statistical data can be analyzed according to the first query question and the determined target data analysis method. For example, according to the statistical dimensions and statistical indicators involved in the first query question, as well as the data characteristics that need to be queried, the target data analysis method is executed on the statistical data to obtain and display the data analysis results corresponding to the first query question to the target object.

[0141] When the first query question input by the target object does not belong to the recommended question displayed to the target object, the first query question can be input into the large language model to perform intent recognition on the first query question through the large language model, and then the statistical data can be analyzed according to the intent recognition result to obtain and display the data analysis result corresponding to the first query question.

[0142] In other words, when the first query question input by the target object does not belong to the recommended questions displayed to the target object, intent recognition is required to determine the target object's true query intent so that the statistical data can be analyzed based on the target object's query intent.

[0143] In some feasible implementations, when the first query question is not a recommended question displayed to the target object, and after intent recognition is performed on the first query question to obtain an intent recognition result, it can be determined whether the query intent of the first query question is recognized based on the intent recognition result.

[0144] Specifically, when it is determined based on the intention recognition result that the large language model recognizes the query intent of the first query question, the target data analysis method that matches the first query question can be determined based on the query intent of the first query question, and then the statistical data can be analyzed based on the first query question and the determined target data analysis method to obtain and display the data analysis results corresponding to the first query question.

[0145] Among them, when determining the target data analysis method that matches the first query question according to the query intent of the first query question, the candidate data analysis method whose data analysis function is associated with the query intent of the first query question can be determined first. That is, the candidate data analysis method whose data analysis function supports the query intent of the first query question is determined first.

[0146] Furthermore, after determining the candidate data analysis method, the key information of the query intent of the first query question can be determined, and when the execution conditions of the candidate data analysis method match the key information, the candidate data analysis method is determined as the target data analysis method that matches the first query question.

[0147] When the key information matches the execution condition of a data analysis method, it indicates that the data analysis method can execute the query requirement of the first query question.

[0148] Among them, the key information of the query intent of the first query question and the process of determining whether the execution conditions of the candidate data analysis method match the key information can be determined by performing slot detection on the first query intent.

[0149] Slot detection plays a vital role in artificial intelligence dialogue systems. It mainly determines whether the input information contains specific information (i.e., slots) based on the input information and intent recognition algorithm, and evaluates whether this information is sufficient to support the system to correctly execute the corresponding request.

[0150] A slot represents the key information that needs to be collected to complete a task in a specific field. For example, in the field of air tickets, there are slots such as departure city, destination city, departure date, etc. Each slot needs to identify the corresponding key information, such as departure city: Beijing, destination city: Shenzhen, departure date: *month*day, etc.

[0151] In an embodiment of the present application, the slot information required to execute each data analysis method can be determined based on the execution conditions of each data analysis method. Then, based on the slots required by each data analysis method, the query intent of the first query question is detected by slot detection, that is, the key information of the query intent is detected. When all the slots corresponding to a certain data analysis method are detected, that is, when it is determined that the execution conditions of a certain data analysis method match the key information of the query intent, it can be determined that the data analysis method can be used to analyze statistical data according to the first query question, and then the data analysis method can be determined as a target data analysis method that matches the first query question.

[0152] Among them, during the slot detection process, the slot identification process can be realized by named entity recognition (NER) technology, which can extract key information in the first query question, such as entity words. In addition, during the slot detection process, the slot can be filled by asking the target object, that is, more slot information is obtained to better understand the query requirements of the first query question. At the same time, the identified slots can also be verified to ensure their accuracy and rationality. For possible errors or ambiguities, error correction can be performed or the target object can be asked to obtain more accurate information.

[0153] For example, if the first query question input by the target object is "check the abnormal situation of indicator cnt", then the large language model can identify that the query intention of the first query question is that the target object needs to perform abnormal indicator detection on the indicator cnt in the statistical data. If there is only a time dimension ds in the statistical data, the query intention of the first query question will be supplemented as "check the abnormal situation of indicator cnt under the time dimension" during the intention recognition process or the slot detection process. If there are multiple time dimensions, the target time dimension that needs to be queried for the first query question can be determined by asking the target object. In the slot detection process, the requirements for statistical indicators and statistical dimensions for executing the candidate analysis method can be determined according to the execution conditions of the candidate analysis method (such as the need for a time dimension and a statistical indicator). When the key information of the query intention is determined to meet the requirement through slot detection, the candidate analysis method can be determined as the target data analysis method that matches the first query question.

[0154] Specifically, when it is determined based on the intent recognition result that the large language model has not recognized the query intent of the first query question, the statistical data and the first query question can be input into the large language model so that the large language model performs data analysis on the statistical data in a preset manner based on the first query question to obtain a data analysis result corresponding to the first query question.

[0155] In this case, the data analysis result corresponding to the first query question may be obtained from the large language model, and the data analysis result may be displayed to the target object.

[0156] After the first query question is input into the large language model, the large language model can convert the first query question into a computer executable code, such as a python code, and execute the code against the statistical data. After the execution is completed, the large language model can process the execution result according to the first query question and return the final data analysis result.

[0157] Optionally, when the execution conditions of the aforementioned candidate data analysis methods do not match the key information, that is, after determining the candidate data analysis method whose data analysis function is associated with the query intent of the first query question, when the key information of the query intent does not meet the execution conditions of the candidate data analysis method, the statistical data and the first query question can also be input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the first query question to obtain the data analysis result corresponding to the first query question.

[0158] In this case, the data analysis result corresponding to the first query question can also be obtained from the large language model, and the data analysis result can be displayed to the target object.

[0159] Combine the following Figure 6The data processing flow after displaying data description information and recommended questions is further explained. Figure 6 Schematic diagram of the process framework for processing query problems provided by the embodiment of the present application. Figure 6 As shown, after the data description information and the recommended questions are displayed to the target object, when the first query question input by the target object is obtained, it is first necessary to determine whether the first query question is a recommended question.

[0160] For example, the first query question is: query the abnormal values ​​in Zhang San’s daily reading volume in the past week, then determine whether the question is included in the recommended questions displayed. If included, determine that the first query question is a recommended question, otherwise determine that the first query question is not a recommended question.

[0161] When the first query question is a recommendation question, the executable data analysis method corresponding to the first query question (such as the data anomaly detection tool Data_Anomaly_Detection) is determined as the target data analysis method matching the first query question. At this time, the statistical data can be analyzed for outliers according to the target data analysis method.

[0162] When the first query question is not a recommendation question, the first query question is input into the large language model to identify the intent of the first query question through the large language model, and the query intent of the target object is obtained as: determine whether there is an abnormality in Zhang San's daily reading volume in the past week. In the case where the large language model identifies the query intent of the first query question, the matching target data analysis method can be determined as a data anomaly detection method (such as a data anomaly detection tool Data_Anomaly_Detection) according to the query intent, and the statistical data can be analyzed according to the target data analysis method.

[0163] When the first query question is not a recommendation question and the large language model does not recognize the query intent of the first query question, such as the first query question is: Zhang San's reading, the statistical data and the first query question can be input into the large language model so that the large language model can perform data analysis on the statistical data in a preset manner.

[0164] Based on the above process, after analyzing the statistical data, the data analysis result corresponding to the first query question can be obtained and displayed.

[0165] In some feasible implementations, when obtaining the statistical data of the target input, the second query question input by the target object is also obtained. Since no recommended questions have been displayed to the target object at this time, the second query question can be input into the large language model to identify the intent of the second query question through the large language model.

[0166] The second query question is also used to query the specified data features of the statistical data.

[0167] Furthermore, when it is determined based on the intent recognition result that the large language model has recognized the query intent of the second query question, a target data analysis method that matches the second query question can be determined based on the query intent of the second query question, and then the statistical data can be analyzed based on the second query question and the determined target data analysis method to obtain and display the data analysis results corresponding to the second query question.

[0168] Among them, when determining the target data analysis method that matches the second query question according to the query intent of the second query question, the candidate data analysis method whose data analysis function is associated with the query intent of the second query question can be determined first. That is, the candidate data analysis method whose data analysis function supports the query intent of the second query question is determined first.

[0169] After determining the candidate data analysis method, the key information of the query intent of the second query question can be determined, and when the execution conditions of the candidate data analysis method match the key information, the candidate data analysis method is determined as the target data analysis method that matches the second query question.

[0170] Among them, when the key information corresponding to the second query question matches the execution condition of a certain data analysis method, it means that the data analysis method can execute the query requirement of the second query question.

[0171] Among them, the key information of the query intent of the second query question and the process of determining whether the execution conditions of the candidate data analysis method match the key information can be determined by performing slot detection on the second query intent. For details, please refer to the implementation method shown above and will not be repeated here.

[0172] When it is determined based on the intention recognition result that the large language model has not recognized the query intention of the second query question, the statistical data and the second query question can be input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner based on the second query question to obtain the data analysis result corresponding to the second query question, and then the data analysis result corresponding to the second query question can be obtained and displayed.

[0173] After the second query question is input into the large language model, the large language model can convert the second query question into a computer executable code, such as a python code, and execute the code against the statistical data. After the execution is completed, the large language model can process the execution result according to the second query question and return the final data analysis result.

[0174] It should be specially noted that the data analysis results returned for the query questions in the embodiments of the present application may be in text form, in chart form, or in a combination of the two, and there is no limitation on this.

[0175] Combine the following Figure 7 The process of the data processing method provided in the embodiment of the present application is further explained. Figure 7 Schematic diagram of the timing of the data processing method adopted in the embodiment of the present application. Figure 7 The data processing method shown can be implemented through the insight service, interpretation algorithm and large language model, and can specifically include the following steps:

[0176] Step 1: Interact with the target object through the data interaction interface to obtain information input by the target object.

[0177] Step 2: Verify the information input by the target object. If the target object only inputs statistical data, execute step 3. If the target object inputs statistical data and query questions, directly execute step 13. If the target object does not input statistical data, it will return "query result has no data, no interpretation for the time being".

[0178] Among them, the query question input by the target object is of string type, and the statistical data is of table type data that needs to be interpreted.

[0179] Step 3: Perform field recognition processing on the statistical data to obtain statistical dimensions and statistical indicators.

[0180] Step 4: Return the field recognition result.

[0181] Step 5: Perform descriptive statistics-describe based on the statistical data and field identification results to obtain data summary information, dimension statistical characteristics, and indicator statistical characteristics.

[0182] Step 6, input the statistical technical bureau, field recognition results, and the information obtained in step 5 into the large language model, and obtain supplementary feature information through the large language model.

[0183] Step 7: Return the supplementary feature information.

[0184] Step 8: Input the field recognition results into the large language model to obtain the recommendation questions.

[0185] Step 9, return to the recommendation question.

[0186] Step 10: Display data description information (data summary information, dimension statistical characteristics, indicator statistical characteristics and supplementary characteristic information) and recommended questions to the target object.

[0187] Step 11: The target object asks questions through the data interaction interface.

[0188] Step 12, determine the source of the question. If the target object directly selects the recommended question, then directly execute step 17. If the target object asks a question on its own, then execute step 13.

[0189] Step 13: Use the large language model to identify the intent of the additional question.

[0190] Step 14, returning the result of intent recognition.

[0191] Step 15: perform slot detection on the query intent.

[0192] Step 16, returning the slot detection result, and determining the data analysis method to be used according to the slot detection result.

[0193] Step 17, analyzing the statistical data according to the determined data analysis method.

[0194] Step 18, return the data analysis results.

[0195] Step 19: Display the data analysis results.

[0196] Among them, the specific implementation methods of the above steps 1-19 can be found in the above related descriptions and will not be repeated here.

[0197] In an embodiment of the present application, after displaying data description information and recommended questions to the target object, the query questions asked by the target object in the subsequent human-computer interaction process can also be obtained. When the query question is a recommended question displayed to the target object, the data analysis method to be used can be quickly determined and the statistical data can be analyzed, which is beneficial to improving the efficiency of data analysis. When the query question is a question raised by the target object itself, valuable information in the query question can be extracted through intent recognition and slot detection, which is beneficial to determine the data analysis method that meets the query requirements of the target object and improve the flexibility of the data analysis method.

[0198] Moreover, when it is impossible to determine a data analysis method that meets the query conditions of the query question through intent recognition and slot detection, the query question can be converted into executable code through the powerful data processing capabilities of the large language model, and the statistical data can be analyzed through the executable code. In this way, even if it is impossible to determine a data analysis method that meets the query conditions of the query question through intent recognition and slot detection, the corresponding data analysis results can still be provided to the target object, further improving the flexibility and scalability of the data analysis method and improving the efficiency of human-computer interaction.

[0199] In some feasible implementations, the target data analysis method that matches the first query question includes but is not limited to attribution analysis methods, outlier detection methods, and trend prediction methods, which can be specifically determined based on the actual application scenario requirements. For example, it can also include methods for analyzing data correlation, regression analysis methods, clustering analysis methods, etc., which are not limited here.

[0200] Among them, attribution analysis methods are used to explain changes or differences in data. Its main purpose is to determine and quantify the contribution of various factors to overall changes. This analysis can help us understand which factors have the greatest impact on the results, so that we can better understand the data and make more informed decisions accordingly. The scales for the impact of quantitative dimensions in attribution problems are contribution and surprise. Attribution analysis methods can specifically include single-dimensional basic indicator analysis and composite indicator analysis.

[0201] When the first query question is used to query the reasons for the changes in various statistical values ​​of the statistical data, the attribution analysis method can be used to analyze the statistical data to obtain data analysis results.

[0202] Specifically, the comparison dimension corresponding to each statistical dimension and the statistical value of each statistical indicator under each comparison dimension may be determined.

[0203] The comparison dimension corresponding to each statistical dimension is the reference dimension of the statistical dimension. For example, when each statistical dimension in the statistical data is a time dimension, the comparison dimension corresponding to each statistical dimension is the historical same period of time.

[0204] Furthermore, for each statistical indicator, the surprise degree of the statistical indicator is determined according to the statistical value of the statistical indicator in each statistical dimension and the corresponding comparison dimension, and the surprise degree is used to characterize the degree of change of the statistical value of the statistical indicator.

[0205] For each statistical indicator, the degree of surprise can be determined in the following way:

[0206]

[0207] Among them, P is the statistical value under the statistical dimension, and Q is the statistical value under the comparison dimension. is the statistical value of the statistical indicator under the statistical dimension i, It is the statistical value of the statistical indicator under the comparison dimension corresponding to the statistical dimension i.

[0208] Furthermore, multiple indicator combinations may be determined, each indicator combination including a statistical indicator and a statistical dimension, that is, one statistical value in the statistical data corresponds to one indicator combination.

[0209] For each statistical indicator and statistical dimension within each indicator combination, the contribution degree corresponding to the indicator combination is determined according to the statistical value of the statistical indicator under the statistical dimension and the corresponding comparative statistical dimension, and the sum of the statistical values ​​of each statistical indicator under the statistical dimension and the corresponding comparative statistical dimension. The contribution degree is used to characterize the contribution degree of the influencing factor corresponding to the statistical indicator to the change of the statistical value under the statistical dimension.

[0210] Among them, the impact factor corresponding to each statistical indicator under each statistical dimension is used to distinguish the differences between the statistical indicators under the statistical dimension. For example, if the statistical dimension is January, and the statistical indicators are the ratings of the young group, the ratings of the elderly group, and the ratings of the middle-aged group, then the impact factor corresponding to each statistical indicator is age.

[0211] Among them, the contribution corresponding to each indicator combination can be determined by the following method:

[0212]

[0213] The current statistical dimension is A, the dimension value is a, and the statistical value of the current statistical indicator under the current statistical dimension is , the current statistical indicator's statistical value under the corresponding comparison dimension is , P is the sum of the statistical values ​​under the statistical dimension, and Q is the sum of the statistical values ​​under the comparison dimension.

[0214] After the degree of surprise and the degree of contribution are determined, the reasons for the changes in the various statistical values ​​of the statistical data can be determined according to at least one of the degree of surprise or the degree of contribution, and the data analysis results can be determined according to the reasons for the changes. For example, the degree of surprise and the degree of contribution can be counted to obtain the maximum value and summary value (such as the mean, median, mode, standard deviation, etc.), and then a single-dimensional attribution analysis can be performed based on the statistical summary value and the maximum value to analyze the reasons for the changes in the statistical values ​​in the statistical data, and finally obtain the data analysis results.

[0215] Trend forecasting method is an important forecasting method, also known as time series forecasting analysis method. It is based on the principle of continuity of business development, arranges past historical data in chronological order by applying mathematical statistics methods, and uses certain digital models to predict and speculate data changes in the planning period.

[0216] Specifically, when the statistical dimension in the statistical data includes a time dimension, and the first query question is used to query the future change trend of the statistical value of the target statistical indicator, a first data sequence can be determined, the first data sequence includes the statistical value of the target statistical indicator in each time dimension, and the arrangement order of each statistical value in the first data sequence is consistent with the arrangement order of the corresponding time dimension. That is, the first data sequence is obtained by arranging each statistical value of the target statistical indicator in time order.

[0217] Furthermore, a preset time interval is determined, and then the statistical value change trend of the target statistical indicator within the preset time interval is predicted according to the relevant data prediction function or model (such as the Prophet model), so as to determine the final data analysis result according to the obtained change trend.

[0218] The preset time interval may be specified by the target object through the first query question, or may be a default preset time interval, which is not limited here.

[0219] Among them, the Prophet model is a widely used time series forecasting model that can handle trends and seasonal changes in time series. The advantages of the Prophet model are that it is easy to use and understand and can automatically detect trend change points.

[0220] Taking the Prophet model as an example, the first data sequence is input into the Prophet model, and a preset time interval is determined. Further, the seasonal parameters, periodic parameters, trend parameters and other related model parameters of the Prophet model are set so that the Prophet model performs sequence decomposition on the first data sequence according to the sequence characteristics of the first data sequence, and then performs fitting processing on the decomposition result according to the preset fitting method, outputs the trend of the statistical value of the target statistical indicator in the preset time interval, and can also input and provide necessary trend indicators, such as fitting curves, upper and lower bounds, etc.

[0221] Specifically, when analyzing statistical data, the target data analysis method used can also be an outlier detection method. When the first query question is used to query abnormal statistical values ​​in statistical data, multiple indicator combinations can be determined, each indicator combination includes a statistical indicator and a statistical dimension, that is, one statistical value in the statistical data corresponds to one indicator combination.

[0222] Furthermore, for each statistical indicator and statistical dimension in each indicator combination, the predicted value of the statistical indicator under the statistical dimension can be determined based on the trend prediction method, and it can be determined whether the predicted value is within a reasonable range. When the predicted value is outside the statistical value range corresponding to the indicator combination, the statistical value corresponding to the indicator combination can be determined as an abnormal statistical value. Based on this, the abnormal statistical value in the statistical data can be determined, and then the data analysis result can be determined according to the determined abnormal statistical value.

[0223] Alternatively, for each statistical indicator and statistical dimension in each indicator combination, a growth factor of the statistical value corresponding to the indicator combination is determined, and abnormal statistical values ​​in the statistical data are determined based on the growth factor, so as to determine the data analysis results based on the determined abnormal statistical values.

[0224] The growth factor may include at least one of a month-on-month growth rate or a year-on-year growth rate. Based on the growth factor, the abnormal growth inflection point and the abnormal decline inflection point in the statistical data can be determined, and then the abnormal statistical value in the statistical data can be determined based on the abnormal growth inflection point and the abnormal decline inflection point.

[0225] Among them, the month-on-month growth rate corresponding to each statistical indicator = (statistical value under the current statistical dimension - statistical value under the previous statistical dimension) / statistical value under the previous statistical dimension × 100%, and the year-on-year growth rate corresponding to each statistical indicator = (statistical value under the current statistical dimension - statistical value under the historical same period dimension) / statistical value under the historical same period dimension × 100%.

[0226] It should be noted that any of the above methods for determining data analysis results are only examples, and any data analysis method in the field of statistics and data analysis can be used, without limitation herein.

[0227] The data processing method provided in the embodiment of the present application can be applied to a data interpretation system in a human-computer dialogue application. Figure 8 The data processing method provided in the embodiment of the present application is further explained. Figure 8 Schematic diagram of the process framework of the data processing method provided in the embodiment of the present application. Figure 8 As shown, the statistical data input by the target object can be obtained through the human-computer interaction interface, and after the data is interpreted, the data description information and recommended questions obtained by the interpretation are displayed through the human-computer interaction interface.

[0228] After obtaining the recommendation question input by the target object again, the data analysis method corresponding to the recommendation question can be directly determined, such as determining the data analysis method to be adopted from data analysis method one to data analysis method n (such as attribution analysis method, trend prediction method, and outlier detection method, etc.), and then interpreting the statistical data according to the determined data analysis method to obtain the corresponding data analysis results.

[0229] After the first query question input by the target object in the form of autonomous questioning is obtained again, the intent of the first query question can be identified through the large language model.

[0230] If the corresponding query intention is not identified, the first query question is directly input into the large language model, and the large language model interprets the statistical data according to a preset method to obtain a data analysis result.

[0231] If the corresponding query intent is identified, the first query question is slot detected to obtain key information, and it is determined whether the key information meets the execution conditions of each method in data analysis method 1 to data analysis method n. If there is a data analysis method that meets the execution conditions, the statistical data is interpreted according to the data analysis method to obtain the data analysis result. If there is no data analysis method that meets the execution conditions, the first query question is directly input into the large language model, and the large language model interprets the statistical data according to a preset method to obtain the data analysis result.

[0232] The embodiments of the present application can support a variety of data analysis methods, can quickly respond to and process query questions raised by the target object according to the corresponding data analysis method, provide timely and accurate data analysis results, and help improve the accuracy and reliability of data analysis. In addition, the data analysis method used in the embodiments of the present application can be replaced and adjusted at any time, which helps to meet the needs of different business scenarios, has higher scalability, meets the data analysis needs of different target objects, and has high applicability.

[0233] On the other hand, the embodiments of the present application combine a large language model with various data analysis methods. When a large language model is used only to generate recommendation questions, and when a data analysis method cannot determine the query conditions that meet the query questions through intent recognition and slot detection, a large language model is used to generate data analysis results. This can reduce dependence on artificial intelligence, thereby reducing the cost and difficulty of data analysis and improving the feasibility and universality of data analysis.

[0234] See also Fig. 9 , Fig. 9 : is a schematic diagram of the structure of a data processing device provided in an embodiment of the present application. The data processing device provided in an embodiment of the present application includes:

[0235] A field recognition module 91 is used for, in response to obtaining statistical data input by a target object, performing field recognition processing on the statistical data to obtain a field recognition result, wherein the field recognition result includes the recognized statistical indicators and statistical dimensions;

[0236] A data analysis module 92, configured to determine data description information according to the statistical data and the field identification result, wherein the data description information is used to describe data features of the statistical data;

[0237] The data analysis module 92 is used to input the field recognition results into a large language model to obtain recommended questions, and the recommended questions are used to guide the target object to query and ask questions about the specified data features of the statistical data;

[0238] The information display module 93 is used to display the above data description information and the above recommended questions.

[0239] In some feasible implementations, the data description information includes at least one of the following:

[0240] Data summary information, the data summary information is used to describe the field characteristics of the field recognition result;

[0241] Dimensional statistical features, the above-mentioned dimensional statistical features include data features of each statistical value under each of the above-mentioned statistical dimensions;

[0242] Indicator statistical characteristics, the above-mentioned indicator statistical characteristics include data characteristics of each statistical value corresponding to each of the above-mentioned statistical indicators;

[0243] Supplementary feature information, the above-mentioned supplementary feature information is used to describe other data features besides the above-mentioned data summary information, the above-mentioned dimension statistical features and the above-mentioned indicator statistical features.

[0244] In some feasible implementations, the field identification result further includes the field type of each of the statistical dimensions; for each of the statistical dimensions, the data analysis module 92 is used to:

[0245] Determine a characteristic statistical function that matches the field type corresponding to the statistical dimension;

[0246] By using the characteristic statistical function, characteristic statistical processing is performed on the statistical values ​​of each of the above statistical indicators under the statistical dimension to obtain data characteristics of the statistical values ​​of each of the above statistical indicators under the statistical dimension.

[0247] In some feasible implementations, the data analysis module 92 is used to:

[0248] Inputting the data summary information, the dimension statistical features, the indicator statistical features, the statistical data, the field recognition result and the first prompt information into the large language model to obtain the supplementary feature information;

[0249] The first prompt information is used to prompt the large language model to generate the supplementary feature information according to the input information.

[0250] In some feasible implementations, the data analysis module 92 is used to:

[0251] Determine second prompt information, input the second prompt information into the large language model to obtain a recommended question, wherein the second prompt information includes the field recognition result, at least one data analysis method, and an execution condition of each of the data analysis methods; the second prompt information is used to prompt the large language model to perform the following operations:

[0252] Determine at least one executable data analysis method capable of performing data analysis on the statistical data according to the execution conditions of each of the data analysis methods and the field identification results;

[0253] According to the above-mentioned field recognition results, a recommended question corresponding to each of the above-mentioned executable data analysis methods is generated, and each of the recommended questions corresponding to the above-mentioned executable data analysis methods meets the execution conditions of the corresponding executable data analysis method.

[0254] In some feasible implementations, the data analysis module 92 is further used to:

[0255] After displaying the data description information and the recommended questions, in response to obtaining a first query question input by the target object, determining whether the first query question belongs to the recommended question, wherein the first query question is used to query a specified data feature of the statistical data;

[0256] In response to the first query question belonging to the recommended question, determining the executable data analysis method corresponding to the first query question as a target data analysis method matching the first query question; analyzing the statistical data according to the first query question and the determined target data analysis method, and obtaining and displaying the data analysis result corresponding to the first query question;

[0257] In response to the first query question not belonging to the recommended question, the first query question is input into the large language model to perform intent recognition on the first query question through the large language model; the statistical data is analyzed according to the intent recognition result to obtain and display the data analysis result corresponding to the first query question.

[0258] In some feasible implementations, the data analysis module 92 is used to:

[0259] In response to the large language model identifying the query intent of the first query question, determining a target data analysis method that matches the first query question according to the query intent of the first query question; analyzing the statistical data according to the first query question and the determined target data analysis method, and obtaining and displaying a data analysis result corresponding to the first query question;

[0260] In response to the large language model failing to recognize the query intent of the first query question, the statistical data and the first query question are input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the first query question to obtain a data analysis result corresponding to the first query question; and the data analysis result corresponding to the first query question is obtained and displayed.

[0261] In some feasible implementations, the data analysis module 92 is used to:

[0262] Determine a candidate data analysis method whose data analysis function is associated with the query intent of the first query question;

[0263] Determine key information of the query intent of the above-mentioned first query question, and in response to the execution conditions of the above-mentioned candidate data analysis method matching the above-mentioned key information, determine the above-mentioned candidate data analysis method as the target data analysis method matching the above-mentioned first query question.

[0264] In some feasible implementations, the data analysis module 92 is used to:

[0265] In response to obtaining the statistical data, a second query question input by the target object is also obtained, the second query question is used to query a specified data feature of the statistical data;

[0266] Inputting the second query question into the large language model to perform intent recognition on the second query question through the large language model;

[0267] In response to the large language model identifying the query intent of the second query question, determining a target data analysis method that matches the second query question according to the query intent of the second query question; analyzing the statistical data according to the second query question and the determined target data analysis method, and obtaining and displaying a data analysis result corresponding to the second query question;

[0268] In response to the large language model failing to recognize the query intent of the second query question, the statistical data and the second query question are input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the second query question to obtain a data analysis result corresponding to the second query question; and the data analysis result corresponding to the second query question is obtained and displayed.

[0269] In some feasible implementations, the first query question is used to query the reasons for the changes in the statistical values ​​of the statistical data;

[0270] The data analysis module 92 is used to:

[0271] Determine the comparison dimension corresponding to each of the above statistical dimensions, and the statistical value of each of the above statistical indicators under each of the above comparison dimensions;

[0272] For each of the above statistical indicators, determine the surprise degree of the statistical indicator according to the statistical value of the statistical indicator in each of the above statistical dimensions and the corresponding comparison dimensions, where the surprise degree is used to characterize the degree of change of the statistical value of the statistical indicator;

[0273] Determine multiple indicator combinations, each of which includes one of the above statistical indicators and one of the above statistical dimensions; for each of the statistical indicators and statistical dimensions in the above indicator combination, determine the contribution corresponding to the indicator combination according to the statistical value of the statistical indicator under the statistical dimension and the corresponding comparative statistical dimension, and the sum of the statistical values ​​of each of the above statistical indicators under the statistical dimension and the corresponding comparative statistical dimension, wherein the contribution is used to characterize the contribution of the influencing factor corresponding to the statistical indicator to the change of the statistical value under the statistical dimension;

[0274] Based on at least one of the above-mentioned surprise degree or the above-mentioned contribution degree, the reasons for the changes in the statistical values ​​of the above-mentioned statistical data are determined, and the data analysis results are determined based on the above-mentioned reasons for the changes.

[0275] In some feasible implementations, the statistical dimension includes a time dimension, and the first query question is used to query the future change trend of the statistical value of the target statistical indicator;

[0276] The data analysis module 92 is used to:

[0277] Determine a first data sequence, wherein the first data sequence includes statistical values ​​of the target statistical indicator in each of the time dimensions, and an arrangement order of the statistical values ​​in the first data sequence is consistent with an arrangement order of the corresponding time dimension;

[0278] Determine a preset time interval, determine the changing trend of the statistical value of the above-mentioned target statistical indicator within the predicted time interval according to the above-mentioned first data sequence, and determine the data analysis result according to the above-mentioned changing trend.

[0279] In some feasible implementations, the first query question is used to query abnormal statistical values ​​in the statistical data; the data analysis module 92 is used to:

[0280] Determine a plurality of indicator combinations, each indicator combination including one of the above statistical indicators and one of the above statistical dimensions;

[0281] For each statistical indicator and statistical dimension in the above indicator combination, determine the predicted value of the statistical indicator under the statistical dimension; in response to the predicted value being outside the statistical value range corresponding to the indicator combination, determine the statistical value corresponding to the indicator combination as an abnormal statistical value; and determine the data analysis result according to the determined abnormal statistical value;

[0282] For each statistical indicator and statistical dimension within the above indicator combination, determine the growth factor of the statistical value corresponding to the indicator combination, and determine the abnormal statistical value in the above statistical data based on the above growth factor. The above growth factor includes at least one of the month-on-month growth rate or the year-on-year growth rate. Determine the data analysis result based on the determined abnormal value.

[0283] In a specific implementation, the above data processing device can execute the above-mentioned Figure 2 For the implementation methods provided in each step, please refer to the implementation methods provided in the above steps for details, which will not be repeated here.

[0284] See also Fig.10 , Fig.10 Schematic diagram of the structure of the electronic device provided in the embodiment of the present application. Fig.10As shown, the electronic device 1000 in this embodiment may include: a processor 1001, a network interface 1004 and a memory 1005. In addition, the above-mentioned electronic device 1000 may also include: an object interface 1003, and at least one communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. Among them, the object interface 1003 may include a display screen (Display), a keyboard (Keyboard), and the object interface 1003 may optionally include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or it may be a non-volatile memory (NVM), such as at least one disk storage. The memory 1005 may optionally also be at least one storage device located away from the aforementioned processor 1001. As Fig.10 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, an object interface module, and a device control application program.

[0285] exist Fig.10 In the electronic device 1000 shown, the network interface 1004 can provide a network communication function; the object interface 1003 is mainly used to provide an input interface for the object; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0286] In response to obtaining the statistical data input by the target object, performing field recognition processing on the statistical data to obtain a field recognition result, wherein the field recognition result includes the recognized statistical indicators and statistical dimensions;

[0287] Determine data description information according to the statistical data and the field identification result, where the data description information is used to describe data characteristics of the statistical data;

[0288] Inputting the field recognition results into a large language model to obtain a recommendation question, wherein the recommendation question is used to guide the target object to ask a query about the specified data features of the statistical data;

[0289] Display the above data description information and the above recommended questions.

[0290] In some feasible implementations, the data description information includes at least one of the following:

[0291] Data summary information, the data summary information is used to describe the field characteristics of the field recognition result;

[0292] Dimensional statistical features, the above-mentioned dimensional statistical features include data features of each statistical value under each of the above-mentioned statistical dimensions;

[0293] Indicator statistical characteristics, the above-mentioned indicator statistical characteristics include data characteristics of each statistical value corresponding to each of the above-mentioned statistical indicators;

[0294] Supplementary feature information, the above-mentioned supplementary feature information is used to describe other data features besides the above-mentioned data summary information, the above-mentioned dimension statistical features and the above-mentioned indicator statistical features.

[0295] In some feasible implementations, the field identification result further includes a field type of each of the statistical dimensions; for each of the statistical dimensions, according to the statistical data and the field identification result, the processor 1001 is configured to:

[0296] Determine a characteristic statistical function that matches the field type corresponding to the statistical dimension;

[0297] By using the characteristic statistical function, characteristic statistical processing is performed on the statistical values ​​of each of the above statistical indicators under the statistical dimension to obtain data characteristics of the statistical values ​​of each of the above statistical indicators under the statistical dimension.

[0298] In some feasible implementations, the processor 1001 is configured to:

[0299] Inputting the data summary information, the dimension statistical features, the indicator statistical features, the statistical data, the field recognition result and the first prompt information into the large language model to obtain the supplementary feature information;

[0300] The first prompt information is used to prompt the large language model to generate the supplementary feature information according to the input information.

[0301] In some feasible implementations, the processor 1001 is configured to:

[0302] Determine second prompt information, input the second prompt information into the large language model to obtain a recommended question, wherein the second prompt information includes the field recognition result, at least one data analysis method, and an execution condition of each of the data analysis methods; the second prompt information is used to prompt the large language model to perform the following operations:

[0303] Determine at least one executable data analysis method capable of performing data analysis on the statistical data according to the execution conditions of each of the data analysis methods and the field identification results;

[0304] According to the above-mentioned field recognition results, a recommended question corresponding to each of the above-mentioned executable data analysis methods is generated, and each of the recommended questions corresponding to the above-mentioned executable data analysis methods meets the execution conditions of the corresponding executable data analysis method.

[0305] In some feasible implementations, the processor 1001 is further configured to:

[0306] After displaying the data description information and the recommended questions, in response to obtaining a first query question input by the target object, determining whether the first query question belongs to the recommended question, wherein the first query question is used to query a specified data feature of the statistical data;

[0307] In response to the first query question belonging to the recommended question, determining the executable data analysis method corresponding to the first query question as a target data analysis method matching the first query question; analyzing the statistical data according to the first query question and the determined target data analysis method, and obtaining and displaying the data analysis result corresponding to the first query question;

[0308] In response to the first query question not belonging to the recommended question, the first query question is input into the large language model to perform intent recognition on the first query question through the large language model; the statistical data is analyzed according to the intent recognition result to obtain and display the data analysis result corresponding to the first query question.

[0309] In some feasible implementations, the processor 1001 is configured to:

[0310] In response to the large language model identifying the query intent of the first query question, determining a target data analysis method that matches the first query question according to the query intent of the first query question; analyzing the statistical data according to the first query question and the determined target data analysis method, and obtaining and displaying a data analysis result corresponding to the first query question;

[0311] In response to the large language model failing to recognize the query intent of the first query question, the statistical data and the first query question are input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the first query question to obtain a data analysis result corresponding to the first query question; and the data analysis result corresponding to the first query question is obtained and displayed.

[0312] In some feasible implementations, the processor 1001 is configured to:

[0313] Determine a candidate data analysis method whose data analysis function is associated with the query intent of the first query question;

[0314] Determine key information of the query intent of the above-mentioned first query question, and in response to the execution conditions of the above-mentioned candidate data analysis method matching the above-mentioned key information, determine the above-mentioned candidate data analysis method as the target data analysis method matching the above-mentioned first query question.

[0315] In some feasible implementations, the processor 1001 is further configured to:

[0316] In response to obtaining the statistical data, a second query question input by the target object is also obtained, the second query question is used to query a specified data feature of the statistical data;

[0317] Inputting the second query question into the large language model to perform intent recognition on the second query question through the large language model;

[0318] In response to the large language model identifying the query intent of the second query question, determining a target data analysis method that matches the second query question according to the query intent of the second query question; analyzing the statistical data according to the second query question and the determined target data analysis method, and obtaining and displaying a data analysis result corresponding to the second query question;

[0319] In response to the large language model failing to recognize the query intent of the second query question, the statistical data and the second query question are input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the second query question to obtain a data analysis result corresponding to the second query question; and the data analysis result corresponding to the second query question is obtained and displayed.

[0320] In some feasible implementations, the first query question is used to query the reasons for the changes in the statistical values ​​of the statistical data;

[0321] The processor 1001 is used for:

[0322] Determine the comparison dimension corresponding to each of the above statistical dimensions, and the statistical value of each of the above statistical indicators under each of the above comparison dimensions;

[0323] For each of the above statistical indicators, determine the surprise degree of the statistical indicator according to the statistical value of the statistical indicator in each of the above statistical dimensions and the corresponding comparison dimensions, where the surprise degree is used to characterize the degree of change of the statistical value of the statistical indicator;

[0324] Determine multiple indicator combinations, each of which includes one of the above statistical indicators and one of the above statistical dimensions; for each of the statistical indicators and statistical dimensions in the above indicator combination, determine the contribution corresponding to the indicator combination according to the statistical value of the statistical indicator under the statistical dimension and the corresponding comparative statistical dimension, and the sum of the statistical values ​​of each of the above statistical indicators under the statistical dimension and the corresponding comparative statistical dimension, wherein the contribution is used to characterize the contribution of the influencing factor corresponding to the statistical indicator to the change of the statistical value under the statistical dimension;

[0325] Based on at least one of the above-mentioned surprise degree or the above-mentioned contribution degree, the reasons for the changes in the statistical values ​​of the above-mentioned statistical data are determined, and the data analysis results are determined based on the above-mentioned reasons for the changes.

[0326] In some feasible implementations, the statistical dimension includes a time dimension, and the first query question is used to query the future change trend of the statistical value of the target statistical indicator;

[0327] The processor 1001 is used for:

[0328] Determine a first data sequence, wherein the first data sequence includes statistical values ​​of the target statistical indicator in each of the time dimensions, and an arrangement order of the statistical values ​​in the first data sequence is consistent with an arrangement order of the corresponding time dimension;

[0329] Determine a preset time interval, determine the changing trend of the statistical value of the above-mentioned target statistical indicator within the predicted time interval according to the above-mentioned first data sequence, and determine the data analysis result according to the above-mentioned changing trend.

[0330] In some feasible implementations, the first query question is used to query abnormal statistical values ​​in the statistical data; the processor 1001 is used to:

[0331] Determine a plurality of indicator combinations, each indicator combination including one of the above statistical indicators and one of the above statistical dimensions;

[0332] For each statistical indicator and statistical dimension in the above indicator combination, determine the predicted value of the statistical indicator under the statistical dimension; in response to the predicted value being outside the statistical value range corresponding to the indicator combination, determine the statistical value corresponding to the indicator combination as an abnormal statistical value; and determine the data analysis result according to the determined abnormal statistical value;

[0333] For each statistical indicator and statistical dimension within the above indicator combination, determine the growth factor of the statistical value corresponding to the indicator combination, and determine the abnormal statistical value in the above statistical data based on the above growth factor. The above growth factor includes at least one of the month-on-month growth rate or the year-on-year growth rate. Determine the data analysis result based on the determined abnormal value.

[0334] It should be understood that in some feasible implementations, the processor 1001 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the type of device.

[0335] In a specific implementation, the electronic device 1000 can execute the above-mentioned functions through its built-in functional modules. Figure 2 For the implementation methods provided in each step, please refer to the implementation methods provided in the above steps for details, which will not be repeated here.

[0336] The present application also provides a computer-readable storage medium, which stores a computer program and is executed by a processor to implement Figure 2 For the methods provided in each step, please refer to the implementation methods provided in the above steps for details, which will not be repeated here.

[0337] The computer-readable storage medium may be an internal storage unit of the data processing device or electronic device provided in any of the aforementioned embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device. The computer-readable storage medium may also include a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc. Further, the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0338] The present invention provides a computer program product, which includes a computer program. The computer program is executed by a processor. Figure 2 The methods provided in each step.

[0339] The terms "first", "second", etc. in the claims, specification and drawings of the present application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or electronic device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or electronic devices. Mentioning "embodiment" in this article means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. Displaying the phrase at various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments. The term "and / or" used in the present application specification and the attached claims refers to any combination of one or more of the items listed in association and all possible combinations, and includes these combinations.

[0340] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Professional and technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0341] The above disclosure is only the preferred embodiment of the present application, which cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A data processing method, characterized in that: The method comprises: In response to obtaining the statistical data input by the target object, performing field recognition processing on the statistical data to obtain a field recognition result, wherein the field recognition result includes the recognized statistical indicators and statistical dimensions; Based on the feature statistical function, data description information is determined according to the statistical data and the field identification result, and the data description information is used to describe the data characteristics of the statistical data; wherein the data description information includes: data summary information, the data summary information is used to describe the field characteristics of the field identification result; dimension statistical characteristics, the dimension statistical characteristics include the data characteristics of each statistical value under each of the statistical dimensions; indicator statistical characteristics, the indicator statistical characteristics include the data characteristics of each statistical value corresponding to each of the statistical indicators; supplementary feature information, the supplementary feature information is used to describe other data characteristics except the data summary information, the dimension statistical characteristics and the indicator statistical characteristics; Inputting the field recognition result into a large language model to obtain a recommended question, wherein the recommended question is used to guide the target object to ask a query about a specified data feature of the statistical data; The data description information and the recommended questions are displayed.

2. The method according to claim 1, characterized in that The field identification result also includes the field type of each of the statistical dimensions; for each of the statistical dimensions, according to the statistical data and the field identification result, determining the data characteristics of the statistical values ​​of each of the statistical indicators under the statistical dimension, including: Determine a characteristic statistical function that matches the field type corresponding to the statistical dimension; By using the characteristic statistical function, characteristic statistical processing is performed on the statistical values ​​of each of the statistical indicators in the statistical dimension to obtain data characteristics of the statistical values ​​of each of the statistical indicators in the statistical dimension.

3. The method according to claim 1 or 2, characterized in that: The supplementary characteristic information is determined based on the following method: Inputting the data summary information, the dimension statistical features, the indicator statistical features, the statistical data, the field recognition result and the first prompt information into the large language model to obtain the supplementary feature information; The first prompt information is used to prompt the large language model to generate the supplementary feature information according to the input information.

4. The method according to claim 1, characterized in that: The step of inputting the field recognition result into a large language model to obtain a recommendation question includes: Determine second prompt information, input the second prompt information into the large language model to obtain a recommendation question, wherein the second prompt information includes the field recognition result, at least one data analysis method, and an execution condition of each of the data analysis methods; the second prompt information is used to prompt the large language model to perform the following operations: Determining at least one executable data analysis method capable of performing data analysis on the statistical data according to the execution condition of each of the data analysis methods and the field identification result; According to the field recognition result, a recommended question corresponding to each of the executable data analysis methods is generated, and the recommended question corresponding to each of the executable data analysis methods meets the execution conditions of the corresponding executable data analysis method.

5. The method according to claim 4, characterized in that The method further comprises: After displaying the data description information and the recommended questions, in response to obtaining a first query question input by the target object, determining whether the first query question belongs to the recommended question, wherein the first query question is used to query a specified data feature of the statistical data; In response to the first query question belonging to the recommended question, determining the executable data analysis method corresponding to the first query question as a target data analysis method matching the first query question; analyzing the statistical data according to the first query question and the determined target data analysis method, and obtaining and displaying the data analysis result corresponding to the first query question; In response to the first query question not belonging to the recommended question, the first query question is input into the large language model to perform intent recognition on the first query question through the large language model; the statistical data is analyzed according to the intent recognition result to obtain and display the data analysis result corresponding to the first query question.

6. The method according to claim 5, characterized in that The analyzing the statistical data according to the intention recognition result to obtain and display the data analysis result corresponding to the first query question includes: In response to the large language model identifying the query intent of the first query question, determining a target data analysis method that matches the first query question according to the query intent of the first query question; analyzing the statistical data according to the first query question and the determined target data analysis method, and obtaining and displaying a data analysis result corresponding to the first query question; In response to the large language model failing to recognize the query intent of the first query question, the statistical data and the first query question are input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the first query question to obtain a data analysis result corresponding to the first query question; and the data analysis result corresponding to the first query question is obtained and displayed.

7. The method according to claim 6, characterized in that The method for analyzing target data that matches the first query question according to the query intent of the first query question includes: Determining candidate data analysis methods whose data analysis functions are associated with the query intent of the first query question; Determine key information of the query intent of the first query question, and in response to an execution condition of the candidate data analysis method matching the key information, determine the candidate data analysis method as a target data analysis method matching the first query question.

8. The method according to claim 1, characterized in that The method further comprises: In response to obtaining the statistical data, a second query question input by the target object is also obtained, the second query question is used to query a specified data feature of the statistical data; Inputting the second query question into the large language model to perform intent recognition on the second query question through the large language model; In response to the large language model identifying the query intent of the second query question, determining a target data analysis method that matches the second query question according to the query intent of the second query question; analyzing the statistical data according to the second query question and the determined target data analysis method, and obtaining and displaying a data analysis result corresponding to the second query question; In response to the large language model failing to recognize the query intent of the second query question, the statistical data and the second query question are input into the large language model, so that the large language model performs data analysis on the statistical data in a preset manner according to the second query question to obtain a data analysis result corresponding to the second query question; and the data analysis result corresponding to the second query question is obtained and displayed.

9. The method according to any one of claims 5 to 7, characterized in that: The first query question is used to query the reasons for the changes in the various statistical values ​​of the statistical data; Determine the results of data analysis, including: Determine a comparison dimension corresponding to each of the statistical dimensions, and a statistical value of each of the statistical indicators under each of the comparison dimensions; For each of the statistical indicators, determine the surprise degree of the statistical indicator according to the statistical value of the statistical indicator in each of the statistical dimensions and the corresponding comparison dimension, where the surprise degree is used to characterize the degree of change of the statistical value of the statistical indicator; Determine a plurality of indicator combinations, each of which includes a statistical indicator and a statistical dimension; for each statistical indicator and statistical dimension in the indicator combination, determine a contribution degree corresponding to the indicator combination according to the statistical value of the statistical indicator under the statistical dimension and the corresponding comparative statistical dimension, and the sum of the statistical values ​​of each statistical indicator under the statistical dimension and the corresponding comparative statistical dimension, wherein the contribution degree is used to characterize the contribution degree of the influencing factor corresponding to the statistical indicator to the change of the statistical value under the statistical dimension; According to at least one of the degree of surprise or the degree of contribution, reasons for changes in various statistical values ​​of the statistical data are determined, and data analysis results are determined according to the reasons for the changes in various statistical values.

10. The method according to any one of claims 5 to 7, characterized in that: The statistical dimension includes a time dimension, and the first query question is used to query the future change trend of the statistical value of the target statistical indicator; Determine the results of data analysis, including: Determine a first data sequence, wherein the first data sequence includes statistical values ​​of the target statistical indicator in each of the time dimensions, and an arrangement order of the statistical values ​​in the first data sequence is consistent with an arrangement order of the corresponding time dimension; Determine a preset time interval, determine a change trend of the statistical value of the target statistical indicator within the preset time interval according to the first data sequence, and determine a data analysis result according to the change trend.

11. The method according to any one of claims 5 to 7, characterized in that: The first query question is used to query abnormal statistical values ​​in the statistical data; the method further includes: Determine a plurality of indicator combinations, each indicator combination comprising one of the statistical indicators and one of the statistical dimensions; Determine the results of the data analysis, including at least one of the following: For each statistical indicator and statistical dimension in the indicator combination, determine the predicted value of the statistical indicator under the statistical dimension; in response to the predicted value being outside the statistical value range corresponding to the indicator combination, determine the statistical value corresponding to the indicator combination as an abnormal statistical value; and determine the data analysis result according to the determined abnormal statistical value; For each statistical indicator and statistical dimension within the indicator combination, determine the growth factor of the statistical value corresponding to the indicator combination, and determine the abnormal statistical value in the statistical data based on the growth factor. The growth factor includes at least one of the month-on-month growth rate or the year-on-year growth rate. Determine the data analysis result based on the determined abnormal statistical value.

12. A data processing device, characterized in that: The device comprises: A field recognition module is used for, in response to obtaining statistical data input by a target object, performing field recognition processing on the statistical data to obtain a field recognition result, wherein the field recognition result includes the recognized statistical indicators and statistical dimensions; A data analysis module, for determining data description information according to the statistical data and the field identification result based on a feature statistical function, wherein the data description information is used to describe the data characteristics of the statistical data; wherein the data description information includes: data summary information, which is used to describe the field characteristics of the field identification result; dimension statistical characteristics, which include data characteristics of each statistical value under each statistical dimension; indicator statistical characteristics, which include data characteristics of each statistical value corresponding to each statistical indicator; supplementary characteristic information, which is used to describe other data characteristics except the data summary information, the dimension statistical characteristics and the indicator statistical characteristics; The data analysis module is used to input the field recognition result into a large language model to obtain a recommended question, and the recommended question is used to guide the target object to query and ask questions about the specified data features of the statistical data; The information display module is used to display the data description information and the recommended questions.

13. An electronic device, characterized in that: comprising a processor and a memory, wherein the processor and the memory are connected to each other; The memory is used to store computer programs; The processor is configured to execute the method according to any one of claims 1 to 11 when calling the computer program.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method according to any one of claims 1 to 11.

15. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Citation Information

Patent Citations

  • Data processing method, device and equipment and computer readable storage medium

    CN116821103A

  • Method and system for solving alignment of complex table and entity by using LLM-driven NL2SQL converter

    CN118410059A

  • Method and system for processing electronic table data based on time sequence

    US20090019350A1