Information processing device, information processing method, and program

Abstract

To generate learning data for improving document classification accuracy of a document classification AI.SOLUTION: An information processing device comprises: an extraction unit which executes natural language processing on first document information to be used for learning of a document classification AI to extract first words from each piece of the first document information; a calculation unit which calculates a degree of importance for each of the first words on the basis of the extracted first words; a determination unit which classifies the first document information by types of documents, arranges the first words in an order of high degrees of importance for every classification, and determines second words which are not included in other classifications for every classification in a preset high level range of the degrees of importance; and a generation unit which generates input data corresponding to an input form of a generative AI by using the extracted second words in order to generate second document information to be used for learning of the document classification AI by using the generative AI.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, and a program for classifying documents. [Background technology]

[0002] To improve the accuracy of document classification AI (Artificial Intelligence), sufficient training data is required. Generally, the more training data there is, the higher the accuracy of document classification AI. However, currently, it is often not possible to prepare sufficient training data, and therefore the accuracy of document classification AI is not high.

[0003] As a related technique, Patent Document 1 discloses an information processing system that generates training sentences (training data) in a manner that has confirmed the effectiveness of learning. According to the information processing system of Patent Document 1, sentences (training sentences) similar to the original sentences are generated by replacing words in the original sentences using thesaurus data. Then, the generated similar sentences (training sentences) are used to train a model, thereby improving the performance of the model. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2021-135839 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the information processing system of Patent Document 1 uses thesaurus data to replace words in the original sentences to generate training sentences (training data), but does not use a generation AI to generate training data. Specifically, it does not cause the generation AI to generate training data necessary to improve the accuracy of document classification by the document classification AI.

[0006] One example of the purpose of the present disclosure is to generate training data for improving the document classification accuracy of a document classification AI. [Means for solving the problem]

[0007] In order to achieve the above object, an information processing device according to one aspect of the present disclosure includes: an extraction unit that performs natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence) and extracts first words from each of the first document information; a calculation unit that calculates the importance of each of the first words based on the extracted first words; a determination unit that classifies the first document information by document type, arranges the first words in order of importance for each classification, and determines, for each classification, second words that are not included in other classifications within a predetermined high range of importance; a generation unit that uses the extracted second words to generate input data corresponding to an input format of a generation AI (Artificial Intelligence) in order to generate second document information to be used for learning of the document classification AI; The present invention is characterized by having the following.

[0008] In order to achieve the above object, an information processing method according to one aspect of the present disclosure includes: The information processing device performing natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence), and extracting first words from each of the first document information; Calculating the importance of each of the first words based on the extracted first words; classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; In order to generate second document information to be used for learning the document classification AI using a generation AI (Artificial Intelligence), input data corresponding to an input format of the generation AI is generated using the extracted second words. It is characterized by:

[0009] Furthermore, in order to achieve the above object, a program according to one aspect of the present disclosure comprises: On the computer, performing natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence), and extracting first words from each of the first document information; Calculating the importance of each of the first words based on the extracted first words; classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; In order to generate second document information to be used for learning the document classification AI using a generation AI (Artificial Intelligence), input data corresponding to an input format of the generation AI is generated using the extracted second words. The present invention is characterized in that it causes the processing to be executed. [Effects of the Invention]

[0010] As described above, according to the present disclosure, it is possible to generate learning data for improving the document classification accuracy of a document classification AI. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram for explaining an example of an information processing device (a learning data generating device for a document classification AI). [Figure 2] FIG. 2 is a diagram illustrating an example of a system including an information processing device. [Figure 3]FIG. 3 is a diagram illustrating an example of calculation of importance. [Figure 4] FIG. 4 is a diagram for explaining an example of keyword extraction. [Figure 5] FIG. 5 is a diagram for explaining the operation in the operation phase. [Figure 6] FIG. 6 is a diagram for explaining an example of the operation of an information processing device (a learning data generating device for a document classification AI). [Figure 7] FIG. 7 is a diagram illustrating an example of a computer that realizes an information processing device (a learning data generation device for a document classification AI) according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments will be described with reference to the drawings. In the drawings described below, elements having the same or corresponding functions are denoted by the same reference numerals, and repeated description thereof may be omitted.

[0013] (Embodiment) The configuration of an information processing device in an embodiment will be described with reference to Fig. 1. Fig. 1 is a diagram for explaining an example of an information processing device (a learning data generation device for a document classification AI).

[0014] [Device configuration] The information processing device shown in Fig. 1 is a device that generates learning data for improving the document classification accuracy of a document classification AI. Also, as shown in Fig. 1, the information processing device (learning data generation device for document classification AI) 10 has an extraction unit (extraction means) 11, a calculation unit (calculation means) 12, a determination unit (determination means) 13, and a generation unit (generation means) 14.

[0015] The extraction unit 11 performs natural language processing on first document information (original learning data) stored in a storage device for use in training the document classification AI, and extracts first words (target words) from each piece of first document information.

[0016] The calculation unit 12 calculates the importance of each of the extracted first words based on the extracted first words.

[0017] The determination unit 13 classifies the first document information by document type, arranges the first words in order of importance for each classification, and determines second words (keywords) for each classification within a predetermined top range of importance that are not included in other classifications.

[0018] The generation unit 14 uses the extracted second words to generate input data (prompts) corresponding to the input format of the generation AI in order to generate second document information (generated learning data) to be used for training the document classification AI using the generation AI.

[0019] In this manner, in the embodiment, the extracted second words (keywords) can be used to generate second document information in addition to the first document information as training data for the document classification AI, thereby increasing the training data for the document classification AI and improving the accuracy of document classification by the document classification AI.

[0020] [System Configuration] Next, the configuration of the information processing device 10 in the embodiment will be described in more detail with reference to Fig. 2. Fig. 2 is a diagram illustrating an example of a system including an information processing device.

[0021] 2, the system 100 in the embodiment includes an information processing device 10, an information processing device (generation AI) 20, and an information processing device (document classification AI) 30. The information processing device 10, the information processing device (generation AI) 20, and the information processing device (document classification AI) 30 are electrically connected via a network.

[0022] The information processing device 10 is, for example, an information processing device such as a CPU (Central Processing Unit), a programmable device such as an FPGA (Field-Programmable Gate Array), a GPU (Graphics Processing Unit), or a circuit equipped with any one or more of these, a server computer, a personal computer, or a mobile terminal.

[0023] The information processing device 20 is an information processing device such as a CPU, a programmable device such as an FPGA, a GPU, or a circuit equipped with one or more of these, a server computer, a personal computer, or the like, equipped with a generating AI.

[0024] The generation AI installed in the information processing device 20 is, for example, an AI that generates text, such as a Generative Pretrained Transformer (GPT), etc. However, it is not limited to the generation AI described above.

[0025] The information processing device 30 is an information processing device such as a CPU, a programmable device such as an FPGA, a GPU, or a circuit equipped with one or more of these, a server computer, a personal computer, or a mobile terminal, equipped with a document classification AI.

[0026] The document classification AI installed in the information processing device 30 is an AI that performs document classification using a natural language processing model such as BERT (Bidirectional Encoder Representations from Transformers), but is not limited to the document classification AI described above.

[0027] The network is a general communication network constructed using communication lines such as the Internet, a LAN (Local Area Network), a dedicated line, a telephone line, an in-house network, a mobile communication network, Bluetooth (registered trademark), or Wi-Fi (Wireless Fidelity) (registered trademark).

[0028] 2 shows a configuration in which the document classification AI is installed in the information processing device 30 and provided outside the information processing device 10, the document classification AI may be provided inside the information processing device 10.

[0029] The learning phase will now be described. ●About the extraction part 2, the extraction unit 11 acquires first document information D1 (original learning data) from a storage device (not shown). The first document information D1 is various documents (electronic data) such as specifications, presentation documents, and email documents.

[0030] The storage device is a database, a server computer, a circuit having a memory, etc. The storage device stores, for example, at least a plurality of pieces of first document information D1. The storage device may be provided outside or inside the information processing device 10.

[0031] The extraction unit 11 performs natural language processing on each of the acquired first document information D1 (original learning data) and extracts, for each piece of first document information D1, the target first word contained in the first document information D1.

[0032] The natural language processing is a process of extracting a first target word from the first document information D1 by, for example, morphological analysis, etc. However, the natural language processing is not limited to the above-mentioned morphological analysis.

[0033] ●About the calculation section The calculation unit 12 calculates the appearance frequency (word frequency) of each extracted first word for each piece of first document information D1. The word frequency is the frequency at which the first word appears in the first document information D1. The word frequency for each piece of first document information D1 can be expressed as in Equation 1.

[0034] (Number 1) Y1=N1 / N2 Y1: Word Frequency N1: Number of occurrences of the first word N2: Number of the first word in the first document information

[0035] The calculation unit 12 also calculates the rarity (inverse document frequency) of the appearance of the first word among the first document information D1. The inverse document frequency is an index that indicates the bias of the first word toward a specific document among a plurality of first document information D1. In other words, the inverse document frequency (index) considers that the more the first word appears in a fewer number of documents, the stronger its role in characterizing the document. The inverse document frequency can be expressed as Equation 2.

[0036] (Number 2) Y2=log(N3 / N4)+1 Y2: Inverse document frequency N3: Total number of first document information N4: Number of first document information containing the first word

[0037] Furthermore, the calculation unit 12 calculates the importance of each first word based on the word frequency and the inverse document frequency. Specifically, for each piece of first document information D1, the calculation unit 12 multiplies the word frequency corresponding to the first word by the inverse document frequency. Then, for each first word, the calculation unit 12 calculates the importance by summing the multiplied values ​​corresponding to the first word.

[0038] The calculation of importance will be described with reference to Fig. 3. In Fig. 3, a case will be described in which documents A and B are each the first document information D1. Fig. 3 is a diagram for explaining an example of the calculation of importance.

[0039] In Figure 3, the content of document A is "Mr. A, please prepare a mid-term plan. In the mid-term plan, please provide a detailed explanation of the technology strategy.", and the content of document B is "Mr. B, please prepare a product proposal. However, please also help with creating the mid-term plan."

[0040] First, when the extraction unit 11 performs natural language processing on documents A and B, six first words are extracted from document A: "Mr. A," "mid-term plan," "set up," "technical strategy," "details," and "explain," and seven first words are extracted from document B: "Mr. B," "product proposal," "set up," "however," "mid-term plan," "create," and "help."

[0041] Next, the calculation unit 12 calculates the word frequency and the inverse document frequency for each of the first words. For example, when the word frequencies for "mid-term plan" and "technical strategy" are calculated, the results are as shown in FIG.

[0042] For "medium-term plan," there are six first words in document A and two "medium-term plan," so the word frequency for "medium-term plan" is 2 / 6 = 0.33. Similarly, there are seven first words in document B and one "medium-term plan," so the word frequency for "medium-term plan" is 1 / 7 = 0.14.

[0043] For "technology strategy," there are six first words in document A and one "technology strategy," so the word frequency for "technology strategy" is 1 / 6 = 0.17. Similarly, there are seven first words in document B and zero "technology strategy," so the word frequency for "technology strategy" is 0 / 7 = 0.00.

[0044] Furthermore, when the inverse document frequency of "medium-term plan" and "technology strategy" between documents A and B is calculated, the results are as shown in Figure 3.

[0045] For "medium-term plan", the total number of first document information items is 2 because the target first document information items are documents A and B, and "medium-term plan" is included in both documents A and B, so the number of first document information items is 2. Therefore, the inverse document frequency for "medium-term plan" is "Log(2 / 2)+1=1.0".

[0046] Furthermore, for "technology strategy," the total number of first document information items is two, in documents A and B, and since "technology strategy" is only included in document A, the number of first document information items is one. Therefore, the inverse document frequency for "technology strategy" is "Log(2 / 1)+1=1.3."

[0047] Next, the calculation unit 12 calculates a multiplication value by multiplying the word frequency corresponding to the first word by the inverse document frequency for each piece of first document information. When the multiplication value is calculated for "Mid-term plan" and "Technology strategy," the result is as shown in FIG.

[0048] For "medium-term plan," the word frequency in document A is 0.33, and the inverse document frequency between documents A and B is 1.0, so the multiplication value for "medium-term plan" in document A is "0.33 x 1.0 = 0.33." Similarly, the word frequency in document B is 0.14, and the inverse document frequency between documents A and B is 1.0, so the multiplication value for "medium-term plan" in document A is "0.14 x 1.0 = 0.14."

[0049] For "technology strategy," the word frequency in document A is 0.17, and the inverse document frequency between documents A and B is 1.2, so the multiplication value for "technology strategy" in document A is "0.17 x 1.3 = 0.22." Similarly, the word frequency in document B is 0.00, and the inverse document frequency between documents A and B is 1.3, so the multiplication value for "technology strategy" in document B is "0.00 x 1.3 = 0.00."

[0050] After that, the calculation unit 12 sums up the multiplication values ​​corresponding to the first words for each piece of first document information. When the sums for "mid-term plan" and "technology strategy" are calculated, the results are as shown in FIG.

[0051] For "medium-term plan", the word frequency in document A is 0.33 and the word frequency in document B is 0.14, so the total value for "medium-term plan" is "0.33 + 0.14 = 0.47".

[0052] For "technology strategy," the word frequency in document A is 0.22 and the word frequency in document B is 0.00, so the total value for "technology strategy" is "0.22+0.00=0.22."

[0053] The importance may be calculated using, for example, TF-IDF (Term Frequency-Inverse Document Frequency), but the importance may also be calculated using other methods.

[0054] ●About the decision section The determination unit 13 first classifies the plurality of first document information D1 according to the type of document. For example, as shown in Fig. 4, the first document information is classified into "highly confidential", "confidential", etc. Fig. 4 is a diagram for explaining an example of keyword extraction.

[0055] Next, the determination unit 13 arranges the first words in order of importance for each of the classifications "highly confidential" and "confidential," as shown in Fig. 4. In the example of Fig. 4, the first words "medium-term plan," "technical strategy," "design document," "organizational chart," "investment," "product proposal," etc. that appear in the first document information D1 classified as "highly confidential" are arranged in order of importance. Also, the first words "confidential," "product proposal," "sales strategy," "technical meeting," "design document," "organizational chart," etc. that appear in the first document information D1 classified as "confidential" are arranged in order of importance.

[0056] Thereafter, the determination unit 13 extracts second words (keywords) that are not included in other categories for each of the top six categories (pre-set upper range of importance), namely, "Important Confidential" and "Confidential." In the example of Fig. 4, "Mid-term Plan," "Technology Strategy," "Investment," "Confidential," "Sales Strategy," and "Technology Meeting" are extracted as second words (keywords).

[0057] However, "design document," "organization chart," and "product proposal" are included in other categories and are therefore not extracted as second words (keywords).The upper range is determined, for example, by experiments, simulations, etc.

[0058] ●About the generation part The generation unit 14 uses the extracted second words (keywords) to generate input data D2 (prompt) corresponding to the input format of the generation AI. The prompt is second document information (generated training data) used for training the document classification AI.

[0059] The prompt is, for example, an instruction to generate second document information D3 (specifications, presentation document, email document) using the second word (keyword) described above. Specifically, the following (prompt example 1: for creating a highly confidential document) and (prompt example 2: for creating a confidential document) are possible. The prompt may be input as text, or the text may be converted into voice and input.

[0060] (Prompt example 1) Please create a presentation document based on the following keywords. "Mid-term plan," "technology strategy," and "investment"

[0061] (Prompt example 2) Please create an email document based on the following keywords: "Confidential," "Sales Strategy," "Technical Meeting"

[0062] ●About information processing devices (generative AI) The information processing device 20 acquires input data D2 (prompt) from the generation unit 14, and when the input data D2 is input to the generation AI, outputs second document information D3 (generated learning data) to be used for training the document classification AI.

[0063] ●About information processing devices (document classification AI) The information processing device 30 acquires first document information D1 (original learning data) and second document information D3 (generated learning data), and inputs the first document information D1 (original learning data) and the second document information D3 (generated learning data) into a document classification AI for learning.

[0064] If the information processing device 30 has already completed learning using the first document information D1 (generated learning data), the information processing device 30 uses the second document information D3 to train the document classification AI.

[0065] In this way, by training the document classification AI using the second document information D3 in addition to the first document information D1 as training data, the accuracy of document classification by the document classification AI can be improved.

[0066] Next, the operation phase will be described using FIG. 5. FIG. 5 is a diagram for explaining the operation of the operation phase. In the operation phase, when third document information D4 is input to the trained document classification AI, a classification result D5 of the third document information D4 is output. For example, the result obtained is that the third document information D4 is classified into documents such as "highly confidential," "confidential," and "general." However, the classification is not limited to "highly confidential," "confidential," and "general."

[0067] [Device operation] Next, the operation of the information processing device in the embodiment will be described with reference to FIG. 6. FIG. 6 is a diagram for explaining an example of the operation of the information processing device (learning data generation device for document classification AI). In the following description, the diagram will be referenced as appropriate. Furthermore, in the embodiment, an information processing method (learning data generation method for document classification AI) is implemented by operating the information processing device 10. Therefore, the description of the information processing method in the embodiment will be replaced by the following description of the operation of the information processing device 10.

[0068] First, the extraction unit 11 performs natural language processing on first document information D1 (original learning data) stored in a storage device to be used for training the document classification AI, and extracts first words (target words) from each piece of first document information (step A1).

[0069] Specifically, in step A1, the extraction unit 11 first acquires first document information D1 (original learning data) from the storage device. Next, in step A1, the extraction unit 11 performs natural language processing on each of the acquired first document information D1 (original learning data) to extract, for each piece of first document information D1, a target first word contained in the first document information D1.

[0070] Next, the calculation unit 12 calculates the importance of each of the first words based on the extracted first words (step A2).

[0071] Specifically, in step A2, the calculation unit 12 first calculates the appearance frequency (word frequency) of each extracted first word for each piece of first document information D1. Also in step A2, the calculation unit 12 calculates the rarity (inverse document frequency) of the appearance of the first word among the pieces of first document information D1.

[0072] Furthermore, in step A2, the calculation unit 12 calculates the importance of each first word based on the word frequency and the inverse document frequency. Specifically, for each piece of first document information D1, the calculation unit 12 multiplies the word frequency corresponding to the first word by the inverse document frequency. Then, for each first word, the calculation unit 12 calculates the importance by summing the multiplied values ​​corresponding to the first word.

[0073] Next, the determination unit 13 classifies the first document information by type of sentence, arranges the first words in order of importance for each classification, and extracts second words (keywords) for each classification within a predetermined top range of importance that are not included in other classifications (step A3).

[0074] Specifically, in step A3, the determination unit 13 first classifies the multiple pieces of first document information D1 by document type. Next, in step A3, the determination unit 13 sorts the first words for each classification in descending order of importance. Thereafter, in step A3, the determination unit 13 extracts, for each classification, second words (keywords) within a predetermined upper range of importance that are not included in other classifications. However, second words (keywords) that are included in other classifications are not extracted.

[0075] Next, the generation unit 14 uses the extracted second words to generate input data (prompts) corresponding to the input format of the generation AI in order to generate second document information (generated learning data) to be used for training the document classification AI using the generation AI (step A4).

[0076] Specifically, in step A4, first, the generation unit 14 uses the extracted second word (keyword) to generate input data D2 (prompt) corresponding to the input format of the generation AI.

[0077] Next, the information processing device 20 acquires input data D2 (prompt) and inputs the input data D2 to the generation AI, which then outputs second document information D3 (generated learning data) to be used for training the document classification AI (step A5).

[0078] Next, the information processing device 30 acquires the first document information D1 (original learning data) and the second document information D3 (generated learning data), and inputs the first document information D1 (original learning data) and the second document information D3 (generated learning data) into the document classification AI for learning (step A6).

[0079] If the information processing device 30 has already completed learning using the first document information D1 (generated learning data), the information processing device 30 uses the second document information D3 to train the document classification AI.

[0080] [Effects of the embodiment] As described above, according to the embodiment, the document classification AI can be trained using the second document information D3 in addition to the first document information D1 as training data, thereby improving the accuracy of document classification by the document classification AI.

[0081] [program] The program in the embodiment may be any program that causes each information processing device 10 (computer) to execute steps A1 to A4 shown in Figure 6. By installing and executing this program on a computer, the device for generating training data for a document classification AI and the method for generating training data for a document classification AI in the embodiment can be realized. In this case, the processor of the information processing device 10 (computer) functions as an extraction unit 11, a calculation unit 12, a determination unit 13, and a generation unit 14 to perform processing.

[0082] The program in the embodiment may be executed by a computer system constructed by a plurality of computers, in which case, for example, each computer may function as one of the extraction unit 11, calculation unit 12, determination unit 13, and generation unit 14.

[0083] 6 is a process of the information processing device (generation AI) 20. The process of step A6 shown in Fig. 6 is a process of the information processing device (document classification AI) 30. Furthermore, if the information processing device 10 includes the information processing device (document classification AI) 30, the process of step A6 is performed using a processor of the information processing device 10 (computer).

[0084] [Physical configuration] Here, a computer that realizes a learning data generation device for a document classification AI by executing a program in the embodiment will be described with reference to Fig. 7. Fig. 7 is a diagram for explaining an example of a computer that realizes an information processing device (learning data generation device for a document classification AI) in the embodiment.

[0085] 7, the computer 110 includes a CPU (Central Processing Unit) 111, a main memory 112, a storage device 113, an input interface 114, a display controller 115, a data reader / writer 116, and a communication interface 117. These components are connected to each other via a bus 121 so as to be able to communicate data with each other. Note that the computer 110 may include a GPU or an FPGA in addition to or instead of the CPU 111.

[0086] The CPU 111 loads a program in the embodiment, which is composed of a group of codes and stored in the storage device 113, into the main memory 112 and executes each code in a predetermined order to perform various calculations. The main memory 112 is typically a volatile storage device such as a DRAM (Dynamic Random Access Memory).

[0087] The program in the embodiment is provided in a state stored in a computer-readable recording medium 120. The program in the embodiment may be distributed over the Internet connected via the communication interface 117.

[0088] Specific examples of the storage device 113 include a hard disk drive and a semiconductor storage device such as a flash memory. The input interface 114 mediates data transmission between the CPU 111 and input devices 118 such as a keyboard and a mouse. The display controller 115 is connected to a display device 119 and controls the display on the display device 119.

[0089] The data reader / writer 116 mediates data transmission between the CPU 111 and the recording medium 120, reads programs from the recording medium 120, and writes processing results from the computer 110 to the recording medium 120. The communication interface 117 mediates data transmission between the CPU 111 and other computers.

[0090] Specific examples of the recording medium 120 include general-purpose semiconductor storage devices such as CF (Compact Flash (registered trademark)) and SD (Secure Digital), magnetic recording media such as flexible disks, or optical recording media such as CD-ROMs (Compact Disk Read Only Memory).

[0091] Note that the information processing device 10 in the embodiment can be realized not by a computer on which a program is installed, but by hardware corresponding to each unit, for example, an electronic circuit. Furthermore, the information processing device 10 may be partially realized by a program and the remaining unit by hardware. In the embodiment, the computer is not limited to the computer shown in FIG. 7.

[0092] [Note] The following supplementary notes are further provided with respect to the above-described embodiments. Some or all of the above-described embodiments can be expressed by (Supplementary Note 1) to (Supplementary Note 15) described below, but are not limited to the following descriptions.

[0093] (Appendix 1) an extraction unit that performs natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence) and extracts first words from each of the first document information; a calculation unit that calculates the importance of each of the first words based on the extracted first words; a determination unit that classifies the first document information by document type, arranges the first words in order of importance for each classification, and determines, for each classification, second words that are not included in other classifications within a predetermined high range of importance; a generation unit that uses the extracted second words to generate input data corresponding to an input format of a generation AI (Artificial Intelligence) in order to generate second document information to be used for learning of the document classification AI; An information processing device having the above.

[0094] (Appendix 2) the calculation unit calculates a word frequency representing the frequency of appearance of the first word and an inverse document frequency representing the rarity of appearance of the first word among the first document information, and calculates the importance of each of the first words based on the word frequency and the inverse document frequency; 10. The information processing device according to claim 1.

[0095] (Appendix 3) The generation AI acquires the input data from the generation unit, and when the input data is input to the generation AI, outputs second document information to be used for learning of the document classification AI. 10. The information processing device according to claim 1.

[0096] (Appendix 4) the document classification AI acquires the first document information and the second document information, and inputs the first document information and the second document information into the document classification AI to learn; 4. The information processing device according to claim 3.

[0097] (Appendix 5) If the document classification AI has already learned using the first document information, it acquires the second document information and inputs the second document information into the document classification AI for learning. 4. The information processing device according to claim 3.

[0098] (Appendix 6) The information processing device performing natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence), and extracting first words from each of the first document information; Calculating the importance of each of the first words based on the extracted first words; classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; In order to generate second document information to be used for learning the document classification AI using a generation AI (Artificial Intelligence), input data corresponding to an input format of the generation AI is generated using the extracted second words. Information processing methods.

[0099] (Appendix 7) In calculating the importance, a word frequency representing the frequency of appearance of the first word and an inverse document frequency representing the rarity of appearance of the first word among the first document information are calculated, and the importance of each of the first words is calculated based on the word frequency and the inverse document frequency. 1. The information processing method described in Appendix 6.

[0100] (Appendix 8) The generation AI acquires the input data, and when the input data is input to the generation AI, outputs second document information to be used for learning of the document classification AI. 1. The information processing method described in Appendix 6.

[0101] (Appendix 9) the document classification AI acquires the first document information and the second document information, and inputs the first document information and the second document information into the document classification AI to learn; 10. The information processing method according to claim 8.

[0102] (Appendix 10) If the document classification AI has already learned using the first document information, it acquires the second document information and inputs the second document information into the document classification AI for learning. 10. The information processing method according to claim 8.

[0103] (Appendix 11) On the computer, performing natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence), and extracting first words from each of the first document information; Calculating the importance of each of the first words based on the extracted first words; classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; In order to generate second document information to be used for learning the document classification AI using a generation AI (Artificial Intelligence), input data corresponding to an input format of the generation AI is generated using the extracted second words. A program that executes a process.

[0104] (Appendix 12) The computer, In calculating the importance, a word frequency representing the frequency of appearance of the first word and an inverse document frequency representing the rarity of appearance of the first word among the first document information are calculated, and the importance of each of the first words is calculated based on the word frequency and the inverse document frequency. 12. The program according to claim 11, which causes the processing to be executed.

[0105] (Appendix 13) The generation AI acquires the input data, and when the input data is input to the generation AI, outputs second document information to be used for learning of the document classification AI. 12. The program described in Appendix 11.

[0106] (Appendix 14) the document classification AI acquires the first document information and the second document information, and inputs the first document information and the second document information into the document classification AI to learn; 13. The program described in Appendix 13.

[0107] (Appendix 15) If the document classification AI has already learned using the first document information, it acquires the second document information and inputs the second document information into the document classification AI for learning. 13. The program described in Appendix 13.

[0108] Although the present invention has been described above with reference to the embodiments, the present invention is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. [Industrial Applicability]

[0109] According to the above description, it is possible to generate training data for improving the document classification accuracy of a document classification AI, and the present invention is useful in the field of document classification AI. [Explanation of symbols]

[0110] 10, 20, 30 Information processing device 11 Extraction part 12 Calculation section 13 Decision Section 14 Generation part 100 systems 110 Computer 111 CPU 112 main memory 113 Storage device 114 Input Interface 115 Display Controller 116 Data Reader / Writer 117 Communication Interface 118 Input Devices 119 Display Device 120 Recording Media 121 Bus

Claims

1. an extraction means for extracting first words from each piece of first document information by performing natural language processing on the first document information to be used for learning document classification AI (Artificial Intelligence); a calculation means for calculating the importance of each of the first words based on the extracted first words; a determining means for classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; a generation means for generating input data corresponding to an input format of a generation AI (Artificial Intelligence) using the extracted second words in order to generate second document information to be used for learning the document classification AI; An information processing device having the above.

2. the calculation means calculates a word frequency representing the frequency of occurrence of the first word and an inverse document frequency representing the rarity of occurrence of the first word among the first document information, and calculates the importance of each of the first words based on the word frequency and the inverse document frequency; The information processing device according to claim 1 .

3. The generation AI acquires the input data from the generation means, and when the input data is input to the generation AI, outputs second document information to be used for learning the document classification AI. The information processing device according to claim 1 .

4. The document classification AI acquires the first document information and the second document information, and inputs the first document information and the second document information into the document classification AI to perform learning. The information processing device according to claim 3 .

5. When the document classification AI has already learned using the first document information, the document classification AI acquires the second document information and inputs the second document information into the document classification AI for learning. The information processing device according to claim 3 .

6. The information processing device performing natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence), and extracting first words from each of the first document information; Calculating the importance of each of the first words based on the extracted first words; classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; In order to generate second document information to be used for learning the document classification AI using a generation AI (Artificial Intelligence), input data corresponding to an input format of the generation AI is generated using the extracted second words. Information processing methods.

7. On the computer, performing natural language processing on first document information to be used for learning document classification AI (Artificial Intelligence), and extracting first words from each of the first document information; Calculating the importance of each of the first words based on the extracted first words; classifying the first document information by document type, arranging the first words in order of importance for each classification, and determining, for each classification, second words that are not included in other classifications within a predetermined high range of importance; In order to generate second document information to be used for learning the document classification AI using a generation AI (Artificial Intelligence), input data corresponding to an input format of the generation AI is generated using the extracted second words. A program that executes a process.

Images