JP7369320B1 - Information processing device, method, program, and system - Google Patents

Abstract

The present invention provides a technology for obtaining appropriate data analysis results at low cost and in a short time, regardless of the level of a user's data analysis skills.
[Solution] A program according to one aspect of the present disclosure provides a means for controlling a computer to manage the progress status of a chat regarding data analysis performed with a user, a means for identifying the purpose of the user's analysis through the progress of the chat, and an analysis method. According to the purpose of , it functions as a means to analyze the target data set and a means to output the results of the analysis of the target data set.
[Selection diagram] Figure 4

Description

The present disclosure relates to an information processing device, method, program, and system.

Conventionally, various BI (Business Intelligence) tools have been known that allow companies to collect, accumulate, analyze, and process the enormous amount of data they have acquired, and to support decision-making for business strategies.

Additionally, Patent Document 1 describes a technology that aims to provide high-quality "managed model self-service BI" in a shorter period of time without relying on individual design skills using a standardized design framework. ideas are disclosed.

Japanese Patent Application Publication No. 2019-175236

However, in many cases, it is not easy for people who do not have advanced data analysis skills, such as those in charge of a company's planning department, to make full use of BI tools on their own, so they need assistance from data analysis experts. There were many cases where this was necessary. Obtaining expert support requires financial costs and the time required to obtain analysis results is not short. In addition, the technical idea of Patent Document 1 also requires input of necessary items into the requirements definition template and detailed analysis model definition document, so it can be understood that it is assumed that design and analysis skills of a certain level or higher are present. be done.

An object of the present disclosure is to provide a technique for obtaining appropriate data analysis results at low cost and in a short time, regardless of the level of data analysis skill of the user.

A program according to an aspect of the present disclosure provides a means for controlling a computer to manage the progress of a chat regarding data analysis performed with a user, a means for identifying a user's purpose of analysis through the progress of the chat, and a means for controlling a computer according to the purpose of analysis. It functions as a means to analyze the target data set and a means to output the results of the analysis of the target data set.

FIG. 1 is a block diagram showing the configuration of an information processing system according to the present embodiment. FIG. 2 is a block diagram showing the configuration of a client device according to the present embodiment. FIG. 2 is a block diagram showing the configuration of a server according to the present embodiment. FIG. 2 is an explanatory diagram of one aspect of the present embodiment. FIG. 3 is a diagram showing the overall flow of information processing according to the present embodiment. 7 is a flowchart of analysis purpose setting processing according to the present embodiment. FIG. 6 is a diagram illustrating an example of a screen displayed during analysis purpose setting processing according to the present embodiment. FIG. 6 is a diagram illustrating an example of a screen displayed during analysis purpose setting processing according to the present embodiment. FIG. 6 is a diagram illustrating an example of a screen displayed during analysis purpose setting processing according to the present embodiment. It is a flowchart of analysis processing of this embodiment. It is a figure which shows the example of a screen displayed in the analysis process of this embodiment. It is a figure which shows the example of a screen displayed in the analysis process of this embodiment.

Hereinafter, one embodiment of the present invention will be described in detail based on the drawings. In addition, in the drawings for explaining the embodiments, the same components are generally designated by the same reference numerals, and repeated explanations thereof will be omitted.

(1) Configuration of information processing system The configuration of the information processing system will be explained. FIG. 1 is a block diagram showing the configuration of an information processing system according to this embodiment.

As shown in FIG. 1, the information processing system 1 includes a client device 10 and a server 30.
The client device 10 and the server 30 are connected via a network (eg, the Internet or an intranet) NW.

The client device 10 is an example of an information processing device that sends a request to the server 30. The client device 10 is, for example, a smartphone, a tablet terminal, or a personal computer. The user of the client device 10 is assumed to be a person who has a need for data analysis but often does not have the advanced skills to perform data analysis, such as a person in a corporate planning department. However, the information processing system 1 of this embodiment is also useful when the user of the client device 10 is a highly skilled data analysis expert.

The server 30 is an example of an information processing device that provides the client device 10 with a response in response to a request transmitted from the client device 10. Server 30 is, for example, a web server.

(1-1) Configuration of client device The configuration of the client device will be explained. FIG. 2 is a block diagram showing the configuration of the client device of this embodiment.

As shown in FIG. 2, the client device 10 includes a storage device 11, a processor 12, an input/output interface 13, and a communication interface 14. Client device 10 is connected to display 21 .

The storage device 11 is configured to store programs and data. The storage device 11 is, for example, a combination of ROM (Read Only Memory), RAM (Random Access Memory), and storage (for example, flash memory or hard disk).

The programs include, for example, the following programs.
・OS (Operating System) programs ・Application programs that execute information processing (e.g., web browsers, data analysis applications, etc.)

The data includes, for example, the following data.
・Databases referenced in information processing ・Data obtained by executing information processing (that is, execution results of information processing)

The processor 12 is a computer that implements the functions of the client device 10 by activating a program stored in the storage device 11. The processor 12 is, for example, at least one of the following.
・CPU (Central Processing Unit)
・GPU (Graphic Processing Unit)
・ASIC (Application Specific Integrated Circuit)
・FPGA (Field Programmable Gate Array)

The input/output interface 13 is configured to acquire information (for example, user instructions) from an input device connected to the client device 10 and output information (for example, an image) to an output device connected to the client device 10. be done.

The input device is, for example, a keyboard, pointing device, touch panel, or a combination thereof.
The output device is, for example, a display 21, a speaker, or a combination thereof.

Communication interface 14 is configured to control communication between client device 10 and an external device (eg, server 30).

The display 21 is configured to display images (still images or moving images). The display 21 is, for example, a liquid crystal display or an organic EL display.

(1-2) Server configuration The server configuration will be explained. FIG. 3 is a block diagram showing the configuration of the server of this embodiment.

As shown in FIG. 3, the server 30 includes a storage device 31, a processor 32, an input/output interface 33, and a communication interface 34.

The storage device 31 is configured to store programs and data. The storage device 31 is, for example, a combination of ROM, RAM, and storage (eg, flash memory or hard disk).

The programs include, for example, the following programs.
・OS program ・Application program that executes information processing

The data includes, for example, the following data.
・Databases referenced in information processing ・Execution results of information processing

The processor 32 is a computer that implements the functions of the server 30 by activating a program stored in the storage device 31. The processor 32 is, for example, at least one of the following.
・CPU
・GPU
・ASIC
・FPGA

The input/output interface 33 is configured to obtain information (e.g., user instructions) from an input device connected to the server 30 and output information (e.g., an image) to an output device connected to the server 30. .

The input device is, for example, a keyboard, pointing device, touch panel, or a combination thereof.
The output device is, for example, a display.

Communication interface 34 is configured to control communication between server 30 and an external device (eg, client device 10). Note that the server 30 is configured to access external systems in order to use some functions, such as natural language models such as large-scale language models (hereinafter simply referred to as "natural language models") or data analysis tools. may be configured.

(2) One aspect of the embodiment One aspect of the present embodiment will be described. FIG. 4 is an explanatory diagram of one aspect of this embodiment.

As shown in FIG. 4, the information processing system 1 advances the data analysis required by the user US1 mainly on a chat basis between the user US1 and the data analysis agent AG2, and finally performs the data analysis required by the user US1. The results are provided to the user US1.

Specifically, the server 30 manages the progress of a chat regarding data analysis between the user US1 and the data analysis agent AG2. Managing the progress of a chat may include understanding the intent of a user instruction, generating a system response based on the intent of the user instruction and a data analysis scenario, outputting the system response, and the like. In managing the progress of the chat, the server 30 can use a natural language model as appropriate. On the other hand, the client device 10 receives a user instruction from the user US1 and outputs it to the server 30, and displays a UI (User Interface) to the user US1 based on a system response received from the server 30.

The server 30 identifies (verbalizes) the purpose of data analysis of the user US1 through the progress of the chat. The verbalized purpose is stored in the storage device 31 and referenced by the server 30 as a guide for subsequent analysis. Further, the server 30 receives a data set from the user US1 via the client device 10. The server 30 analyzes a target data set based on the received data set, and outputs a system response including the result of the analysis to the client device 10. The client device 10 displays a UI showing the results of the analysis to the user US1.

As a result, the user US1 only needs to respond to questions from the data analysis agent AG2 or perform actions requested by the data analysis agent AG2 (for example, specifying a dataset). (including potentially needed analysis results) can be obtained. In particular, before starting the chat, user US1 is looking for solutions such as what purpose the analysis should be carried out, what kind of analysis method should be chosen, and what kind of data set should be prepared. These elements can be determined through chat and the necessary analysis can be performed without having to prepare them in advance.

(3) Information processing Information processing of this embodiment will be explained. FIG. 5 is a diagram showing the overall flow of information processing in this embodiment.

The information processing of this embodiment starts in response to establishment of a predetermined start condition. The starting condition may be, for example, that the user starts a new analysis project.

As shown in FIG. 5, the server 30 executes data set acquisition (S100).
Specifically, the server 30 obtains a system response (for example, a chat message) requesting the user to provide a data set as analysis material, and transmits it to the client device 10 .
The client device 10 displays a UI based on the received system response on the display 21. The client device 10 receives a data set designation from a user and acquires the designated data set. The client device 10 transmits the acquired data set to the server 30.

After step S100, the server 30 executes setting of the purpose of analysis (S110).
Specifically, the server 30 starts a chat to identify the user's purpose of analysis. The server 30 identifies (verbalizes) the user's purpose of analysis based on the exchange of messages between the user and the data analysis agent. The server 30 stores the identified purpose of analysis in the storage device 31 and transmits it to the client device 10. The client device 10 provides the user with a UI that allows the user to refer to the purpose of the received analysis.
Note that details of setting the purpose of analysis (S110) will be described later.

After step S110, the server 30 performs analysis (S120).
Specifically, the server 30 analyzes the data set acquired in step S100 or the data set obtained by processing the data set (for example, preprocessing such as data cleansing) according to the purpose of analysis set in step S110. The server 30 transmits the analysis results to the client device 10. The client device 10 displays a UI on the display 21 that shows the received analysis results.
Note that details of the analysis (S120) will be described later.

(3-1) Analysis purpose setting process The analysis purpose setting process of this embodiment will be explained. FIG. 6 is a flowchart of analysis purpose setting processing according to this embodiment. FIG. 7 is a diagram showing an example of a screen displayed in the analysis purpose setting process of this embodiment. FIG. 8 is a diagram showing an example of a screen displayed in the analysis purpose setting process of this embodiment. FIG. 9 is a diagram showing an example of a screen displayed in the analysis purpose setting process of this embodiment.

As shown in FIG. 6, the server 30 executes system response generation (S111).
Specifically, the server 30 generates a system response (message) that asks the user what to analyze, and sends it to the client device 10.

As a first example of system response generation (S111), the server 30 identifies (extracts) candidates for analysis based on the data item names (variable names) included in the data set acquired in step S100, and A system response may be generated that includes questions about the candidates. The question may ask the user which of the candidates is the target of analysis, or may ask the user whether each candidate is the target of analysis.

In the following description, unless otherwise specified, the server 30 may generate a system response using one or both of a predetermined template and a natural language model. For example, the server 30 may generate the system response by inserting text generated by a natural language model into a template.

As a second example of system response generation (S111), the server 30 may generate a standard message as the system response, regardless of the data item name included in the data set acquired in step S100.

After step S111, the client device 10 executes screen display (S211).
Specifically, the client device 10 receives the system response sent from the server 30 in step S111. The client device 10 displays a screen based on the received system response on the display 21. An example of such a screen is shown in FIG.

The screen in FIG. 7 includes objects J20 to J27.
Object J20 displays which phase is currently in among a plurality of phases that define the progress state of the chat (that is, the progress state of the analysis project). In the example of FIG. 7, object J20 indicates that the chat progress status is "Setting analysis purpose."

Object J21 displays the main UI. Object J21 displays the chat between the user and the data analysis agent, the results of analysis, etc. (in other words, system responses and user instructions). Object J21 is configured to be able to display at least the latest system response or user instruction. Furthermore, the object J21 may be configured to be able to display past system responses or user instructions, for example, in response to a scroll operation.

In the example of FIG. 7, object J21 displays the system response "What do you want to analyze?" Furthermore, in the example of FIG. 7, objects J210a to J210b are arranged in object J21.

Object J210a accepts a user instruction to select a candidate for analysis. In the example of FIG. 7, the object J210a corresponds to a drop-down list, but it can be replaced with any selection-based GUI component such as a list box, check box, or radio button.

Object J210b accepts a user instruction to confirm the selection of object J210a. When object J210b is selected, client device 10 transmits information that can identify the candidate selected by object J210a to server 30.

Object J22 accepts user instructions. The client device 10 transmits the user instruction received by the object J22 to the server 30. In the example of FIG. 7, by using the object J22, the user can directly specify the analysis target without selecting the object J210a.

Object J23 displays a sub-UI. The object J23 is configured to be able to display information related to the main points of the analysis project, such as the dataset used for analysis, the purpose of analysis (text), the results of analysis, and the history of chats. Objects J24 to J27 are arranged in object J23.

Object J24 accepts a user instruction to switch between displaying and non-displaying information regarding the data set used for analysis. When the object J24 is selected, the client device 10 expands and displays information regarding the data set used for analysis on the object J23, or closes the information that has already been expanded and displayed. Information regarding the data set used for analysis can include, for example, at least one of the following information.
・Name of the file where the dataset is stored ・Name of the data items (variables) included in the dataset (e.g. sales, date and time, etc.)
・Types of data items (variables) included in the dataset (e.g., integer type, date type, etc.)
・Examples of values of data items (variables) included in the dataset

Object J25 accepts a user instruction to switch between displaying and non-displaying the purpose of analysis. When the object J25 is selected, the client device 10 expands and displays the purpose of analysis on the object J23, or closes the expanded and displayed information. However, if the purpose of analysis has not been set, the client device 10 can use a fixed text or image such as "Undetermined" instead of the purpose of analysis. Further, as an option, when the object J25 is selected, in addition to the purpose of analysis, an object that accepts a user instruction to edit the purpose may be expanded and displayed.

Object J26 accepts a user instruction to switch between displaying and hiding the results of analyzes that have already been performed in the current analysis project. When the object J26 is selected, the client device 10 expands and displays the analysis results on the object J23, or closes the expanded and displayed information. However, if no analysis has been performed, the client device 10 can use a fixed text or image, such as "none", instead of the analysis result. Further, as an option, when object J26 is selected, an object that accepts user instructions for editing the visual representation of the results (eg, graph format, etc.) in addition to the analysis results may be expanded and displayed.

Object J27 accepts a user instruction to switch between displaying and non-displaying the chat history. When object J27 is selected, client device 10 expands and displays the chat history on object J23, or closes the expanded and displayed information.

After step S211, the client device 10 accepts a user instruction (S212).
Specifically, the client device 10 receives a user instruction to select or specify an analysis target from the user. In the example of FIG. 7, the user can select or specify an analysis target by operating objects J210a to J210b or object J22. The client device 10 transmits the accepted user instruction to the server 30.

After step S212, the server 30 identifies the analysis target (S112).
Specifically, the server 30 receives the user instruction sent from the client device 10 in step S212. The server 30 identifies the analysis target based on the received user instruction. Here, for example, if a user instruction to specify the analysis target in text is obtained, the server 30 may specify the analysis target by providing model input based on the text to the natural language model. Optionally, the server 30 may re-execute steps S111-S112 if sufficient information is not obtained from the user instructions to identify the analysis target.

After step S111, the server 30 generates a system response (S113).
Specifically, the server 30 generates a system response (message) that asks the user what kind of analysis they would like to perform on the analysis target (that is, the direction of the analysis), and transmits it to the client device 10 . For example, the server 30 inserts a text (which may include a text generated by a natural language model based on the text) corresponding to the analysis target identified in step S112 into a predefined message template. May generate a response. Further, the system response may include information indicating possible directions of analysis for the analysis target. Information indicating a candidate direction of analysis may include text that verbalizes the direction, an image (eg, a graph) that helps understand the concept of the direction, or a combination thereof. Candidates for the direction of analysis may be defined in advance according to, for example, the type of data item to be analyzed, the type of data item other than the target to be analyzed, other elements, or a combination thereof.

After step S113, the client device 10 executes screen display (S213).
Specifically, the client device 10 receives the system response sent from the server 30 in step S113. The client device 10 displays a screen based on the received system response on the display 21. An example of such a screen is shown in FIG.

The screen in FIG. 8 includes objects J20 to J27.
In the example of FIG. 8, object J21 says, ``The analysis target is ``sales.'' Next, choose what kind of analysis you would like to do. If none of the options apply to you, please enter them in the chat. ” is displayed as the system response. Furthermore, in the example of FIG. 8, objects J211a to J211b are arranged in object J21.

Object J211b accepts a user instruction to select a candidate for the direction of analysis. In the example of FIG. 8, the object J211a corresponds to a radio button, but it can be replaced with any selection-based GUI component such as a drop-down list, a list box, or a radio button.

Object J211b accepts a user instruction to confirm the selection of object J211a. When the object J211b is selected, the client device 10 transmits information that can identify the candidate selected by the object J211a to the server 30.

In the example of FIG. 8, by using the object J22, the user can directly specify the direction of the analysis without depending on the selection of the object J211a.

After step S213, the client device 10 accepts a user instruction (S214).
Specifically, the client device 10 receives a user instruction for selecting or specifying the direction of analysis from the user. In the example of FIG. 8, the user can select or specify the direction of analysis by operating objects J211a to J211b or object J22. The client device 10 transmits the accepted user instruction to the server 30.

After step S214, the server 30 identifies the purpose of analysis (S114).
Specifically, the server 30 receives the user instruction sent from the client device 10 in step S214. The server 30 specifies the purpose of the analysis (that is, the combination of the analysis target and the direction of the analysis) based on the received user instruction. That is, the server 30 obtains text indicating the purpose of the analysis. Here, for example, when a user instruction specifying the direction of analysis in text is obtained, the server 30 provides model input based on the text (which may further include text indicating the analysis target) to the natural language model. You may also specify the purpose of the analysis. Optionally, server 30 may re-perform steps S113-S114 if sufficient information is not obtained from the user instructions to determine the purpose of the analysis.

After step S114, the server 30 generates a system response (S115).
Specifically, the server 30 generates a system response that includes the purpose of the analysis specified (verbalized) in step S114, and transmits it to the client device 10. For example, the server 30 may generate the system response by inserting text corresponding to the purpose of the analysis identified in step S114 into a predefined message template.

After step S115, the client device 10 executes screen display (S215).
Specifically, the client device 10 receives the system response sent from the server 30 in step S115. The client device 10 displays a screen based on the received system response on the display 21. An example of such a screen is shown in FIG.

The screen in FIG. 9 includes objects J20 to J27.
In the example of FIG. 9, object J212 is placed in object J21.

Object J212 displays a system response that informs the user of the purpose of the analysis identified in step S114. Thereafter, the purpose of this analysis can be expanded and displayed on the object J23 via the object J25, if necessary.

(3-2) Analysis Processing The analysis processing of this embodiment will be explained. FIG. 10 is a flowchart of the analysis process of this embodiment. FIG. 11 is a diagram showing an example of a screen displayed in the analysis process of this embodiment. FIG. 12 is a diagram showing an example of a screen displayed in the analysis process of this embodiment.

As shown in FIG. 10, the server 30 executes prediction using a suitable analysis method (S121).
Specifically, the server 30 predicts an analysis method suitable for the analysis purpose specified in step S110, based on the user input on the chat and the system response.

As an example, the server 30 may provide model input based on information obtained through user input to the natural language model, and cause the natural language model to infer an analysis method suitable for the purpose of the analysis. The analysis technique may be, for example, regression (which may include future prediction), classification, clustering, dimensionality reduction (eg, principal component analysis), or a combination thereof on the object of analysis. Further, specifying the analysis method may include specifying data items to be used as explanatory variables.

As another example, an analysis method suitable for the purpose of analysis may be specified, and usable methods may be narrowed down from the analysis method. An analysis technique suitable for the purpose of analysis may be defined in advance, for example. Furthermore, whether or not each analysis method is usable is determined based on the name or type of data item included in the dataset acquired in step S100, the number of data samples included in the dataset, etc. good.

As an option for predicting a suitable analysis method (S121), when there is insufficient judgment material for predicting an analysis method suitable for the purpose of analysis, the server 30 may ask the user a question or prompt a predetermined action. may generate a system response and output it to the chat. As an example, server 30 may generate a system response to inquire about the availability of additional data sets relevant to the analysis, or to prompt the provision of additional data sets.

After step S121, the server 30 generates a system response (S122).
Specifically, the server 30 generates a system response (message) asking whether to adopt the analysis method predicted in step S121. As an example, such a system response may ask the user which of multiple analysis methods to select (multiple selections may be possible), or may ask the user whether to adopt each analysis method. good. For example, the server 30 may generate the system response by inserting text corresponding to the analysis method predicted in step S121 into a predefined message template. In addition, the system response may include, as information indicating possible analysis method candidates, text that verbalizes the method, images (e.g., graphs) that help understand the concept of the method, or a combination thereof. .

After step S122, the client device 10 executes screen display (S221).
Specifically, the client device 10 receives the system response sent from the server 30 in step S122. The client device 10 displays a screen based on the received system response on the display 21.

After step S221, the client device 10 accepts a user instruction (S222).
Specifically, the client device 10 receives a user instruction for selecting or specifying an analysis method from the user. The client device 10 transmits the accepted user instruction to the server 30.

After step S222, the server 30 determines an analysis method (S123).
Specifically, the server 30 receives the user instruction sent from the client device 10 in step S222. The server 30 determines an analysis method based on the received user instruction. Here, for example, when a user instruction specifying an analysis method in text is obtained, the server 30 inputs a model based on the text (which may further include text indicating the analysis target, the direction of analysis, or a combination thereof). The analysis method may be determined by giving the following to the natural language model. Optionally, the server 30 may re-execute steps S121-S123 if sufficient information is not obtained from the user instructions to determine an analysis technique.

After step S123, the server 30 executes a dataset check (S124).
Specifically, the server 30 examines the data set provided by the user using a predetermined algorithm in step S100 or at other timings. The algorithm may be determined differently depending on the analysis method determined in step S123. Inspection of the data set can include, for example, at least one of the following:
・Detection of data deficiencies (e.g., missing values, outliers, inappropriate variations, mismatched granularity, or combinations thereof) ・Determination of whether pre-processing conditions are met

After step S125, the server 30 generates a system response (S125).
Specifically, the server 30 generates a system response based on the result of the test in step S124. The system response can include, for example, at least one of the following:
- A system response that prompts correction of deficiencies in the data set (for example, removal of samples containing missing values or outliers) discovered by the inspection in step S124 - A system response that suggests implementation of a predetermined preprocessing (step S124) (If it is determined that the application conditions for pre-processing are met)
- A system response that searches for another dataset from an external source to compensate for the deficiencies in the dataset discovered by the inspection in step S124, and proposes the use of the dataset discovered from the external source.

The predetermined pretreatment can include, for example, at least one of the following.
・Name identification・Exclusion of missing values・Feature engineering (e.g., variable conversion, exclusion of similar (highly correlated) variables)
Here, the server 30 obtains inference results regarding the types of variables, hypotheses of relationships between variables, etc. from the natural language model, and performs preprocessing (especially feature engineering) based on the inference results. good.

After step S125, the client device 10 executes screen display (S223).
Specifically, the client device 10 receives the system response sent from the server 30 in step S125. The client device 10 displays a screen based on the received system response on the display 21.

After step S223, the client device 10 accepts a user instruction (S224).
Specifically, the client device 10 receives a user instruction corresponding to a response from the user to the screen displayed in step S223. The client device 10 transmits the accepted user instruction to the server 30.

As an example, the client device 10 can accept at least one of the following user instructions.
・User instructions to correct the deficiencies pointed out by the system response (for example, edit the dataset) ・User instructions to provide (upload) a dataset with the deficiencies pointed out by the system response corrected ・User instructions to correct the deficiencies pointed out by the system response ・User instructions to correct the deficiencies pointed out by the system response User instructions to perform or not perform pre-processing

After step S224, the server 30 acquires the target data set (S126).
Specifically, the server 30 receives the user instruction sent from the client device 10 in step S224. The server 30 acquires a data set actually used for analysis (hereinafter referred to as "target data set") in response to the received user instruction.

As a first example of acquiring the target data set (S126), when a user instruction to correct the deficiencies pointed out by the system response is obtained, the server 30 acquires the dataset inspected in step S124 according to the user instruction. Obtain the target dataset by editing.

As a second example of acquiring the target data set (S126), when a user instruction to provide a data set with the deficiencies pointed out by the system response corrected is obtained, the server 30 acquires the data set provided by the user instruction. Obtain as the target dataset. Alternatively, the server 30 may re-execute the data set inspection (S124) on the data set provided by the user's instruction.

As a third example of obtaining the target data set (S126), when a user instruction requesting implementation of the preprocessing proposed by the system response is obtained, the server 30 performs a process on the data set inspected in step S124. The target data set is obtained by performing the preprocessing.

As a fourth example of acquiring the target dataset (S126), when a user instruction requesting non-implementation of the proposed preprocessing is obtained from the system response, the server 30 acquires the dataset inspected in step S124. Obtain it as is as a target dataset.

Note that if the inspection in step S124 determines that no defects are found in the data set and that the data set does not satisfy any preprocessing application conditions, steps S125 to S126 can be omitted. That is, the server 30 can acquire the data set inspected in step S124 as it is as the target data set.

After step S126, the server 30 performs analysis (S127).
Specifically, the server 30 analyzes the target data set acquired in step S126 using the analysis method determined in step S123.

The server 30 may also generate a graph illustrating the results of the analysis. As an example, the server 30 may obtain a graph by providing commands generated by a natural language model to a data analysis tool. The type of graph (for example, line graph, bar graph, scatter diagram, pie graph, bubble chart, radar chart, stacked graph, box plot, etc.) may be determined depending on the purpose of the analysis and the analysis method. The server 30 may determine important locations in the graph depending on the purpose of the analysis and the analysis method.

Furthermore, the server 30 may generate text representing suggestions obtained from the results of the analysis. As an example, the server 30 may obtain suggestions by providing model input based on the results of the analysis to the natural language model. The suggestion can include, for example, at least one of the following information:
- Information that explains the objective variable, explanatory variable, or a combination thereof (that is, variables selected according to the purpose of the analysis) (for example, information summarizing variables such as distribution spread and representative value)
・Information that explains the relationship between the objective variable and explanatory variables (i.e., the results of the analysis) ・Information that explains the relationship between different explanatory variables ・Information that explains the statistical treatment of the results of the analysis (e.g., If the result of this analysis is not statistically significant, a message that the result of this analysis may be due to chance, or if the result of this analysis is statistically significant, a message that the result of this analysis may be due to chance. message that the reliability is low (that is, there is a certain level of reliability)
- Information that explains a statistically important explanatory variable (e.g., the explanatory variable with the lowest significance level, or the explanatory variable with the largest absolute value of the correlation coefficient among the significant explanatory variables) (e.g., the explanatory variable with the highest correlation coefficient) (correlation coefficient value with objective variable)
・Information explaining the outlier (e.g., where the outlier occurs, what value it is, etc.)
・Information that explains important points in the graph ・Information that recommends analysis using other analysis methods

The topics covered by the suggestions may be narrowed down depending on the purpose of the analysis. Specifically, the server 30 may score topics that can be addressed by suggestions using a natural language model, and generate suggestions only for topics whose scores meet the criteria.

Note that the suggestion may be generated, for example, by inserting the analysis results (numeric values) into a template prepared in advance. Preferably, the template is a text that describes the results of the analysis without using statistical or data analysis terminology. This helps users understand and translate suggestions into concrete business actions.

After step S127, the server 30 generates a system response (S128).
Specifically, the server 30 generates a system response based on the result of the analysis generated in step S127. The system response can include, for example, a graph, a portion of the graph determined to be important, a suggestion, or a combination thereof.

After step S125, the client device 10 executes screen display (S225).
Specifically, the client device 10 receives the system response sent from the server 30 in step S127. The client device 10 displays a screen based on the received system response on the display 21. An example of such a screen is shown in FIGS. 11 and 12.

The screen in FIG. 11 includes objects J20 to J27.
In the example of FIG. 11, object J20 indicates that the progress state of the chat is "analysis."

In the example of FIG. 11, objects J213a, J213a1, and J213b are arranged in object J21.

Object J213a displays a line graph showing the relationship between the objective variable "number of sales" and the explanatory variable "date".
The object J213a1 displays information indicating outliers as important points in the graph shown in the object J213a.

Object J213b displays suggestions. Specifically, Object J213b says, ``The number of sales generally fluctuates at about 100 to 400 pieces/day.The number of sales increases to about 600 pieces/day as an outlier during the year-end and New Year holidays and Golden Week. '' includes information regarding a summary of the objective variable and information explaining outliers.

In the example of FIG. 12, objects J214a, J214a1, and J214b are arranged in object J21.

Object J214a displays a boxplot showing the relationship between the objective variable "number of sales" and the explanatory variable "neighborhood event".

Object J214a1 is an important point in the graph shown in object J214a, which shows the value of interest (the number of sales for nearby events "Nationwide") with respect to the baseline (in the example of FIG. Display information showing the difference between the median values of

Object J214b displays suggestions. Specifically, the object J214b tends to have a larger number of sales when there is a nearby event "Nationwide" compared to when there is no nearby event. It includes information that explains the relationship between the objective variable and explanatory variables, such as ``when compared with the median value, the increase in sales is 30 units/day.''

Note that the server 30 may perform the analysis (S127) in stages over multiple phases (for example, aggregation, simple analysis, machine learning, etc.). That is, after step S225, the client device 10 accepts a user instruction regarding the analysis result, and the server 30 performs further analysis (S127), generates a system response (S128), and displays the screen (S225). good. In this case, the server 30 may change at least one of the graph and the suggestion according to the phase transition.

(4) Summary As explained above, the server 30 of this embodiment manages the progress of chat regarding data analysis performed with the user, and identifies the purpose of the user's analysis through the progress of the chat. The server 30 analyzes the target data set according to the purpose of the analysis, and outputs the results of the analysis of the target data set. Thereby, the user can obtain the analysis results that the user needs simply by answering questions given on the chat or taking actions requested on the chat. Therefore, even users who are reluctant to use CUI (Character User Interface) or GUI (Graphical User Interface), which are often found in conventional BI tools, can use the approach they are interested in without experiencing any technical or psychological hurdles. You can freely try out the analysis.

The progress state of a chat may be defined using a plurality of predetermined phases. In each of the plurality of phases, the server 30 may provide predetermined information or output a response to the chat to prompt the user to take a predetermined action, without relying on input from the user. This allows users to find out before starting a chat, such as what the purpose of the analysis should be, what kind of analysis method to choose, or what kind of data set to prepare. These elements can be determined through chat and the necessary analysis can be performed without having to prepare them in advance.

The server 30 may predict an analysis method suitable for the purpose of the analysis based on the user input on the chat and the system response, and output information on the predicted analysis method suitable for the purpose of the analysis. As a result, even if the user does not have sufficient skill to incorporate the purpose of analysis into analysis design, the user can perform analysis in accordance with the intended purpose and obtain the desired results.

When the server 30 lacks judgment materials for predicting an analysis method suitable for the purpose of analysis, the server 30 may output a system response to the chat to ask the user a question or prompt a predetermined action. Thereby, it is possible to appropriately predict the analysis method required by the user.

The server 30 acquires the first data set that is the data set provided by the user, detects deficiencies in the first data set based on the purpose of analysis, and provides the user with the first data according to the detection result. A response that prompts correction of deficiencies in the set may be output to the chat. As a result, if the data set provided by the user has a flaw or an appropriate data set is not provided, and highly accurate analysis cannot be obtained as is, the flaw can be corrected and the data set can be improved. Accurate analysis can be performed.

The server 30 obtains a first data set that is a data set provided by the user, and when the first data set matches a predetermined condition, a response that proposes implementation of preprocessing corresponding to the condition is sent. may be output to chat. This allows the user to select necessary preprocessing even if the user does not have sufficient knowledge regarding preprocessing such as data cleansing. In other words, the number of steps for preprocessing can be reduced.

The server 30 may output information for displaying an area where the chat is displayed and an area where the analysis results of the target data set are displayed on one screen or in a switchable manner. This allows the user to refer to the analysis results and the chat simultaneously or alternately, to consider the next instruction to the system, and to deepen their understanding of the analysis results.

The results of the analysis of the target dataset may include a graph illustrating the results of the analysis of the target dataset and text representing suggestions based on the results of the analysis of the target dataset. This allows the user to check the graph without performing any command operations, and to understand the gist of the analysis results even if the user does not have sufficient statistical knowledge. In other words, it is possible to prevent the user from subjectively interpreting the results of the analysis, interpreting as if something that cannot be shown to be objectively correct is true, or overlooking something that can be shown to be objectively correct.

The suggestion may indicate an explanation of variables selected according to the purpose of the analysis, or an explanation of a relationship between two or more variables. This allows the user to deepen their understanding of the variables (data items) that are the targets of analysis.

The server 30 may determine the type of graph depending on the purpose of the analysis and the analysis method used for the analysis. Thereby, the user can check a graph of a type suitable for checking the analysis results without performing any command operations or the like.

The server 30 may determine important locations in the graph according to the purpose of analysis and the analysis method used in the analysis, and may output information indicating the locations determined to be important in the graph. This allows the user to understand the gist of the graph even if the user does not have sufficient background knowledge to properly interpret the graph.

Analysis of the target data set may be performed in stages over multiple phases. The server 30 may change at least one of the graph and the suggestion according to the transition of the phases. This provides step-by-step results, even for complex analyses, so users can step-by-step through the key points of an analysis, even if they do not have sufficient background knowledge to properly interpret the results. be able to understand.

(5) Other Modifications The storage device 11 may be connected to the client device 10 via the network NW. Display 21 may be integrated with client device 10 . The storage device 31 may be connected to the server 30 via the network NW.

Each step of the above information processing can be executed by either the client device 10 or the server 30. For example, processing described as being performed by one device may be performed by another device, or processing described as being performed by interaction of multiple devices may be performed by a single device.

In the above description, an example was shown in which each step is executed in a specific order in each process, but the execution order of each step is not limited to the example described as long as there is no dependency relationship. For example, in the above explanation, an example was shown in which the purpose of analysis is set after acquiring the data set. However, both may be performed in the reverse order or in parallel.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited to the above embodiments. Moreover, various improvements and changes can be made to the embodiments described above without departing from the spirit of the present invention. Furthermore, the above embodiments and modifications can be combined.

1 : Information processing system 10 : Client device 11 : Storage device 12 : Processor 13 : Input/output interface 14 : Communication interface 21 : Display 30 : Server 31 : Storage device 32 : Processor 33 : Input/output interface 34 : Communication interface

Claims (15)

computer,
A means for managing the progress of a chat regarding data analysis conducted with a user;
means for identifying the purpose of analysis of the user through the progress of the chat;
means for analyzing the target data set according to the purpose of said analysis;
means for outputting the results of the analysis of the target data set;
function as
The means for specifying the purpose of said analysis includes:
means for outputting a first system response that asks the user about an analysis target, and identifying the analysis target based on the user's answer to the first system response;
outputting a second system response that includes information indicating possible directions of analysis for the analysis target and asking the user about the direction of the analysis; means for identifying the purpose of the analysis based on the user's answer;
means for outputting a third system response including the identified purpose of the analysis;
Equipped with
program. The progress state of the chat is defined using a plurality of predetermined phases,
The computer is used as a means for providing predetermined information or outputting a response to the chat to prompt the user to take a predetermined action without input from the user in each of the plurality of phases. make it work,
The program according to claim 1. The computer,
means for predicting an analysis method suitable for the purpose of the analysis based on user input and system responses on the chat;
means for outputting information on an analysis method suitable for the predicted purpose of the analysis;
to function as
The program according to claim 2. means for outputting a system response to the chat for asking the user a question or prompting a predetermined action when the computer lacks judgment materials for predicting an analysis method suitable for the purpose of the analysis; to function as
The program according to claim 3. The computer,
means for obtaining a first data set that is a data set provided by the user;
means for detecting deficiencies in the first data set based on the purpose of the analysis;
means for outputting a response to the chat, prompting the user to correct deficiencies in the first data set, according to the detection result;
to function as
The program according to claim 2. The computer,
means for obtaining a first data set that is a data set provided by the user;
means for outputting, to the chat, a response suggesting implementation of preprocessing corresponding to the condition when the first data set matches a predetermined condition;
to function as
The program according to claim 2. The output means outputs information for displaying an area where the chat is displayed and an area where the analysis results of the target data set are displayed on one screen or in a switchable manner.
The program according to claim 1. The results of the analysis of the target data set include a graph illustrating the results of the analysis of the target data set, and text representing suggestions based on the results of the analysis of the target data set.
The program according to claim 7. The suggestion indicates an explanation of the variables selected according to the purpose of the analysis, or an explanation of the relationship between two or more variables.
The program according to claim 8. causing the computer to function as means for determining the type of the graph according to the purpose of the analysis and the analysis method used in the analysis;
The program according to claim 8. The computer,
means for determining important points in the graph according to the purpose of the analysis and the analysis method used in the analysis;
means for outputting information indicating locations determined to be important in the graph;
to function as
The program according to claim 8. The analysis of the target data set is performed stepwise over multiple phases,
The outputting means changes at least one of the graph and the suggestion according to the transition of the phase.
The program according to claim 8. The computer is
managing the progress of a chat regarding data analysis with a user;
identifying the purpose of analysis of the user through the progress of the chat;
analyzing the target dataset according to the purpose of said analysis;
outputting a result of the analysis of the target dataset ;
The step of identifying the purpose of said analysis comprises:
outputting a first system response that asks the user about the target of analysis, and identifying the target of analysis based on the user's answer to the first system response;
outputting a second system response that includes information indicating possible directions of analysis for the analysis target and asking the user about the direction of the analysis; identifying the purpose of the analysis based on the user's answers;
outputting a third system response including the identified purpose of the analysis;
Equipped with
Method A means for managing the progress of a chat regarding data analysis conducted with a user;
means for identifying the purpose of analysis of the user through the progress of the chat;
means for analyzing the target data set according to the purpose of said analysis;
and means for outputting the results of the analysis of the target data set,
The means for specifying the purpose of said analysis includes:
means for outputting a first system response that asks the user about an analysis target, and identifying the analysis target based on the user's answer to the first system response;
outputting a second system response that includes information indicating possible directions of analysis for the analysis target and asking the user about the direction of the analysis; means for identifying the purpose of the analysis based on the user's answer;
means for outputting a third system response including the identified purpose of the analysis;
Equipped with
Information processing device. A system composed of multiple computers,
A means for managing the progress of a chat regarding data analysis conducted with a user;
means for identifying the purpose of analysis of the user through the progress of the chat;
means for analyzing the target data set according to the purpose of said analysis;
and means for outputting the results of the analysis of the target data set,
The means for specifying the purpose of said analysis includes:
means for outputting a first system response that asks the user about an analysis target, and identifying the analysis target based on the user's answer to the first system response;
outputting a second system response that includes information indicating possible directions of analysis for the analysis target and asking the user about the direction of the analysis; means for identifying the purpose of the analysis based on the user's answer;
means for outputting a third system response including the identified purpose of the analysis;
Equipped with
system.