Information processing device, information processing method, and information processing system

The information processing system enhances data analysis and machine learning design by reproducing execution environments, converting source code, and visualizing dependencies, addressing the inefficiencies of conventional methods and improving the design process through effective workflow management.

WO2024247705A1PCT designated stage expired Publication Date: 2024-12-05SONY GROUP CORP
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2024/017717
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-31
Filing Date
2024-05-14
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

Conventional data analysis and machine learning design processes are complex and inefficient due to their reliance on trial and error, with a lack of effective methods to visualize and manage the dependencies between data and processing nodes, leading to suboptimal results and increased time required for execution.

Method used

An information processing system that includes an environment reproduction unit, a source code converter, a dynamic analysis execution unit, and a log analysis unit to reproduce and analyze the execution environment, convert source code, execute and log the analysis, and output behavior and operation logs, thereby visualizing the dependency relationships between data and processing nodes, and a workflow management system that supports user operations to select and display differences in workflows.

Benefits of technology

This approach improves the efficiency and quality of data analysis and machine learning design by enabling the visualization of workflows, identifying dependencies, and facilitating the comparison of changes, thereby optimizing the design process and reducing execution time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2024017717_05122024_PF_FP_ABST
    Figure JP2024017717_05122024_PF_FP_ABST
Patent Text Reader

Abstract

The present disclosure relates to an information processing device, an information processing method, and an information processing system that enable improvements in efficiency and quality, for example, in data analysis and machine learning design. According to the present invention, an analysis processing unit comprises: an environment reproduction unit that reproduces an environment for analysis on the basis of an environment in which source code to be analyzed is executed; a source code converting unit that converts the source code to be analyzed to generate source code for analysis; a dynamic analysis execution unit that executes the source code for analysis in the environment for analysis and outputs an operation log that is the result of the execution; and a log analyzing unit that outputs a work flow indicating a dependency between data nodes and processing nodes as operation characteristic data obtained as a result of analyzing the operation log. The present technology can be applied to a work flow management system, for example.
Need to check novelty before this filing date? Find Prior Art

Description

Information processing device, information processing method, and information processing system

[0001] The present disclosure relates to an information processing device, an information processing method, and an information processing system, and in particular to an information processing device, an information processing method, and an information processing system that enable efficiency and quality improvement in data analysis and machine learning design.

[0002] Traditionally, data analysis and machine learning design has been an exploratory implementation design, characterized by trial and error. For example, unlike typical software design, in data analysis and machine learning design, the results depend not only on the processing but also on the data. Therefore, when designing a process, multiple branching processes are usually tried, the results are visualized, and then the next process is designed. However, the process does not form a single flow, but rather a flow with branches and merges, making it necessary to compare the results at the branching points.

[0003] As described above, the design of data analysis and machine learning becomes complicated, so it is necessary to support it by visualizing the workflow and to detect the characteristics of each process in the flow. As related techniques, techniques for detecting malware are disclosed in Patent Literature 1 and 2.

[0004] For example, Patent Document 1 discloses a method for detecting bugs in source code. In this method, in the first stage, a location that could potentially be a bug is identified by statically analyzing the source code, and the source code is edited (instructions are added) so as to record the values ​​of program variables at the identified location. In the second stage, the edited source code is executed, and the values ​​of program variables are dynamically analyzed to determine whether the location contains a bug.

[0005] Furthermore, Patent Document 2 discloses a method for speeding up dynamic analysis for debugging, assuming a compiler-type programming language that generates a machine-language program directly executable on a computer from source code. This method is based on the premise that when source code is edited, new bugs can be discovered by monitoring the changes. In this method, in the first stage, changes are collected from the source code edit history, and debugging instructions for monitoring those changes, the memory areas used by those changes, and the machine-language instruction areas generated from those changes are identified. In the second stage, the source code is compiled and the resulting program is executed, and when dynamic analysis is performed, the analysis is limited to the areas related to the edited portions of the source code based on the information obtained in the first stage, thereby realizing faster dynamic analysis.

[0006] JP 2011-133966 A JP 2013-206266 A

[0007] However, the method disclosed in the above-mentioned Patent Document 1 aims to narrow down the locations to be monitored, and simply adds (inserts) debugging instructions to the locations. Therefore, there is a demand for improving quality by monitoring all instruction locations, while also improving efficiency by making changes to the program processing flow itself, such as iteration and recursion, for example, by limiting the time required to execute source code to an allowable upper limit.

[0008] Furthermore, the technology disclosed in the aforementioned Patent Document 2 assumes a compiler-type language and only adds or deletes debug instructions to a program converted into machine language and narrows down the areas to be monitored. Therefore, regardless of whether it is a compiler-type or interpreter-type program, there is a demand for making analytical changes in human-readable source code and for improving efficiency and quality by not narrowing down the areas to be monitored based on the editing history.

[0009] The present disclosure has been made in light of these circumstances, and is intended to enable efficiency and quality improvements in data analysis and machine learning design.

[0010] An information processing device according to a first aspect of the present disclosure includes an environment reproduction unit that reproduces an environment for analysis based on an environment in which a source code to be analyzed is executed, a source code conversion unit that converts the source code to be analyzed to generate source code for analysis, a dynamic analysis execution unit that executes the source code to be analyzed in the environment for analysis and outputs an operation log resulting from the execution, and a log analysis unit that outputs a workflow indicating the dependency between data nodes and processing nodes as operation characteristic data obtained as a result of analyzing the operation log.

[0011] An information processing method according to a first aspect of the present disclosure includes an information processing device reproducing an analysis environment based on an environment in which a source code to be analyzed is executed, converting the source code to be analyzed to generate source code for analysis, executing the source code to be analyzed in the analysis environment and outputting an operation log resulting from the execution, and outputting a workflow indicating the dependency between data and processing as operation characteristic data obtained as a result of analyzing the operation log.

[0012] In a first aspect of the present disclosure, an analysis environment is reproduced based on the environment in which the source code to be analyzed is executed, the source code to be analyzed is converted to generate source code to be analyzed, the source code to be analyzed is executed in the analysis environment, an operation log resulting from the execution is output, and a workflow showing the dependency between data and processing is output as operation characteristic data obtained as a result of analyzing the operation log.

[0013] An information processing device according to a second aspect of the present disclosure includes a workflow operation unit that acquires user operations on a workflow that indicates the dependency between a data node and a processing node, a comparison target selection unit that selects a comparison target for which a difference display is to be performed based on the user operation or the immediately preceding user interface display, and a difference content display unit that displays the difference content of the comparison target.

[0014] An information processing method according to a second aspect of the present disclosure includes an information processing device acquiring a user operation on a workflow indicating a dependency relationship between a data node and a processing node, selecting a comparison target for displaying a difference based on the user operation or the immediately preceding user interface display, and displaying the difference contents of the comparison target.

[0015] In a second aspect of the present disclosure, a user operation on a workflow indicating a dependency relationship between a data node and a processing node is acquired, and a comparison object for displaying differences is selected based on the user operation or the immediately preceding user interface display, and the difference contents of the comparison object are displayed.

[0016] An information processing system according to a third aspect of the present disclosure includes an analysis processing unit having an environment reproduction unit that reproduces an analysis environment based on an environment in which source code to be analyzed is executed, a source code conversion unit that converts the source code to be analyzed to generate source code to be analyzed, a dynamic analysis execution unit that executes the source code to be analyzed in the analysis environment and outputs an operation log resulting from the execution, and a log analysis unit that outputs a workflow indicating the dependency relationships between data nodes and processing nodes as operation characteristic data obtained as a result of analyzing the operation log, a workflow operation unit that acquires user operations on the workflow, and a visualization processing unit having a comparison target selection unit that selects a comparison target for which a difference display is to be performed based on the user operation or the immediately preceding user interface display, and a difference content display unit that displays the difference content of the comparison targets.

[0017] In a third aspect of the present disclosure, an analysis environment is reproduced based on the environment in which the source code to be analyzed is executed, the source code to be analyzed is converted to generate source code to be analyzed, the source code to be analyzed is executed in the analysis environment, an operation log resulting from the execution is output, a workflow indicating the dependency between data and processing is output as operation characteristic data obtained as a result of analyzing the operation log, user operations on that workflow are obtained, a comparison object for which a difference display is to be performed is selected based on the user operation or the immediately preceding user interface display, and the difference contents of the comparison object are displayed.

[0018] 14 is a block diagram showing an example of the configuration of an embodiment of a workflow management system to which the present technology is applied. FIG. 15 is a diagram showing an example of a method of expressing a workflow. FIG. 16 is a diagram showing an example of data and processes having dependencies regarding handover tasks visualized as a workflow. FIG. 17 is a diagram showing an example of a workflow in which a current month's process is added by reusing an existing process. FIG. 18 is a diagram explaining comparison and difference confirmation at any point in a workflow. FIG. 19 is a diagram showing an example of baseline creation data and processes visualized as a workflow in predictive analysis. FIG. 19 is a diagram showing an example of a workflow in which a process is added to a baseline, the workflow and process content are reconfigured, and predictive analysis is re-executed. FIG. 19 is a diagram showing an example of history display and comparison and difference confirmation. FIG. 19 is a diagram showing an example of a comparison of prediction results. FIG. 20 is a diagram showing an example of a best solution mark. FIG. 21 is a diagram showing an example of advantages and disadvantages of static analysis and dynamic monitoring. FIG. 22 is a diagram showing an example of pseudocode. FIG. 23 is a block diagram showing an example of the configuration of an analysis processing unit that performs behavior characteristic analysis of source code in a compiler-type language. FIG. 24 is a block diagram showing a modified example of the dynamic analysis environment source code execution unit of FIG. 23. FIG. 25 is a block diagram showing an example of the configuration of an analysis processing unit that performs behavior characteristic analysis of source code in an interpreter-type language. FIG. 26 is a block diagram showing a modified example of the dynamic analysis environment source code execution unit of FIG. 25. FIG. 27 is a flowchart explaining a first processing example of analysis processing. FIG. 10 is a flowchart illustrating a second processing example of the analysis processing. FIG. 11 is a diagram illustrating the effect of the analysis processing by the analysis processing unit. FIG. 12 is a diagram illustrating an example of a difference display of "entire workflow". FIG. 13 is a diagram illustrating an example of a difference display of "node properties". FIG. 14 is a diagram illustrating an example of a difference display of "node contents". FIG. 15 is a flowchart illustrating a first processing example of the visualization processing. FIG. 16 is a flowchart illustrating a second processing example of the visualization processing. FIG. 17 is a block diagram illustrating an example of the configuration of an embodiment of a computer to which the present technology is applied.

[0019] Hereinafter, specific embodiments to which the present technology is applied will be described in detail with reference to the drawings.

[0020] <Configuration Example of Workflow Management System> FIG. 1 is a block diagram showing a configuration example of an embodiment of a workflow management system to which the present technology is applied.

[0021] 1 is configured to include an analysis processing unit 12 and a visualization processing unit 13, and the analysis processing unit 12 can perform analysis processing by referring to an external resource 14. The external resource 14 is a server that provides various resources used when the analysis target environment source code execution unit 21 of the analysis processing unit 12 executes the analysis target source code.

[0022] In the workflow management system 11, the analysis processing unit 12 imports and automatically analyzes existing assets (e.g., data, scripts, etc.), automatically explores the relationship between data and processes in linear time, and constructs a workflow. The visualization processing unit 13 visualizes the workflow and presents differences in the processes. Note that if the time required to execute source code can be kept within the allowable time, which is the upper limit of the allowable time, i.e., if the allowable time is more relaxed than linear time, it is not necessary to keep the time required to execute source code to linear time. For example, the allowable time (for processing) can be adjusted using the parameters MAX_ITER and MAX_RECUR shown in FIG. 12 (described later), and linear time is required as a minimum. Of course, the same applies to linear time, as described below.

[0023] The analysis processing unit 12 includes an analysis target environment source code execution unit 21 , a dynamic analysis environment source code execution unit 22 , an environment reproduction unit 23 , a source code conversion unit 24 , and a log analysis unit 25 .

[0024] The analysis target environment source code execution unit 21 executes, in a predetermined environment, analysis target source code, which is source code to be analyzed in the analysis processing unit 12. Note that the analysis target source code may be input to the analysis processing unit 12 from the analysis target environment source code execution unit 21 provided outside the analysis processing unit 12.

[0025] The dynamic analysis environment source code execution unit 22 executes the analysis source code supplied from the source code conversion unit 24 in the analysis environment reproduced by the environment reproduction unit 23, and outputs an operation log representing the execution result.

[0026] The environment reproduction unit 23 reproduces the analysis environment in the dynamic analysis environment source code execution unit 22 based on the environment (for example, folder structure) in which the analysis target environment source code execution unit 21 executes the analysis target source code.

[0027] The source code conversion unit 24 converts the source code to be analyzed into analysis source code and supplies it to the dynamic analysis environment source code execution unit 22. Furthermore, when an error log is supplied from the log analysis unit 25, the source code conversion unit 24 may refer to the error log, reconvert the source code so as to avoid the error, and supply it to the dynamic analysis environment source code execution unit 22.

[0028] The log analysis unit 25 outputs a workflow indicating the dependency relationship between the data node and the processing node as operation characteristic data obtained as a result of analyzing the operation log output from the dynamic analysis environment source code execution unit 22. Furthermore, if the analysis of the operation log includes an error log caused by source code conversion by the source code conversion unit 24 or an environmental difference between the analysis target environment source code execution unit 21 and the dynamic analysis environment source code execution unit 22, the log analysis unit 25 may notify the source code conversion unit 24 of the error log.

[0029] In this way, the analysis processing unit 12 can combine the conversion of source code through static analysis with dynamic monitoring by executing the source code in a dynamic analysis environment, and can automatically explore the relationship between data and processing in linear time to construct a workflow.

[0030] The visualization processing unit 13 includes a workflow information storage unit 31 , a user operation history storage unit 32 , a workflow display unit 33 , a workflow operation unit 34 , a comparison target selection unit 35 , a difference content acquisition unit 36 ​​, and a difference content display unit 37 .

[0031] The workflow information storage unit 31 stores workflow information (for example, the configuration and history of processes and data in the workflow) relating to the workflow constructed in the analysis processing unit 12.

[0032] The user operation history storage unit 32 stores the history of various user operations when editing the workflow constructed by the analysis processing unit 12 .

[0033] The workflow display unit 33 displays, for example, the workflow to be edited in a workflow view in accordance with the workflow information stored in the workflow information storage unit 31 .

[0034] The workflow operation unit 34 acquires user operations on the workflow, edits the workflow by reflecting the user operations in the workflow information stored in the workflow information storage unit 31, and stores the user operations in the user operation history storage unit 32. Furthermore, when a user operation is performed to select a workflow or node to be compared and instruct a difference display, for example, the workflow operation unit 34 notifies the comparison target selection unit 35 that the user operation has been performed.

[0035] In accordance with a user operation to select a workflow or node to be compared and instruct a difference display, the comparison target selection unit 35 refers to the user operation history accumulation unit 32 and checks the previous user interface display. Then, the comparison target selection unit 35 selects a comparison target for which a difference display is to be performed based on the previous user interface display, and notifies the difference content acquisition unit 36 ​​of the comparison target. Note that the criteria by which the comparison target selection unit 35 selects a comparison target for which a difference display is to be performed will be described later with reference to the flowcharts shown in FIGS. 22 and 23 .

[0036] The difference content acquisition unit 36 ​​acquires workflow information required to display the difference content of the comparison target from the workflow information storage unit 31 in accordance with the comparison target notified by the comparison target selection unit 35, and supplies it to the difference content display unit 37.

[0037] The difference content display unit 37 displays the difference content of the comparison target in accordance with the workflow information supplied from the difference content acquisition unit 36 ​​.

[0038] In this way, the visualization processing unit 13 can visualize the workflow and display the difference contents of the comparison targets in accordance with the user's operation.

[0039] In the workflow management system 11 configured as described above, the analysis processing unit 12 can analyze the dependency relationships between related data from processes created using general-purpose scripts, processing codes, etc., and analyze the relationships between data and processes. As a result, the workflow management system 11 can extract units such as processes and data as nodes, and build a workflow by structuring the entire process according to the dependency relationships between these nodes, thereby improving the visibility and readability of the workflow.

[0040] This allows the workflow management system 11 to make the most effective use of past assets and display them as a workflow, which is a node dependency relationship between structured data and processes. Furthermore, the workflow management system 11 can modify, edit, and design a workflow by adding new branching processes based on the constructed workflow or by changing input / output data and processing content.

[0041] Furthermore, in the workflow management system 11, when editing a workflow through trial and error, the differences resulting from changes to the workflow can be confirmed by the visualization processing unit 13. For example, a workflow contains information of various granularities, such as the structure, arrangement, and node contents of the workflow, and the visualization processing unit 13 can visualize the differences in an appropriate location, granularity, and presentation method in accordance with the changes to the workflow.

[0042] For example, if there is a change difference in the processing within a processing node, the visualization processing unit 13 highlights the processing node in the workflow, and when viewing the details, it highlights the before and after changes at the line level of code, thereby visualizing the difference content. In this way, the visualization processing unit 13 can present the difference content in accordance with the granularity of the user's operation or focus, and when showing change differences, it can present the difference content in a necessary and sufficient manner.

[0043] As described above, the workflow management system 11 can improve the efficiency and quality of data analysis and machine learning design more than ever before by analyzing the source code using the analysis processing unit 12 and visualizing the difference contents using the visualization processing unit 13.

[0044] <Use Cases of Workflow Management System> Specific use cases using the workflow management system 11 will be described with reference to FIGS. 2 to 10. FIG.

[0045] FIG. 2 shows an example of a representation method for visualizing a workflow in the workflow management system 11.

[0046] For example, the workflow management system 11 reads a folder containing data to be input to a process and processing code to execute the process, and searches for data files and code from the folder.The workflow management system 11 can then display these as data nodes and processing nodes in the workflow view.

[0047] As shown in FIG. 2, the workflow management system 11 displays node types such as processing nodes and data nodes by differentiating them using colors (types of hatching in the illustrated example) or icons, and displays multiple data nodes as data group nodes.

[0048] The workflow management system 11 uses a technology for identifying dependencies through source code analysis to analyze file input / output for processing nodes, and displays the data nodes and processing nodes corresponding to the input / output by connecting them with edges. That is, as shown in Figure 2, the workflow management system 11 displays the dependency relationships between data nodes and processing nodes by arrows that represent a flow.

[0049] As shown in FIG. 2, when a change occurs to a processing node, data node, or the like, the workflow management system 11 displays a change indication mark (in the illustrated example, a black circle displayed in the upper left corner of the icon) on the node where the change occurred.

[0050] 3 to 5 are diagrams illustrating a first use case in which a current person in charge takes over the monthly report creation work from a previous person in charge.

[0051] For example, in the process of creating a monthly report, the previous person in charge would first read the monthly data stored on the shared server, then process the necessary data using a processing script in a specified programming language. Next, the previous person in charge would visualize the processed data for visualization using a visualization script in a specified programming language, and output the image file of the graph obtained by the visualization to the reporting server. This is the procedure for creating the monthly report.

[0052] When the current person in charge takes over this monthly report creation work, the workflow management system 11 reads the processing script, visualization script, and folder containing the data used in the work described above. Then, the workflow management system 11 automatically analyzes the workflow and visualizes the dependent data and processes related to the handed over work as a workflow using the representation method described with reference to FIG.

[0053] As a result, a workflow that visualizes the monthly report creation work is displayed in the workflow view, as shown in FIG.

[0054] The workflow shown in Figure 3 visualizes the monthly report creation process, in which the data group node of the monthly data file is acquired by the data acquisition / processing script, the processed data processed by the data acquisition / processing script is input into the visualization script, the processed data is visualized by the visualization script to create a monthly report chart, which is input into the chart sharing script, and the monthly report chart is output to the reporting server by the chart sharing script and shared.

[0055] Therefore, by checking this workflow, the current person in charge can get an overview of the monthly report creation work of the previous person in charge.The current person in charge can then reuse the previous month's processing for this month's monthly report, allowing them to carry out their work efficiently.

[0056] For example, the current person in charge copies a "processing node" corresponding to the data acquisition and processing script, and performs a user operation to modify only the relevant portion so that the current month's monthly data is input into the copied "processing node." In response to this user operation, the workflow management system 11 automatically searches for and analyzes files, adds a "data node" corresponding to the current month's monthly data to the workflow, and connects the flow so that the current month's monthly data and monthly data file are input into the copied "processing node." The workflow management system 11 then executes the data acquisition and processing script, connecting the flow so that the "data node" corresponding to the current month's processed data is output from the "processing node" corresponding to the data acquisition and processing script.

[0057] Similarly, the current person in charge copies the "processing node" corresponding to the visualization script and performs a user operation to modify only the relevant portion so that the processed data for this month is input into the copied "processing node." In response to this user operation, the workflow management system 11 re-executes the visualization script using the processed data for this month as input, thereby outputting a monthly report chart.

[0058] As a result, as shown in FIG. 4, a workflow in which the processing for this month is added by reusing an existing workflow is displayed in the workflow view.

[0059] The workflow shown in Figure 4 visualizes the monthly report creation process, in which newly added monthly data for the current month and a monthly data file are input into the data acquisition and processing script, the processed data processed by the data acquisition and processing script is input into the visualization script, the processed data is visualized by the visualization script and the monthly report chart that has been overwritten is input into the chart sharing script, and the monthly report chart is output to the reporting server by the chart sharing script and shared.

[0060] Furthermore, the current person in charge can check the differences between the workflow before and after the handover in the workflow management system 11. For example, the workflow management system 11 allows the person in charge to check whether there are differences in the workflow as a whole or in the content of the nodes that make up the workflow. Furthermore, if there are differences in the content of the nodes, the workflow management system 11 allows the person in charge to drill down and check where the differences lie in the content.

[0061] Comparison and difference confirmation at any point in the workflow will be described with reference to FIG.

[0062] For example, the workflow management system 11 can compare the monthly data for the current month input to the data acquisition and processing script with the monthly data file to check for differences, or compare processed data output from the data acquisition and processing script to check for differences. In other words, the workflow management system 11 can easily compare data at any point in the workflow to check for differences.

[0063] The workflow management system 11 also makes it possible to check the differences by comparing the processing of the workflows themselves, and by comparing the "code" of data acquisition and processing scripts. Furthermore, the workflow management system 11 makes it possible to check the differences by comparing the history of past monthly report charts with the latest monthly report chart for overwritten monthly report charts.

[0064] In this way, in the workflow management system 11, for example, if there seems to be an error in the latest monthly report chart, it is possible to check the workflow by tracing back from the monthly report chart, or to check the workflow in order starting from the newly added monthly data for the current month. This makes it possible to efficiently check whether the data, processing, etc. are as expected in the workflow management system 11.

[0065] Furthermore, the workflow management system 11 allows the current month's monthly report creation work to be carried out efficiently and reliably by checking the status at each point in the workflow while overlooking the entire workflow. This allows the current person in charge to easily take over the monthly report creation work from the previous person in charge.

[0066] As described above, in the first use case, by using the workflow management system 11, the current person in charge can easily grasp the dependencies related to the monthly report creation work as a workflow by automatically analyzing the source code and processing files. Furthermore, the workflow difference identification and display function of the workflow management system 11 allows the current person in charge to check the difference information at a granularity appropriate for the difference while getting an overview of the whole. Furthermore, since the workflow management system 11 allows the current person in charge to check and compare data at each point in the workflow, the current person in charge can check the validity of the processing at any point in the workflow.

[0067] 6 to 10 are diagrams illustrating a second use case in which, in machine learning design, feature quantities that serve as explanatory variables are designed through trial and error, and predictive analysis is performed.

[0068] For example, when conducting predictive analysis using machine learning, it is extremely important to design features that serve as explanatory variables to be input into the machine learning algorithm, in addition to the machine learning algorithm and learning parameters, etc. However, the degree of predictive accuracy that can be achieved depends on the problem, data, and machine learning algorithm being solved, so there is not necessarily a single ideal answer, and trial and error in feature design and the search for machine learning parameters are required.

[0069] The workflow management system 11 can efficiently carry out trial and error in feature design. Note that the search for machine learning parameters is called hyperparameter tuning, and the search and optimization are carried out using a separate general technique.

[0070] First, the designer creates a "baseline" by writing the minimum data processing and machine learning processing in code, etc. Then, the workflow management system 11 reads this processing code, and a workflow is constructed in the same way as in the first use case described above.

[0071] 6 shows an example of a workflow in which the baseline creation data and processing in predictive analysis are visualized. In this workflow, the results after predictive analysis can also be confirmed as the results of the processing node, and these are used as the baseline results.

[0072] The workflow shown in Figure 6 visualizes the baseline for predictive analysis, in which the data node of the source data is input to the feature creation script, feature data A and feature data B created by the feature creation script are input to the learning / prediction script, and a result aggregation file is output.

[0073] Next, the designer copies the "processing nodes" corresponding to the feature creation script and the learning / prediction script, reconfigures the workflow and processing content, and then re-runs the predictive analysis and compares the results with the baseline. By repeatedly reconfiguring the workflow and processing content and re-running the predictive analysis in this way, the designer can record various data processing operations and their results in the workflow as a trial-and-error history and compare those results.

[0074] FIG. 7 shows an example of a workflow in which processing is added to the baseline shown in FIG. 6, the workflow and processing content are reconfigured, and predictive analysis is re-executed.

[0075] In the workflow shown in Fig. 7 , a feature creation script 2 and a learning / prediction script 2 are added to the baseline shown in Fig. 6 . As a result, the data node of the source data is input to the feature creation script 2, feature data A2 and feature data B2 created by the feature creation script 2 are input to the learning / prediction script 2, and a result aggregation file 2 is output. For example, the feature creation script 2 is configured to improve the feature data A and feature data B of the baseline, and predictive analysis is re-executed using the improved feature data A2 and feature data B2.

[0076] Therefore, the designer can compare the result summary file, which is the result of the predictive analysis at the baseline, with result summary file 2, which is the result of the predictive analysis using the improved feature data A2 and feature data B2.

[0077] Furthermore, designers may wish to review processes that have already been created during the process of trial and error in the design, comparison of results, etc. In such cases, the workflow management system 11 can check any processing node in the workflow, change the processing, and then re-execute the changed processing node or the entire subsequent processing, thereby implementing the changed processing.

[0078] Of course, the workflow management system 11 can compare not only the final results of the predictive analysis, but also the differences before and after a processing change for any processing node and data node in the middle of the workflow, as in the first use case described above with reference to Figure 5.

[0079] For example, the workflow management system 11 can save the change history in chronological order as a history for the baseline result tally file shown in Fig. 7 and the result tally file 2 in which the processing contents have been revised, as shown in Fig. 8. This allows the designer to compare these histories and check the differences between them.

[0080] Furthermore, designers can check the results of each trial and error in the workflow as they proceed, and once sufficient trial and error has been completed. For example, since the final trial does not necessarily produce the optimal result, it is possible to check and adopt the workflow that provides the best solution, including intermediate results.

[0081] For example, the workflow management system 11 can display the result tally file 1 and result tally file 2 shown in Fig. 7 side by side as shown in Fig. 9. This allows the designer to compare the predicted results.

[0082] After carrying out the above-described trial and error and comparison, the designer can select the predicted result that provides the best solution.

[0083] For example, as shown in FIG. 10, if the result aggregation file 3 is selected by the designer as the best solution, the workflow management system 11 can mark the result aggregation file 3 as the best solution and record the part of the workflow that became the best solution.

[0084] In this way, by using the workflow management system 11, designers can always understand the entire process being designed as a workflow, including trial and error during the design process, and can carry out machine learning design while reviewing and reflecting on the results of trial and error.

[0085] As described above, in the second use case, by using the workflow management system 11, trial and error of processing content in machine learning design can be expressed as a workflow. This allows the designer to grasp the overall picture of the trial and error, compare the content and results of intermediate processing, and as a result, easily search for the best solution. Furthermore, the designer can change processing at any point in the workflow after grasping the entire workflow. Furthermore, the workflow management system 11 allows the difference history and change confirmation to check change differences due to trial and error in the design process not only in the results but also in the intermediate steps within the workflow. This allows the designer to efficiently understand and review the processing content and its results.

[0086] <Analysis of Source Code> The process in which the analysis processing unit 12 analyzes source code and identifies dependencies will be described with reference to FIGS.

[0087] Generally, there are two main methods for understanding processing characteristics (such as input / output): static analysis and dynamic monitoring (dynamic analysis). Static analysis is a method of analyzing all or part of the code to determine whether the code matches a pre-prepared pattern definition. Dynamic monitoring is a method of running the code in a monitoring environment and understanding its characteristics through records of system calls and library function calls.

[0088] 11, static analysis has the advantage of being constant time or linear time relative to the amount of code in terms of time complexity, but has the disadvantage of only being able to detect static characteristics. In contrast, dynamic monitoring has the disadvantage of being unable to grasp the characteristics in advance due to nonlinear time complexity, but has the advantage of being able to detect both static and dynamic characteristics.

[0089] To support the user's trial-and-error data analysis, the analysis processing unit 12 is required to quickly analyze the source code created by the user and extract as many characteristics as possible (especially dependencies with data). To meet this requirement, the analysis processing unit 12 employs a configuration that combines source code conversion by static analysis with dynamic monitoring using the converted code.

[0090] In the first step, the source code conversion unit 24 in the analysis processing unit 12 statically analyzes the source code and imposes a limit on the number of iterations and recursion within the code according to the conversion pattern definition. By imposing an upper limit on iterations and recursion within the same location, it becomes possible to predict in advance the time required to execute the source code. In particular, if the upper limit on the number of iterations and recursion is set to one, it is possible to guarantee that the execution of the source code will be completed in linear time.

[0091] In the next step, the analysis processing unit 12 detects the characteristics of the converted source code by having the dynamic analysis environment source code execution unit 22 execute the converted source code in a dynamic monitoring environment. At this time, the source code conversion is completed in a time linear with respect to the amount of code, so the amount of time calculation required for analysis by the analysis processing unit 12 can be known in advance. For example, when parameters are set so that the execution of the converted code is completed in a linear time, the series of analyses is completed in a time linear with respect to the amount of code. Furthermore, by using dynamic monitoring in addition to static characteristics, the analysis processing unit 12 can detect dynamic characteristics as well.

[0092] This section provides a detailed explanation of source code transformation using static analysis.

[0093] In the first step of the analysis processing unit 12, the source code conversion unit 24 statically analyzes the source code to be analyzed, thereby converting the source code to be analyzed so that the time complexity required to execute the source code is finite and can be known in advance. Here, the factors that require a nonlinear time complexity to execute the source code to be analyzed can be divided into iterative processing and recursive processing.

[0094] In this first step, the source code conversion unit 24 identifies points in the analysis source code where iteration or recursion may occur, and adds a counter to the code to record the number of times that point has actually been executed. The source code conversion unit 24 also adds a process to forcibly exit the iteration or recursion point when the counter reaches a certain value, thereby ensuring that the converted code will complete execution in linear time.

[0095] For example, for a code amount N and an upper limit number of iterations / recursions M, the worst case calculation amount is O(Σi←1toN M i ) can be obtained, so when the upper limit number of iterations / recursions M is greater than 1 (1<M), the worst case calculation amount is O(M N ) These parameters can be adjusted depending on the time cost that is acceptable for the analysis. In particular, when the upper limit number of iterations / recursions M is set to 1, it is guaranteed that the analysis will be completed in linear time with respect to the amount of code N.

[0096] Fig. 12 shows, in pseudo-code, an example of conversion by the source code conversion unit 24. Note that the pseudo-code shown in Fig. 12 shows conversion using a variable "counter" and a system call stack, but this is not a requirement and is merely an example.

[0097] This section explains how to deal with the deterioration of comprehensiveness caused by source code conversion.

[0098] As explained above, while the conversion by the source code conversion unit 24 makes the execution time of the analysis source code finite and predictable in advance, this conversion aborts iterative and recursive processing midway, which means that measures must be taken to address the resulting differences from the processing before and after conversion. As mentioned above, in the workflow management system 11, the analysis processing unit 12 analyzes the dependency relationships between processing and data, and can clarify whether or not various resources (for example, local files, web servers, databases, etc.) are accessed from processing based on source code.

[0099] For example, when one or more various resources are accessed in an iterative / recursive process in the analysis source code, it is expected that each iterative / recursive process will handle similar or nearby resources (e.g., sequential images, daily data, etc.) Therefore, when the source code conversion unit 24 detects that one or more various resources are accessed in an iterative / recursive process in the analysis source code, it can automatically group and handle those similar or nearby resources, and when one or more resources in the group are accessed, it sets an edge between the data group node and the processing node.

[0100] This approach allows dependencies between processes and data to be displayed as dependencies between processes and abstract data groups, thereby reducing the impact of source code conversion on the loss of coverage.In this way, the impact of source code conversion can be reduced by implementing post-processing and other techniques under certain assumptions.

[0101] FIG. 13 is a block diagram showing an example of the configuration of the analysis processing unit 12A that analyzes the behavior characteristics of source code written in a compiler-type language.

[0102] In the analysis target environment source code execution unit 21A, the analysis target source code is input to a language processing system, which compiles the analysis target source code into executable code by referencing a static link library. The analysis target environment source code execution unit 21A then executes the executable code using a DLL (Dynamic Linking Library), an API (Application Programming Interface), internal resources (files), and an external resource (server) 14.

[0103] Meanwhile, the environment of the analysis target environment source code execution unit 21A is reproduced in the dynamic analysis environment source code execution unit 22A by the environment reproduction unit 23, and the analysis target source code is converted into analysis source code by the source code conversion unit 24.

[0104] In the dynamic analysis environment source code execution unit 22A, the analysis source code is input to a language processing system, and the language processing system compiles the analysis source code into analysis execution code by referring to an analysis static link library. Then, the dynamic analysis environment source code execution unit 22A executes the analysis execution code using the analysis DLL and API, pseudo internal resources, and pseudo external resources.

[0105] Then, the operation log output from the analysis execution code of the dynamic analysis environment source code execution unit 22A, the analysis DLL, and the API is supplied to the log analysis unit 25.

[0106] In the analysis processing unit 12A, the language processing system (compiler) of the analysis target environment source code execution unit 21A may be used in the dynamic analysis environment source code execution unit 22A. Furthermore, it is not essential that resources inside and outside the environment, such as files, be reproduced in the dynamic analysis environment source code execution unit 22A.

[0107] Furthermore, in the analysis processing unit 12A, the source code conversion unit 24 may reconvert the source code by referring to the result of analysis of the operation log by the log analysis unit 25, and the dynamic analysis may be performed again by the dynamic analysis environment source code execution unit 22A. For example, errors that occur due to source code conversion or differences in the environment may be detected from the operation log, and the occurrence of errors may be avoided by taking measures when the source code conversion unit 24 reconverts the source code.

[0108] FIG. 14 is a block diagram showing a modified example of the dynamic analysis environment source code execution unit 22A.

[0109] The dynamic analysis environment source code execution unit 22A-1 shown in A of Figure 14 can compile and execute analysis source code, and the library group is virtual for the monitoring environment and can output call records.

[0110] Such a dynamic analysis environment source code execution unit 22A-1 can quickly acquire the characteristics of source code written in a compiler-type language, including dynamic characteristics.

[0111] The dynamic analysis environment source code execution unit 22A-2 shown in FIG. 14B can compile and execute analysis source code, and can monitor API calls within the environment to grasp behavior characteristics.

[0112] Such a dynamic analysis environment source code execution unit 22A-2 can quickly acquire the characteristics of source code written in a compiler-type language, including dynamic characteristics, and can extract a wider variety of characteristics with higher accuracy than the dynamic analysis environment source code execution unit 22A-1.

[0113] FIG. 15 is a block diagram showing an example of the configuration of the analysis processing unit 12B that analyzes the behavior characteristics of source code written in an interpreted language.

[0114] In the analysis target environment source code execution unit 21B, the source code to be analyzed is input into the language processing system, and the language processing system reads and executes the source code to be analyzed using libraries and APIs, internal resources (files), and external resources (servers) 14.

[0115] Meanwhile, the environment of the analysis target environment source code execution unit 21B is reproduced in the dynamic analysis environment source code execution unit 22B by the environment reproduction unit 23, and the analysis target source code is converted into analysis source code by the source code conversion unit 24.

[0116] In the dynamic analysis environment source code execution unit 22B, the analysis source code is input into the language processing system, and the language processing system reads and executes the analysis source code using the analysis library and API, pseudo internal resources, and pseudo external resources.

[0117] Then, the operation logs output from the language processing system, analysis library, and API of the dynamic analysis environment source code execution unit 22B are supplied to the log analysis unit 25.

[0118] In the analysis processing unit 12B, the language processing system (interpreter) of the analysis target environment source code execution unit 21B may be used in the dynamic analysis environment source code execution unit 22B. Furthermore, it is not essential that resources inside and outside the environment, such as files, be reproduced in the dynamic analysis environment source code execution unit 22B.

[0119] Furthermore, in the analysis processing unit 12B, the source code conversion unit 24 may convert the source code by referring to the result of analysis of the operation log by the log analysis unit 25, and the dynamic analysis may be performed again by the dynamic analysis environment source code execution unit 22B. For example, errors that occur due to source code conversion or differences in the environment may be detected from the operation log, and the occurrence of errors may be avoided by devising measures when the source code conversion unit 24 reconverts the source code.

[0120] FIG. 16 is a block diagram showing a modified example of the dynamic analysis environment source code execution unit 22B.

[0121] The dynamic analysis environment source code execution unit 22B-1 shown in A of Figure 16 can execute processing using an interpreter within the dynamic monitoring environment, and the library group is virtual for the monitoring environment and can output call records.

[0122] Such a dynamic analysis environment source code execution unit 22B-1 can quickly acquire the characteristics of source code written in an interpreter language, including dynamic characteristics.

[0123] The dynamic analysis environment source code execution unit 22B-2 shown in B of Figure 16 can execute processing using an interpreter within the dynamic monitoring environment, and can monitor API calls within the same environment to grasp behavior characteristics.

[0124] Such a dynamic analysis environment source code execution unit 22B-2 can quickly acquire the characteristics of source code written in an interpreted language, including dynamic characteristics, and can extract a wider variety of characteristics with higher accuracy than the dynamic analysis environment source code execution unit 22B-1.

[0125] <Example of Analysis Processing> An example of analysis processing executed in the analysis processing unit 12 will be described with reference to FIGS. 17 and 18 .

[0126] FIG. 17 is a flowchart illustrating a first processing example of the analysis processing.

[0127] In step S11 , the environment reproducing unit 23 reproduces, in the dynamic analysis environment source code executing unit 22 , the environment in which the analysis target environment source code executing unit 21 executes the analysis target source code.

[0128] In step S12 , the source code conversion unit 24 converts the source code to be analyzed to generate analysis source code, and supplies the analysis source code to the dynamic analysis environment source code execution unit 22 .

[0129] In step S13, the dynamic analysis environment source code execution unit 22 uses a language processing system to execute the analysis source code supplied from the source code conversion unit 24, and generates an operation log as the execution result.

[0130] In step S14, the log analysis unit 25 analyzes the operation log generated by the dynamic analysis environment source code execution unit 22 and generates operation characteristic data, after which the analysis process is terminated.

[0131] FIG. 18 is a flowchart illustrating a second processing example of the analysis processing.

[0132] In steps S21 to S23, the same processes as in steps S11 to S13 in FIG. 17 are performed, and in step S24, the log analysis unit 25 analyzes the operation log generated in the dynamic analysis environment source code execution unit 22.

[0133] In step S25, the log analysis unit 25 analyzes the operation log and determines whether it contains any error logs caused by source code conversion by the source code conversion unit 24 or differences in the environment between the analysis target environment source code execution unit 21 and the dynamic analysis environment source code execution unit 22.

[0134] In step S25, if the log analysis unit 25 determines that the log contains an error log caused by source code conversion by the source code conversion unit 24 or differences in the environment between the analysis target environment source code execution unit 21 and the dynamic analysis environment source code execution unit 22, processing proceeds to step S26.

[0135] In step S26, the log analysis unit 25 notifies the source code conversion unit 24 of the error log, and the source code conversion unit 24 refers to the error log, reconverts the source code so as to avoid the error, and supplies the reconverted source code to the dynamic analysis environment source code execution unit 22. Then, the process returns to step S23, and the same processes are repeated thereafter.

[0136] On the other hand, in step S25, if the log analysis unit 25 determines that the log does not contain any error logs caused by source code conversion by the source code conversion unit 24 or differences in the environment between the analysis target environment source code execution unit 21 and the dynamic analysis environment source code execution unit 22, processing proceeds to step S27.

[0137] In step S27, the log analysis unit 25 generates operation characteristic data based on the analysis of the operation log in step S24, after which the analysis process is terminated.

[0138] The effect of the analysis process performed by the analysis processing unit 12 will be described with reference to FIG.

[0139] By performing the above-described analysis process, the analysis processing unit 12 can construct a workflow as shown in Fig. 19. That is, when the analysis processing unit 12 acquires behavior characteristic data regarding a plurality of processes, it visualizes the dependencies between them in the form of a graph or the like and presents the workflow.

[0140] 19, data 0 is input to process A, and data 1, data 2, and data 3 are output by executing process A. Furthermore, data 1 is input to process B, and data 4 is output by executing process B, and data 2 and data 3 are input to process C, and data 5 is output by executing process C.

[0141] The analysis processing unit 12 then analyzes the source code to quickly and comprehensively clarify the dependencies between processes and data, allowing users of the workflow management system 11 to easily reproduce and optimize their own work. Furthermore, from the dependency information obtained by the analysis by the analysis processing unit 12, the workflow management system 11 can generate a directed graph (workflow) with processes and data as nodes.

[0142] For example, in the example of the workflow shown in Fig. 19, processes that can be executed in parallel can be identified by the analysis processing by the analysis processing unit 12, such as processes B and C that can be executed in parallel. Also, in the example of the workflow shown in Fig. 19, processes required to obtain a specific result can be identified by the analysis processing by the analysis processing unit 12, such as process C not needing to be executed when data 4 is to be obtained.

[0143] <Visualization of Difference Content> Visualization of difference content by the visualization processing unit 13 will be described with reference to FIGS. 20 to 24. FIG.

[0144] In the workflow management system 11, when a user views or edits a workflow constructed by the analysis processing unit 12, the visualization processing unit 13 can automatically identify the user's focus range in accordance with the user's operations on the workflow, the user interface display, etc. This allows the workflow management system 11 to automatically switch between "difference displays" of various granularities and present them to the user.

[0145] For example, the timing at which a user wants to check the difference can be divided into two main timings, as described below: Timing 1 and Timing 2. Here, because both timings have in common the desire to check the difference between two comparison targets, the differences will be presented in the same way in both Timing 1 and Timing 2 below.

[0146] In the first case, the user explicitly selects the workflow or node to be compared and displays the differences between them. In the second case, the difference between the workflow or node currently being compared and the difference between the updated workflow or node from a snapshot at a certain point in the past is displayed.

[0147] Furthermore, the user's focus range in the user interface display of the workflow management system 11 can be mainly divided into first to third focus ranges as described below.

[0148] The first focus area is the entire workflow (addition / deletion of nodes and edges, change of positions, etc.), the second focus area is the node properties (attribute information such as descriptions and file paths), and the third focus area is the node contents (code and table data in files, etc.).

[0149] The visualization processing unit 13 uses the user's operation and the immediately preceding user interface display to automatically determine the user's attention range.

[0150] For example, if the previous user interface display was for viewing the properties of a node in a workflow or editing the contents of the node, the visualization processing unit 13 determines that the user's interest is in the node and displays difference information related to the node. Also, if the previous user interface display was for viewing the entire workflow or editing the node arrangement, the visualization processing unit 13 determines that the user's interest is in the entire workflow and displays difference information related to the workflow.

[0151] The difference display for the "entire workflow" needs to allow the user to understand that nodes and edges have been added or deleted, node positions have been changed, and node properties and contents have been changed between the workflow before and after the change.

[0152] FIG. 20 shows an example of the difference display for the "entire workflow."

[0153] For example, A in Fig. 20 shows a workflow before the change in which "data.csv" is input to "analysis.py" and "analysis.py" is executed to output "results.csv." Also, B in Fig. 20 shows a workflow after the change in which "data.csv" is input to "analysis.py" and "analysis.py" is executed to output "report.csv."

[0154] Therefore, as shown in C of Figure 20, a change is made to rewrite the code of the "analysis.py" file, and a difference display is displayed showing that "results.csv" has been deleted and "report.csv" has been added as the output of "analysis.py".

[0155] The diff display in the "Node Properties" highlights the attributes that have changed to indicate that there are differences. The highlighting should allow the user to see the contents of the update (addition, modification, deletion).

[0156] FIG. 21 shows, as an example of a difference display of "node properties," the difference in the property information of "analysis.py" in the example of the difference display of the entire workflow shown in C of FIG.

[0157] A in Fig. 21 shows the node properties before the change, and B in Fig. 21 shows the node properties after the change. Therefore, as shown in C in Fig. 21, a difference display is displayed in which "Analyze the data and create a report" has been added to the description, and "Results.csv" has been deleted from the output and "Report.csv" has been added.

[0158] When the contents of the "node contents" are table data, the difference display must be such that the user can grasp additions, changes, and deletions for each cell, as shown in FIG.

[0159] FIG. 22 shows an example of a difference display of "node contents."

[0160] A in Fig. 22 shows the table data before the change, and B in Fig. 22 shows the table data after the change. Therefore, as shown in C in Fig. 22, a difference display is displayed in which the description of id "0001" has been changed to "hogehoge", id "0002" has been deleted, and "baz" and "piyopiyo" have been added to the name and description of id "0003".

[0161] In the workflow management system 11, if the contents of a node are source code or text data, the visualization processing unit 13 needs to visualize it so that the user can understand additions, changes, and deletions on a line-by-line basis, for example.

[0162] In this way, the visualization processing unit 13 can automatically select an appropriate method for displaying differences depending on the data type of the nodes to be compared, and present it to the user.

[0163] <Example of Visualization Processing> An example of the visualization processing executed in the visualization processing unit 13 will be described with reference to FIGS. 23 and 24 .

[0164] FIG. 23 is a flowchart illustrating a first processing example of the visualization processing.

[0165] In step S31, the workflow operation unit 34 selects a node to be compared in accordance with a user operation.

[0166] In step S32, the workflow operation unit 34 determines whether or not a user operation to instruct a difference display has been performed.

[0167] If the workflow operation unit 34 determines in step S32 that a user operation to instruct a difference display has not been performed, the process returns to step S31, and the same processes are repeated thereafter. On the other hand, if the workflow operation unit 34 determines in step S32 that a user operation to instruct a difference display has been performed, the comparison target selection unit 35 is notified that the user operation has been performed, and the process proceeds to step S33.

[0168] In step S33, the comparison target selection unit 35 determines whether the immediately preceding user interface display is a workflow screen, a node attribute information screen, or a node content screen.

[0169] If the comparison target selection unit 35 determines in step S33 that the immediately preceding user interface display was a workflow screen, the process proceeds to step S34. In step S34, the comparison target selection unit 35 notifies the difference content acquisition unit 36 ​​that a node has been selected as a comparison target, and the difference content acquisition unit 36 ​​determines whether there has been a change in the node selected in step S31 based on the workflow information acquired from the workflow information accumulation unit 31. If there has been a change in the node selected in step S31, the difference content display unit 37 displays that there has been a change in the node to be compared, and then the visualization process ends.

[0170] On the other hand, if the comparison target selection unit 35 determines in step S33 that the immediately preceding user interface display was a node attribute information screen, the process proceeds to step S35. In step S35, the comparison target selection unit 35 notifies the difference content acquisition unit 36 ​​that it has selected the attribute information of the node as the comparison target, and the difference content acquisition unit 36 ​​determines whether there has been a change in the attribute information of the node selected in step S31, based on the workflow information acquired from the workflow information accumulation unit 31. Then, if there has been a change in the attribute information of the node selected in step S31, the difference content display unit 37 displays the difference content of the attribute information of the node being compared, and then the visualization process ends.

[0171] On the other hand, if the comparison target selection unit 35 determines in step S33 that the immediately preceding user interface display was a node content screen, the process proceeds to step S36. In step S36, the comparison target selection unit 35 notifies the difference content acquisition unit 36 ​​that the node content has been selected as the comparison target, and the difference content acquisition unit 36 ​​determines whether there has been a change in the content of the node selected in step S31, based on the workflow information acquired from the workflow information accumulation unit 31. Then, if there has been a change in the content of the node selected in step S31, the difference content display unit 37 displays the difference content of the node to be compared, and the visualization process is then terminated.

[0172] FIG. 24 is a flowchart illustrating a second processing example of the visualization processing.

[0173] In step S41, the workflow operation unit 34 selects a workflow to be compared in accordance with a user operation.

[0174] In step S42, the workflow operation section 34 determines whether or not a user operation to instruct a difference display has been performed.

[0175] If the workflow operation unit 34 determines in step S42 that a user operation to instruct a difference display has not been performed, the process returns to step S41, and the same processes are repeated thereafter. On the other hand, if the workflow operation unit 34 determines in step S42 that a user operation to instruct a difference display has been performed, the comparison target selection unit 35 is notified that the user operation has been performed, and the process proceeds to step S43.

[0176] In step S43, the comparison target selection unit 35 determines whether the immediately preceding user interface display is a workflow screen, a node attribute information screen, or a node content screen.

[0177] If the comparison target selection unit 35 determines in step S43 that the immediately preceding user interface display was a workflow screen, the process proceeds to step S44. In step S44, the comparison target selection unit 35 notifies the difference content acquisition unit 36 ​​that a workflow has been selected as a comparison target, and the difference content acquisition unit 36 ​​determines whether there has been a change in the workflow selected in step S41 based on the workflow information acquired from the workflow information accumulation unit 31. Then, the difference content display unit 37 displays the difference content of the workflow selected in step S41, and the visualization process ends.

[0178] On the other hand, if the comparison target selection unit 35 determines in step S43 that the previous user interface display is a node attribute information screen, the process proceeds to step S45. In step S45, the comparison target selection unit 35 notifies the difference content acquisition unit 36 ​​that it has selected the attribute information of the node previously displayed as the comparison target, and the difference content acquisition unit 36 ​​determines whether there has been a change in the attribute information of the node previously displayed, based on the workflow information acquired from the workflow information accumulation unit 31. Then, if there has been a change in the attribute information of the node previously displayed, the difference content display unit 37 displays the difference content, and the visualization process is then terminated.

[0179] On the other hand, if the comparison target selection unit 35 determines in step S43 that the immediately preceding user interface display is a node content screen, the process proceeds to step S46. In step S46, the comparison target selection unit 35 notifies the difference content acquisition unit 36 ​​that it has selected the content of the node previously displayed as the comparison target, and the difference content acquisition unit 36 ​​determines whether there has been a change in the content of the node previously displayed, based on the workflow information acquired from the workflow information accumulation unit 31. Then, if there has been a change in the content of the node previously displayed, the difference content display unit 37 displays the difference content of the node content, and the visualization process is then terminated.

[0180] As described above, by displaying the difference contents based on the user operation or the immediately preceding user interface display through the visualization processing by the visualization processing unit 13, the user operation required for displaying the desired difference contents can be reduced, and the work policy for the data analysis work can be efficiently reviewed. Because reviewing the work policy for such data analysis work is also an important element in considering future policies, the workflow management system 11 can contribute to making the overall data analysis work more efficient and smooth.

[0181] <Example of Computer Configuration> Next, the above-described series of processes (information processing method) can be performed by hardware or software. When the series of processes is performed by software, a program constituting the software is installed in a general-purpose computer or the like.

[0182] FIG. 25 is a block diagram showing an example of the configuration of an embodiment of a computer in which a program for executing the above-described series of processes is installed.

[0183] The program can be recorded in advance on the hard disk 105 or ROM 103 as a recording medium built into the computer.

[0184] Alternatively, the program can be stored (recorded) on a removable recording medium 111 driven by the drive 109. Such a removable recording medium 111 can be provided as a so-called package software. Here, examples of the removable recording medium 111 include a flexible disk, a CD-ROM (Compact Disc Read Only Memory), an MO (Magneto Optical) disk, a DVD (Digital Versatile Disc), a magnetic disk, and a semiconductor memory.

[0185] The program can be installed into the computer from the removable recording medium 111 as described above, or can be downloaded to the computer via a communication network or a broadcasting network and installed on the built-in hard disk 105. That is, the program can be transferred to the computer wirelessly from a download site via an artificial satellite for digital satellite broadcasting, or transferred to the computer by wire via a network such as a LAN (Local Area Network) or the Internet.

[0186] The computer includes a CPU (Central Processing Unit) 102 , to which an input / output interface 110 is connected via a bus 101 .

[0187] When a user inputs a command via the input / output interface 110 by operating the input unit 107, the CPU 102 executes a program stored in a read-only memory (ROM) 103 in accordance with the command. Alternatively, the CPU 102 loads a program stored on a hard disk 105 into a random access memory (RAM) 104 and executes the program.

[0188] As a result, the CPU 102 performs processing according to the flowchart described above or processing performed by the configuration of the block diagram described above. Then, the CPU 102 outputs the processing results from the output unit 106 via the input / output interface 110, transmits them from the communication unit 108, or records them on the hard disk 105, as necessary.

[0189] The input unit 107 is made up of a keyboard, a mouse, a microphone, etc. The output unit 106 is made up of an LCD (Liquid Crystal Display), a speaker, etc.

[0190] In this specification, the processing performed by a computer according to a program does not necessarily have to be performed in chronological order according to the order described in the flowchart. In other words, the processing performed by a computer according to a program also includes processing that is executed in parallel or individually (for example, parallel processing or object-based processing).

[0191] The program may be processed by a single computer (processor), or may be distributed among multiple computers. Furthermore, the program may be transferred to and executed on a remote computer.

[0192] Furthermore, in this specification, a system refers to a collection of multiple components (devices, modules (components), etc.), regardless of whether all of the components are contained in the same housing. Therefore, multiple devices housed in separate housings and connected via a network, and a single device housed in a single housing with multiple modules, are both systems.

[0193] Also, for example, a configuration described as one device (or processing unit) may be divided and configured as multiple devices (or processing units). Conversely, configurations described above as multiple devices (or processing units) may be combined and configured as one device (or processing unit). Of course, configurations other than those described above may be added to the configuration of each device (or each processing unit). Furthermore, as long as the configuration and operation of the entire system are substantially the same, part of the configuration of one device (or processing unit) may be included in the configuration of another device (or other processing unit).

[0194] Furthermore, for example, the present technology can be configured as a cloud computing system in which a single function is shared and processed collaboratively by a plurality of devices via a network.

[0195] Furthermore, for example, the above-described program can be executed in any device, as long as the device has the necessary functions (functional blocks, etc.) and can obtain the necessary information.

[0196] Also, for example, each step described in the above flowchart can be executed by one device or can be shared and executed by multiple devices. Furthermore, if one step includes multiple processes, the multiple processes included in that one step can be executed by one device or can be shared and executed by multiple devices. In other words, multiple processes included in one step can be executed as multiple step processes. Conversely, processes described as multiple steps can be executed collectively as a single step.

[0197] In addition, the processing of the steps of a program executed by a computer may be executed in chronological order according to the order described in this specification, or may be executed in parallel or individually at the required timing, such as when a call is made. In other words, as long as no contradiction occurs, the processing of each step may be executed in an order different from the order described above. Furthermore, the processing of the steps of this program may be executed in parallel with the processing of another program, or may be executed in combination with the processing of another program.

[0198] It should be noted that the present technologies described in this specification can be implemented independently and singly, unless a contradiction arises. Of course, any two or more of the present technologies can also be implemented in combination. For example, part or all of the present technologies described in any embodiment can be implemented in combination with part or all of the present technologies described in other embodiments. Furthermore, part or all of any of the present technologies described above can also be implemented in combination with other technologies not described above.

[0199] <Examples of Combinations of Configurations> The present technology can also be configured as follows. (1) An information processing device comprising: an environment reproduction unit that reproduces an analysis environment based on an environment in which analysis target source code is executed; a source code conversion unit that converts the analysis target source code to generate analysis target source code; a dynamic analysis execution unit that executes the analysis target source code in the analysis environment and outputs an operation log resulting from the execution; and a log analysis unit that outputs a workflow indicating dependencies between data nodes and processing nodes as operation characteristic data obtained as a result of analyzing the operation log. (2) The information processing device described in (1) above, wherein the source code conversion unit statically analyzes the analysis target source code to convert it so that the time complexity required to execute the analysis target source code is finite and can be known in advance. (3) The information processing device described in (2) above, wherein the source code conversion unit, when detecting that one or more resources are accessed in an iterative / recursive process in the analysis target source code, automatically groups and handles the resources. (4) The information processing device according to any one of (1) to (3), wherein the dynamic analysis execution unit analyzes behavioral characteristics of the analysis source code using a compiler-type language. (5) The information processing device according to any one of (1) to (3), wherein the dynamic analysis execution unit analyzes behavioral characteristics of the analysis source code using an interpreter-type language. (6) The information processing device according to any one of (1) to (5), wherein, when an error log is included as a result of analyzing the operation log due to source code conversion by the source code conversion unit or a difference between an environment in which the analysis target source code is executed and an environment for analysis, the log analysis unit notifies the source code conversion unit of the error log, and the source code conversion unit re-converts the analysis target source code with reference to the error log to regenerate the analysis source code and supply the regenerated analysis source code to the dynamic analysis execution unit.(7) An information processing method including: an information processing device reproducing an analysis environment based on an environment in which source code to be analyzed is executed, converting the source code to be analyzed to generate source code to be analyzed, executing the source code to be analyzed in the analysis environment and outputting an operation log that is the execution result, and outputting a workflow that indicates a dependency relationship between data and processing as operation characteristic data obtained as a result of analyzing the operation log. (8) An information processing device comprising: a workflow operation unit that acquires a user operation on a workflow that indicates a dependency relationship between a data node and a processing node, a comparison target selection unit that selects a comparison target for which a difference display is to be performed based on the user operation or an immediately previous user interface display, and a difference content display unit that displays a difference content of the comparison target. (9) The information processing device according to (8), wherein the comparison target selection unit selects a node as the comparison target when a node to be compared has been selected by the user operation and the immediately previous user interface display is a workflow screen, and the difference content display unit displays whether there has been a change in the node to be compared. (10) The information processing device according to (8), wherein, when a node to be compared is selected by the user operation and the previous user interface display is an attribute information screen of the node, the comparison target selection unit selects attribute information of the node as the comparison target, and the difference content display unit displays difference content of the attribute information of the node to be compared. (11) The information processing device according to (8), wherein, when a node to be compared is selected by the user operation and the previous user interface display is a node content screen, the comparison target selection unit selects the content of the node as the comparison target, and the difference content display unit displays difference content of the content of the node to be compared. (12) The information processing device according to (8), wherein, when a workflow to be compared is selected by the user operation and the previous user interface display is a workflow screen, the comparison target selection unit selects a workflow as the comparison target, and the difference content display unit displays difference content of the workflow to be compared.(13) The information processing device according to (8), wherein, when a workflow to be compared is selected by the user operation and the previous user interface display is a node attribute information screen, the comparison target selection unit selects attribute information of the node as the comparison target, and the difference content display unit displays the difference content if there is a change in the attribute information of the node that was displayed last time and is the comparison target. (14) The information processing device according to (8), wherein, when a workflow to be compared is selected by the user operation and the previous user interface display is a node content screen, the comparison target selection unit selects the content of the node as the comparison target, and the difference content display unit displays the difference content if there is a change in the content of the node that was displayed last time and is the comparison target. (15) An information processing method, comprising: an information processing device acquiring a user operation on a workflow that indicates a dependency relationship between a data node and a processing node; selecting a comparison target for which a difference display is to be performed based on the user operation or the previous user interface display; and displaying the difference content of the comparison target. (16) An information processing system comprising: an analysis processing unit having an environment reproduction unit that reproduces an analysis environment based on an environment in which a source code to be analyzed is executed; a source code conversion unit that converts the source code to be analyzed to generate source code to be analyzed; a dynamic analysis execution unit that executes the source code to be analyzed in the analysis environment and outputs an operation log that is the execution result; and a log analysis unit that outputs a workflow that indicates the dependency relationship between a data node and a processing node as operation characteristic data obtained as a result of analyzing the operation log; a workflow operation unit that acquires user operations on the workflow; a comparison target selection unit that selects a comparison target for which a difference display is to be performed based on the user operation or the immediately preceding user interface display; and a difference content display unit that displays the difference content of the comparison targets.

[0200] It should be noted that the present embodiment is not limited to the above-described embodiment, and various modifications are possible within the scope of the gist of the present disclosure. Furthermore, the effects described in this specification are merely examples and are not intended to be limiting, and other effects may also be obtained.

[0201] REFERENCE SIGNS LIST 11 Workflow management system, 12 Analysis processing unit, 13 Visualization processing unit, 14 External resource, 21 Analysis target environment source code execution unit, 22 Dynamic analysis environment source code execution unit, 23 Environment reproduction unit, 24 Source code conversion unit, 25 Log analysis unit, 31 Workflow information accumulation unit, 32 User operation history accumulation unit, 33 Workflow display unit, 34 Workflow operation unit, 35 Comparison target selection unit, 36 Difference content acquisition unit, 37 Difference content display unit

Claims

1. An information processing device comprising: an environment reproduction unit that reproduces an environment for analysis based on the environment in which a source code to be analyzed is executed; a source code conversion unit that converts the source code to be analyzed to generate source code for analysis; a dynamic analysis execution unit that executes the source code to be analyzed in the analysis environment and outputs an operation log that is the result of the execution; and a log analysis unit that outputs a workflow that indicates the dependency between data nodes and processing nodes as operation characteristic data obtained as a result of analyzing the operation log.

2. The information processing device according to claim 1, wherein the source code conversion unit performs static analysis of the analysis target source code to convert the source code so that the time complexity required to execute the analysis source code is finite and can be known in advance.

3. The information processing device according to claim 2, wherein when the source code conversion unit detects that one or more resources are being accessed in an iterative or recursive process in the source code for analysis, the source code conversion unit automatically groups and handles those resources.

4. The information processing device according to claim 1, wherein the dynamic analysis execution unit performs an analysis of the operating characteristics of the analysis source code using a compiler-type language.

5. The information processing device according to claim 1, wherein the dynamic analysis execution unit performs an analysis of the operating characteristics of the analysis source code using an interpreter language.

6. The information processing device of claim 1, wherein if the analysis of the operation log contains an error log caused by the source code conversion by the source code conversion unit or a difference between the environment in which the source code to be analyzed is executed and the environment for analysis, the log analysis unit notifies the source code conversion unit of the error log, and the source code conversion unit re-converts the source code to be analyzed by referring to the error log, thereby regenerating the source code for analysis, and supplies the regenerated source code to the dynamic analysis execution unit.

7. An information processing method comprising: an information processing device reproducing an analysis environment based on an environment in which a source code to be analyzed is executed; converting the source code to be analyzed to generate source code for analysis; executing the source code for analysis in the analysis environment and outputting an operation log that is the result of the execution; and outputting a workflow indicating dependencies between data and processing as operation characteristic data obtained as a result of analyzing the operation log.

8. An information processing device comprising: a workflow operation unit that acquires user operations on a workflow indicating a dependency relationship between a data node and a processing node; a comparison target selection unit that selects a comparison target for which a difference display is to be performed based on the user operations or the immediately preceding user interface display; and a difference content display unit that displays the difference content of the comparison targets.

9. The information processing device according to claim 8, wherein the comparison target selection unit selects a node as the comparison target when a node to be compared is selected by the user operation and the immediately preceding user interface display is a workflow screen, and the difference content display unit displays whether there are any changes to the node to be compared.

10. The information processing device according to claim 8, wherein when a node to be compared is selected by the user operation and the immediately preceding user interface display is a node attribute information screen, the comparison target selection unit selects attribute information of the node as the comparison target, and the difference content display unit displays difference content of the attribute information of the node to be compared.

11. The information processing device according to claim 8, wherein when a node to be compared is selected by the user operation and the immediately preceding user interface display is a node content screen, the comparison target selection unit selects the contents of the node as the comparison target, and the difference content display unit displays the difference content of the contents of the node to be compared.

12. The information processing device according to claim 8, wherein the comparison target selection unit selects a workflow as the comparison target when a workflow to be compared is selected by the user operation and the immediately preceding user interface display is a workflow screen, and the difference content display unit displays the difference content of the workflow to be compared.

13. The information processing device according to claim 8, wherein the comparison target selection unit selects node attribute information as the comparison target when a workflow to be compared is selected by the user operation and the immediately previous user interface display is a node attribute information screen, and the difference content display unit displays the difference content if there is a change in the attribute information of the node that was previously displayed and is the comparison target.

14. The information processing device according to claim 8, wherein the comparison target selection unit selects the contents of a node as the comparison target when a workflow to be compared is selected by the user operation and the immediately previous user interface display is a node content screen, and the difference content display unit displays the difference content if there is a change in the content of the node that was previously displayed as the comparison target.

15. An information processing method including an information processing device: acquiring a user operation on a workflow indicating a dependency relationship between a data node and a processing node; selecting a comparison target for displaying differences based on the user operation or the immediately preceding user interface display; and displaying the difference contents of the comparison targets.

16. An information processing system comprising: an analysis processing unit having an environment reproduction unit that reproduces an environment for analysis based on an environment in which a source code to be analyzed is executed; a source code conversion unit that converts the source code to be analyzed to generate source code for analysis; a dynamic analysis execution unit that executes the source code to be analyzed in the environment for analysis and outputs an operation log resulting from the execution; and a log analysis unit that outputs a workflow indicating the dependency between a data node and a processing node as operation characteristic data obtained as a result of analyzing the operation log; and a visualization processing unit having a workflow operation unit that acquires user operations on the workflow; a comparison target selection unit that selects a comparison target for which a difference display is to be performed based on the user operation or the immediately preceding user interface display; and a difference content display unit that displays the difference content of the comparison targets.

Citation Information

Patent Citations

  • Debug supporting device, parallel execution information generation device and recording medium for recording program applied to the same

    JP1999194957A

  • Distributed processing computer system equipped with software monitor device

    JP2000215085A

  • Log information analysis device and log information analysis method

    JP2017199126A

  • Design assistance device for control software

    WO2013190644A1