Document creation support device, document creation support method, and computer program

Abstract

We support the creation of compelling job application documents. [Solution] A document creation support device 1 comprising: an annotation data storage unit 1B that stores text composed of predetermined decomposition granularities as annotated text by associating the semantic elements of the text; a correct answer information storage unit 1D that stores the decomposition granularity of the text, semantic elements, and correct answer content in association; a reception processing unit 11 that receives input of document data to be supported; an analysis processing unit 12 that decomposes the text contained in the document data into blocks of various decomposition granularities and associates the semantic elements associated with the annotated text with blocks that include at least a portion of the annotated text; an extraction processing unit 13 that extracts correct answer information from the correct answer information storage unit for each block that is associated with the same decomposition granularity and semantic elements as the block; and a generation processing unit 14 that calculates the difference between the extracted correct answer information and the block and generates correction information that includes information about the difference.

Description

Technical Field

[0001] The present invention relates to a document creation support device, a document creation support method, and a computer program.

Background Art

[0002] Conventionally, in situations such as job hunting, job changing, and recruitment, the job seeker side and the employer side have been conducting activities based on documents such as resumes and job offers. Such documents have been created by people with experience and know-how, or have been created based on the advice of such people.

[0003] In Patent Document 1, in a support system for creating sentences used in a specific document, an input means for acquiring sentence data input into a predetermined form for the specific document, and based on a legibility evaluation rule, analyzing the legibility of the input sentence to generate first evaluation score data and first evaluation comment data, and extracting at least one of the sentences, words, and symbols used in the input sentence, and based on a meaning determination rule, analyzing the suitability of the extracted sentence, word, and symbol to generate a second evaluation score and second evaluation comment data. A system equipped with a meaning analysis means has been proposed.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the system described in Patent Document 1 above, it only evaluates legibility based on a predetermined rule, and in employment-related documents such as resumes and job offers, how the content appeals to the other party is not evaluated.

[0006] The present invention aims to support the creation of employment-related documents, such as resumes and job postings, that are appealing to both recruiting companies and job seekers. [Means for solving the problem]

[0007] To achieve the above objective, a document creation support device according to one aspect of the present invention includes: an annotation data storage unit that stores text composed of a predetermined decomposition granularity as annotated text by associating the semantic elements of the text; a correct answer information storage unit that stores the decomposition granularity of the text, semantic elements, and correct answer content in association; a reception processing unit that receives input of document data to be supported; an analysis processing unit that decomposes the text contained in the document data into blocks of various decomposition granularities and associates semantic elements associated with the annotated text with each block, which includes at least a portion of the annotated text; an extraction processing unit that extracts correct answer information from the correct answer information storage unit for each block that is associated with the same decomposition granularity and semantic elements as the block; and a generation processing unit that calculates the difference between the extracted correct answer information and the block and generates correction information including information about the difference.

[0008] The correct answer information may include information relating to the amount of text to be written and the position of text on the document data, in association with the semantic elements, and the correction information may include information relating to the difference in the amount of text to be written or the difference in the position of text between the block and the correct answer information.

[0009] The generation processing unit may generate a document containing information relating to the difference and present the document as advice to the user terminal via an appropriate communication processing unit.

[0010] The system further includes an example sentence information storage unit that stores document description items constituting a document, semantic elements of example sentences, and example sentence text in association with each other, and the extraction processing unit may further extract the example sentence text associated with the same semantic elements as the block from the example sentence information storage unit and present the example sentence text to the user terminal via a communication processing unit.

[0011] At least a portion of the aforementioned correct answer information may be created based on annotated text obtained by analyzing other document data, and the correct answer information may be associated with performance information indicating whether or not the other document data is valid, and the reception processing unit may receive the specification of the performance information from the user as a search condition for extracting the aforementioned correct answer information, and the extraction processing unit may extract, for each block, the correct answer information that matches the specified performance information as the correct answer information to be compared and analyzed with the block, from the correct answer information storage unit.

[0012] A document creation support method according to another aspect of the present invention involves a computer having: an annotation data storage unit that stores information relating to semantic elements of text composed of a predetermined decomposition granularity as an annotated text; and a correct answer information storage unit that stores the decomposition granularity of text, semantic elements, and correct answer content in association with each other; a reception process that accepts input of document data to be supported; an analysis process that decomposes the text contained in the document data into blocks of various decomposition granularities and associates semantic elements associated with the annotated text with each block, which includes at least a portion of the annotated text; an extraction process that extracts correct answer information from the correct answer information storage unit for each block that is associated with the same decomposition granularity and semantic elements as the block; and a generation process that calculates the difference between the extracted correct answer information and the block and generates correction information including information regarding the difference.

[0013] A computer program according to yet another aspect of the present invention includes an annotated data storage unit that stores annotated text by associating information relating to semantic elements of text composed of a predetermined decomposition granularity with the text, and a correct answer information storage unit that stores the decomposition granularity of text, semantic elements, and correct answer content in association with the text, and causes the computer to perform an input reception process that accepts input of document data to be supported, an analysis process that decomposes the text contained in the document data into blocks of various decomposition granularities and associates semantic elements associated with the annotated text with each block which includes at least a part of the annotated text, an extraction process that extracts correct answer information from the correct answer information storage unit that is associated with the same decomposition granularity and semantic elements as the block for each block, and a generation process that calculates the difference between the extracted correct answer information and the block and generates correction information that includes information regarding the difference.

[0014] Computer programs can be provided by storing them on various data-readable storage media, or by making them available for download via networks such as the Internet. [Effects of the Invention]

[0015] According to the present invention, it is possible to support the creation of employment-related documents such as resumes and job postings that are appealing to both recruiting companies and job seekers. [Brief explanation of the drawing]

[0016] [Figure 1] This figure shows the functional configuration of a document creation support device according to the first embodiment of the present invention. [Figure 2] This diagram shows the structure of the document data to be supported by the above-mentioned document creation support device. [Figure 3] This is an example of an annotated data table stored in the annotated data storage unit of the above-mentioned document creation support device. [Figure 4]This is an example of an attribute information data table stored in the attribute information storage unit included in the above document creation support device. [Figure 5] This is an example of a correct answer information data table stored in the correct answer information storage unit included in the above document creation support device. [Figure 6] This is a conceptual diagram explaining the statistical value of the annotated text. [Figure 7] This is an example of an example sentence information data table stored in the example sentence information storage unit included in the above document creation support device. [Figure 8] This is a diagram explaining the outline of the first to fourth data sets generated by the above document creation support device. [Figure 9] This is a diagram explaining the processing result data and the correct answer information compared and analyzed by the above document creation support device, showing (a) the processing result data (the second data set) and (b) the correct answer information. [Figure 10] This is a processing flow diagram showing the flow of processing executed by the above document creation support device. [Figure 11] These are (a) the first example and (b) the second example of the screen displayed on the user terminal by the above document creation support device. [Figure 12] This is a diagram showing the functional configuration of the document creation support device according to the second embodiment of the present invention. [Figure 13] This is a sequence diagram showing the flow of processing executed by the above document creation support device.

Embodiments for Carrying Out the Invention

[0017] ●First Embodiment● Hereinafter, the document creation support device 1 according to the first embodiment of the present invention will be described with reference to the drawings. Document creation support device 1 is a device that assists users in creating documents, such as proofreading the content of documents and suggesting specific drafts of sentences. In particular, document creation support device 1 assists in creating employment-related documents necessary for job hunting, career changes, recruitment, and hiring activities, such as resumes and job postings. In the following, when "documents" are referred to, unless otherwise specified, it means "employment-related documents." Here, the documents that support the creation of applications include those from both the job seeker and the employer. Furthermore, document creation support includes both the support of editing and revising created documents to improve their quality, and the support of creating documents from scratch or from a partially created state. The document creation support device 1 according to this embodiment can handle both of these tasks.

[0018] ●Functional Configuration As shown in Figure 1, the document creation support device 1 is configured to communicate with the user terminal 2 via a network NW such as the Internet. In this embodiment, communication between the document creation support device 1 and the user terminal 2 is wireless, but some or all of the connections may be wired. Furthermore, the document creation support device 101 may be composed of multiple hardware configurations. In this case, these multiple hardware configurations may be connected by wired or wireless connections, and information may be sent and received from each other.

[0019] User terminal 2 is a terminal used by users such as job seekers and employers who benefit from the services provided by the functions of document creation support device 1. User terminal 2 can be implemented by various terminals capable of executing computer programs, such as portable terminals and tablet terminals that can send and receive data, as well as so-called personal computers. In this embodiment, document creation support device 1 is given as an example of an online service such as SaaS, but it is not limited to this and may be used by being incorporated into user terminal 2.

[0020] This user terminal 2 includes at least communication processing means, such as a browser program for sending and receiving data with the document creation support device 1, and input / output means, such as a touch display for inputting and outputting data. Furthermore, user terminal 2 can create, edit, and view employment-related documents such as job postings and resumes, and view information created by document creation support device 1, through applications for creating, editing, and viewing documents. The application for creating, editing, and viewing documents may be a dedicated application, but it may also be possible to perform the above-mentioned processes via a general-purpose web browser.

[0021] The document creation support device 1 is implemented by, for example, a server equipped with a processing unit such as a CPU (Central Processing Unit), a computer program executed by the CPU, and internal memory such as RAM (Random Access Memory) and ROM (Read Only Memory). As a result, the document creation support device 1 has a functional configuration consisting of a document data storage unit 1A, annotated data storage unit 1B, attribute information storage unit 1C, correct answer information storage unit 1D, example sentence information storage unit 1E, reception processing unit 11, analysis processing unit 12, extraction processing unit 13, generation processing unit 14, and communication processing unit 15, as shown in Figure 1.

[0022] <Document Data Storage Unit 1A> The document data storage unit 1A is a storage unit that stores document data, which is digitized from documents created by the user or documents prepared by the provider of a service using the document creation support device 1 (hereinafter sometimes referred to as "this service"). Each piece of document data is assigned a unique document ID.

[0023] Documents include, for example, resumes prepared by job seekers and job postings prepared by employers. These documents are typically organized into designated sections. These sections divide the document into meaningful units; for example, a document might be divided into sections such as job description and working conditions. Each section contains sentences with various meanings related to that section. A single section may contain sentences with multiple meanings, and sometimes a single sentence may contain multiple meanings.

[0024] Figure 2 shows an example of the structure of document data that the document creation support device 1 accepts as input. The document data consists of one or more document description items, each of which contains one or more paragraphs. Furthermore, each paragraph contains one or more sentences. Each sentence contains multiple phrases or words. Note that the document data storage unit 1A may store documents that are not categorized into document description items. In this case, the analysis processing unit 12, which will be described later, may categorize the document data.

[0025] Furthermore, document data may be associated with information about the document's owner (creator), so-called basic owner information. The items of information about the owner may vary depending on the owner's attributes. For example, if the owner is a job seeker, the information about the owner may include the job seeker's name, age, occupation, and annual income. If the owner is a company, the information about the owner may include the company name, industry, occupation, company profile, and employee annual income.

[0026] However, the document data storage unit 1A may choose not to store some or all of the basic information of the owner of the document data among the information held by the document data registered by the user. For example, if the document data is a resume or other document data from a job seeker, the owner's name and other basic information may not be stored from the standpoint of protecting personal information. In addition, the document data storage unit 1A may associate information with certain pre-set items among the information held by the document registered by the user, indicating that those items should be excluded from the analysis by the analysis processing unit 12 described later. In this case, the analysis processing unit 12 does not use the information of those items. The analysis processing unit 12 refers to the accessibility information set for each item included in the document data and uses the information set to be accessible for analysis.

[0027] On the other hand, such basic owner information may be stored as attribute information such as age group, gender, or income category, which does not identify an individual. In this case, the attribute information may be stored in association with the document ID in the annotated data storage unit 1B, the attribute information storage unit 1C, or the example sentence information storage unit 1E, which will be described later.

[0028] Regarding employment documents and other documents from the employer's side, if the information is publicly available, the name of the employing company, etc., may be treated as usable information. Furthermore, it may be possible to accept settings for usability for each document, and information indicating, for example, that confidential job postings will be excluded from various references may be stored associated with the document data.

[0029] Furthermore, the document data storage unit 1A may be configured to register only document data that has been selected and filtered by the provider of this service.

[0030] Furthermore, the document data storage unit 1A may also store performance information linked to the document ID when the document data is used for job-related activities, namely, job-seeking activities in job postings and job-seeking activities in resumes. Performance information indicates whether or not the information contained in the document data is valid, for example, for a job posting, it may be information such as "an application was received within 60 days of the job posting being made public." For a resume, it may be information such as "a job offer was received." Conversely, performance information may also be information indicating that there were no applications for a job posting or that a job offer was not received for a resume. Performance information is entered by the user or the creator of each document at a time different from when the document data is stored. Such performance information is linked to correct answer information and example sentence information via the document ID.

[0031] <Annotated Data Storage Unit 1B> The annotated data storage unit 1B is a storage unit that stores annotated data, which is data in which text data and semantic elements that indicate the meaning of the text data are associated. In this description, text data to which semantic elements are associated is also called annotated text.

[0032] Figure 3 shows an example of an annotated data table T1 stored in the annotated data storage unit 1B. As shown in the figure, the annotated data storage unit 1B stores information consisting of, for example, a document description item name, annotated text, and semantic elements, associated with a document ID. A single document description item contains multiple pairs of annotated text and semantic elements.

[0033] As mentioned above, document description items categorize the content of the document data. Examples of document description item names include "Job Description Section."

[0034] Annotated text is text data included in document data, and in the annotated data storage unit 1B, semantic elements are associated and stored. This annotated text is broken down into several predetermined levels of decomposition, such as paragraph blocks, sentence blocks, and phrase / word blocks. A paragraph block is a block consisting of a set of sentences with a specific meaning, and may be a block separated by predetermined paragraph symbols, etc. A sentence block or phrase / word block is a block consisting of one sentence or one phrase or word, respectively. Annotated text may be a noun phrase such as "consulting sales," or it may be a sentence or group of sentences such as "The job content is...."

[0035] Semantic elements are pieces of information assigned to each annotated text, and they are information that annotates the meaning of the annotated text. In the example shown, the annotated text "consulting sales" has the semantic element "job title" labeled as an annotation, and the annotated text "bonus for achieving sales targets" has the semantic element "incentive" labeled as an annotation.

[0036] The annotated data storage unit 1B may store annotated text with semantic elements that have been artificially added in advance, along with the semantic elements themselves, or it may store annotated data on which semantic elements have been added by the document creation support device 1. The process of adding semantic elements by the document creation support device 1 will be described later.

[0037] The information stored in this annotated data storage unit 1B is referenced in the execution of the functions of the analysis processing unit 12 and the generation processing unit 14, which will be described later. In addition, the annotated data storage unit 1B stores the second and third datasets, which will be described later, generated in accordance with the semantic element assignment process performed by the analysis processing unit 12. As mentioned above, the annotated text is linked to the document ID of the document data from which the text was extracted, so by referring to the document data storage unit 1A, annotated text with predetermined performance information can be extracted. With this configuration, as a large amount of document data is received and supported, information is accumulated, and the accuracy of document data analysis can be improved.

[0038] <Attribute information storage section 1C> The attribute information storage unit 1C is a storage unit that stores information indicating the attributes of the document data. Figure 4 shows an example of an attribute information table T2 stored in the attribute information storage unit 1C. As shown in Figure 4, the attribute information storage unit 1C stores information relating to each other, such as document IDs and document attributes. Document attributes include, for example, whether the document is a job application or a resume. Document attributes also include the industry and occupation targeted when the document data is used in job-related activities. Furthermore, document attributes may include information such as the number of employees at the recruiting company or the company to which the applicant is applying. Document attributes are used to identify the target attributes to be referenced as correct answer information for document correction in the reference of the correct answer information storage unit 1D, which will be described later.

[0039] <Correct Answer Information Storage Unit 1D> The correct answer information storage unit 1D is a storage unit that stores correct or exemplary correct answer information in the creation of document data. The correct answer information storage unit 1D is referenced by the generation processing unit 14, which will be described later, in a comparison process that compares the correct answer information with the processing result data obtained by the analysis processing unit 12 that analyzes the document data to be supported, and in a presentation process that presents correction information to the user.

[0040] Figure 5 shows an example of a correct answer information table T3 stored in the correct answer information storage unit 1D. As shown in Figure 5, the correct answer information storage unit 1D stores, for example, the target document description item, correct answer classification, correct answer granularity, and correct answer content, all associated with each other for each attribute of the correct answer information. The target attributes of the correct answer information correspond to, for example, the document attributes of the document data, indicating which document data the correct answer information applies to. These target attributes include, for example, information on document type, industry, and job title. The target document description item indicates which document description item in the document data the correct information applies to.

[0041] The correct answer classification is information that describes the nature of the correct answer information. Examples of correct answer classifications include "statistical values" and "know-how."

[0042] The correct answer information designated as "statistical value" in the correct answer classification is information obtained by aggregating annotated text associated with document attributes that match the target attributes of the correct answer information. Statistical values ​​may also be obtained, for example, by extracting annotated text that matches the conditions of a proofreading order received from a user, and then aggregating that annotated text.

[0043] Figure 6 is a conceptual diagram illustrating the statistical values ​​of annotated text. The correct information, defined as "statistical values" in the correct classification, stores distribution information such as that shown in the figure. For example, the upper left of Figure 6 shows a histogram illustrating the distribution of the location of the semantic element "customer description" in a job posting. In this example, the figure includes information indicating that the paragraph block containing the semantic element "customer description" is most frequently found in the third paragraph of a job posting. It also includes information such as the fact that in approximately 75% of job postings, the element is located in the third paragraph or later. The lower left of Figure 6 shows a histogram illustrating the distribution of the number of characters in paragraph blocks containing the semantic element "customer description." The right column of Figure 6 shows a histogram illustrating the distribution of the location and number of characters in paragraph blocks containing the semantic element "description of case / product."

[0044] Furthermore, the correct answer information designated as "know-how" in the correct answer classification is information registered based on knowledge and rules of thumb in document creation. Know-how is defined, for example, by an appropriate threshold. It should be noted that this correct answer information may also contain fictitious distribution information that reflects know-how, or information that could be included in distribution information.

[0045] The correct answer granularity indicates the decomposition granularity of the block to which the correct answer information is applied, and the decomposition granularities for paragraph blocks, sentence blocks, phrase / word blocks, etc., as described above are defined. The comparison between the blocks obtained by decomposing the document data and the correct answer information is performed at decomposition granularities that match each other.

[0046] The correct answer represents the content of the correct answer and consists of information that quantifies the entry format (meaning elements, amount of information to be entered, location of entry, etc.) and information that quantifies the content that has been verbalized as know-how. Note that a meaning element is not necessarily composed of only one meaning element, but may be composed of multiple meaning elements, in which case the meaning element is represented by the probability of each meaning element being applicable. The correct answer for a meaning element means, for example, the meaning element that should be entered in the description item of the document.

[0047] The correct answer information to be applied to the document data may be selected according to the performance information of the document data received via the reception processing unit 11. For example, a dataset of document data associated with performance information indicating effectiveness may be stored as the correct answer information. In addition, the amount and position of entries of blocks having predetermined semantic elements obtained from document data associated with performance information indicating effectiveness may be added to the source data of the statistical values ​​of the correct answer information.

[0048] At least some of the correct answer information is stored in conjunction with performance information. For example, the correct answer information stored from the received document data is linked to the performance information via the document ID. With this configuration, the correct answer information to be presented to the user can be extracted from the correct answer information linked to the specified performance information, according to the correction specification information received from the user along with the document data. Furthermore, statistical values ​​for the amount of writing and the position of writing can be calculated using only the correct answer information linked to the specified performance information specified by the correction specification information.

[0049] <Example Sentence Information Storage Unit 1E> The example sentence information storage unit 1E is a storage unit that stores example sentences to be suggested to the user. Figure 7 shows an example of an example sentence information table T4 stored in the example sentence information storage unit 1E. As shown in Figure 7, the example sentence information storage unit 1E stores, for example, the target document description item, example sentence granularity, example sentence classification, and example sentence text, associated with the document ID of the document data from which the example sentence was created.

[0050] The target document description items indicate the items among the document description items that make up the document data to be supported, for which example sentence information may be proposed. The example sentence granularity indicates the granularity of the example sentences in the associated example text. The example sentence classification represents information that describes the semantic elements of the example sentences, and corresponds to the semantic elements in the annotated data table T1. The example text is the data that constitutes the content of the example sentence, and is the text suggested to the user. This example text can be suggested to blocks of the same granularity as the example sentence itself.

[0051] Furthermore, the example sentences stored in the example sentence information storage unit 1E may be created from document data stored in the document data storage unit 1A based on annotated data. That is, by using at least a portion of the annotated text as keywords, the semantic elements of a predetermined text constituting the document data are identified, the item to which the text belongs is designated as the target document description item, the identified semantic elements are classified as example sentences, and the predetermined text is stored as the example sentence text. Alternatively, text extracted from multiple document data sets based on selection criteria such as having a predetermined frequency of appearance may be stored as example sentence text.

[0052] Furthermore, the document data from which example sentences are extracted may be document data to which predetermined performance information is linked. In addition, the example sentences extracted from the document data may be stored with the performance information of that document data linked via the document ID, and by referring to the correction request information received from the user along with the document data, only example sentences with performance information that meet the conditions may be presented to the document data. With such a configuration, it is possible to present example sentences contained in document data with a proven track record that meet the user's requests.

[0053] Furthermore, the example sentences stored in the example sentence information storage unit 1E do not necessarily have to be created from document data stored in the document data storage unit 1A; they may also be original example sentences prepared by the service provider.

[0054] With this configuration, example sentences can be presented for the corresponding document description items in the supported document data, specifically for text that has similar semantic elements within those document description items.

[0055] <Reception Processing 11> The reception processing unit 11 is a functional unit that receives information regarding the editing of documents to be supported. For example, the reception processing unit 11 may receive editing specification information from the user, which defines the editing method of the document, along with the document data to be supported. The editing specification information is information related to the assumptions and policies when providing support for document creation, or the direction and nature of the document. The editing specification information is referenced as a search condition when searching the correct answer information to be compared from the correct answer information storage unit 1D in the process of extracting correct answer information used for comparison with the document data. In addition, the editing specification information is referenced as a search condition when searching the example sentence information storage unit 1E in the process of extracting example sentence text for the document data.

[0056] The editing specifications may include, for example, one or more of the target information, owner information, and editing order for the document in question.

[0057] Target information refers to information about the goals or objectives that the user wishes to achieve through the document, such as employment, job offer, interview, application, job type, annual salary, and work history. In addition, if the user is a job seeker, the target information may include the specific names of companies the user is interested in and information about job postings, and if the user is an employer, it may include information such as the profile and attributes of the person they are looking for.

[0058] Owner information refers to information about the owner of the document. For example, if the owner is a company, this would include company details such as the number of employees, whether it is publicly listed, industry, and job titles.

[0059] A proofreading order is information that represents the user's proofreading request. For example, the user might want to compare their own application documents with those of people in the same age group and occupation who have received job offers from companies in their desired field. This target information, owner information, and editing order details all or part of them correspond to the aforementioned performance information, and the editing specification information also serves as information that specifies the performance information as a search criterion. Furthermore, the reception processing unit 11 can also receive instructions from the user regarding whether or not they wish to be shown example sentences in the proofreading results.

[0060] <Analysis Processing Unit 12> The analysis processing unit 12 is a functional unit that performs structural analysis of the document data to be supported and decomposes the text contained in the document data into blocks of various decomposition granularities. Furthermore, the analysis processing unit 12 estimates the contents of the blocks obtained by the decomposition process by searching the annotated text stored in the annotated data storage unit 1B, and assigns semantic elements to each block. The analysis processing unit 12 mainly comprises an item analysis unit 121, a paragraph analysis unit 122, and a sentence / phrase / word block analysis unit 123.

[0061] Figure 8 shows the flow of structural analysis processing by the analysis processing unit 12 and an overview of the dataset converted or generated from document data.

[0062] The item analysis unit 121 converts the document data into data (also called the "first data set") that is divided into predetermined document description items. For example, if the document data is structured into document description items according to a predetermined form, the item analysis unit 121 divides the document data into document description items according to the metadata contained in the form.

[0063] The item analysis unit 121 performs a sorting process for document description items if the document data is not organized by description. In this case, for example, semantic elements that may be described in different document description items are pre-stored in an appropriate memory unit. Then, the item analysis unit 121 identifies the semantic elements contained in the document data and estimates the document description items according to the semantic elements. In this case, the semantic elements may be estimated by the paragraph analysis unit 122 or the sentence / phrase / word block analysis unit 123 before the sorting process for document description items is performed, prior to the analysis by the item analysis unit 121.

[0064] The paragraph analysis unit 122 decomposes the text data into paragraph blocks and generates processing result data (also called the "second data set") that shows the composition of each paragraph block. A paragraph block is an example of a block in the claims. As shown in Figure 8, the second dataset is a dataset in which, for each paragraph block, data is associated with the type of text, the level of decomposition, the position of the text, the amount of text, and the semantic elements, as well as the probability of matching the target paragraph block to those semantic elements.

[0065] The paragraph analysis unit 122 refers to the annotated data storage unit 1B and searches for locations where semantic elements can be determined, using at least a portion of the annotated text pre-stored in the annotated data table T1 (see Figure 3) of the annotated data storage unit 1B as keywords. More specifically, for example, the paragraph analysis unit 122 searches for the keyword in the text data and identifies a group of sentences containing the keyword as a paragraph block. The keyword may include regular expressions, and for example, so-called fuzzy searches may be possible, or conditional searches may be possible, such as extracting only keywords that appear at the beginning, middle, or end of a sentence.

[0066] Furthermore, if the paragraph analysis unit 122 detects a keyword from the text data, it may also search for synonyms of that keyword that are stored in advance, and identify groups of sentences containing synonyms as the same paragraph block.

[0067] The paragraph analysis unit 122 may also determine a paragraph block based on editing mark information, such as paragraph marks, that are pre-included in the text.

[0068] For the identified group of sentences, semantic elements associated with keywords in the annotated data storage unit 1B are linked. Then, the similarity of the content of each sentence is determined using these keywords to identify multiple sentences with the same content as a single paragraph block, or similar expressions are searched for from the paragraph blocks that have been semantically assigned, and they are merged or separated as appropriate to determine the final paragraph block. Note that these processes may be performed individually or in combination as appropriate.

[0069] Furthermore, the paragraph analysis unit 122 may use appropriate artificial intelligence technology to decompose the text data into paragraph blocks and then link the semantic elements. The specific configuration will be described in the second embodiment described later.

[0070] Furthermore, the second dataset includes information on the decomposition granularity of the paragraph block, its location, the amount of text, and its semantic elements and corresponding probabilities. The text decomposition granularity indicates the unit of decomposition of the text data. In the second dataset, the text data is decomposed into paragraph blocks, so the text decomposition granularity in the second dataset is paragraph.

[0071] The location of the entry is information indicating where the target paragraph block is located within the document data or within a specified document description item. For example, it includes information indicating the start and end positions of the paragraph block within the entire document data or within a specified document description item. Alternatively, the location of the entry may be, for example, the paragraph number in the document data, or information regarding the number of paragraphs and paragraph numbers contained in the document data. The length of the text refers to the number of characters in the paragraph block in question.

[0072] The semantic element and matching probability indicate the semantic element that fits the target paragraph block and the likelihood that the semantic element is applicable. A paragraph block is not necessarily composed of a single semantic element; it may be determined to be composed of multiple semantic elements. If it is composed of multiple semantic elements, the proportion of each semantic element that makes up the paragraph block is shown as the matching probability. The matching probability is calculated based on the frequency (probability of occurrence) of keywords that represent a given semantic element, and the weights set for each semantic element. In other words, if many keywords are detected in the paragraph block, the matching probability will be calculated to be relatively high. Also, if keywords with high weights are detected based on the weights set for each keyword, the matching probability will be calculated to be relatively high.

[0073] Furthermore, when presenting the probability to the user, it is also possible to selectively present only the semantic element with the highest probability of being relevant.

[0074] The text / phrase / word block analysis unit 123 decomposes paragraph blocks into text blocks or phrase / word blocks, and generates processing result data (also called the "third data set") that shows the composition of each text block and phrase / word block. Text blocks and phrase / word blocks are, respectively, other examples of blocks in the claims. The sentence / phrase / word block analysis unit 123, for example, decomposes paragraph blocks into sentence blocks by treating periods in a sentence as delimiters for sentence blocks. Furthermore, the sentence / phrase / word block analysis unit 123 also decomposes paragraph blocks into phrases and word blocks by referring to words identified as such using a word dictionary.

[0075] Furthermore, the semantic elements of each sentence block or phrase / word block are identified, for example, by using the annotated text stored in the annotated data table T1 (see Figure 3) of the annotated data storage unit 1B as keywords, similar to the processing in the paragraph analysis unit 122, and by identifying the annotated text contained in each sentence block, phrase, or word block. The keywords may include regular expressions, similar to the processing in the paragraph analysis unit 122, or processing may be performed to link semantic elements using artificial intelligence technology. Processing using artificial intelligence technology will be described in the second embodiment described later.

[0076] The third dataset, like the second dataset, consists of items such as the decomposition granularity of text that constitutes sentence blocks or phrase / word blocks, the location of the text, the amount of text, and the semantic elements and probability of relevance. The text decomposition granularity indicates the decomposition unit of the text data; in the third dataset, the text decomposition granularity is either sentence blocks or phrase / word blocks.

[0077] In this way, the analysis processing unit 12 generates the first dataset, the second dataset, and the third dataset. That is, the analysis processing unit 12 decomposes the document data into blocks of various decomposition granularities, such as document description items, paragraph blocks, sentence blocks, and phrase / word blocks, and assigns semantic elements to each block.

[0078] <Extraction Processing Unit 13> The extraction processing unit 13 extracts correct answer information from the correct answer information storage unit 1D (see Figures 1 and 5) to compare and analyze with the blocks obtained by the analysis processing unit 12. In the extraction process, for each block of text data that has been decomposed to a predetermined decomposition granularity, correct answer information with the same decomposition granularity as the block and a high degree of semantic element agreement is extracted. Furthermore, if the reception processing unit 11 has received correction specification information from the user, the extraction processing unit 13 also extracts correct answer information that matches that correction specification information.

[0079] Furthermore, the system may determine whether each of the multiple pieces of information included in the correction specification information matches, and extract the correct information with the most matches. The decomposition granularity, semantic elements, and correction specification information referenced during extraction may all be treated as equal conditions, or one of them may be treated as a condition that takes precedence over the others. In addition, the extraction processing unit 13 may extract the correct information to compare with the document data from the correct information associated with predetermined performance information, according to the correction specification information received from the user along with the document data. The extraction processing unit 13 may also use statistical values ​​of the amount and location of entries calculated only from the correct information associated with predetermined performance information specified by the correction specification information as the correct information to apply to the document data.

[0080] Furthermore, the extraction processing unit 13 may extract correct information that matches the correction specification information based on the document attributes of the document data to be supported, and then determine whether or not the semantic elements of the correct information are included in the document data.

[0081] Furthermore, if the correction specification information includes an order for example sentences, the system extracts example sentence texts corresponding to the areas to be improved from the example sentence information storage unit 1E, based on the document description item to which the target block belongs, and its decomposition granularity and semantic elements. In extracting example sentence texts, the correction specification information may also be taken into consideration, and based on the attribute information and performance information of the document associated with the example sentence text, the system may present example sentence texts created from documents made by owners in an age group specified by the user, or from documents with performance information demonstrating effectiveness. Furthermore, the extraction processing unit 13 can also extract example text for each paragraph block, sentence block, or phrase / word block and present it to the user, regardless of whether it is an area for improvement or not.

[0082] Figure 9 shows an example of such analysis results. Figure 9(a) is the processing result data of a block that underwent structural analysis, and is an example of a second dataset where the block decomposition granularity consists of "paragraph blocks". Figure 9(b) is the ground truth information compared to this.

[0083] The processing result data shown in Figure 9(a) shows the block decomposition granularity, description position, description quantity, semantic element, and corresponding probability. The correct answer information shows the same decomposition granularity as in Figure 9(a) and the same semantic element (in the example in the figure, "Company Introduction").

[0084] <Generation Processing Unit 14> The generation processing unit 14 analyzes the second and third datasets by referring to the correct answer information and generates correction information (also called the "fourth dataset") in which areas for improvement have been extracted. Specifically, the process involves comparing and analyzing the processed document data converted to the second and third datasets, as illustrated in Figure 9(a), with the correct information shown in Figure 9(b), and generating correction information based on the results. In the example shown in the figure, the "Company Introduction" section is located at position "3" in the supported document data, while it is located at position "1" in the correct information, suggesting that the location is inappropriate. Furthermore, regarding the length of the text, it is 100 characters in the supported document data, while it is 50 characters in the correct information, suggesting that the company introduction section is too long. Figure 9 shows a comparison between the second dataset, in which the block decomposition granularity consists of "paragraph blocks," and the ground truth information. The third dataset is similarly compared for each annotated text.

[0085] This allows us to identify any similarities or differences between the analysis results and the correct answer information. The parts with differences can be extracted as areas for improvement, and these can be combined with the correct answer information to create revised information.

[0086] The editing information includes suggestions for adding document description items, suggesting additional semantic elements, suggesting changes in text length, and suggesting changes in the order of entries. For example, if the text is placed in a different position than the correct position, the program will identify the paragraph block containing the correct position as an area for improvement and show the correct position. In the example in Figure 9, for a given paragraph block presumed to be related to "company introduction," the program can generate and present a sentence recommending changes to the placement from "3 to 1" and the length from "100 to 50," specifically, a sentence such as "Why not place it at the beginning and limit the length to around 50 characters?"

[0087] In addition, along with the correction information, the following may be output: identification information or attribute information of the document data to be supported, or document data with similar attribute information or correction specification information to the document data to be supported.

[0088] Furthermore, when presenting areas for improvement / correct answers, if there are differences in the probability of semantic elements being present, these differences may be pointed out, or they may only be pointed out if the difference exceeds a predetermined threshold. By pointing out differences in the probability of semantic elements being present, users can understand the direction of the content that should be written in the designated paragraph block.

[0089] Furthermore, if there is an order for example sentences to be provided in the correction specification information, example sentences for the designated paragraph block, sentence block, or phrase / word block are extracted from the example sentence information storage unit 1E (see also the example sentence information table T4 in Figure 6), and the extracted example sentences are added to the correction information.

[0090] Furthermore, in addition to presenting the fourth dataset to the user as correction information, the system may also present, in a comparative manner, the structural analysis processing results (Figure 9(a)) for each paragraph block, sentence block, and phrase / word block, and the correct information (Figure 9(b)) that is compared with the said structural analysis processing results.

[0091] Furthermore, the editing information may include editing results that conform to the search criteria. For example, if the user is a job seeker, and the editing request information received includes the specific name of the company the user is interested in or target information regarding the job posting, then, in light of the job posting and the company's requirements, the system may determine whether the document data conforms to the format required in the job posting, whether the amount and placement of each paragraph block, sentence block, and phrase / word block are appropriate, and include the results of this determination in the editing results. Additionally, the editing information may include editing results when compared with document data owned by the same owner as the document data being supported, or with document data that has the same target information. If such editing results do not necessarily match the editing results generated by comparing with the correct information, one may be prioritized over the other. Which to prioritize may be set in advance, or it may be possible to switch based on the user's specification.

[0092] The generation processing unit 14 may generate example sentences for document description items and semantic elements for which the user has not provided text. With this configuration, the user can obtain example sentences for document description items and semantic elements for which no text has been entered. Consequently, it is also possible to automatically generate employment-related documents.

[0093] <Communication Processing Unit 15> The communication processing unit 15 performs data transmission and reception processing with the user terminal 2 via a network NW such as the Internet, in accordance with a predetermined communication protocol. The document creation support device 1 uses this communication processing unit 15 to send correction information to the user terminal 2 and to receive document data from the user terminal 2.

[0094] ●Processing flow The following describes the processing flow performed by the document creation support device 1 according to an embodiment of the present invention, with reference to Figure 10. First, the document creation support device 1 receives document data input from the user via the reception processing unit 11 (S101). The reception processing unit 11 also receives correction specification information from the user via the reception processing unit 11 (S102). In response, the document creation support device 1 performs structural analysis processing of the document data via the analysis processing unit 12.

[0095] First, the item analysis unit 121 converts the document data into a first dataset consisting of predetermined document description items and text data for each of those predetermined document description items (S103). Next, the paragraph analysis unit 122 refers to the annotated data storage unit 1B and decomposes the text data for each document description item into paragraph blocks (S104) and adds semantic elements as annotations to each paragraph block (S105). In this way, a second dataset is generated. Next, the sentence / phrase / word block analysis unit 123 decomposes the paragraph blocks into sentence blocks (S106) and adds semantic elements as annotations to the sentence blocks (S107). Next, the paragraph blocks or sentence blocks are decomposed into phrase / word blocks (S108) and add semantic elements as annotations to the phrase / word blocks (S109). In this way, a third dataset is generated. The second and third datasets generated by the paragraph analysis unit 122 and the sentence / phrase / word block analysis unit 123 are stored in the annotated data table T1 of the annotated data storage unit 1B.

[0096] The extraction processing unit 13 refers to the correct answer information storage unit 1D and extracts correct answer information from the correct answer information storage unit 1D to be compared with the processing result data of the text that has been decomposed into paragraph blocks, sentence blocks, or phrase / word blocks (second dataset, third dataset) (S110). The generation processing unit 14 compares and analyzes the processing result data with the correct answer information and generates correction information (S111). Steps S110 and S111 are repeated until the generation of correction information for all processing result data is complete. Once the generation of all correction information is complete, the process moves to step S112.

[0097] If there is no order for example sentences to be provided in the received correction request information (N in S112), the correction information is presented to the user and the process ends (S113). On the other hand, if there is an order for example sentences to be provided in the received correction request information (Y in S112), the generation processing unit 14 further refers to the example sentence information storage unit 1E to extract example sentences (S114). Then, it presents the correction information along with the example sentences to the user and terminates the process (S115).

[0098] Furthermore, the order for displaying example sentences may be set to display example sentences by default, regardless of any conditions specified by the user.

[0099] Furthermore, while the processing flow shown in Figure 10 describes the process of correcting document data, the document creation support device 1 may also suggest example sentences to be written in each document description item even if a predetermined document description item in the document data is blank or the entire document data is not yet created. That is, the generation processing unit 14 presents example sentences in accordance with the correct content for each target document description item based on the correct answer information and example sentence information. For example, for a target document description item, the example sentence classification extracts example sentences corresponding to the semantic elements shown in the correct answer content and writes them in the writing positions shown in the correct answer content. It is also possible to perform a process to adjust the extracted example sentences to the length shown in the correct answer content. With such a configuration, it is possible to create sentences even from a blank state, and the user can use or refer to these sentences as appropriate to complete the document data.

[0100] In the document creation support device 1 according to this embodiment, the functional configuration of each terminal or device is just an example, and the functional units shown in this example can also be provided in terminals or devices different from those shown in this example.

[0101] Figure 11 shows an example of a screen displayed on user terminal 2. Figure 11(a) is an example of input screen G1 showing the document data entered into the user terminal. In the example of input screen G1, the user is a recruiting company, and the document description items of the document data include "Job Title," "Job Description," and "Application Requirements." Below each document description item, an input field linked to the document description item is displayed, ready to accept text input from the user. Multiple paragraphs can be entered in this input field, and each paragraph can contain multiple sentences, phrases, or words. Additionally, an improvement suggestion button is displayed on the input screen G1. When the user selects the improvement suggestion button via the user terminal 2, the improvement suggestion screen G2 shown in Figure 11(b) is displayed.

[0102] As shown in Figure 11(b), the improvement suggestion screen G2 primarily displays "advice," "reasons," and "example sentences." The advice section extracts semantic elements not included in the document data and presents this information. The advice section also presents the recommended amount of information to include for those semantic elements. The reasons section presents the basis for the content mentioned in the advice section. The example sentence section displays example sentences obtained by referring to the example sentence information storage unit 1E.

[0103] According to the document creation support device 1 of this embodiment, when creating employment-related documents such as resumes and job postings, it is possible to support the creation of documents with appealing content to recruiting companies and job seekers based on accumulated data, without relying on people with experience and know-how. In other words, resumes supported by the document creation support device 1 are appealing to recruiting companies, and job postings are appealing to job seekers. Furthermore, revisions can be made in accordance with the meaning and context that should be written by revising according to semantic elements. Moreover, by changing the correct or model information according to the attribute information of the document data and the revision request information from the user, revision results can be flexibly proposed by a single document creation support device 1. Furthermore, by displaying the reason for revision based on the correct information, it is possible to provide users with advice that they can understand. Furthermore, by analyzing the actually created documents and incorporating the analysis content into document creation, the correct information is updated as needed, so documents that are more in line with reality can be created.

[0104] ●Second Embodiment● The following description focuses on the differences between the document creation support device 101 of the second embodiment of the present invention and the first embodiment. The document creation support device 101 of the second embodiment differs from the first embodiment in that it performs some or all of the following processes: decomposition into paragraph blocks, decomposition into sentence blocks or phrase / word blocks, and example sentence extraction, by sending instructions to the artificial intelligence unit and obtaining output from the artificial intelligence unit. The following description also includes a configuration in which some of the functions or processes are implemented in the first embodiment, as part of the technical concept of the present invention. Components similar to those in the first embodiment are denoted by the same reference numerals, and their descriptions are omitted.

[0105] ●Document creation support device 101 As shown in Figure 12, the document creation support device 101 is configured to communicate with the user terminal 2 used by the user and the artificial intelligence unit 100 via a network NW. The document creation support device 101 may be composed of hardware devices, or some or all of its functions may be implemented by a cloud computer. Furthermore, each component of the document creation support device 101 may be implemented by an API (Application Programming Interface).

[0106] Furthermore, the document creation support device 101 is configured to communicate with the artificial intelligence unit 100. The artificial intelligence unit 100 has AI (Artificial Intelligence) capabilities. The artificial intelligence unit 100 is also connected to a database DB. The database DB has an annotated data storage unit 1B and an example sentence information storage unit 1E, and the artificial intelligence unit 100 acquires information from these storage units. The document creation support device 101 transmits appropriate instructions to the artificial intelligence unit 100 and acquires information output from the artificial intelligence unit 100.

[0107] The document creation support device 101 includes an AI control unit 102. The AI ​​control unit 102 is a functional unit that controls the input to the artificial intelligence unit 100. The various instructions generated by the AI ​​control unit 102 are, for example, prompts, but are not limited to any instructions that the artificial intelligence unit 100 can interpret.

[0108] The AI ​​control unit 102 mainly comprises an analysis instruction unit 112 and an example sentence generation instruction unit 113.

[0109] The analysis instruction unit 112 generates analysis instructions for the artificial intelligence unit 100, which perform structural analysis of the document data to be supported and decompose the text contained in the document data into blocks of various decomposition granularities. The analysis instructions may also be instructions for the artificial intelligence unit 100 to extract correct information by referring to the annotated data storage unit 1B (see Figures 1 and 3).

[0110] The example sentence generation instruction unit 113 generates an example sentence generation instruction to generate an example sentence corresponding to a block. The example sentence generation instruction may also be an instruction to the artificial intelligence unit 100 to extract example sentence information by referring to the example sentence information storage unit 1E (see Figures 1 and 6). Alternatively, the example sentence generation instruction may also be an instruction to the artificial intelligence unit 100 to generate example sentences using a large-scale language model or the like, as described later.

[0111] In analysis instructions, extraction instructions, or example sentence generation instructions, the data in the annotated data storage unit 1B, the correct answer information storage unit 1D, or the example sentence information storage unit 1E may be specified as an attached document to be retrieved by so-called RAG (Retrieval-Augmented Generation), which retrieves information from a predetermined attached document and generates answers or sentences based on that information.

[0112] ● Artificial Intelligence Department 100 The artificial intelligence unit 100 is an artificial intelligence (AI) equipped with an appropriate learning model. A learning model (also called a machine learning model) refers to a learning model based on a machine learning algorithm. Specific machine learning algorithms include the nearest neighbor method, Naive Bayes method, decision trees, and support vector machines. Another example is deep learning, which uses neural networks to generate features and connection weights for learning. The artificial intelligence unit 100 can apply the above algorithms as appropriate.

[0113] Furthermore, the artificial intelligence unit 100 may be an artificial intelligence equipped with a language model such as a transformer including BART (Bidirectional and Auto-regressive Transformer), BERT (Bidirectional Encoder Representations from Transformers), or GPT (Generative Pretrained Transformer, including GPT-1, GPT-2, GPT-3, GPT-4), and especially a learning model such as a large language model (LLM). Using such a learning model, the artificial intelligence unit 100 generates and outputs example sentences in response to an instruction to generate example sentences.

[0114] The artificial intelligence unit 100 has a pre-trained model that has been appropriately machine-learned. The training data may be provided by an administrator or the like, or it may include information collected from the internet or the like. For example, the artificial intelligence unit 100 has a first pre-trained model that takes document data as input and outputs text data broken down to a predetermined granularity, i.e., blocks, and the semantic elements of those blocks, and a second pre-trained model that takes blocks and their semantic elements as input and outputs example sentences. A pre-trained model has already obtained appropriate training data. This training data consists of annotated text and semantic element data. Furthermore, the pre-trained model can undergo additional training as needed. Specifically, the pre-trained model is retrained using document data and performance information received by the reception processing unit 11.

[0115] Based on the analysis instructions generated by the analysis instruction unit 112, the artificial intelligence unit 100 uses a first trained model to take document data as input and outputs a second dataset and a third dataset consisting of multiple blocks and their semantic elements.

[0116] Based on the example sentence generation instructions generated by the example sentence generation instruction unit 113, the artificial intelligence unit 100 uses a second trained model to output example sentence text, taking the second and third datasets, i.e., blocks of various granularities and their semantic elements, as well as correct answer information, as input.

[0117] ● Sequence diagram Figure 13 illustrates the flow of information in the document creation support device 101, user terminal 2, artificial intelligence unit 100, and database DB according to the second embodiment. First, the document creation support device 101 receives input of the document data to be supported and the correction specification information from the user terminal 2 (S201). The document creation support device 101 acquires this document data and correction specification information using the reception processing unit 11.

[0118] Next, the document creation support device 101 generates an analysis instruction for the artificial intelligence unit 100 using the analysis instruction unit 112 and transmits it to the artificial intelligence unit 100 along with at least the document data (S202). The artificial intelligence unit 100 decomposes the document data into blocks of various decomposition granularities, generates a dataset in which semantic elements are assigned to each block, transmits it to the document creation support device 101, and the document creation support device 101 receives it (S203).

[0119] Next, the document creation support device 101, using the extraction processing unit 13, appropriately refers to the correct answer information storage unit 1D for each block of the dataset and extracts the correct answer information (S204).

[0120] Next, the document creation support device 101 uses the generation processing unit 14 to compare and analyze the dataset with the correct answer information and generate correction information (S205).

[0121] Next, the document creation support device 101 generates an instruction to the artificial intelligence unit 100 to generate example sentences using the example sentence generation instruction unit 113, and transmits it to the artificial intelligence unit 100 along with the dataset and correct answer information (S206). The artificial intelligence unit 100 appropriately refers to the example sentence information storage unit 1E of the database DB, extracts example sentences for each block of the dataset, and transmits them to the document creation support device 101. The document creation support device 101 receives the example sentences in association with the blocks included in the transmitted dataset (S207).

[0122] Steps S206 and S207 may be performed before step S205, or they may be performed simultaneously.

[0123] Next, the document creation support device 101 outputs correction information, including example sentences, to the user terminal 2 via the communication processing unit 15 (S208).

[0124] This structure also allows for the creation of compelling job application documents based on accumulated data, without relying on individuals with experience and expertise. Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes are possible within the scope of its gist. For example, the present invention may include a configuration in which instructions are sent to an artificial intelligence unit for correct answer information extraction processing, comparative analysis processing, or correction information generation processing, and processing results obtained from the artificial intelligence unit are received. In this case, the artificial intelligence unit may have, for example, a third trained model that takes blocks and their semantic elements as input and outputs correct answer information to be compared, or a fourth trained model that takes blocks and correct answer information as input and outputs correction information for the blocks. As correct answer information extraction processing, the artificial intelligence unit outputs correct answer information using the third trained model, based on instructions from the document creation support device, with at least blocks and their semantic elements as input. In this case, correction specification information, attribute information, or performance information may also be input. Furthermore, as correction information generation processing, the artificial intelligence unit may output correction information using the fourth trained model, based on instructions from the document creation support device, with blocks and their semantic elements, as well as correct answer information as input. [Explanation of symbols]

[0125] 1: Document creation support device 1A: Document data storage unit 1B: Annotated data storage unit 1C: Attribute information storage section 1D: Correct Answer Information Storage Unit 1E: Example Sentence Information Storage Unit 2: User terminal 11: Reception Processing Section 12: Analysis Processing Unit 13: Extraction Processing Unit 14: Generation Processing Unit 15: Communication Processing Unit NW: Network

Claims

1. An annotated data storage unit that stores text composed of predetermined decomposition granularities as annotated text by associating the semantic elements of the text, A correct answer information storage unit that stores the decomposition granularity of the text, semantic elements, and correct answer content in association with each other, A reception processing unit that accepts input of document data to be supported, An analysis processing unit that decomposes the text contained in the document data into blocks of various decomposition granularities, and associates semantic elements associated with the annotated text with each block, which includes at least a portion of the annotated text. For each of the aforementioned blocks, an extraction processing unit extracts correct answer information from the correct answer information storage unit that is associated with the same decomposition granularity and semantic elements as the aforementioned block, A generation processing unit calculates the difference between the extracted correct answer information and the block, and generates correction information that includes information about the difference, Having, Document creation support device.

2. The aforementioned correct answer information includes, in relation to the aforementioned semantic elements, information relating to the amount of information to be written and the location of the information on the document data. The correction information includes information relating to the difference in the amount of text or the position of text between the block and the correct answer information. The document creation support device according to claim 1.

3. The generation processing unit generates a document containing information relating to the difference, and presents the document as advice to the user terminal via an appropriate communication processing unit. The document creation support device according to claim 2.

4. It further comprises an example sentence information storage unit that stores the document description items that make up the document, the semantic elements of the example sentences, and the example sentence text in association with each other. The extraction processing unit further extracts the example sentence text associated with the same semantic element as the block from the example sentence information storage unit, and presents the example sentence text to the user terminal via the communication processing unit. A document creation support device according to claim 1 or 2.

5. At least a portion of the aforementioned correct answer information is created based on annotated text obtained by analyzing other document data, and the aforementioned correct answer information is associated with performance information indicating whether or not the other document data is valid. The reception processing unit receives the user's specification of the performance information as a search condition for extracting the correct answer information. The extraction processing unit extracts, for each block, the correct answer information that matches the specified performance information as correct answer information to be compared and analyzed with the block, from the correct answer information storage unit. The document creation support device according to claim 1.

6. An annotated data storage unit stores information relating to the semantic elements of text composed of text at a predetermined decomposition granularity as annotated text, A computer having a correct answer information storage unit that stores the decomposition granularity of text, semantic elements, and correct answer content in association, The reception process accepts input of document data to be supported, The analysis process involves decomposing the text contained in the document data into blocks of various decomposition granularities, and associating semantic elements associated with the annotated text with each block, which includes at least a portion of the annotated text. For each of the aforementioned blocks, an extraction process is performed to extract from the correct answer information storage unit the correct answer information associated with the same decomposition granularity and semantic elements as the aforementioned block. A generation process that calculates the difference between the extracted correct answer information and the block, and generates correction information that includes information about the difference, Execute Document creation support methods.

7. An annotated data storage unit stores information relating to the semantic elements of text composed of text at a predetermined decomposition granularity as annotated text, A computer having a correct answer information storage unit that stores the decomposition granularity of text, semantic elements, and correct answer content in association, An input acceptance process that receives data from documents to be supported, The analysis process involves decomposing the text contained in the document data into blocks of various decomposition granularities, and associating semantic elements associated with the annotated text with each block, which includes at least a portion of the annotated text. For each of the aforementioned blocks, an extraction process is performed to extract from the correct answer information storage unit the correct answer information associated with the same decomposition granularity and semantic elements as the aforementioned block. A generation process that calculates the difference between the extracted correct answer information and the block, and generates correction information that includes information about the difference, To execute Computer program.

Images