Interactive data processing device, interactive data processing method, and interactive data processing program

Abstract

This invention provides an interactive data processing device, an interactive data processing method, and an interactive data processing program that improve usability by making it easier to obtain output results that match the user's intentions. [Solution] The interactive data processing device comprises a reception unit and an embodiment unit. The reception unit receives user input. The embodiment unit materializes the data processing necessary to output results corresponding to the user input. The embodiment unit has a determination unit, a question generation unit, and a decision unit. The determination unit determines the level of abstraction of the user input. If the question generation unit determines that the level of abstraction is high, it generates a question for the user to reduce the level of abstraction. If the decision unit determines that the level of abstraction is low, it determines the data processing necessary to output results corresponding to the user input.

Description

Technical Field

[0001] Embodiments of the present invention relate to an interactive data processing apparatus, an interactive data processing method, and an interactive data processing program.

Background Art

[0002] Recently, DX (Digital Transformation) has been promoted in the fields of manufacturing and social infrastructure. Along with this, various types of data such as text, sensor data, images, and voices are being collected and accumulated. Useful information leading to decision-making is obtained from the accumulated data. The obtained information is used to maintain and enhance the competitiveness of companies.

[0003] In the process of utilizing data, various data processing such as data analysis is performed. Here, for example, data analysis often requires specialized knowledge such as statistics and machine learning, and advanced skills such as programming. Therefore, there are few human resources with the expertise to engage in data analysis work, and the cost required for data analysis tends to be high because it is necessary to secure such human resources. Against such a background, in recent years, the technological development of text-to-text machine learning models such as large language models (LLMs) has advanced, and it is becoming possible to generate another natural language or source code by inputting natural language. The development of a mechanism that can realize advanced data processing such as data analysis without requiring experts through an interactive interaction with a system based on such technology is expected.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Users without specialized knowledge often provide vague, highly abstract natural language input. For example, in data analysis, user input such as "Please tell me the trend of the data" is highly abstract because it is unclear whether the user wants to find representative values, visualize a graph, or find characteristic points of time-series changes. With highly abstract natural language input, it is possible that no output result will be obtained, or even if an output result is obtained, the result may not match the user's intention. In such cases, the user must look at the output result, judge whether it matches their intention, determine what specific information should be added as needed, and add that information. The user repeats this process until the output result matches their intention. The effort involved in interacting with the system during this process, as well as the waiting time for responses from the system, impairs usability.

[0006] The embodiment provides an interactive data processing device, an interactive data processing method, and an interactive data processing program that improve usability by making it easier to obtain output results that match the user's intentions. [Means for solving the problem]

[0007] One embodiment of an interactive data processing device comprises a reception unit and an embodiment unit. The reception unit receives user input. The embodiment unit materializes the data processing necessary to output a result corresponding to the user input. The embodiment unit includes a determination unit, a question generation unit, and a decision unit. The determination unit determines the level of abstraction of the user input. If the question generation unit determines that the level of abstraction is high, it generates a question for the user to reduce the level of abstraction. If the decision unit determines that the level of abstraction is low, it determines the data processing necessary to output a result corresponding to the user input. [Brief explanation of the drawing]

[0008] [Figure 1] Figure 1 is a block diagram showing the configuration of an interactive data device according to the first embodiment. [Figure 2] Figure 2 is a flowchart showing the operation of the interactive data processing device in the first embodiment. [Figure 3] Figure 3 shows an example of an abstraction level determination prompt. [Figure 4] Figure 4 shows an example of a question generation prompt. [Figure 5] Figure 5 shows an example of a code generation prompt. [Figure 6] Figure 6 shows an example of a predefined process. [Figure 7] Figure 7 shows an example of the hardware configuration of the interactive data processing device according to each embodiment. [Modes for carrying out the invention]

[0009] (First Embodiment) Embodiments will be described below with reference to the drawings. Figure 1 is a block diagram showing the configuration of an interactive data device according to the first embodiment. The interactive data processing device 1 includes a reception unit 2, an embodiment unit 3, an execution unit 4, a presentation unit 5, and a database 6. The interactive data processing device 1 can be installed on a computer such as a personal computer, tablet terminal, or smartphone. Functions equivalent to those of the interactive data processing device 1 may be implemented on a server provided on a network that is accessible to user U. This server may be configured as a cloud.

[0010] The reception unit 2 receives user input from user U. User input includes, for example, instructions or questions to the interactive data processing device 1 in natural language. Furthermore, user input includes the user's answers to questions presented to the user by the interactive data processing device 1 based on the level of abstraction of the previous user input. Input is not necessarily limited to natural language input. User input may be other forms of input, such as images, as long as the instructions or questions to the interactive data processing device 1 can be identified.

[0011] The concretization unit 3 concretizes the data processing necessary to output results corresponding to user input received by the reception unit 2. The concretization unit 3 includes a determination unit 31, a question generation unit 32, and a decision unit 33.

[0012] The determination unit 31 determines whether the user input is highly abstract or not. In the first embodiment, the determination unit 31 determines whether the input itself is highly abstract or not. The determination unit 31 makes the determination using a machine learning model. The machine learning model is, for example, a language model. A language model is a model that predicts the probability of a word following a given word by learning text data in natural language. A large-scale language model (LLM) is assumed as the language model, but it does not necessarily have to be an LLM, and any language model may be used. In the first embodiment, the determination unit 31 first creates an abstraction determination prompt for the user input. The abstraction determination prompt is a prompt to cause the language model to answer whether the user input is highly abstract or not. The abstraction determination prompt may include examples of yes or no answers. Also, if the user input includes instructions or questions regarding specific data processing such as data analysis, the abstraction determination prompt may include information about the type of data to be processed. The data type information is, for example, text that represents the type of data, such as "image data," "tabular data," "time series data," or "document data." Furthermore, the abstraction level determination prompt may include information relating to the content of the data to be processed. This information relating to the data content may include, for example, text describing the attribute names and column names of the data if the data is tabular data or time-series data. The determination unit 31 queries the language model using the abstraction level determination prompt and obtains the abstraction level determination result. If the determination unit 31 receives a determination result indicating that the user input has a high level of abstraction, it requests the question generation unit 32 to generate a question. If the determination unit 31 receives a determination result indicating that the user input has a low level of abstraction, it requests the decision unit 33 to make a decision on the data processing necessary to output a result corresponding to the user input.

[0013] The question generation unit 32 generates questions for the user in order to reduce the level of abstraction of the user input. The question generation unit 32 generates questions using a machine learning model. The machine learning model is, for example, a language model. The language model is assumed to be an LLM, but it does not necessarily have to be an LLM and can be any language model. In the first embodiment, the question generation unit 32 first creates a question generation prompt. The question generation prompt is a prompt that causes the language model to generate questions that should be asked of the user in order to reduce the level of abstraction of the user input. If the user input includes instructions or questions regarding specific data processing such as data analysis, the question generation prompt may include information about the type of data to be processed. Furthermore, the abstraction determination prompt may include information related to the content of the data to be processed. The question generation unit 32 queries the language model using the question generation prompt and obtains a question. Then, the question generation unit 32 outputs the obtained question to the presentation unit 5.

[0014] The determination unit 33 determines the specific data processing necessary to output results corresponding to user input. In the first embodiment, the determination unit 33 outputs the initial user input and the user responses entered until it is determined that the level of abstraction is low to the execution unit 4.

[0015] The execution unit 4 executes the data processing determined by the decision unit 33. In the first embodiment, the execution unit 4 queries the language model using, for example, the initial user input received from the decision unit 33 and the user's responses as needed, and obtains the processing results. The execution unit 4 then outputs the obtained processing results to the presentation unit 5. Here, the execution unit 4 may perform data analysis as data processing. Data analysis includes processes such as data extraction, statistical processing, and graph plotting. The execution unit 4 may also perform machine learning of the language model for data analysis, or other machine learning as data processing.

[0016] The prompting unit 5 prompts the user U with the question input from the question generation unit 32. Further, the prompting unit 5 prompts the user U with the processing result input from the execution unit 4. The prompting unit 5 prompts the question, for example, by displaying text representing the question on a display. Further, the prompting unit 5 prompts the processing result, for example, by displaying text, tables, graphs, etc. representing the data analysis result on the display. The prompting of the question and the processing result does not necessarily have to be performed by displaying on the display. The prompting of the question and the processing result may be performed by any method such as voice prompting or printing on paper.

[0017] The database 6 stores a language model 61 as a machine learning model. Here, the language model 61 may be divided into a language model for abstraction degree determination, a language model for question generation, and a language model for executing data processing. The language model for abstraction degree determination, the language model for question generation, and the language model for executing data processing according to user input may be language models individually learned for each. On the other hand, the language model 61 may be one language model that executes abstraction degree determination, question generation, and execution of data processing according to user input in response to a query. Further, the database 6 may store various types of data other than the language model 61 such as data used for data analysis. Furthermore, the database 6 may be provided separately from the interactive data processing apparatus 1. In this case, the interactive data processing apparatus 1 exchanges necessary information by communicating with the database 6.

[0018] Next, the operation of the interactive data processing apparatus 1 in the first embodiment will be described. FIG. 2 is a flowchart showing the operation of the interactive data processing apparatus 1 in the first embodiment.

[0019] In step S1, the reception unit 2 receives a user input from the user U. For example, in the case of data analysis, the user U inputs text in natural language indicating what kind of analysis result is required for what kind of data, for example, "Please tell me the trend of data A."

[0020] In step S2, the determination unit 31 of the concretization unit 3 determines the abstraction level of the user input. In the first embodiment, the determination unit 31 creates an abstraction level determination prompt based on the user input. Then, the determination unit 31 queries the language model using the abstraction level determination prompt and receives the abstraction level determination result from the language model. FIG. 3 is a diagram showing an example of the abstraction level determination prompt. Here, the abstraction level determination prompt in FIG. 3 includes examples for each of the user inputs with a high abstraction level and a low abstraction level.

[0021] In step S3, the determination unit 31 determines whether the abstraction level of the user input is high. If it is determined in step S3 that the abstraction level of the user input is high, the process proceeds to step S4. If it is determined in step S3 that the abstraction level of the user input is low, the process proceeds to step S6.

[0022] In step S4, the question generation unit 32 generates a question in response to a request from the determination unit 31. In the first embodiment, the question generation unit 32 generates a question generation prompt based on the user input. Then, the question generation unit 32 queries the language model using the question generation prompt and receives a question from the language model. FIG. 4 is a diagram showing an example of the question generation prompt. Here, the question generation prompt in FIG. 4 includes the contents of the columns included in data A as information related to the contents of data A.

[0023] In step S5, the presentation unit 5 presents the question received from the question generation unit 32 to the user U. Then, the process returns to step S1. The user U, for example, checks the question displayed on the display and inputs an answer to the question. This answer is received by the reception unit 2 in the process of step S1 again, and then in the process of step S2, the determination unit 31 determines the abstraction level of the user input including the answer from the user U.

[0024] In step S6, the decision unit 33 determines the processing to generate the processing result for the user input. For example, the decision unit 33 outputs the initial user input and the user responses entered up to the point where it is determined that the level of abstraction is low to the execution unit 4.

[0025] In step S7, the execution unit 4 executes the data processing determined by the decision unit 33. In the first embodiment, the execution unit 4 queries the language model using, for example, the initial user input received from the decision unit 33 and the user's responses as needed, and receives the processing results.

[0026] In step S8, the presentation unit 5 presents the processing results received from the execution unit 4 to the user U. After that, the processing shown in Figure 2 is completed. However, depending on the processing performed by the execution unit 4, the presentation of the processing results to the user in step S8 may be omitted. For example, the results of machine learning of the machine learning model by the execution unit 4 may not be presented to the user U.

[0027] As described above, according to the first embodiment, in an interactive data processing device that performs processing in a dialogue format with the user, the level of abstraction of the user input is determined, and if the level of abstraction of the user input is determined to be high, questions are generated to reduce the level of abstraction of the user input. If the level of abstraction of the user input is determined to be low, data processing is performed. The processing result performed when the level of abstraction of the user input is low is likely to reflect the user's intention. Therefore, it is expected that the number of times the user has to look at the output result to determine if it matches their intention, determine what specific information should be added as needed, and add that information will be reduced. As a result, the effort involved in interacting with the system during such tasks and the waiting time until a response from the system is received during the interaction will be reduced, which is expected to improve usability.

[0028] In the first embodiment, user input includes instructions or questions to the interactive data processing device 1, and the user's answers to questions presented by the interactive data processing device 1 to the user. Here, the sequence in which the user first makes user input to the interactive data processing device 1, the interactive data processing device 1 asks the user a question in response to that user input, and the user answers that question is considered to be the history of the dialogue between the user and the interactive data processing device 1. Therefore, the determination unit 31 and the question generation unit 32 may create an abstraction level determination prompt and a question generation prompt by including the questions asked to the user in addition to the user's answers in the user input. Alternatively, past user inputs, questions, answers, and finally determined data processing may be stored as a dialogue history in, for example, a database 6, and when similar user inputs are received, the determination unit 31 and the question generation unit 32 may create an abstraction level determination prompt and a question generation prompt by including the dialogue history of past similar user inputs in the user input. Alternatively, the determination unit 31 may initially create an abstraction determination prompt using user input without dialogue history to determine the abstraction level of the user input. If it determines that the user input without dialogue history is highly abstract, it may then create another abstraction determination prompt using user input including dialogue history to re-determine the abstraction level of the user input.

[0029] (Second embodiment) Next, a second embodiment will be described. The basic configuration of the interactive data device in the second embodiment may be the one shown in Figure 1. However, in the second embodiment, the determination unit 31 performs a different level of abstraction determination than in the first embodiment.

[0030] In the second embodiment, the determination unit 31 instructs the machine learning model to generate code multiple times for data processing based on user input, and determines the level of abstraction based on the variation in the results of each instruction. The machine learning model is, for example, a language model. The language model is assumed to be a large-scale language model (LLM), but it does not necessarily have to be an LLM and may be any language model. In the second embodiment, the determination unit 31 creates a code generation prompt for user input. The code generation prompt is a prompt to instruct the language model to create code based on user input. The code generation prompt may be expressed differently each time a query is made to the machine learning model, as long as the meaning does not change. Also, as in the first embodiment, the code generation prompt may include information about the type of data to be processed and information related to its content. Also, as in the first embodiment, the user input of the code generation prompt may include the dialogue history. Furthermore, as a method for causing the language model to generate multiple codes, the determination unit 31 may instruct it to generate multiple codes in a single execution, or it may execute an instruction to generate one code multiple times. Also, the instruction to generate code may be given to multiple language models. Once multiple codes are obtained, the determination unit 31 determines whether the variation in those codes is large. The determination unit 31 determines that the level of abstraction is high if the variation in the codes is large, and that the level of abstraction is low if the variation is small. This determination is based on the idea that if the level of abstraction of the user input is high, the codes generated for the same user input are likely to be different, and if the level of abstraction of the user input is low, the codes generated for the same user input are likely to be similar. If the determination unit 31 determines that the level of abstraction of the user input is high, it requests the question generation unit 32 to generate a question. If the determination unit 31 determines that the level of abstraction of the user input is low, it requests the decision unit 33 to decide on the data processing necessary to output a result corresponding to the user input.

[0031] Next, the operation of the interactive data processing device 1 in the second embodiment will be described. The operation of the second embodiment basically follows the flowchart shown in Figure 2. In the second embodiment, the determination of the level of abstraction in step S2 differs from that of the first embodiment. The determination of the level of abstraction in the second embodiment will be described below.

[0032] In step S2, the determination unit 31 of the materialization unit 3 determines the level of abstraction of the user input. In the second embodiment, the determination unit 31 creates a code generation prompt based on the user input. The determination unit 31 then queries the language model using the code generation prompt and receives multiple codes from the language model. Figure 5 shows an example of a code generation prompt. Here, the abstraction determination prompt in Figure 5 includes information about the columns contained in the data. After receiving multiple codes from the language model, the determination unit 31 determines the level of abstraction of the user input from the variation in the codes. As a method for determining the magnitude of the code variation, a method may be used in which the received codes are further input into the language model and a query is made as to whether or not the code variation is large. In addition, a method may be used in which the execution unit 4 actually executes the code and compares the resulting value or type, or other existing methods for detecting code similarity may be used. Furthermore, a combination of methods for determining the variation of multiple codes may be used, and the final magnitude of the code variation may be determined by integrating their results. Integration is performed, for example, by calculating the average value if the code variation is given as a numerical value. Furthermore, after determining the magnitude of code variation, i.e., the level of abstraction, the process proceeds to step S3. Thereafter, the process is carried out in the same manner as in the first embodiment. However, in the second embodiment, the determination unit 33 may determine that the data processing for generating processing results for user input is a process that executes the code generated when it is determined that the level of abstraction is low.

[0033] As explained above, in the second embodiment, in an interactive data processing device that performs processing in a dialogue format with the user, code for processing based on user input is actually generated, and the level of abstraction of the user input is determined by the variation in that code. In other words, if the level of abstraction of the user input is high, there is a high possibility that the code generated for the same user input will be different, and if the level of abstraction of the user input is low, there is a high possibility that the code generated for the same user input will be similar, so the level of abstraction of the user input can also be determined by this method. For this reason, an improvement in usability can be expected in the second embodiment as well.

[0034] Furthermore, in the second embodiment, the code generated when it is determined that the level of abstraction is low can be used directly for data processing in the execution unit 4.

[0035] (Third embodiment) Next, a third embodiment will be described. The basic configuration of the interactive data device in the second embodiment may be the one shown in Figure 1. On the other hand, in the third embodiment, the determination unit 31 performs a different level of abstraction determination than in the first and second embodiments.

[0036] In the third embodiment, the embodiment unit 3 is configured to perform data processing by calling a predefined process in the machine learning model and having the execution unit 4 execute the predefined process that was successfully called. In this case, the determination unit 31 queries the machine learning model for user input and has it call a predefined process. The determination unit 31 then determines the level of abstraction based on the result of calling the predefined process. The machine learning model is, for example, a language model. The language model is assumed to be a large-scale language model (LLM), but it does not necessarily have to be an LLM and can be any language model. In the third embodiment, the determination unit 31 queries the machine learning model for user input. The determination unit 31 then determines that the information contained in the user input is insufficient, i.e., the level of abstraction of the user input is high, if it receives a response from the language model indicating that it failed to call a predefined process, for example, because there is no definition of an appropriate process for the user input or it failed to extract something equivalent to the argument of the process from the user input. On the other hand, the determination unit 31 determines that the level of abstraction of the user input is low when it receives a response from the language model indicating that it has successfully called a predefined process. If the determination unit 31 determines that the level of abstraction of the user input is high, it requests the question generation unit 32 to generate a question. If the determination unit 31 determines that the level of abstraction of the user input is low, it requests the decision unit 33 to decide on the data processing necessary to output a result corresponding to the user input.

[0037] Next, the operation of the interactive data processing device 1 in the third embodiment will be described. The operation of the third embodiment is also basically carried out according to the flowchart shown in Figure 2. In the third embodiment, the determination of the level of abstraction in step S3 differs from that of the first and second embodiments. The determination of the level of abstraction in the third embodiment will be described below.

[0038] In step S2, the determination unit 31 of the concretization unit 3 determines the level of abstraction of the user input. In the third embodiment, the determination unit 31 queries the language model using the user input, causes the language model to execute a call to a predefined process, and receives a result indicating whether the call to the predefined process was successful or not. Figure 6 shows an example of a predefined process. After receiving the result indicating whether the call to the predefined process was successful or not, the process moves to step S3. After that, the process is carried out in the same manner as in the first embodiment. However, in the third embodiment, the decision unit 33 may decide that the data processing for generating the processing result for the user input is a process that causes the successfully called predefined process to be executed. Alternatively, the decision unit 33 may decide that the processing for generating the processing result for the user input is another process using the user input at the time the call to the predefined process was successful. This other process includes, for example, the generation of code based on the user input as described in the second embodiment.

[0039] As described above, according to the third embodiment, in an interactive data processing device that performs processing in a dialogue format with the user, a predefined call is made based on the user input, and the level of abstraction of the user input is determined by the success or failure of the call to the predefined process. In other words, if the level of abstraction of the user input is high, there is a high probability that the call to the predefined process will succeed, and if the level of abstraction of the user input is low, there is a high probability that the call to the predefined process will fail, so the level of abstraction of the user input can also be determined by this method. For this reason, an improvement in usability can be expected in the third embodiment as well.

[0040] Furthermore, in the third embodiment, a predefined process that is successfully called when it is determined to have a low level of abstraction can be used directly for data processing in the execution unit 4.

[0041] Figure 7 shows an example of the hardware configuration of the interactive data processing device 1 according to each embodiment. The interactive data processing device 1 may be a computer having, for example, a processor 101, memory 102, storage 103, input interface 104, display 105, and communication device 106 as hardware. The processor 101, memory 102, storage 103, input interface 104, display 105, and communication device 106 are connected to a bus 107. As mentioned above, the interactive data processing device 1 can be installed in a computer such as a personal computer, tablet terminal, or smartphone. Functions equivalent to those of the interactive data processing device 1 may be implemented on a server provided on a network that is accessible to user U. This server may be configured as a cloud.

[0042] The processor 101 is a processor that controls the overall operation of the interactive data processing device 1. The processor 101 operates as the receiving unit 2, the materialization unit 3, the execution unit 4, and the presentation unit 5 by executing the interactive data processing program 1031 stored in the storage 103, for example. The processor 101 is, for example, a CPU. The processor 101 may also be an MPU, GPU, ASIC, FPGA, etc. The processor 101 may be a single CPU, etc., or multiple CPUs, etc.

[0043] Memory 102 includes ROM and RAM. ROM is non-volatile memory. ROM stores the startup program for the interactive data processing device 1, etc. RAM is volatile memory. RAM is used, for example, as working memory during processing in the processor 101.

[0044] The storage 103 is, for example, a flash memory, a hard disk drive, or a solid-state drive. The storage 103 stores various programs executed by the processor 101, such as the interactive data processing program 1031. The storage 103 may also store the language model 1032. The language model 1032 does not necessarily have to be stored in the storage 103. For example, the language model 1032 may be stored in external storage provided separately from the interactive data processing device 1. In this case, the interactive data processing device 1 obtains the necessary information from the external storage using the communication device 106.

[0045] The input interface 104 is an input device such as a touch panel, keyboard, mouse, or microphone. When the input interface 104 is operated, a signal corresponding to the operation is input to the processor 101 via the bus 107. The processor 101 performs various data processing in response to this signal. The input interface 104 can be used for user input by user U.

[0046] Display 105 is a display device such as a liquid crystal display or an organic EL display. Display 105 performs various types of displays. These displays may include the display of a GUI (Graphical User Interface) for user input, the display of questions after user input, and the display of processing results of processes performed in response to user input.

[0047] The communication device 106 is a communication device for the interactive data processing device 1 to communicate with external devices. The communication device 106 may be a communication device for wired communication or a communication device for wireless communication.

[0048] (modified version) The following describes modifications of the embodiment. In the embodiment, both the determination of whether the user input is highly abstract by the determination unit 31 and the generation of questions by the question generation unit 32 are performed using a language model. However, one or both of the determination of whether the user input is highly abstract by the determination unit 31 and the generation of questions by the question generation unit 32 do not have to be performed using a language model. For example, the abstraction determination prompt described above may include examples where the answer is yes or no. The determination unit 31 may determine the abstraction level of the user input using a rule-based method with these examples as rules.

[0049] Furthermore, the abstraction level determination methods described in the first, second, and third embodiments above may be used in combination. For example, the determination unit 31 may initially determine the abstraction level of the user input using the abstraction level determination method of the first embodiment, and then, if it determines that the abstraction level is high, perform the abstraction level determination method of the second embodiment.

[0050] Furthermore, the instructions shown in the processing procedure described in the above-described embodiment can be executed based on a software program. A general-purpose computer system can store this program in advance and, by reading this program, can obtain the same effects as those of the interactive data processing device described above. The instructions described in the above-described embodiment are recorded as a program that can be executed by a computer on a magnetic disk (flexible disk, hard disk, etc.), optical disk (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, Blu-ray® Disc, etc.), semiconductor memory, or similar recording medium. Any storage format is acceptable as long as it is a recording medium that can be read by a computer or embedded system. The computer can read the program from this recording medium and, based on this program, have the CPU execute the instructions described in the program, thereby achieving the same operation as the interactive data processing device of the above-described embodiment. Of course, when the computer acquires or reads the program, it may do so via a network. Furthermore, an operating system (OS) running on a computer, a database management software, a network, or other middleware (MW) operating on a computer, based on instructions from a program installed on a computer or embedded system from a recording medium, may execute some of the processes necessary to realize this embodiment. Furthermore, the recording medium in this embodiment is not limited to a medium independent of the computer or embedded system, but also includes recording media that store or temporarily store programs downloaded via LAN, the Internet, etc. Furthermore, the recording medium is not limited to one; even when the processing in this embodiment is performed from multiple media, these are also included as recording media in this embodiment, and the configuration of the media may be any configuration.

[0051] In this embodiment, the computer or embedded system is used to execute each process in this embodiment based on a program stored on a recording medium, and may be configured as any of the following: a single device such as a personal computer or microcomputer, or a system in which multiple devices are connected via a network. Furthermore, the term "computer" in this embodiment is not limited to personal computers, but also includes arithmetic processing units, microcontrollers, and the like included in information processing equipment, and refers collectively to any equipment or device capable of realizing the functions of this embodiment through a program.

[0052] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents. [Explanation of symbols]

[0053] 1 Interactive data processing device, 2 Reception unit, 3 Materialization unit, 4 Execution unit, 5 Presentation unit, 6 Database, 31 Judgment unit, 32 Question generation unit, 33 Decision unit, 61 Language model, 101 Processor, 102 Memory, 103 Storage, 104 Input interface, 105 Display, 106 Communication device, 107 Bus, 1031 Interactive data processing program, 1032 Language model.

Claims

1. A reception area that accepts user input, A concretization unit that concretizes the data processing necessary to output the results corresponding to the user input, It is equipped with, The aforementioned concrete part is, A determination unit for determining the level of abstraction of the user input, A question generation unit that generates questions for the user to reduce the level of abstraction when it is determined that the level of abstraction is high, A determination unit that determines the data processing necessary to output the result corresponding to the user input when the level of abstraction is determined to be low, Having, Interactive data processing device.

2. Either the determination unit or the question generation unit uses a machine learning model. The interactive data processing device according to claim 1.

3. The aforementioned machine learning model is a language model. The interactive data processing device according to claim 2.

4. The language model includes the user input and determines the level of abstraction by inputting an abstraction determination prompt that prompts for the user input to indicate the level of abstraction. The interactive data processing device according to claim 3.

5. The language model generates multiple codes by a code generation prompt that includes the user input and generates code for data processing necessary to output a result corresponding to the user input. The determination unit determines the level of abstraction based on the magnitude of the variation in the plurality of codes. The interactive data processing device according to claim 3.

6. The language model calls predefined processes necessary to output a result corresponding to the user input, The determination unit determines the level of abstraction based on whether the call to the predefined process by the language model was successful or not. The interactive data processing device according to claim 3.

7. The language model includes the user input and generates the question by inputting a question generation prompt that generates a question to the user to reduce the level of abstraction. The interactive data processing device according to claim 3.

8. Either the determination unit or the question generation unit uses information about the data subject to the data processing. The interactive data processing device according to claim 1.

9. The information relating to the target data includes information relating to the type of data, The interactive data processing device according to claim 8.

10. The aforementioned data is in tabular format or time series format. The information relating to the target data includes the column names contained in the tabular data or time-series data. The interactive data processing device according to claim 8.

11. Either the determination unit or the question generation unit uses dialogue history, which is historical information relating to one or more of the user input, the results of the data processing, or the question. The interactive data processing device according to claim 1.

12. If the determination unit determines, without using the dialogue history, that the level of abstraction of the user input is high, it re-determines the level of abstraction of the user input using the dialogue history. The interactive data processing device according to claim 11.

13. The interactive data processing device according to claim 1, wherein the data processing includes one or more of the following: data extraction processing, statistical processing, graph plotting, and machine learning.

14. The reception desk accepts user input, The determination unit determines the level of abstraction of the user input, If the level of abstraction is determined to be high, the question generation unit generates a question for the user to reduce the level of abstraction, If the level of abstraction is determined to be low, the determination unit determines the data processing necessary to output the result corresponding to the user input, Equipped with, Interactive data processing method.

15. Accepting user input, Determining the level of abstraction of the user input, If the aforementioned level of abstraction is determined to be high, the system generates questions for the user to reduce the level of abstraction, If the level of abstraction is determined to be low, the data processing necessary to output the result corresponding to the user input is determined, An interactive data processing program that causes a processor to execute a command.

Images