Data processing device and data processing method

Abstract

To reduce the effort involved in preparing reports. [Solution] The data processing device is characterized by comprising: an explanatory text data generation unit that generates explanatory text data that explains the content of target data; a first database that stores explanatory text data that has been generated in the past; and a determination unit that compares new explanatory text data generated by the explanatory text data generation unit with past explanatory text data stored in the first database and determines whether or not to consider it as an output candidate.

Description

Technical Field

[0001] The present invention relates to a data processing apparatus and a data processing method.

Background Art

[0002] Conventionally, in order to efficiently manage data such as research records, there is a technique described in Japanese Patent Translation Publication No. 2023-552683 (Patent Document 1). This publication states that "a standardized research record data automatic generation method, apparatus, and computer program for learning an artificial intelligence model are provided. The standardized research record data automatic generation method for learning an artificial intelligence model according to various embodiments of the present invention includes, in a method performed by a computing device, a step of obtaining research record information related to an experiment, a step of processing the obtained research record information based on data related to the experiment stored in advance, and a step of generating standardized research record data using the processed research record information."

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] According to the above prior art, it is possible to process non-standardized research records and automatically generate standardized research record data. Here, when reporting research or the like, the data obtained as a result is evaluated, the necessity of the report is determined, and a report is created. For example, table data is created from handwritten notes, drawn as an appropriate graph, the meaning is read from the graph, compared with past research data, and if it is determined to be necessary, a report is created, following such a procedure. Conventional technologies do not take into account the effort required to prepare such reports. Therefore, the present invention aims to reduce the effort involved in preparing reports. [Means for solving the problem]

[0005] To achieve the above objective, one representative data processing device of the present invention is characterized by comprising: an explanatory text data generation unit that generates explanatory text data describing the content of target data; a first database that stores explanatory text data generated in the past; and a determination unit that compares new explanatory text data generated by the explanatory text data generation unit with past explanatory text data stored in the first database and determines whether or not to consider it as an output candidate. Furthermore, one representative data processing method of the present invention is characterized in that a data processing device having a first database that stores descriptive text data describing the content of target data includes a generation step of generating new descriptive text data describing the content of new target data, a comparison step of comparing the new descriptive text data with past descriptive text data stored in the first database, and a determination step of determining whether or not to set the new descriptive text data as an output candidate based on the results of the comparison. [Effects of the Invention]

[0006] According to the present invention, the effort required to prepare reports can be reduced. Other issues, configurations, and effects will be clarified by the following description of embodiments. [Brief explanation of the drawing]

[0007] [Figure 1] Diagram illustrating the configuration of a data processing device. [Figure 2] A flowchart illustrating the processing steps of a data processing device. [Figure 3] A diagram illustrating how to read input data. [Figure 4] A diagram illustrating how parameter inputs are received. [Figure 5]An explanatory diagram regarding the determination of the graph's parameter set. [Figure 6] An explanatory diagram for graph generation. [Figure 7] An explanatory diagram regarding the generation of descriptive text data. [Figure 8] A flowchart illustrating the procedure for generating descriptive text data. [Figure 9] A concrete example of descriptive data. [Figure 10] Diagram illustrating the selection of similarity scores for descriptive data. [Figure 11] A flowchart showing the procedure for determining the output candidates. [Figure 12] An explanatory diagram about report generation. [Modes for carrying out the invention]

[0008] The following describes an example using drawings. [Examples]

[0009] Figure 1 is an explanatory diagram of the configuration of the data processing device. The data processing device 10 takes handwritten experimental results as input data. The input data may be scanned paper data or data received via electronic paper.

[0010] The data processing device 10 includes, as processing units, a reading unit 21, a graph parameter determination unit 22, a graph generation unit 23, an explanatory text data generation unit 24, a determination unit 25, and a report output unit 26. These processing units may be implemented, for example, by a CPU (Central Processing Unit) executing a predetermined program.

[0011] The data processing device 10 uses a graph selection model 31, an explanatory text generation model 32, and a report selection model 33. These models may be, for example, AI (artificial intelligence) that has undergone machine learning. Furthermore, these models may be configured to reside outside the data processing device 10.

[0012] The data processing device 10 has an explanatory text database 41 and a report database 42. These databases are configured, for example, by storing data in a hard disk drive. Also, these databases may be configured to exist outside the data processing device 10.

[0013] The reading unit 21 is a processing unit that reads the described content from the input data. The reading of the described content may be performed, for example, using OCR (Optical Character Recognition). The reading unit 21 can read handwritten experimental notes and characters described in tabular form. The reading unit 21 sends the read experimental notes to the report output unit 26. The reading unit 21 generates table data from the read tabular characters and sends it to the graph parameter determination unit 22.

[0014] The graph parameter determination unit 22 determines a parameter set related to the generation of a graph. Specifically, the graph parameter determination unit 22 first extracts items that are candidates for the axes and / or values of the graph from the table data sent from the reading unit 21. The graph parameter determination unit 22 accepts an input for specifying the items to be used as axes and the items to be used as values from the extracted items. Also, the graph parameter determination unit 22 accepts an input for specifying the shape of the graph. These specifications are made using the display unit and the input unit of the data processing device 10. The display unit and the input unit may be, for example, a touch panel display.

[0015] The graph parameter determination unit 22 is a processing unit that generates a plurality of parameter set candidates by combining the specified items and shape. The graph parameter determination unit 22 uses the graph selection model 31 to select an appropriate parameter set from the plurality of parameter set candidates. The graph parameter determination unit 22 sends the selected parameter set to the graph generation unit 23. Here, the graph selection model 31 is, for example, an AI that has performed machine learning on combinations of graph axes, data, shapes, etc.

[0016] The graph generation unit 23 is a processing unit that generates a graph using the parameter set and tabular data sent from the graph parameter determination unit 22. The graph generation unit 23 sends the generated graph to the explanatory text data generation unit 24.

[0017] The explanatory text data generation unit 24 is a processing unit that generates explanatory text data that explains the content of the target data. The explanatory text data generation unit 24 uses the graph sent from the graph generation unit 23 as the target data. An explanatory text generation model 32 is used to generate the explanatory text data. An existing text creation AI can be used as the explanatory text generation model 32. For one target data, the explanatory text data generation unit 24 generates explanatory text data multiple times while increasing the amount of text, and terminates the generation of explanatory text data when the increase in content relative to the increase in text amount becomes small. The explanatory text data generation unit 24 sends the graph and explanatory text data to the determination unit 25.

[0018] The determination unit 25 compares the new explanatory data generated by the explanatory data generation unit 24 with past explanatory data stored in the explanatory database 41 and determines whether or not to select the new explanatory data as an output candidate.

[0019] The description database 41 is the first database that stores multiple sets of description data generated in the past. The determination unit 25 calculates the similarity of the new descriptive text data to the descriptive text database 41 and compares the similarity to two thresholds. The larger of the two thresholds is called the upper threshold, and the smaller one is called the lower threshold. The determination unit 25 selects the new descriptive text data as an output candidate if the similarity is greater than or equal to the upper threshold or less than the lower threshold. The reason why new descriptive text data is selected as an output candidate when the similarity is greater than or equal to the upper threshold is that the experimental results are strongly similar to past experimental results and are useful as representative data for the experiment. The reason why new descriptive text data is selected as an output candidate when the similarity is less than the lower threshold is that the experimental results are significantly different from past experimental results and may indicate a new discovery.

[0020] A method for calculating the similarity of new descriptive text data to the descriptive text database 41 is described below. In the first method, the determination unit 25 repeatedly calculates the similarity between the new descriptive text data and any of the past descriptive text data, and sets the highest similarity as the similarity of the new descriptive text data to the descriptive text database 41. In the second method, the determination unit 25 obtains features for the new descriptive text data and compares them with the distribution of features for the past descriptive text data to calculate the similarity of the new descriptive text data to the descriptive text database 41.

[0021] The determination unit 25 sends the new descriptive text data selected as an output candidate, along with the graph corresponding to that descriptive text data, to the report output unit 26.

[0022] The report output unit 26 is a processing unit that creates and outputs report data. First, the report output unit 26 determines whether output is necessary for new report data generated from new explanatory text data, graphs corresponding to the explanatory text data, experimental notes, etc., which are candidates for output, based on the report selection model 33.

[0023] The report selection model 33 is an AI that has undergone machine learning using past report data stored in the second database, the report database 42. The report selection model 33 can determine whether or not to output report data based on what kinds of reports have been submitted in the past. Therefore, it is possible to make determinations that match the requirements of the recipient of the report, such as what kind of report is required by a certain research institute. The report output unit 26 uses the report selection model 33, which has been trained using the report database 42, to decide whether or not to output new report data by comparing it with past report data.

[0024] When the report output unit 26 determines that new report data should be output, it appropriately determines the arrangement of explanatory text data, graphs, experimental notes, etc., generates report data, and outputs it.

[0025] Figure 2 is a flowchart showing the processing procedure of the data processing device 10. The data processing device 10 sequentially executes the following steps S101 to S110. Step S101: The reading unit 21 receives the input data. Then, the process proceeds to step S102. In step S102, the reading unit 21 recognizes the characters in the input data and generates experimental notes, table data, etc. Then, the process proceeds to step S103.

[0026] In step S103, the graph parameter determination unit 22 extracts candidate items from the table data and accepts the specifications for the items and graph shape. The process then proceeds to step S104. Step S104 The graph parameter determination unit 22 generates candidate parameter sets for the graph based on the received input and determines an appropriate parameter set using the graph selection model 31. Then, the process proceeds to step S105.

[0027] In step S105, the graph generation unit 23 generates a graph from the determined parameter set and tabular data. Then, the process proceeds to step S106. In step S106, the explanatory text data generation unit 24 generates new explanatory text data for the generated graph. Then, the process proceeds to step S107. Step S107: The determination unit 25 calculates the similarity between the new description data and the past description data stored in the description database 41. Then, the process proceeds to step S108. In step S108, the determination unit 25 determines whether or not to include the new descriptive text data as an output candidate based on the similarity. The process then proceeds to step S109.

[0028] In step S109, the report output unit 26 uses the report selection model 33 to determine whether or not to output the new explanatory text data that is a candidate for output. After that, the process proceeds to step S110. In step S110, the report output unit 26 generates and outputs report data corresponding to the explanatory text data that it determined needed to be output. After that, the process ends.

[0029] Figure 3 is an explanatory diagram of the input data reading process. The reading unit 21 accepts handwritten experimental data in a table format as input data. The reading unit uses existing technologies such as OCR to extract structured data from the handwritten notes. In Figure 3, the reading unit 21 reads the display data and the experimental notes. The table data has the following items: "Experiment No.", "Experiment Site", "Material A", "Material B", "Material C", "Evaluation X", and "Evaluation Y", with values ​​entered for each item. The experimental notes are text data that states, "We will experiment to see how evaluations X and Y change by trying various combinations of materials A, B, and C. We will try 3 patterns for material A, 3 patterns for material B, and 2 patterns for material C."

[0030] Figure 4 is an explanatory diagram of the input reception process for parameters. The graph parameter determination unit 22 extracts items from the table data that can be candidates for the axes and / or values ​​of the graph. The graph parameter determination unit 22 generates a specification reception screen using the extracted items and displays it on a predetermined display unit. In Figure 4, the graph parameter determination unit 22 extracts the item names of each column from the table data and generates a specification reception screen in checkbox format that allows input of options such as "whether to treat each column as an axis or as data," "whether to use ratio values ​​or sums for axes and data," and "whether to use a line graph or a pie graph." Alternatively, the table data may be input to a generation AI, etc., and the initial values ​​of the checkboxes may be output to reduce the effort required for checkbox input.

[0031] Figure 5 is an explanatory diagram of the determination of the graph parameter set. The graph parameter determination unit 22 generates all possible combinations of items and graph types specified on the input screen as parameter set candidates. Then, the graph parameter determination unit 22 uses the graph selection model 31 to determine whether each of the parameter set candidates is appropriate. The graph parameter determination unit 22 selects the parameter set candidates that it has determined to be appropriate as the parameter set and outputs them.

[0032] In the designation screen shown in Figure 5, the experimental site, material A, material B, and material C are specified as axes. Evaluation X and evaluation Y are specified as data. Furthermore, it is specified that the axes should display ratios, and the data should be displayed as either ratios or sums. The graph types specified are line, bar, and scatter plot.

[0033] Based on these specifications, the graph parameter determination unit 22 determines: {Graph type: line, Graph elements: {Data column: [Evaluation X], Horizontal axis: Experiment field}}, {Graph type: line, Graph elements: {Data column: [Evaluation Y], Horizontal axis: Material C}}, Find combinations such as {Graph type: Scatter plot, Graph elements: {Horizontal data: Evaluation X, Vertical data: Evaluation Y}}.

[0034] The results of the evaluation of the obtained combinations using the graph selection model 31 were as follows: The graph {Graph type: line, Graph elements: {Data column: [Evaluation X], Horizontal axis: Experiment field}} has been determined to be inappropriate. The graph {Graph type: line, Graph elements: {Data column: [Evaluation Y], Horizontal axis: Material C}} has been determined to be appropriate. The graph type {scatter plot, graph elements: {horizontal data: evaluation X, vertical data: evaluation Y}} has been determined to be appropriate. Therefore, the graph parameter determination unit 22, {Graph type: line, Graph elements: {Data column: [Evaluation Y], Horizontal axis: Material C}}, The system outputs a parameter set containing combinations of appropriate types such as {graph type: scatter plot, graph elements: {horizontal data: evaluation X, vertical data: evaluation Y}}.

[0035] Here, we will explain the advantages of accepting specifications using checkboxes. For example, if a user has a preconceived notion such as "the amount of material C is irrelevant to evaluation X," they are unlikely to plot a graph of material C and evaluation X. In contrast, with the checkbox method, if the user believes that "something (such as evaluation B) changes with a change in material C," they will check "axis_material C," and if the user believes that "evaluation X changes due to a change in something (such as the experimental field or material A)," they will check "data_evaluation X." As a result, combinations of material C and evaluation X can be generated as candidates. In this way, by generating graph candidates from item-specific specifications, it is possible to avoid situations where useful graphs are omitted due to user preconceptions or oversights.

[0036] Figure 6 is an explanatory diagram of graph generation. The graph generation unit 23 generates corresponding graphs from the parameter sets output by the graph parameter determination unit 22. In Figure 6, a line graph with "Evaluation Y" on the vertical axis and "Material C" on the horizontal axis, and a scatter graph with "Evaluation Y" on the vertical axis and "Evaluation X" on the horizontal axis are generated.

[0037] Figure 7 is an explanatory diagram of the generation of descriptive text data. The descriptive text data generation unit 24 generates descriptive text data using a graph and a parameter set. In Figure 7, a line graph with "Evaluation Y" on the vertical axis and "Material C" on the horizontal axis is used to generate the following descriptive text data: "A graph showing the amount of Material C on the horizontal axis and the score of Evaluation Y on the vertical axis. When Material C is 1 [g], Evaluation Y tends to be about 30 [points] higher compared to when it is 10 [g]. Even with the same 1 [g] of Material C, Evaluation Y can vary by about 20 [points]..."

[0038] Figure 8 is a flowchart illustrating the procedure for generating explanatory text data. The explanatory text data generation unit 24 sequentially executes the following steps S201 to S207. In step S201, the descriptive text data generation unit 24 sets the length of the descriptive text to an initial value. This initial value is set to a sufficiently small value. Then, the process proceeds to step S202. In step S202, the descriptive text data generation unit 24 generates descriptive text data using the descriptive text generation model 32 with an initial text length. After that, the process proceeds to step S203.

[0039] In step S203, the explanatory text data generation unit 24 increases the amount of text in the explanatory text. Then, the process proceeds to step S204. In step S204, the explanatory text data generation unit 24 generates explanatory text data with the increased amount of text using the explanatory text generation model 32. After that, the process proceeds to step S205. In step S205, the description data generation unit 24 calculates the similarity α between the latest description and the previous description. Then, the process proceeds to step S206.

[0040] In step S206, the descriptive text data generation unit 24 determines whether the similarity α is equal to or greater than the threshold. If the similarity α is less than the threshold (step S206; No), the process returns to step S203. If the similarity α is equal to or greater than the threshold (step S206; Yes), the process proceeds to step S207. In step S207, the explanatory text data generation unit 24 outputs the latest explanatory text and terminates the process.

[0041] Figure 9 shows a specific example of explanatory text data. The explanatory text data T1 generated with the initial text length is: This is a graph showing the amount of ingredient C on the horizontal axis and the evaluation score Y on the vertical axis. The explanatory text data T2, generated by increasing the amount of text from the initial value, "This graph shows the amount of ingredient C on the horizontal axis and the evaluation score Y on the vertical axis. When ingredient C is 1 g, the evaluation Y tends to be approximately 30 points higher compared to when it is 10 g." Furthermore, the explanatory text data T2, which has an increased amount of text, "This graph shows the amount of ingredient C on the horizontal axis and the evaluation score Y on the vertical axis. When ingredient C is 1 g, the evaluation Y tends to be approximately 30 points higher compared to when it is 10 g." Furthermore, the explanatory text data T3, which has an increased amount of text, "This graph shows the amount of ingredient C on the horizontal axis and the evaluation score Y on the vertical axis. When ingredient C is 1 g, the evaluation Y tends to be about 30 points higher compared to when it is 10 g. Even with the same 1 g of ingredient C, the evaluation Y can vary by about 20 points." Furthermore, the explanatory text data T4, which has an increased amount of text, "This graph shows the amount of ingredient C on the horizontal axis and the evaluation score Y on the vertical axis. When ingredient C is 1 g, the evaluation Y tends to be about 30 points higher compared to when it is 10 g. Even with the same 1 g of ingredient C, the evaluation Y can vary by about 20 points. The evaluation Y is drawn with an orange line."

[0042] Comparing the descriptive text data T1 to T4, the increase in information from T3 to T4 is small. In other words, the amount of text in descriptive text data T3 is sufficient to explain the matters that need to be explained, and it is thought that increasing the amount of text further will not increase the amount of information much. In this way, if the text describing the graph is made longer and the amount of information does not seem to increase compared to the previous one, it is determined that the information of the graph has been converted into text without any omissions and that there is no loss of information. It is preferable to output either the latest descriptive text data T4 or the previous descriptive text data T3.

[0043] The data processing unit 10 employs a configuration that generates graphs and converts graph images into text. This is because the subsequent determination unit 25 performs similarity calculations between texts. When calculating similarity using something other than text, for example, if similarity is calculated between graph images, even if the images are almost identical, the similarity will be low if the pixels are shifted. Also, even with the same value, the similarity will be low between a bar graph and a line graph. Furthermore, if similarity is calculated between tabular data, the similarity calculation cannot be performed if the number (dimensions) of data differs. For these reasons, instead of calculating similarity directly from tables or graphs, the similarity is calculated after converting them to text.

[0044] Figure 10 is an explanatory diagram for selecting similarity of descriptive text data. The determination unit 25 calculates the similarity β of the new descriptive text data to the descriptive text database 41. The descriptive text database 41 stores multiple descriptive text data that have been generated in the past.

[0045] As already explained, in the first method for calculating the similarity β, the determination unit 25 performs the process of calculating the similarity between the new description data and the description data in the description database 41 for all the description data in the description database 41. The maximum value of the calculated similarity is then set as the similarity β.

[0046] In the second method for calculating the similarity β, features are obtained for each descriptive text in the descriptive text database 41, and a probability distribution model, such as a GMM (Gaussian mixture model), is created. Then, features are obtained for new descriptive text data, and the likelihood for the model is taken as the similarity β.

[0047] Here, multiple explanatory text databases 41 may be established, and explanatory text data may be classified and stored according to predetermined criteria. For example, an explanatory text database 41 can be established for each experimental team or project. If an explanatory text database 41 is established for each project, the similarity β of new explanatory text data can be calculated not only for the explanatory text database 41 of the project to which the explanatory text data belongs, but also for the explanatory text databases 41 of other projects. If the similarity with the explanatory text database 41 of another project is higher, the priority of report output is increased, as it may be possible to utilize the knowledge gained from that project. Specifically, a flag can be set, and text data such as "Relevance to reports from other projects was observed" can be added.

[0048] Figure 11 is a flowchart showing the procedure for determining the output candidates. The determination unit 25 sequentially executes the following steps S301 to S306. Step S301: The determination unit 25 calculates the similarity β between the new description data and the past description database. Then, the process proceeds to step S302.

[0049] Step S302 The determination unit 25 compares the similarity β with the upper threshold. If the similarity β is greater than or equal to the upper threshold, the process proceeds to step S303. If the similarity β is less than the upper threshold, the process proceeds to step S304. In step S303, the determination unit 25 determines that the graph is strongly similar to one previously reported and is representative data, therefore it is appropriate as an output candidate. The process is then terminated.

[0050] Step S304 The determination unit 25 compares the similarity β with the lower threshold. If the similarity β is less than the lower threshold, the process proceeds to step S305. If the similarity β is equal to or greater than the lower threshold, the process proceeds to step S306. In step S305, the determination unit 25 determines that the graph differs significantly from those previously reported and may represent a new discovery, therefore it is appropriate as an output candidate. The process is then terminated. Step S306: The determination unit 25 determines that the output candidate is inappropriate. The process then terminates.

[0051] Figure 12 is an explanatory diagram of the report generation process. The report output unit 26 generates report data using candidate descriptive text data, a graph corresponding to the descriptive text data, experimental notes corresponding to the descriptive text data, and flags indicating potential relevance to other projects. In Figure 12, line graphs of material C and evaluation Y, descriptive text data "The amount of material C on the horizontal axis, and... on the vertical axis," experimental notes "The proportions of materials A, B, and C...," and flags "Related to the report of project P2..." are arranged on the report page to create the report data and format the report.

[0052] As described above, the data processing device 10 disclosed in the embodiment is characterized by comprising: an explanatory text data generation unit 24 that generates explanatory text data that explains the content of target data; an explanatory text database 41 which is a first database that stores explanatory text data that has been generated in the past; and a determination unit 25 that compares new explanatory text data generated by the explanatory text data generation unit 24 with past explanatory text data stored in the first database and determines whether or not to consider it as an output candidate. With this configuration and operation, the data processing device 10 automatically determines the output target, thereby reducing the effort required to create reports.

[0053] Furthermore, the system includes a reading unit 21 that reads characters written in a tabular format from the input data, a graph parameter determination unit 22 that extracts items that can be candidates for the axes and / or values ​​of the graph based on the reading results of the reading unit 21 and determines a parameter set for graph generation, and a graph generation unit 23 that generates a graph using the parameter set, and the explanatory text data generation unit 24 generates the explanatory text data using the graph as the target data. With this configuration and operation, the data processing device 10 can generate a graph from handwritten characters, generate descriptive data for the graph, and determine whether it should be considered an output candidate.

[0054] Furthermore, the graph parameter determination unit 22 accepts input specifying the items to be used as axes and the items to be used as values ​​from the extracted items, as well as input specifying the shape of the graph. It generates a plurality of parameter set candidates by combining the specified items and shape, and selects an appropriate parameter set from the plurality of parameter set candidates. This configuration and operation allows for comprehensive collection of necessary graphs and highly accurate determination of whether output is required.

[0055] Furthermore, the explanatory text data generation unit 24 generates the explanatory text data multiple times for a single target data, increasing the amount of text each time, and terminates the generation of the explanatory text data when the increase in content relative to the increase in text amount becomes small. This configuration and operation enable the generation of comprehensive explanatory texts about graphs, and allow for highly accurate determination of whether output is necessary.

[0056] As an example, the determination unit 25 repeatedly calculates the similarity between the new description data and any of the past description data, and sets the highest similarity as the similarity of the new description data to the first database. As an example, the determination unit 25 obtains features for the new descriptive text data and compares them with the distribution of features for the past descriptive text data to calculate the similarity of the new descriptive text data to the first database. With this configuration and operation, the determination unit 25 can determine the similarity between one descriptive text data and multiple descriptive text data and decide whether or not to include it as an output candidate.

[0057] Furthermore, the determination unit 25 selects the new descriptive text data as an output candidate if the similarity of the new descriptive text data to the first database is lower than the first threshold, or higher than the second threshold which is greater than the first threshold. This configuration and operation allows for output candidates to include both cases where the new descriptive data represents the content of existing descriptive data, and cases where the new descriptive data includes content not present in existing descriptive data.

[0058] Furthermore, the data processing device 10 is characterized by further comprising a report database 42, which is a second database that stores past report data, and a report output unit 26 that compares new report data generated from new explanatory text data, which is an output candidate, and the reading results and graphs corresponding to said explanatory text data, with past report data stored in the second database to determine whether or not to output the new report data. This configuration allows you to refer to past reports to determine whether or not a report needs to be generated.

[0059] Furthermore, the data processing device 10 includes a plurality of first databases corresponding to past descriptive text data classified according to predetermined criteria. The report output unit 26 increases the output priority and includes a comment in the new report data indicating a strong relationship with other classifications if the similarity obtained for the new descriptive text data against the first database of a different classification is higher than the similarity obtained against the first database of the corresponding classification. This configuration allows for the output of reports even if the descriptive data is not in the relevant field, but is strongly related to other fields.

[0060] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are explained in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace or add configurations, not just delete them. For example, the above example illustrates the creation of an experimental results report, but it can be applied to any data, such as work reports or document summaries. [Explanation of Symbols]

[0061] 10: Data processing unit, 21: Reading unit, 22: Graph parameter determination unit, 23: Graph generation unit, 24: Explanatory text data generation unit, 25: Judgment unit, 26: Report output unit, 31: Graph selection model, 32: Explanatory text generation model, 33: Report selection model, 41: Explanatory text database, 42: Report database

Claims

1. An explanatory text data generation unit generates explanatory text data that explains the content of the target data, The first database stores explanatory text data generated in the past, A determination unit compares the new descriptive data generated by the descriptive data generation unit with past descriptive data stored in the first database and determines whether or not to consider it as an output candidate. A data processing device characterized by having the following features.

2. A data processing device according to claim 1, A reading unit that reads characters written in a tabular format from the input data, Based on the reading results from the aforementioned reading unit, a graph parameter determination unit extracts items that can be candidates for the axes and / or values ​​of the graph and determines a parameter set for generating the graph. A graph generation unit that generates a graph using the aforementioned parameter set, Furthermore, The data processing device is characterized in that the explanatory text data generation unit generates the explanatory text data using the graph as the target data.

3. A data processing device according to claim 2, The graph parameter determination unit accepts input specifying items to be used as axes and items to be used as values ​​from the extracted items, accepts input specifying the shape of the graph, generates a plurality of parameter set candidates by combining the specified items and shape, and selects an appropriate parameter set from the plurality of parameter set candidates.

4. A data processing device according to claim 1, The data processing device is characterized in that the explanatory text data generation unit generates the explanatory text data multiple times for a single target data while increasing the amount of text, and terminates the generation of the explanatory text data when the increase in content relative to the increase in the amount of text becomes small.

5. A data processing device according to claim 1, The data processing device is characterized in that the determination unit repeatedly calculates the similarity between the new descriptive text data and any of the past descriptive text data, and sets the highest similarity as the similarity of the new descriptive text data to the first database.

6. A data processing device according to claim 1, The data processing device is characterized in that the determination unit obtains features for the new descriptive text data, compares them with the distribution of features for the past descriptive text data, and calculates the similarity of the new descriptive text data to the first database.

7. A data processing device according to claim 1, The data processing device is characterized in that the determination unit designates the new descriptive text data as an output candidate when the similarity of the new descriptive text data to the first database is lower than a first threshold and higher than a second threshold which is greater than the first threshold.

8. A data processing device according to claim 2, A second database containing past report data, A data processing device further comprising a report output unit that compares new report data generated from new explanatory text data, which is a candidate for output, and the reading results and graphs corresponding to the explanatory text data, with past report data stored in the second database, and determines whether or not to output the new report data.

9. A data processing device according to claim 8, It comprises multiple first databases corresponding to past descriptive text data classified according to predetermined criteria, The data processing device is characterized in that, when the similarity obtained for the new descriptive text data against a first database of another classification is higher than the similarity obtained against a first database of the corresponding classification, the output priority is increased, and a comment indicating a high relationship with the other classification is included in the new report data.

10. A data processing device having a first database that stores explanatory text data describing the content of the target data, A generation step that generates new descriptive data that explains the content of the new target data, A comparison step involves comparing the new descriptive text data with past descriptive text data stored in the first database. A determination step is to determine whether or not to include the new explanatory text data as an output candidate based on the results of the comparison. A data processing method characterized by including

Images