Information processing device, information processing method, and program

Abstract

To solve the problem in which: conventionally, it is not easy to determine whether a patent is important or not.SOLUTION: An information processing apparatus A is provided with an enterprise information storage unit 11 that stores enterprise information that is related to business carried out by an enterprise, and that includes the set of sentences; an association degree acquisition unit 31 that acquires the degree of association between the enterprise information and patent information for each of one or more pieces of patent information of the enterprise: a determination unit 32 that determines one or more pieces of patent information in which the degree of association satisfies an extraction condition; and a selected patent output unit 41 that outputs selected patent information related to the one or more pieces of patent information determined by the determination unit 32. The information processing apparatus can determine an important patent by using the enterprise information related to the business carried out by the enterprise.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an information processing apparatus and the like for determining patent information related to corporate information. Note that patent information related to corporate information is usually information on patents strongly related to important businesses of a company.

Background Art

[0002] Conventionally, various technologies related to patent analysis have existed. For example, there has been a technology for automatically assigning F terms using important words in patent documents (see Non-Patent Document 1). Also, there has been a technology for automatically extracting information necessary for a technical survey from patents by giving it as a type of clue expression, such as expressions like "can do" and "is possible" (see Non-Patent Document 2).

[0003] In addition, there has been a technology for treating related patent search as a binary classification problem (see Non-Patent Document 3). Also, there has been a technology for extracting the actions and effects of inventions in patent documents (see Non-Patent Document 4). Furthermore, there has been a technology for a similar patent search method focusing on the functional expressions of patents (see Non-Patent Document 5).

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

[0005] However, in the prior art, it was not easy to determine whether a company's patents were important to that company. [Means for solving the problem]

[0006] The first information processing device of this invention comprises: a corporate information storage unit that stores corporate information including a set of sentences, which is information relating to the business conducted by a company; a relevance acquisition unit that acquires the degree of relevance between the corporate information and the patent information for each of the one or more patent information of a company; a determination unit that determines one or more patent information whose degree of relevance satisfies the extraction conditions; and a selected patent output unit that outputs selected patent information relating to the one or more patent information determined by the determination unit.

[0007] This configuration allows for the determination of important patents using corporate information related to the business activities conducted by a company.

[0008] Furthermore, the information processing device of the second invention, compared to the first invention, is an information processing device comprising: a relevance acquisition unit comprising: a means for acquiring corporate important information which acquires one or more corporate important information which is important information that satisfies corporate important conditions from corporate information; a means for acquiring patent important information which is important information that satisfies patent important conditions for each patent information using one or more patent information; and a relevance acquisition means for acquiring the degree of relevance between corporate information and patent information for each patent information using one or more corporate important information and one or more patent important information for each patent information.

[0009] This configuration allows for the determination of important patents using corporate information related to the business activities conducted by a company.

[0010] Furthermore, in contrast to the second invention, the information processing device of this third invention is characterized in that the corporate information is corporate financial report information, and the means for acquiring important corporate information is an information processing device that acquires performance factor statements from the financial report information.

[0011] With this configuration, important patents can be identified using financial statements.

[0012] Furthermore, the information processing device of the fourth invention, compared to the third invention, is an information processing device in which the patent information includes one or more patent classification codes, the patent important information acquisition means acquires one or more patent classification codes contained in the patent information for each piece of patent information, and acquires descriptive information containing one or more descriptive terms corresponding to one or more patent classification codes from a code dictionary having one or more pairs of patent classification codes and descriptive information, the relevance acquisition means acquires distributed representations of one or more descriptive terms acquired by the patent important information acquisition means for each piece of patent information, acquires a patent vector which is a representative value of the distributed representations, the corporate important information acquisition means acquires distributed representations of one or more terms contained in the performance factor statement acquired, acquires a business vector which is a representative value of the distributed representations of one or more terms, and calculates a relevance score which is the similarity between the patent vector and the business vector for each piece of patent information.

[0013] With this configuration, important patents can be identified using financial statements.

[0014] Furthermore, the information processing device of this fifth invention, in contrast to the second invention, is an information processing device in which the corporate information is press release article information of a company's product announcement.

[0015] With this configuration, important patents can be determined using product announcement press releases.

[0016] Furthermore, the information processing device of the sixth invention, compared to the fifth invention, is an information processing device in which the corporate important information acquisition means acquires one or more corporate important sentences contained in press release article information, the patent important information acquisition means acquires one or more patent important sentences containing sentences in the section on the effects of the invention contained in the patent information, the relevance acquisition means acquires a business vector which is a distributed representation using one or more corporate important sentences, and acquires a patent vector which is a distributed representation using one or more patent important sentences, and calculates a relevance score which is the similarity between the patent vector and the business vector for each piece of patent information.

[0017] With this configuration, important patents can be determined using product announcement press releases.

[0018] Further, the information processing apparatus of the seventh invention further includes a model storage unit that stores a learning model constructed by learning processing of machine learning, with the sentences in the item of the effect of the invention as positive examples and the sentences in the items other than the item of the effect of the invention as negative examples with respect to the sixth invention. The enterprise important information acquisition means applies the learning model to each of one or more sentences included in the press release article information, performs prediction processing of machine learning, and acquires one or more enterprise important sentences corresponding to positive examples from the one or more sentences by the prediction processing.

[0019] With such a configuration, important patents can be determined using product release press releases.

[0020] Further, the information processing apparatus of the eighth invention further includes a value information acquisition unit that acquires value information regarding the value of patent information corresponding to the selected patent information with respect to the first invention. The output unit is an information processing apparatus that outputs value information instead of or in addition to the selected patent information.

[0021] With such a configuration, value information regarding the value of a patent can be acquired using enterprise information regarding the business conducted by an enterprise.

[0022] Further, the information processing apparatus of the ninth invention, with respect to the eighth invention, the value information acquisition unit is an information processing apparatus that acquires numerical information which is a numerical value regarding the performance of an enterprise from enterprise information and acquires value information using the numerical information.

[0023] With such a configuration, appropriate value information of a patent can be acquired using enterprise information regarding the business conducted by an enterprise.

[0024] Further, in the information processing apparatus of the tenth invention, with respect to the ninth invention, the value information acquisition unit acquires first numerical information, which is a numerical value related to the performance corresponding to one business category from enterprise information, acquires second numerical information, which is a numerical value related to the overall performance of the enterprise's business, uses the first numerical information and the second numerical information to acquire third numerical information, which is a ratio indicating the contribution of the enterprise in the business category, and uses the third numerical information to acquire value information.

[0025] With such a configuration, appropriate value information of a patent can be acquired by using enterprise information related to the business conducted by an enterprise.

[0026] Further, in the information processing apparatus of the eleventh invention, with respect to the eighth invention, the enterprise information is the enterprise's settlement short message information, the enterprise important information acquisition means acquires a performance factor sentence from the settlement short message information, acquires a performance result sentence corresponding to the performance factor sentence, and the value information acquisition unit acquires numerical information, which is a numerical value related to the performance of the enterprise, from the performance result sentence, and uses the numerical information to acquire value information.

[0027] With such a configuration, the value information of a patent can be acquired by using the settlement short message.

[0028] Further, in the information processing apparatus of the twelfth invention, with respect to the eleventh invention, the value information acquisition unit acquires numerical information, which is sales, operating profit or net profit, from the performance result sentence, and uses the numerical information to acquire value information.

[0029] With such a configuration, the value information of a patent can be acquired by using the settlement short message.

Effect of the Invention

[0030] According to the information processing apparatus of the present invention, important patents can be determined by using enterprise information related to the business conducted by an enterprise.

Brief Description of the Drawings

[0031] [Figure 1]Block diagram of information processing device A in Embodiment 1 [Figure 2] Diagram showing the code explanation management table. [Figure 3] A flowchart illustrating an example of the operation of the information processing device A. [Figure 4] A flowchart illustrating an example of the process for acquiring important corporate information. [Figure 5] A flowchart illustrating an example of the same decision-making process. [Figure 6] A flowchart illustrating an example of the learning process. [Figure 7] A flowchart illustrating an example of the process for obtaining correlation scores. [Figure 8] A flowchart illustrating an example of the process for obtaining important information related to the patent. [Figure 9] A flowchart illustrating a detailed example of the process for obtaining relevance scores. [Figure 10] A flowchart illustrating an example of the process for obtaining equivalent value information. [Figure 11] A diagram showing the financial results summary information. [Figure 12] This diagram illustrates the overview of the processing performed by the information processing device A. [Figure 13] A diagram showing an example of the same important sentence. [Figure 14] A diagram showing an example of the same important sentence. [Figure 15] Overview of the computer system [Figure 16] Block diagram of the computer system [Modes for carrying out the invention]

[0032] The embodiments of the information processing device, etc., will be described below with reference to the drawings. In the embodiments, components that are denoted by the same reference numerals perform the same operation, and therefore, their description may be omitted again.

[0033] (Embodiment 1) This embodiment describes an information processing device that acquires the degree of relevance between company information and the company's patent information, and uses this degree of relevance to determine important patents. The company information includes, for example, financial statements and product announcement press releases.

[0034] Furthermore, in this embodiment, important information is extracted from company information and important information is obtained using patent information. Next, an information processing device that calculates the similarity between important pieces of information and determines important patents using this similarity will be described.

[0035] Furthermore, in this embodiment, we will describe an information processing device that acquires and outputs value information of important patents.

[0036] In this embodiment, the association of information X with information Y means that information Y can be obtained from information X, or information X can be obtained from information Y, and the method of association is not limited. Information X and information Y may be linked, may exist in the same buffer, may information X be contained in information Y, or information Y may be contained in information X, and so on.

[0037] Figure 1 is a block diagram of the information processing device A in this embodiment. The information processing device A comprises a storage unit 1, a receiving unit 2, a processing unit 3, and an output unit 4. The storage unit 1 comprises a company information storage unit 11, a patent information storage unit 12, a code dictionary 13, and a model storage unit 14. The processing unit 3 comprises a relevance acquisition unit 31, a determination unit 32, and a value information acquisition unit 33. The relevance acquisition unit 31 comprises a company important information acquisition means 311, a patent important information acquisition means 312, and a relevance acquisition means 313. The output unit 4 comprises a selected patent output unit 41 and a value information output unit 42.

[0038] The storage unit 1, which constitutes the information processing device A, stores various types of information. These types of information include, for example, company information, patent information, code description information, and learning models, which will be described later.

[0039] The corporate information storage unit 11 stores one or more pieces of corporate information. Corporate information refers to information about the business conducted by a company. Corporate information includes a set of one or more sentences. Examples of corporate information include financial statements, securities reports, press releases, company web pages, and newspaper articles. Financial statements refer to information that includes financial statements. Financial statements are documents disclosed in accordance with the rules of the stock exchange, and are currently required to be disclosed within 45 days of the closing date. Securities reports refer to information that includes securities reports. Press releases refer to articles of press releases issued by a company. Examples of press releases refer to articles of press releases announcing a company's products. A company web page is the company's homepage. Newspaper articles refer to newspaper articles that contain information about the business conducted by a company. Financial statements, securities reports, press releases, and newspaper articles are, for example, files, but their structure is not specified.

[0040] The patent information storage unit 12 stores one or more patent information entries. Patent information refers to information about patents. One patent information entry usually refers to information about one patent. Preferably, the patent information in the patent information storage unit 12 is a patent in which the company corresponding to the company information is the applicant or rights holder. The patent here may be a registered patent, a patent in the process of being filed, etc. The status of the patent is irrelevant. The patent information usually includes a patent identifier. A patent identifier is information that identifies a patent. Examples of patent identifiers include the registration number, application number, publication number, or a reference number or ID assigned by the company. Preferably, the patent information includes the claims and specification, but may also include only an abstract, only drawings, or only one or more patent classification codes. In other words, preferably, the patent information includes a document describing the content of the invention. However, the patent information may also include only the patent identifier and patent classification codes.

[0041] The code dictionary 13 stores one or more code description information. Each code description information is a pair of a patent classification code and a description.

[0042] Patent classification codes include, for example, IPC codes, F-terms, FI codes, and CPC codes. The type and depth of the branching of the patent classification code are not restricted. If the patent classification code is an IPC, it can be limited to sections only, up to the main class (or class), up to the subclass, up to the main group, up to the subgroup, etc. However, it is preferable for the IPC patent classification code to include everything from sections to subgroups.

[0043] Descriptive information refers to a sentence that explains the meaning of a patent classification code or one or more terms that correspond to a patent classification code. A pair of a patent classification code and descriptive information is shown, for example, in Figure 2. Figure 2 is a code description management table. The code description management table manages two or more records that have "patent classification code," "depth," and "descriptive information."

[0044] The model storage unit 14 stores one or more learning models. A learning model is information used in the prediction process of machine learning. A learning model can also be called a learner, classifier, or classification model.

[0045] A learning model is, for example, a model for determining important information (e.g., important sentences) from a set of one or more sentences.

[0046] The important sentences here are, for example, sentences that describe the effects of the invention. The learning model was constructed using machine learning, with sentences included in the tag "Effects of the Invention" in the patent document as positive examples and sentences in items other than "Effects of the Invention" (for example, embodiments) as negative examples.

[0047] This learning process is to be performed by a learning unit (not shown). The learning unit may be provided by information processing device A or by another device. Furthermore, a device equipped with a learning unit has a storage unit that stores one or more positive examples and one or more negative examples. In addition, techniques for obtaining positive examples, which are sentences contained in the tag "Effects of the Invention," from patent documents (e.g., published patent gazettes, patent gazettes), and techniques for obtaining negative examples, which are sentences in items other than the "Effects of the Invention" item (e.g., embodiments), are known technologies.

[0048] Furthermore, the machine learning algorithm can be anything from deep learning, random forest, decision tree, SVM, etc. In addition, various machine learning functions and existing libraries can be used, such as the TensorFlow library, the R language's random forest module, fastText, and TinySVM.

[0049] Reception Unit 2 receives various instructions and information. These instructions and information include, for example, start instructions, company information, patent information, code description information, and learning models.

[0050] Any means of inputting instructions and information is acceptable, such as a touch panel, keyboard, mouse, or menu screen.

[0051] The processing unit 3 performs various processes. These processes include, for example, those performed by the relevance acquisition unit 31, the determination unit 32, and the value information acquisition unit 33.

[0052] The relevance acquisition unit 31 acquires the degree of relevance between company information and patent information for each of the company's one or more patent information entries. The degree of relevance is, for example, the similarity described later. The degree of relevance may also be a numerical value based on the number of identical terms in the company information and the patent information, or the proportion of identical terms. In such cases, generally, the greater the number and proportion of identical terms, the higher the degree of relevance.

[0053] The similarity is, for example, the similarity between a business vector and a patent vector. A business vector is a vector constructed using company information. A patent vector is a vector constructed using patent information. A vector is a set of two or more elements (information), and its structure is not specified. Furthermore, a company's patent information refers to information on patents in which the company is the applicant or rights holder.

[0054] The corporate important information acquisition means 311 acquires one or more corporate important information items that satisfy the corporate important criteria from the corporate information storage unit 11.

[0055] Corporate critical conditions are the criteria that determine whether a company's information qualifies as corporate critical conditions. For example, corporate critical conditions are the criteria for obtaining performance factor statements (described later) from company information. For example, corporate critical conditions are the criteria for obtaining performance result statements (described later) from company information. For example, corporate critical conditions are the criteria for obtaining effect statements (described later) from company information.

[0056] A key corporate condition is, for example, matching a cue expression used to retrieve a sentence that corresponds to key corporate information. Another key corporate condition is, for example, using a learning model to retrieve sentences that correspond to key corporate information, performing a machine learning prediction process and obtaining a prediction result that it corresponds to a positive example (that it is key corporate information).

[0057] The corporate important information acquisition means 311, for example, acquires one or more corporate important information items that match a clue expression from corporate information that includes one or more sentences.

[0058] The corporate important information acquisition means 311, for example, uses a learning model to perform machine learning prediction processing on each of the one or more sentences contained in the corporate information, obtains a prediction result of whether or not it corresponds to a positive example (that it is corporate important information), and obtains one or more corporate important information corresponding to the prediction result that it corresponds to a positive example.

[0059] Furthermore, the learning model shown here is a model obtained by a learning unit (not shown) that performs machine learning training using one or more positive example sentences and one or more negative example sentences.

[0060] The corporate information acquisition means 311, for example, acquires performance factor statements from the financial results summary information in the corporate information storage unit 11. Performance factor statements are statements that describe the reasons why a company's performance was strong or why a company's performance was poor.

[0061] The corporate important information acquisition means 311 acquires performance factor statements by, for example, one of the following two processes. (1) When using clues

[0062] The corporate information acquisition means 311 acquires performance factor sentences, which are sentences containing performance factor cue expressions, from corporate information containing one or more sentences. A performance factor cue expression is a cue expression used to acquire performance factor sentences. Examples of performance factor cue expressions are "is performing well" and "is performing poorly."

[0063] Furthermore, it is preferable that the performance factor cues be information automatically collected from one or more financial statements using a known bootstrap method. (2) When using a learning model

[0064] The corporate information acquisition means 311 uses a learning model and each sentence of corporate information containing one or more sentences to perform machine learning prediction processing and obtain prediction results for each sentence. The prediction results are information indicating whether or not a sentence is a performance factor sentence (whether or not it is a positive example).

[0065] Furthermore, the learning model here is a model obtained by performing machine learning training on a learning unit (not shown in the diagram) using one or more performance factor sentences as positive examples and one or more non-performance factor sentences as negative examples. A non-performance factor sentence is a sentence that is not a performance factor sentence.

[0066] Furthermore, while BERT is a preferred machine learning algorithm here, it goes without saying that other algorithms and modules mentioned above can also be used.

[0067] The corporate information acquisition means 311 acquires, for example, one or more corporate important sentences contained in the press release article information of the corporate information storage unit 11. Here, corporate important sentences are sentences written in the press release article, such as effect statements. An effect statement is a sentence that describes the effects of the technology.

[0068] The corporate information acquisition means 311 acquires important sentences from a press release article by, for example, the following two processes: (1) When using a learning model

[0069] The corporate information acquisition means 311 applies a learning model to each of the one or more sentences in the press release article information of the corporate information storage unit 11, performs machine learning prediction processing to determine whether one or more sentences are positive examples (corporate important sentences), and acquires one or more corporate important sentences that correspond to positive examples.

[0070] The learning model described here is, for example, a model obtained by a learning unit (not shown) that uses one or more sentences in the "[Effects of the Invention]" section of the specification of one or more patent information as positive examples, and one or more sentences in sections other than the "[Effects of the Invention]" section (e.g., examples, embodiments of the invention) as negative examples, and performs machine learning training. The learning model can be any model obtained by performing machine learning training using one or more corporate-important sentences as positive examples and one or more non-corporate-important sentences as negative examples; the method of obtaining the training data (positive and negative examples) is not specified. The training data may be, for example, sentences created manually, or it may be positive examples containing cue expressions (e.g., "can be done," "improved") and negative examples not containing cue expressions from one or more sentences in one or more press release articles.

[0071] Furthermore, while BERT is a preferred machine learning algorithm here, it goes without saying that other algorithms and modules mentioned above can also be used. (2) When using clue expressions

[0072] The corporate important information acquisition means 311 acquires important sentences from each of the one or more sentences in the press release article information of the corporate information storage unit 11 that match the important sentence cue expression.

[0073] Key sentence clues are clues used to identify important sentences. Examples of key sentence clues include "can do" and "improved." Preferably, key sentence clues are information automatically collected from one or more press release articles or patent information using a known bootstrap method.

[0074] The corporate important information acquisition means 311 acquires one or more performance result statements that satisfy the corporate important criteria from the corporate information storage unit 11. The corporate important information acquisition means 311 acquires performance factor statements from the financial report information and acquires performance result statements corresponding to those performance factor statements.

[0075] A performance result statement corresponding to a performance factor statement is, for example, a set number of sentences that follow a performance factor statement and contain specific clue expressions. In this case, the specific clue expressions are "sales," "operating profit," "net profit," "ten thousand yen," and "hundred million yen."

[0076] Corporate critical conditions are the criteria for something to qualify as corporate critical information. For example, corporate critical conditions are the criteria for obtaining performance factor statements, as described later. For example, corporate critical conditions are the criteria for obtaining performance result statements.

[0077] A performance results statement is a document that contains information about the sales of the business segment to which the performance factors belong.

[0078] The corporate important information acquisition means 311 obtains performance result statements from corporate information, for example, using performance factor cue expressions. A performance factor cue expression is a cue expression used to obtain performance factor statements. Examples of performance factor cue expressions are "sales," "operating profit," "net profit," and "ten thousand yen" and "hundred million yen." The corporate important conditions here are, for example, "contains any of "sales," "operating profit," or "net profit" AND contains either "ten thousand yen" or "hundred million yen." That is the case.

[0079] The corporate information acquisition means 311 extracts a predetermined number of sentences (for example, the last 5 sentences) following the performance factor sentence with the highest similarity to the target patent information. Next, the corporate information acquisition means 311 performs a keyword search using a specific keyword (for example, "ten thousand yen" or "hundred million yen") as the key, and extracts sentences containing that keyword as performance result sentences. If no sentences containing the keyword exist, the corporate information acquisition means 311 performs a keyword search on a predetermined number of sentences following the next most similar performance factor sentence, and extracts sentences containing that keyword as performance result sentences. If no sentences containing the keyword exist, the process is repeated based on the next most similar performance factor sentence until a sentence containing the keyword is found.

[0080] The patent material information acquisition means 312 uses one or more pieces of patent information to acquire one or more pieces of patent material information for each piece of patent information, which are pieces of important information that satisfy the patent material conditions.

[0081] Patent materiality refers to the conditions for something to qualify as patent materiality. For example, patent materiality may be a descriptive term corresponding to a patent classification code. For example, patent materiality may be a sentence in the section describing the effects of the invention in the specification of the patent information. For example, patent materiality may be a sentence containing a specific clue expression. A specific clue expression is, for example, "by doing so, it is possible to do so."

[0082] The patent important information acquisition means 312, for example, acquires one or more patent classification codes contained in each of the one or more patent information items in the patent information storage unit 12, and acquires descriptive information containing one or more descriptive terms corresponding to each of the one or more patent classification codes from the code dictionary 13. In such cases, one or more descriptive information items or one or more descriptive terms constitute patent important information.

[0083] The patent information acquisition means 312 acquires one or more patent information sentences, each containing a sentence describing the effects of the invention, for each of the one or more patent information items in the patent information storage unit 12. In such cases, one or more patent information sentences or terms contained in one or more patent information sentences constitute the patent information.

[0084] The relevance acquisition means 313 uses one or more corporate important information and one or more patent important information to acquire the degree of relevance between corporate information and patent information for each piece of patent information. The degree of relevance may also be the degree of similarity between the corporate information and the patent information.

[0085] The relevance acquisition means 313 acquires the degree of relevance between company information and patent information for each piece of patent information, using one or more patent-related important information and one or more company-related important information.

[0086] Below, we will explain two specific examples of how to obtain the degree of relevance between company information and patent information, for cases where the company information is from financial statements and for cases where it is from press release articles. Note that methods other than the two described below are also acceptable for obtaining the degree of relevance between company information and patent information. (A) When the company information is financial results summary information (1) When using distributed representations

[0087] The relevance acquisition means 313 obtains, for example, one or more distributed representations of each descriptive term obtained by the patent important information acquisition means 312 for each piece of patent information. Next, the relevance acquisition means 313 obtains, for example, a patent vector which is a representative value of the one or more distributed representations. The descriptive terms are preferably nouns, but they may also be independent words containing nouns, etc. Furthermore, the technique for obtaining distributed representations of descriptive terms is publicly known. Furthermore, the representative value of two or more distributed representations is, for example, a vector whose elements are the average values ​​of each element of the vector which is the two or more distributed representations. However, instead of the average value of each element, the median of each element may be used, etc.

[0088] Furthermore, the relevance acquisition means 313 obtains, for example, distributed representations of one or more terms included in the performance factor statement acquired by the corporate important information acquisition means 311. Next, the relevance acquisition means 313 obtains, for example, a business vector which is a representative value of the distributed representations of the one or more terms.

[0089] Next, the relevance acquisition means 313 calculates a relevance score for each piece of patent information, which is the similarity between the patent vector and the business vector. Note that the technique for acquiring the similarity between the two vectors is publicly known. (2) When using the same number or proportion of terms

[0090] The relevance acquisition means 313, for example, obtains the number of identical terms among the one or more descriptive terms obtained by the patent important information acquisition means 312 and the one or more terms included in the performance factor statement obtained by the corporate important information acquisition means 311, for each piece of patent information. Alternatively, the relevance acquisition means 313 may use the number of identical terms to obtain the proportion of identical terms.

[0091] Next, the relevance acquisition means 313 calculates the relevance for each piece of patent information using, for example, an increasing function with parameters such as the number of identical terms or the proportion of identical terms. The relevance acquisition means 313 may also acquire the relevance corresponding to the number of identical terms or the proportion of identical terms from a correspondence table. The correspondence table has two or more correspondence pieces, each having a range for the number of identical terms or the proportion of identical terms and a relevance score. (B) When the company information is from a press release article. (1) When using distributed representations

[0092] The relevance acquisition means 313 acquires a business vector, which is a distributed representation using one or more corporate important sentences acquired by the corporate important information acquisition means 311. Note that the technique for acquiring distributed representations of sentences is publicly known.

[0093] Furthermore, the relevance acquisition means 313 acquires a patent vector, which is a distributed representation using one or more patent-important sentences acquired by the patent-important information acquisition means 312.

[0094] Next, the relevance acquisition means 313 calculates, for example, a relevance score, which is the similarity between the patent vector and the business vector for each piece of patent information. (2) When using the same number or proportion of terms

[0095] The relevance acquisition means 313 detects identical terms between the terms contained in one or more corporate important sentences acquired by the corporate important information acquisition means 311 and the terms contained in one or more patent important sentences acquired by the patent important information acquisition means 312, and obtains the number of such identical terms. The relevance acquisition means 313 may also use the number of identical terms to obtain the proportion of identical terms.

[0096] Next, the relevance acquisition means 313 calculates the relevance for each piece of patent information using, for example, an increasing function with parameters such as the number of identical terms or the proportion of identical terms.

[0097] The determination unit 32 determines one or more patent information items whose relevance scores, obtained by the relevance acquisition unit 31, satisfy the extraction conditions.

[0098] Extraction criteria are the conditions used to determine which patent information to select. For example, an extraction criterion might be that the relevance is above a certain threshold, or that the relevance is greater than a certain threshold.

[0099] The value information acquisition unit 33 acquires value information for each of the one or more selected patent information determined by the determination unit 32. Value information refers to information relating to the value of the patent corresponding to the patent information. While it is preferable for the value information to be information indicating quantitative value, it may also be information indicating the presence or absence of value, information indicating the value rank, or information indicating the value category.

[0100] The value information acquisition unit 33 acquires value information by, for example, one of the following two methods. (1) When using numerical information corresponding to company information

[0101] The value information acquisition unit 33 acquires numerical information from company information and uses that numerical information to acquire value information. The value information acquisition unit 33 typically acquires numerical information that shows a large number and value information that indicates a high value.

[0102] Numerical information refers to figures related to a company's performance. Examples of numerical information include sales figures, operating profit figures, and net profit figures.

[0103] More specifically, the value information acquisition unit 33 acquires numerical information from, for example, performance result statements, and uses that numerical information to acquire value information. Typically, the value information acquisition unit 33 acquires value information that indicates a higher value the larger the numerical information.

[0104] The value information acquisition unit 33 acquires numerical information from, for example, the performance result statement acquired by the corporate important information acquisition means 311, and uses this numerical information to acquire value information. The numerical information is information obtained by concatenating the number immediately preceding the term "ten thousand yen" or "hundred million yen" in the performance result statement with "ten thousand yen" or "hundred million yen".

[0105] Next, the value information acquisition unit 33 acquires value information, for example, by using an increasing function that takes the acquired numerical information as a parameter. (2) When using ratios

[0106] The value information acquisition means 332 acquires first numerical information, which is a numerical value relating to performance corresponding to a business category, from the company information. The first numerical information is the numerical information in (1).

[0107] Furthermore, the value information acquisition means 332 acquires second numerical information, which is numerical data relating to the overall business performance of the company. This second numerical information may include, for example, the company's total sales and total profit.

[0108] The value information acquisition means 332 acquires second numerical information from the performance result statement, for example, using a cue expression. Here, the cue expression is, for example, "Total sales.*(billion yen|ten thousand yen)" and "Total profit.*(billion yen|ten thousand yen)". The value information acquisition means 332 acquires second numerical information, which is a string formed by concatenating the number immediately preceding the term "ten thousand yen" or "billion yen" with "ten thousand yen" or "billion yen", for example, using the cue expression.

[0109] The value information acquisition means 332 may, for example, acquire second numerical information corresponding to the company identifier of a company from a company database (not shown). The company identifier is, for example, the company name or company ID (for example, securities code or stock code).

[0110] Next, the value information acquisition means 332 uses, for example, the first numerical information and the second numerical information to acquire a third numerical information, which is the percentage representing the company's contribution to the business category.

[0111] Next, the value information acquisition means 332 acquires value information, for example, using third numerical information. The value information acquisition means 332 acquires value information of higher value the larger the third numerical information is. The value information acquisition means 332 calculates value information, for example, using an increasing function with the third numerical information as a parameter. (3) When using the degree of relevance

[0112] The value information acquisition unit 33 acquires value information, for example, using an increasing function that takes as a parameter the maximum relevance acquired by the relevance acquisition means 313. The value information acquisition unit 33 acquires value information of higher value the greater the relevance.

[0113] Output unit 4 outputs various types of information. These types of information include, for example, selected patent information and value information.

[0114] Here, "output" is a concept that includes display on a screen, projection using a projector, printing with a printer, sound output, transmission to an external device, storage on a recording medium, and transfer of processing results to other processing devices or other programs.

[0115] The selected patent output unit 41 outputs selected patent information relating to one or more patent information determined by the determination unit 32. The selected patent information is, for example, the patent number and the application number. The selected patent output unit 41 passes the selected patent information to, for example, the value information acquisition unit 33.

[0116] The value information output unit 42 outputs the value information acquired by the value information acquisition unit 33. It is preferable for the value information output unit 42 to output the value information in association with the selected patent information.

[0117] The storage unit 1, the corporate information storage unit 11, the patent information storage unit 12, the code dictionary 13, and the model storage unit 14 are preferably made of non-volatile recording media, but can also be made of volatile recording media.

[0118] The process by which information is stored in the storage unit 1, etc. is not relevant. For example, information may be stored in the storage unit 1, etc. via a recording medium, information transmitted via a communication line, etc. may be stored in the storage unit 1, etc., or information input via an input device may be stored in the storage unit 1, etc.

[0119] The reception unit 2 can be implemented using device drivers for input means such as touch panels and keyboards, or control software for menu screens.

[0120] The processing unit 3, relevance acquisition unit 31, determination unit 32, value information acquisition unit 33, corporate important information acquisition means 311, patent important information acquisition means 312, relevance acquisition means 313, category determination means 331, and value information acquisition means 332 can typically be implemented using a processor, memory, etc. The processing procedures of the processing unit 3, etc., are typically implemented in software, and this software is recorded on a recording medium such as ROM. However, it may also be implemented in hardware (dedicated circuitry). The processor can be a CPU, MPU, GPU, etc., and the type is not limited.

[0121] The output unit 4 may or may not include the selected patent output unit 41, the value information output unit 42, and output devices such as a display or speakers. The output unit 4, etc., can be implemented using driver software for the output device, or the driver software for the output device and the output device, etc.

[0122] Next, an example of the operation of the information processing device A will be explained using the flowchart in Figure 3. Note that the following processes are initiated, for example, upon receipt of a start command.

[0123] (Step S301) The corporate important information acquisition means 311 acquires one or more corporate information from the corporate information storage unit 11.

[0124] (Step S302) The corporate important information acquisition means 311 acquires corporate important information from one or more corporate information acquired in step S301. An example of such corporate important information acquisition process will be explained using the flowchart in Figure 4.

[0125] (Step S303) The relevance acquisition unit 31 assigns 1 to counter i.

[0126] (Step S304) The relevance acquisition unit 31 determines whether the i-th patent information exists in the patent information storage unit 12. If the i-th patent information exists, the process proceeds to step S305; otherwise, the process proceeds to step S308.

[0127] (Step S305) The relevance acquisition unit 31 acquires the i-th patent information from the patent information storage unit 12.

[0128] (Step S306) The relevance acquisition unit 31 performs the relevance acquisition process. An example of the relevance acquisition process will be explained using the flowchart in Figure 7.

[0129] (Step S307) The relevance acquisition unit 31 increments the counter i by 1. Return to step S304.

[0130] (Step S308) The determination unit 32 determines one or more patent information whose relevance scores obtained by the relevance acquisition unit 31 satisfy the extraction conditions. The determined patent information is the selected patent information and may be referred to as the selected patent information as appropriate.

[0131] (Step S309) The value information acquisition unit 33 determines whether or not to output value information for the selected patent information. If value information is to be output, the unit proceeds to step S310; otherwise, it proceeds to step S315. Note that whether or not to output value information is predetermined, for example.

[0132] (Step S310) The relevance acquisition unit 31 assigns 1 to counter j.

[0133] (Step S311) The value information acquisition unit 33 determines whether or not the j-th selected patent information exists among the selected patent information acquired in step S308. If the j-th selected patent information exists, the process proceeds to step S312; otherwise, the process ends.

[0134] (Step S312) The value information acquisition unit 33 performs a process to acquire value information for the j-th selected patent information. An example of such value information acquisition process will be explained using the flowchart in Figure 10.

[0135] (Step S313) The value information output unit 42 outputs the value information acquired in step S312. It is preferable for the value information output unit 42 to output the value information in pairs with the corresponding selected patent information.

[0136] (Step S314) The relevance acquisition unit 31 increments counter j by 1. Return to step S311.

[0137] (Step S315) The selected patent output unit 41 outputs one or more selected patent information obtained in step S308. The process ends.

[0138] Next, an example of the corporate information acquisition process in step S302 will be explained using the flowchart in Figure 4.

[0139] (Step S401) The corporate important information acquisition means 311 assigns 1 to counter i.

[0140] (Step S402) The corporate important information acquisition means 311 determines whether or not the i-th sentence exists among the one or more corporate information obtained in step S302. If the i-th sentence exists, the process proceeds to step S403; otherwise, it returns to the higher-level process.

[0141] (Step S403) The corporate important information acquisition means 311 determines whether the i-th sentence satisfies the corporate important conditions. An example of such determination process will be explained using the flowchart in Figure 5.

[0142] (Step S404) If the company meets the important criteria in step S403, proceed to step S404; otherwise, proceed to step S406.

[0143] (Step S405) The corporate important information acquisition means 311 temporarily stores the i-th sentence in a buffer (not shown). The i-th sentence is corporate important information. The i-th sentence is, for example, a performance factor sentence.

[0144] (Step S406) The corporate important information acquisition means 311 increments counter i by 1. Return to step S402.

[0145] Next, an example of the decision-making process in step S403 will be explained using the flowchart in Figure 5.

[0146] (Step S501) The corporate important information acquisition means 311 acquires a learning model from the model storage unit 14. This learning model is a binary classification model for determining whether a sentence is corporate important information or not.

[0147] (Step S502) The corporate important information acquisition means 311 acquires the sentence to be inspected (the i-th sentence in Figure 4).

[0148] (Step S503) The corporate important information acquisition means 311 provides the learning model and the sentence to be inspected to a machine learning prediction processing module and executes the module.

[0149] (Step S504) The corporate important information acquisition means 311 returns the prediction result from step S503 as a return value to the higher-level processing.

[0150] Furthermore, in step S503 of the flowchart in Figure 5, as described above, the corporate important information acquisition means 311 may use a cue expression (for example, a performance factor cue expression) to determine whether or not the sentence to be examined is corporate important information.

[0151] Next, we will explain an example of the learning process used to construct the learning model used in the process shown in Figure 5, using the flowchart in Figure 6.

[0152] (Step S601) The learning unit (not shown) assigns 1 to counter i.

[0153] (Step S602) The learning unit determines whether or not the i-th sentence exists in the set of sentences to be learned. If the i-th sentence exists, proceed to step S603; otherwise, proceed to step S608.

[0154] (Step S603) The learning unit retrieves the i-th sentence from the set of sentences to be learned.

[0155] (Step S604) The learning unit determines whether the i-th sentence meets the company's important criteria. If it meets the company's important criteria, proceed to step S605; otherwise, proceed to step S606.

[0156] (Step S605) The learning unit associates the i-th sentence with a positive example flag. Proceed to Step S607.

[0157] (Step S606) The learning unit associates a negative example flag with the i-th sentence.

[0158] (Step S607) The learning unit increments counter i to 1.

[0159] (Step S608) The learning unit uses one or more sentences corresponding to the positive example flag as positive examples and one or more sentences corresponding to the negative example flag as negative examples, and performs machine learning training to obtain a learning model.

[0160] (Step S609) The learning unit stores the learning model acquired in step S608. The process ends.

[0161] Next, an example of the relevance acquisition process in step S306 will be explained using the flowchart in Figure 7.

[0162] (Step S701) The relevance acquisition means 313 acquires corporate important information to be used for acquiring the relevance. This corporate important information is the information acquired in step S302.

[0163] (Step S702) The patent important information acquisition means 312 performs a process to acquire patent important information of the target patent. An example of such patent important information acquisition process will be explained using the flowchart in Figure 8.

[0164] (Step S703) The relevance acquisition means 313 uses the corporate important information acquired in step S701 and the patent important information acquired in step S702 to acquire the degree of relevance between the corporate information and the patent information. It then returns to the higher-level processing. An example of the details of acquiring such relevance will be explained using the flowchart in Figure 9. This process is referred to as the relevance detail processing.

[0165] Next, an example of the patent information acquisition process in step S702 will be explained using the flowchart in Figure 8.

[0166] (Step S801) The patent important information acquisition means 312 acquires one or more patent classification codes from the patent information.

[0167] (Step S802) The patent important information acquisition means 312 assigns 1 to counter i.

[0168] (Step S803) The patent important information acquisition means 312 determines whether or not the i-th patent classification code exists. If the i-th patent classification code exists, the process proceeds to step S804; otherwise, it returns to the higher-level process.

[0169] (Step S804) The patent important information acquisition means 312 acquires one or more descriptive pieces of information corresponding to the i-th patent classification code from the code dictionary 13 and temporarily stores them in a buffer (not shown). The descriptive information here is patent important information.

[0170] (Step S805) The patent important information acquisition means 312 increments counter i by 1. Return to step S803.

[0171] In addition, in the flowchart of Figure 8, the patent important information acquisition means 312 may acquire one or more sentences included in the [Effects of the Invention] section of the patent information as patent important information.

[0172] Next, an example of the relevance detail processing in step S703 will be explained using the flowchart in Figure 9.

[0173] (Step S801) The relevance acquisition means 313 acquires one or more terms from the corporate important information. These terms may be all terms (words) contained in the corporate important information, or they may be only terms that satisfy predetermined conditions. The conditions may be, for example, that the term is a noun and that the tf / idf value is above a threshold.

[0174] (Step S802) The relevance acquisition means 313 assigns 1 to counter i.

[0175] (Step S803) The relevance acquisition means 313 determines whether or not the i-th term exists among the terms acquired in step S801. If the i-th term exists, the process proceeds to step S804; otherwise, the process proceeds to step S806.

[0176] (Step S804) The relevance acquisition means 313 acquires the distributed representation of the i-th term. The distributed representation is acquired by, for example, Word2Vec, but the acquisition method is not limited.

[0177] (Step S805) The relevance acquisition means 313 increments the counter i by 1. Return to step S803.

[0178] (Step S806) The relevance acquisition means 313 acquires a representative value of one or more distributed representations acquired in step S804. The vector which is the representative value of such distributed representations is the business vector.

[0179] (Step S807) The relevance acquisition means 313 acquires one or more terms from the patent material information. These terms may be all terms (words) contained in the corporate material information, or they may be only terms that satisfy predetermined conditions. For example, the condition is that the term is a noun.

[0180] (Step S808) The relevance acquisition means 313 assigns 1 to counter j.

[0181] (Step S809) The relevance acquisition means 313 determines whether or not the j-th term exists among the terms acquired in step S807. If the j-th term exists, proceed to step S810; otherwise, proceed to step S812.

[0182] (Step S810) The relevance acquisition means 313 obtains the distributed representation of the j-th term.

[0183] (Step S811) The relevance acquisition means 313 increments the counter j by 1. Return to step S809.

[0184] (Step S812) The relevance acquisition means 313 acquires a representative value of one or more distributed representations acquired in step S810. The vector which is such a representative value of distributed representations is a patent vector.

[0185] (Step S813) The relevance acquisition means 313 calculates the similarity between the business vector obtained in step S806 and the patent vector obtained in step S812. It then returns to the higher-level processing.

[0186] Note that the similarity referred to here is the degree of relevance. Furthermore, the relevance acquisition means 313 typically associates the degree of relevance with the corporate key information and corporate information that formed the basis of the business vector, and with the patent key information and patent information that formed the basis of the patent vector.

[0187] In the flowchart of Figure 8, if the corporate important information consists of one or more sentences, the relevance acquisition means 313 may obtain a distributed representation from those one or more sentences and use that distributed representation as the business vector. Similarly, if the patent important information consists of one or more sentences, the relevance acquisition means 313 may obtain a distributed representation from those one or more sentences and use that distributed representation as the patent vector. The relevance acquisition means 313 may also calculate the similarity between the business vector and the patent vector.

[0188] An example of the value information acquisition process in step S312 will be explained using the flowchart in Figure 10.

[0189] (Step S1001) The value information acquisition unit 33 acquires one or more performance factor sentences corresponding to the selected patent information. Each of the one or more performance factor sentences is associated with a degree of relevance.

[0190] (Step S1002) The value information acquisition unit 33 sorts the performance factor sentences in descending order using relevance as the key.

[0191] (Step S1003) The value information acquisition unit 33 assigns 1 to counter i.

[0192] (Step S1004) The value information acquisition unit 33 determines whether or not the i-th performance factor statement exists. If the i-th performance factor statement exists, the unit proceeds to step S1005; otherwise, it proceeds to step S1014.

[0193] (Step S1005) The value information acquisition unit 33 acquires one or more candidate performance result sentences corresponding to the i-th performance factor sentence.

[0194] (Step S1006) The value information acquisition unit 33 assigns 1 to counter j.

[0195] (Step S1007) The value information acquisition unit 33 determines whether or not the j-th sentence exists among the sentences acquired in step S1005. If the j-th sentence exists, the unit proceeds to step S1008; otherwise, it proceeds to step S1013.

[0196] (Step S1008) The value information acquisition unit 33 determines whether the j-th statement satisfies the condition. If the j-th statement satisfies the condition, the process proceeds to step S1009; otherwise, it proceeds to step S1012. The condition is the condition for acquiring the performance result statement.

[0197] (Step S1009) The value information acquisition unit 33 acquires the j-th sentence and temporarily stores it in a buffer (not shown). The j-th sentence is a performance result sentence corresponding to a performance factor sentence.

[0198] (Step S1010) The value information acquisition unit 33 acquires numerical information from the performance result statement acquired in step S1009.

[0199] (Step S1011) The value information acquisition unit 33 acquires value information using the numerical information acquired in step S1010. The value information acquisition unit 33 associates the value information with the patent information. The process returns to the higher-level processing.

[0200] (Step S1012) The value information acquisition unit 33 increments counter j by 1. Return to step S1007.

[0201] (Step S1013) The value information acquisition unit 33 increments counter i by 1. Return to step S1004.

[0202] (Step S1014) The value information acquisition unit 33 substitutes information indicating "unknown" into the value information. The value information acquisition unit 33 associates the value information with the patent information. It returns to the higher-level processing.

[0203] Note that the method for obtaining the performance results statement in the flowchart of Figure 10 is not specified. In the flowchart of Figure 10, it is sufficient to obtain the numerical information contained in the performance results statement and then use that numerical information to obtain value information.

[0204] Furthermore, in the flowchart of Figure 10, the process of obtaining performance result statements may be performed by the corporate important information acquisition means 311.

[0205] The following describes specific examples of the operation of the information processing device A in this embodiment. Two specific examples will be described. Specific example 1 is the case where the company information is financial results summary information. Specific example 2 is the case where the company information is product announcement press release information.

[0206] (Specific example 1) Let's assume that the corporate information storage unit 11 of the information processing device A currently stores the financial results summary information of "AAA Corporation" as shown in Figure 11.

[0207] Furthermore, the patent information storage unit 12 stores numerous patent information for which "AAA Corporation" is the rights holder or applicant. The patent information includes one or more IPC codes and the title of the invention.

[0208] Furthermore, the code dictionary 13 is assumed to store a set of pairs of patent classification codes and explanatory information, as shown in Figure 2.

[0209] Furthermore, the model storage unit 14 stores a learning model that extracts performance factor sentences from financial report information. The learning model is a model for determining whether each of the one or more sentences in the financial report information is a positive example (performance factor sentence).

[0210] In the above situation, suppose, for example, that the user inputs an operation start command to the information processing device A. Then, the receiving unit 2 receives the operation start command.

[0211] Then, through the process described above, the corporate important information acquisition means 311 of the information processing device A acquires the performance factor statement from the financial report information in Figure 11: "For the domestic food business "AAA Corporation," sales increased compared to the same period of the previous year, due to strong performance of major products such as the designated quasi-drug "○○ Tablets" and the supplement "XXX Natura," as well as the diet support food "BBB Up Slim" and the nutritional supplement "CCC Bar."

[0212] Furthermore, through the processing described above, the corporate information acquisition means 311 obtains the performance results statement "As a result of the above, although sales in the food business were affected by the Great East Japan Earthquake, sales increased by 4.5% year-on-year to 47.633 billion yen, thanks to the efforts of each group company to strengthen the brands of their main products." from the financial results summary information in Figure 11.

[0213] Furthermore, the patent important information acquisition means 312 acquires one or more IPC codes from each of the two or more patent pieces of information. Then, for each of the two or more patent pieces of information, the patent important information acquisition means 312 acquires one or more terms corresponding to each of the one or more IPC codes from the code dictionary 13. Here, it is assumed that the patent important information acquisition means 312 acquires all the terms in the explanatory information of each level of the hierarchical IPC code. In other words, the IPC code is subgroup If it is "A01B 1 / 02", the patent important information acquisition means 312 acquires all terms corresponding to the above nodes (main group, subclass, class, section) of "A01B 1 / 02": "daily necessities, agriculture, forestry, livestock, hunting, capture, fishing, agriculture, forestry, soil work, agricultural machinery, tools, parts, details, accessories in general, hand tools, plows, shovels".

[0214] Next, the relevance acquisition means 313 obtains a number of nouns from the performance factor sentences, which are important company information. Next, the relevance acquisition means 313 obtains the distributed representation of each noun. Next, the relevance acquisition means 313 obtains a business vector, which is a vector of the average values ​​of the distributed representations of each noun.

[0215] Furthermore, the relevance acquisition means 313 acquires distributed representations of one or more terms acquired by the patent important information acquisition means 312 for each of the two or more patent information. Next, the relevance acquisition means 313 acquires a patent vector, which is a vector of the average values ​​of the distributed representations of each term.

[0216] Next, the relevance acquisition means 313 calculates the similarity between the business vector and the patent vector for each of the two or more patent information items. Then, the relevance acquisition means 313 associates this similarity with the patent information.

[0217] Next, the determination unit 32 obtains the patent information corresponding to the highest relevance. Here, let's assume that the determination unit 32 has obtained the patent information "<Title of Invention> Blood Acetaldehyde Reducing Agent". It is preferable that the patent information includes a patent number and application number.

[0218] Next, the value information acquisition unit 33 acquires the numerical information "47,633 million yen" from the performance result statement acquired by the corporate important information acquisition means 311, which matches the numerical information condition ("a string of characters that constitutes a number including "ten thousand yen" and "billion yen"). The numerical information condition is the condition for acquiring numerical information.

[0219] Furthermore, the Value Information Acquisition Unit 33 acquires AAA Corporation's total sales of "651,661 million yen" for the fiscal year ending December 2023 from the financial results summary information in Figure 11.

[0220] Next, the value information acquisition unit 33 acquires the percentage (7.3%) of the numerical information "47.633 billion yen" relative to the total sales of "651.661 million yen".

[0221] Next, the value information acquisition unit 33 calculates value information (V) using an increasing function with the said ratio as a parameter.

[0222] Next, the selected patent output unit 41 outputs the patent information "<Title of Invention> Blood Acetaldehyde Reducing Agent". The value information output unit 42 stores the value information (V) calculated by the value information acquisition unit 33, associating it with the patent information.

[0223] In summary, as shown in Specific Example 1, important patents could be selected from the financial statements. Furthermore, the value of those patents could be obtained from the financial statements.

[0224] (Specific example 2) In Specific Example 2, we will use Figure 12, which illustrates the overview of the processing of information processing device A, to explain the process. Currently, the storage unit 1 of information processing device A stores a large amount of patent information for constructing a learning model to extract important sentences from patent information and press release article information. In this case, there are 12,000 patent entries. Furthermore, the patent information here includes patent specifications.

[0225] The learning unit (not shown) of the information processing device A then retrieves sentences from the "[Effects of the Invention]" section of each of the 12,000 patent specifications. Such sentences are positive examples. The learning unit also retrieves sentences from the "[Examples]" (embodiments) section of each of the 12,000 patent specifications. Such sentences are negative examples. In other words, here, sentences constituting the effects of the invention are important information, and sentences constituting the examples are non-important information. Next, the learning unit uses the positive and negative example sentences as training data (1201 in Figure 12) to perform machine learning training, obtain a trained model, and store it in the model storage unit 14. The trained model created by this machine learning training is the so-called BERT (Bidirectional Encoder Representations from Transformers) (1202 in Figure 12).

[0226] Furthermore, the corporate information storage unit 11 of the information processing device A stores press release article information about S Company's cameras.

[0227] Furthermore, the patent information storage unit 12 stores one or more patent information entries for S Company's patents. The patent information includes [effects of the invention].

[0228] In the above situation, suppose, for example, that the user inputs an operation start command to the information processing device A. Then, the receiving unit 2 receives the operation start command.

[0229] Then, through the process described above, the corporate important information acquisition means 311 uses a learning model (BERT) to perform machine learning prediction processing on each sentence of the press release article information to determine whether or not it is an important sentence. The corporate important information acquisition means 311 then obtains one or more important sentences 1203 from the press release article information. The important sentences 1203 are, for example, each sentence in Figure 13.

[0230] Furthermore, the patent important information acquisition means 312 uses a learning model (BERT) to perform machine learning prediction processing on each sentence in each of the one or more patent pieces of patent information to determine whether or not it is an important sentence. The patent important information acquisition means 312 then acquires one or more important sentences 1204 from each of the one or more patent pieces of patent information. The important sentences 1204 are, for example, the sentences in Figure 14.

[0231] Note that, in this context, the learning model used to extract key sentences from press release articles and the learning model used to extract key sentences from patent information are the same.

[0232] Next, the relevance acquisition means 313 acquires a distributed representation 1205 of a document containing two or more sentences, as shown in Figure 13. Such a distributed representation is a business vector. The relevance acquisition means 313 also acquires a distributed representation 1206 of a document containing one or more sentences (for example, Figure 14) for each piece of patent information. Such a distributed representation is a patent vector.

[0233] Next, the relevance acquisition means 313 calculates the similarity between the business vector 1205 and the patent vector 1206 for each piece of patent information (1207). This similarity is the relevance.

[0234] Based on the above, we were able to obtain the degree of relevance of each patent information item to press release article information.

[0235] Next, the determination unit 32 determines patent information whose relevance satisfies the extraction criteria. Here, the extraction criteria are that the relevance is equal to or greater than a threshold, and the relevance rank is 5th or higher. The determination unit 32 then determines five pieces of patent information.

[0236] Next, the selected patent output unit 41 outputs the patent number or application number associated with each of the five patent pieces of information determined by the determination unit 32.

[0237] Next, the value information acquisition unit 33 acquires the patent classification code (for example, the IPC code) for each of the five patent pieces. Next, for each of the five patent pieces, the value information acquisition unit 33 acquires descriptive information corresponding to the patent classification code from the code dictionary 13. Next, for each of the five patent pieces, the value information acquisition unit 33 performs machine learning prediction processing using the acquired descriptive information and the business field learning model to acquire the business field. Next, for each of the five patent pieces, the value information acquisition unit 33 acquires numerical information (for example, sales) that is paired with the business field from the sales management table.

[0238] The business field learning model is a learning model obtained by performing machine learning training using two or more training data sets that have descriptive information containing one or more terms and a business field. The business field learning model is stored in the model storage unit 14.

[0239] Furthermore, the sales management table is a table that manages the sales revenue for each business area of ​​Company S. The sales management table is a table that manages two or more records that have a business area and sales revenue. Assume that the sales management table is stored in storage unit 1.

[0240] Next, the value information acquisition unit 33 acquires value information for each of the five patent information items using an increasing function that uses the acquired numerical information (for example, sales revenue) as a parameter.

[0241] Next, the value information output unit 42 outputs value information for each of the five patent information entries. For example, the value information output unit 42 outputs the patent number and value information in association for each of the five patent information entries.

[0242] As described above, according to this embodiment, important patents can be determined using corporate information related to the business conducted by a company.

[0243] Furthermore, according to this embodiment, important patents can be determined using financial statement information.

[0244] Furthermore, according to this embodiment, important patents can be determined using product announcement press release information.

[0245] Furthermore, according to this embodiment, value information regarding the value of a patent can be obtained using corporate information related to the business conducted by the company.

[0246] Furthermore, according to this embodiment, value information regarding the value of a patent can be obtained using financial statement information.

[0247] Furthermore, according to this embodiment, value information regarding the value of a patent can be obtained using product announcement press release information.

[0248] In this embodiment, the information processing device A may also be a server that outputs selected patent information or patent value information. In this case, the information processing device A receives, for example, company information (e.g., financial results summary information), determines one or more selected patent information items that are highly related to the company information, and outputs them. The information processing device A also outputs value information for one or more selected patent information items.

[0249] Furthermore, the processing in this embodiment may be implemented in software. This software may be distributed by software download or the like. Alternatively, this software may be recorded on a recording medium such as a CD-ROM and distributed. This also applies to other embodiments in this specification. The software that implements the information processing device A in this embodiment is the following program. In other words, this program is a program that causes a computer that can access a corporate information storage unit, which stores corporate information including a set of sentences relating to the business conducted by a company, to function as a correlation acquisition unit that acquires the correlation between the corporate information and the patent information for each of the one or more patent pieces of information of the company, a determination unit that determines one or more patent pieces of information whose correlation satisfies the extraction conditions, and a selected patent output unit that outputs selected patent information relating to the one or more patent pieces of information determined by the determination unit.

[0250] Figure 15 also shows the appearance of a computer that executes the program described herein to realize the various embodiments of the information processing device A described above. The embodiments described above can be realized with computer hardware and computer programs executed thereon. Figure 15 is an overview of this computer system 300, and Figure 16 is a block diagram of the system 300.

[0251] In Figure 15, the computer system 300 includes a computer 301 with a CD-ROM drive, a keyboard 302, a mouse 303, and a monitor 304.

[0252] In Figure 16, the computer 301 includes, in addition to the CD-ROM drive 3012, an MPU 3013, a bus 3014 connected to the CD-ROM drive 3012, a ROM 3015 for storing programs such as boot-up programs, a RAM 3016 connected to the MPU 3013 for temporarily storing application program instructions and providing temporary storage space, and a hard disk 3017 for storing application programs, system programs, and data. Although not shown here, the computer 301 may further include a network card for providing connectivity to a LAN.

[0253] The program that causes the computer system 300 to execute the functions of the information processing device A of the above embodiment may be stored on the CD-ROM 3101, inserted into the CD-ROM drive 3012, and then transferred to the hard disk 3017. Alternatively, the program may be transmitted to the computer 301 via a network (not shown) and stored on the hard disk 3017. The program is loaded into the RAM 3016 during execution. The program may also be loaded directly from the CD-ROM 3101 or the network.

[0254] The program does not necessarily have to include an operating system (OS) or third-party program that causes the computer 301 to execute the functions of the information processing device A of the above embodiment. The program only needs to include the instruction portion that calls the appropriate function (module) in a controlled manner and obtains the desired result. How the computer system 300 operates is well known, so a detailed explanation is omitted.

[0255] In the above program, steps such as sending information and receiving information do not include hardware-based processing, such as processing performed by a modem or interface card in the transmission step (processing that can only be performed by hardware).

[0256] Furthermore, the computer running the above program may be a single computer or multiple computers. In other words, it may perform centralized processing or distributed processing.

[0257] Furthermore, in each of the above embodiments, each process may be implemented by centralized processing by a single device, or by distributed processing by multiple devices.

[0258] It goes without saying that the present invention is not limited to the embodiments described above, and various modifications are possible, all of which are also included within the scope of the present invention. [Industrial applicability]

[0259] As described above, the information processing device A according to the present invention has the effect of being able to determine important patents using corporate information related to the business conducted by a company, and is useful as an information processing device, etc. [Explanation of symbols]

[0260] A Information Processing Device 1 Storage Unit 2. Reception Department 3 Processing Unit 4 Output section 11. Corporate Information Storage Unit 12 Patent Information Storage Unit 13 Code Dictionary 14 Model Storage Unit 31. Relevance Acquisition Section 32 Decision Section 33 Value Information Acquisition Unit 41 Output section of the selected patent 42 Value Information Output Unit 311 Means of obtaining important corporate information 312 Means for acquiring important patent information 313. Methods for obtaining relevance 331 Category Determination Methods 332 Means for acquiring value information

Claims

1. A corporate information storage unit stores corporate information that includes information about the business conducted by a company, corporate information that includes a set of sentences, and corporate information that includes press release articles about the company's product announcements. A correlation acquisition unit acquires the degree of correlation between the company information and the patent information for each of the one or more patent information of the said company, A determination unit that determines one or more patent information items that satisfy the extraction criteria in terms of relevance, The system comprises an output unit that outputs selected patent information relating to the one or more patent information determined by the determination unit, The aforementioned relevance acquisition unit, A means for acquiring corporate important information that acquires one or more pieces of corporate important information that satisfy corporate important criteria from the aforementioned corporate information, A means for acquiring patent important information, which uses the one or more patent information described above to acquire one or more patent important pieces of information for each of the patent information described above, which are important pieces of information that satisfy the patent material conditions, The system comprises a relevance acquisition means that uses the one or more corporate important information and the one or more patent important information for each of the patent information to acquire the degree of relevance between the corporate information and the patent information for each of the patent information, The aforementioned means for acquiring important corporate information is: Obtain one or more important company statements from the aforementioned press release article information, The aforementioned patent information acquisition means is Obtain one or more important patent sentences, including a sentence in the section describing the effects of the invention as contained in the aforementioned patent information. The aforementioned relevance acquisition means is An information processing device that obtains a business vector, which is a distributed representation using one or more corporate key sentences, obtains a patent vector, which is a distributed representation using one or more patent key sentences, and calculates a correlation score, which is the similarity between the patent vector and the business vector, for each piece of patent information.

2. The system further comprises a model storage unit in which a learning model constructed by machine learning processing is stored, with sentences in the section describing the effects of the invention as positive examples and sentences in sections other than the effects of the invention as negative examples. The aforementioned means for acquiring important corporate information is: The information processing apparatus according to claim 1, which applies the learning model to each of the one or more sentences contained in the press release article information, performs machine learning prediction processing, and obtains one or more important corporate sentences that correspond to positive examples from the one or more sentences by means of the prediction processing.

3. The system further comprises a value information acquisition unit that acquires value information relating to the value of patent information corresponding to the selected patent information. The output unit is, The information processing apparatus according to claim 1, which outputs the value information in place of, or in addition to, the selected patent information.

4. The aforementioned value information acquisition unit, The information processing apparatus according to claim 3, which obtains numerical information relating to a company's performance from the aforementioned company information and uses said numerical information to obtain said value information.

5. The aforementioned value information acquisition unit, The information processing apparatus according to claim 4, which obtains first numerical information, which is a numerical value relating to the performance of one business category, from the aforementioned company information; second numerical information, which is a numerical value relating to the overall business performance of the aforementioned company; third numerical information, which is a percentage indicating the contribution of the aforementioned business category to the company, using the first numerical information and the second numerical information; and third numerical information, which obtains the aforementioned value information, using the third numerical information.

6. A corporate information storage unit that stores corporate information which is information relating to the business conducted by a company, corporate information which includes a set of sentences, and corporate financial report information of a company, A correlation acquisition unit acquires the degree of correlation between the company information and the patent information for each of the one or more patent information of the said company, A determination unit that determines one or more patent information items that satisfy the extraction criteria in terms of relevance, An output unit that outputs selected patent information relating to the one or more patent information determined by the determination unit, The system comprises a value information acquisition unit that acquires value information relating to the value of patent information corresponding to the selected patent information, From the aforementioned financial report information, obtain the performance factor statement, and obtain the performance result statement corresponding to that performance factor statement. The aforementioned value information acquisition unit, An information processing device that obtains numerical information relating to a company's performance from the aforementioned performance results statement, and uses said numerical information to obtain the aforementioned value information.

7. The aforementioned value information acquisition unit, The information processing apparatus according to claim 6, which obtains numerical information such as sales revenue, operating profit, or net profit from the aforementioned performance results statement, and uses said numerical information to obtain said value information.

8. An information processing method for causing a computer to perform all the processing performed by the information processing device described in any one of Claims 1 to 7.

9. Computers, A program for causing an information processing device to function as described in any one of claims 1 to 7.

Images