JP6165957B1 - Accounting processing apparatus, accounting processing system, accounting processing method, and accounting processing program - Google Patents

Abstract

An accounting processing apparatus, an accounting processing system, an accounting processing method, and an accounting processing program capable of improving journaling accuracy in automatic journal entry from a voucher are provided. An accounting processing apparatus extracts a journal entry element of a voucher by an image analysis unit, and assigns classification information corresponding to the journal element from a classification DB group by a data adding unit, and digitizes / vectorizes the information. After the journal elements are vectorized by the section 13, the journal data is generated by selecting the account item by the journal AI in which the journal determination section 14 has previously learned the journal by machine learning. [Selection] Figure 1

Description

  The present invention relates to an accounting processing apparatus, an accounting processing system, an accounting processing method, and an accounting processing program that take in voucher information such as receipts and receipts and automatically make journals.

  Traditionally, as a voucher accounting process, tax accountants, accountants, and bookkeepers read the information on the voucher one by one, enter account entries based on empirical judgments.

  In this way, when a voucher reads a voucher one by one and makes a journal entry, work efficiency is poor, and the accuracy of the journal entry largely depends on the experience of the person in charge, resulting in variations in the journal entry accuracy. It was.

  Therefore, by reading the contents of the voucher as electronic data using an OCR (Optical Character Reader) device and sending it to the journal analysis center system via the Internet, the result of the journal entry concerning the bookkeeping transaction indicated in the voucher. So-called cloud-type accounting processing systems that allow users to refer to the above have been developed (see Patent Documents 1 and 2).

  Specifically, with the technique described in Patent Document 1, for registered users who own accounting software, refer to the past history of the registered user and extract journals of transactions described in vouchers and similar transactions, and use frequency Suggest recommended journals accordingly. On the other hand, for unregistered users, a journal with the largest number of users of all users (a large number of individuals and companies nationwide) is presented to the user as a recommended journal. In addition, in Patent Document 1, when a new journal is generated or a journal is corrected on the user side, the database used for journaling is updated by the learning means, so that the result of the journal is increased as each user uses it. Optimized.

  In addition, in the technique described in Patent Document 2, information on transactions that are manually input or digitized by OCR is inferred from general business transactions (general reasoning) and journal entries that have been input in the past. Journal entries are made based on three inferences: reasoning (history reasoning) and reasoning from accounting events that occurred in advance, such as accounts receivable and borrowings (consumption reasoning).

JP 2014-235484 A JP 2007-304643 A

  In the above-mentioned patent document 1, the account item and the merchandise group are paired in the mapping table, and the number of journal entries processed in the pair is stored. Further, in the above-mentioned patent document 2, in general reasoning, for “summary” which is an element of the journal, account items are judged using a general database in which words used for judgment of the summary are stored.

  However, in practice, it is difficult to set a pair of account items based on the product or description. For example, even if the product (summary) is listed as “note” on the voucher, if the user is a general company, the account will be “office supplies cost”, and if the user is a retail store, the account will be “ "On". That is, even for the same product (summary), the account item may change depending on the user's business type and transaction status. In addition, even in the case of a transaction for the same product, the account item may change depending on the transaction status such as the business partner, date, and amount.

  On the other hand, in Patent Documents 1 and 2, optimization is performed only for individual users by performing learning based on the history information of individual users. It is not possible to improve the accuracy of journal entries for registered users, journal entries for unregistered business partners, and journal entries based on general reasoning in Patent Document 2.

  Moreover, in the journal based on the number of users of all users with respect to non-registered users in Patent Document 1, not all users have expertise in accounting, and inaccurate journals are included. If learning including such an inaccurate journal is performed, the accuracy of the journal may be reduced.

  In addition, the appearance and description of vouchers differ depending on the issuer, and it is difficult to accurately read all the elements necessary for journalizing by OCR. Reading handwritten characters, company logos, etc. This is not possible with the technical improvement of the conventional method described above.

  For these reasons, the journal accuracy in automatic journal entry from vouchers is still not sufficient.

  The present invention has been made to solve such problems, and an object of the present invention is to provide an accounting processing apparatus, an accounting processing system, and an accounting processing method capable of improving journaling accuracy in automatic journal entry from a voucher. And providing an accounting program.

  In order to achieve the above-described object, the accounting processing apparatus according to the present invention is an accounting processing apparatus that performs automatic journaling, and extracts journal elements including at least a date, a customer, an amount, and a description from image data of a voucher. Journal determination that generates journal data in which the account item for the journal element extracted by the image analysis unit is selected by the image analysis unit and the journal entry AI learned in advance to select the account item corresponding to the journal element by machine learning A section.

  Alternatively, the accounting processing apparatus according to the present invention is an accounting processing apparatus that performs automatic journaling, and extracts an journal element including at least a date, a customer, an amount, and a description from image data of a voucher, and a machine in advance From the journal entry tendency of the journal entry AI learned to select the account item corresponding to the journal element by learning, the journal element necessary for selecting the account item is set as a criterion for each account item, and the image analysis unit A journal determination unit that generates journal data by selecting an account item for the extracted journal element based on the set determination criterion;

  Furthermore, a classification database that stores classification information set corresponding to each journal element, and classification information corresponding to the journal element from the classification database is assigned to the journal element extracted by the image analysis unit. And a data adding unit to be sent to the journal determination unit.

  Furthermore, a journalizing element extracted by the image analysis unit may be digitized and vectorized, and sent to the journal determination unit.

  Further, the journal data generated by the journal determination unit may include an erroneous determination detection unit that detects an erroneous determination of the journal content.

  Further, the journal element may include information on a transaction source, and the journal AI may learn in advance a journal for selecting an account item for the journal element including the transaction source by machine learning.

  In addition, the image analysis unit extracts journal elements from the graphic described in the voucher, the classification database stores classification information based on the graphic, and the data adding unit is configured for the journal journal elements of the graphic, Similar graphics may be detected from the classification database, and classification information corresponding to the graphics may be assigned.

  Further, the image analysis unit extracts journal elements from the appearance including at least the size and color of the voucher, the classification database stores classification information based on the appearance of the voucher, and the data providing unit includes: For the journal elements having the appearance, a similar appearance may be detected from the classification database, and classification information corresponding to the appearance may be given.

  The image analysis unit also extracts journal elements from a corporate number described in a voucher, classification information based on the corporate number is stored in the classification database, and the data adding unit is a journal of the corporate number For the element, a corresponding corporation number may be detected from the classification database, and classification information corresponding to the corporation number may be given.

  In order to achieve the above object, in the accounting processing system according to the present invention, the above-described accounting processing device, a reading device capable of reading the voucher as image data, and the accounting processing device are connected via a communication network. And an information terminal capable of transmitting image data read by the reading device to the accounting processing device and receiving journal data generated by the accounting processing device.

  The information terminal is capable of correcting the received journal data and feeding back the corrected journal data to the accounting processing device, and the journal determination unit of the accounting processing device is based on the fed-back corrected journal data. AI may be further learned.

  Further, a plurality of information terminals are connected to the accounting processing apparatus via the communication network, and information terminals capable of feeding back the corrected journal data are limited to specific information terminals.

  In order to achieve the above-described object, the accounting method according to the present invention is an accounting method that performs automatic journaling, and extracts journal elements including at least a date, a customer, an amount, and a description from image data of a voucher. Journals that generate journal data that selects the account items for the journal elements extracted in the image analysis process by the image analysis process and journal entries AI that have previously learned to select account items corresponding to journal elements by machine learning A determination step.

  Alternatively, in order to achieve the above-described object, the accounting method according to the present invention is an accounting method for performing automatic journalizing, and includes journal elements including at least date, business partner, amount, and description from image data of a voucher. Based on the image analysis process to be extracted and the journal entry tendency of the journal entry AI learned in advance to select the account item corresponding to the journal element by machine learning, the criteria for determining the journal element necessary for selecting the account item for each account item And a journal determination step for generating journal data by selecting an account item based on the set criteria for the journal elements extracted in the image analysis step.

  Further, a data providing step for assigning classification information corresponding to the journal element from the classification database storing the classification information set corresponding to each journal element for the journal element extracted in the image analysis step. You may prepare.

  Furthermore, a journalizing element extracted in the image analysis step may be digitized and vectorized, and a digitizing / vectorizing step for proceeding to the journal determination step may be provided.

  Further, the journal data generated in the journal determination step may include an erroneous determination detection step for detecting an erroneous determination of the journal content.

  Further, the journal element may include information on a transaction source, and the journal AI may learn in advance a journal for selecting an account item for the journal element including the transaction source by machine learning.

  Further, in the image analysis step, journal elements are extracted from the graphic described in the voucher, classification information based on the graphic is stored in the classification database, and in the data adding step, the journal element of the graphic is Similar graphics may be detected from the classification database, and classification information corresponding to the graphics may be assigned.

  Further, in the image analysis step, at least the size of the voucher, the journal element is extracted from the appearance including the color, and the classification database stores the classification information based on the appearance of the voucher. For the journal elements having the appearance, a similar appearance may be detected from the classification database, and classification information corresponding to the appearance may be given.

  Further, in the image analysis step, journal elements are also extracted from the corporate number described in the voucher, classification information based on the corporate number is stored in the classification database, and in the data granting step, the journal number of the corporate number is stored. For the element, a corresponding corporation number may be detected from the classification database, and classification information corresponding to the corporation number may be given.

  In order to achieve the above-described object, the accounting processing program according to the present invention causes a computer to execute the above-described journal processing method.

  According to the present invention using the above means, it is possible to improve the accuracy of journal entry in automatic journal entry from a voucher.

It is a system configuration figure showing an accounting processing system concerning one embodiment of the present invention. It is explanatory drawing which shows the structural example of a classification database group. It is a system configuration figure showing a learning system for forming journal AI. It is a flowchart showing the automatic journal entry control routine in the accounting processing system concerning this embodiment. It is explanatory drawing which shows data structural example (a)-(e).

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a system configuration diagram showing an accounting processing system including an accounting processing apparatus according to an embodiment of the present invention.

  As shown in FIG. 1, the accounting processing system according to the present embodiment includes various devices on the user side and accounting processing devices on the accounting processing service provider side via a communication network 1 such as the Internet or VPN (Virtual Private Network). 10 are connected to each other. For simplification of explanation, only one user is shown in FIG. 1, but the accounting processing apparatus 10 can be connected to a plurality of users via the communication network 1.

  The user is, for example, an expert such as a tax accountant and an accountant, or a corporation or an individual who directly performs accounting processing, and includes at least the reading device 2 and the information terminal 3.

  The reading device 2 is an optical device such as a scanner or a camera, for example, and is a device that can capture a voucher as image data. The term “voucher” in this embodiment and claims refers to receipts, receipts, other receipts, invoices, invoices, documents for proof of financial transfer, IC cards such as electronic money, etc. Also includes transaction information.

  The information terminal 3 is a portable terminal such as a personal computer (hereinafter referred to as a PC), a smartphone, a tablet PC, and a mobile phone, and is a terminal that can display at least web information.

  The user can acquire the image data of the voucher by the reading device 2 and transmit it to the accounting processing device 10 by the information terminal 3 and can receive information from the accounting processing device 10. In FIG. 1, the reading device 2 and the information terminal 3 are shown as separate bodies, but the reading device 2 and the information terminal 3 may be integrated like a mobile terminal with a camera.

  On the other hand, the accounting processing service provider is a business operator who provides accounting processing service by so-called cloud computing and manages the accounting processing apparatus 10.

  The accounting processing apparatus 10 has one or a plurality of servers that execute a journalizing process based on a program, and functionally mainly includes an image analyzing unit 11, a data adding unit 12, a numerical / vectorizing unit 13, and a journal determining unit. 14 and an erroneous determination detection unit 15. Further, the accounting processing apparatus 10 has a database (hereinafter, referred to as DB) in which various information is stored, and specifically has a classification DB group 16.

  The image analysis unit 11 has a function of receiving image data of a voucher sent from a user and extracting information (unformatted data) that is a journal element from the image data. Journal elements include, for example, date, amount, business partner, description (including proviso and product name), business source (including address), and numbers, letters, and figures (for example, logo marks, seals, and other companies) corresponding to these. The identifiable pattern) and the appearance of the voucher (for example, the size and color of the receipt) are extracted by OCR or an image analysis function.

  For example, for dates, numbers before and after characters such as “year”, “month”, and “day” and symbols such as “/” are extracted. For the amount of money, the numbers before and after the symbol “¥” and “Yen” are extracted. In addition, for the business partner name, the characters before and after the letters “corporation” and “(stock)”, the logo mark, the telephone number, and the appearance of the voucher are extracted. For the summary, the character following the “but” character is extracted. For the transaction source, the character preceding the character “sama” is extracted.

  Here, for business partners, when only logo marks, seals, or telephone numbers are extracted, although not shown, information retrieved from a company information DB in which company information is stored in advance, or information published on the Internet is used. Acquire business partner name by searching. In addition, as for the transaction source, when there is no address description such as a receipt, the user who has transmitted the image data may be the transaction source, or the company that is the customer of the user is set in advance and the user The image data transmitted from may be set as the trading company.

  The journal element is not limited to this, and the numbers, characters, and figures used for extracting the journal element are not limited to this. For example, if the voucher indicates the quantity of the purchased item, the quantity may be included as a journal element, and if the information such as the name and number of people present is indicated, May be included as a journal element. Moreover, you may extract the number (corporate number, establishment number) set in order to identify each company.

  Thus, the unformatted data of the journal elements extracted by the image analysis unit is sent to the data adding unit 12 in such a state that at this point in time, the data corresponding to which journal element is not organized.

  The data assigning unit 12 is connected to the classification DB group 16 and assigns the classification information corresponding to each journal element stored in the classification DB group 16 to the unformatted data, so that the formatted data of the journal element is obtained. It has a function to generate.

  Here, referring to FIG. 2, an example of the configuration of the classification DB group 16 is shown. As shown in FIG. 2, the classification DB group 16 includes a date DB group 20 corresponding to the date of the journal element and a transaction corresponding to the business partner. It has a destination DB group 21, a summary DB group 22 corresponding to the summary, and a transaction source DB group 23 corresponding to the transaction source.

  The date DB group 20 includes a holiday DB 20a in which holiday information is stored, a day DB 20b in which day information is stored, and a season DB 20c in which four season information is stored. The supplier DB group 21 is classified according to the type of business of the company, and has, for example, a food and drink DB 21a, a traffic DB 21b, a retail DB 21c, and the like. The summary DB group 33 is classified according to the types of products and services (services), and includes, for example, a food DB 22a, a pharmaceutical DB 22b, an electronic device DB 22c, a daily goods DB 22d, a book DB 22e, and the like. The transaction source DB group 23 is classified according to user information contracted with a service provider as a transaction source, and includes, for example, a user business DB 23a, an address DB 23b, an employee number DB 23c, and the like. The DB is not limited to this, and various DBs exist depending on the journal elements. For example, as the supplier DB group 21, a logo mark DB storing a company logo mark (graphic), an appearance DB storing a characteristic voucher appearance according to the type of business, a company name and a type of business corresponding to the corporate number, etc. A stored corporate number DB or the like may be provided.

  The data adding unit 12 converts the unformatted data corresponding to the date from the date DB group 21 and the unformatted data corresponding to the supplier from the supplier DB group 21 to the unformatted data corresponding to the summary. On the other hand, for the unformatted data corresponding to the transaction source from the summary DB group 22, the corresponding classification data is detected from the transaction source DB group 23, and one or a plurality of classification data is assigned.

  For unformatted data corresponding to monetary amounts, for example, 0 to 5,000 yen, 5,001 yen to 10,000 yen, 10,001 yen to 200,000 yen, 200,001 yen, etc., are assigned in stages according to the amount of money. To do.

  The digitizing / vectorizing unit 13 converts the shaped data provided with the classification data corresponding to the journal elements by the data providing unit 12 into vector data suitable for the journal. The shaped data is data composed of characters and numbers, which is digitized into a data format that can be easily recognized by the journal entry determination unit, and vectorized to generate vector data.

  The journal determination unit 14 has a journal AI (Artificial Intelligence), and generates journal data from the vector data of journal elements using the journal AI. The journal entry AI is an AI specialized in journal entries learned by using specialist journals such as tax accountants and high-precision journal data carried out in the past as teacher data.

  Specifically, referring to FIG. 3, a system configuration diagram showing a learning system for forming the journal entry AI is shown.

  The learning system 30 shown in FIG. 1 includes one or a plurality of servers (computers), and functionally includes a numerical / vectorization unit 31, a learning unit 32, and a learned model DB 33. Then, journal element data based on the voucher information and the corresponding journal data are input to the learning system 30. As the combination of the journal element data and the journal data to be input, highly accurate data previously journalized by a specialist or a company is used.

  The digitizing / vectorizing unit 31 converts the input journal element data and journal data into vector data in the same manner as the vectorization of the journal processing apparatus.

  The learning unit 32 learns the relationship between combinations of vectorized journal element data and journal data, thereby determining the journal element data in a composite manner, and outputting journal data in which an appropriate account item is selected. A journal entry AI is formed.

  The learned model DB 33 stores journal entries AI formed by the learning unit 32. The journal AI is not limited to one. For example, by inputting and learning journal element data and journal data specialized for the user's business type or the user alone, the journal AI optimized for the user business type or the user alone is formed. May be. Thereby, it is possible to realize a journal with higher accuracy for the user.

  The journal entry AI formed in this way is characterized by the journal element that is the key to determine the account item among the journal elements such as business partner, date, amount, description, and transaction source, depending on the account item. The account item is selected based on the journal element that is a feature point.

  For example, if it is an account item for entertainment expenses, it is characterized by the business partner and the business partner, if it is travel expense, it is characterized by the business source information, if it is office supplies, it is characterized by the business source and description. For example, one or more feature points are set according to the account item. As a result, the accuracy of journal entry determination can be improved, and efficient journal entry determination can be performed using the minimum necessary elements for account item selection. When a plurality of feature points are set, priority may be set between feature points.

  Returning to FIG. 1, the journal determination unit 14 uses the journal AI formed as described above to select an appropriate account item for the journal element that is vector data, and outputs it as journal data. If the journal element is a journal element that cannot be journaled by the journal entry AI, journal data that does not correspond to the account item is output.

  The erroneous determination detection unit 15 has a function of detecting an erroneous determination on the journal data generated by the journal determination unit 14. For example, an abnormality determination value is set for a journal element or account item, and an erroneous determination is made based on whether or not the abnormality determination value is exceeded. The abnormality determination value is set statistically, for example, a normal distribution is created, and a predetermined σ value or a predetermined value such as ± 1σ (= 68.27%), ± 2σ (95.45%), ± 3σ (99.73%) is set. Set to the corresponding value.

  For example, because the image analysis is inaccurate, the summary is “coffee”, but the amount is “10 yen”, “100,000 yen”, etc. When a specific numerical value or account item is selected, such as being selected, the misjudgment detection unit 15 regards it as abnormal.

  The misjudgment detection unit 15 provides data for displaying an abnormality different from the normal time, such as displaying in red for example, in order to prompt the user to confirm the possibility of erroneous determination for the journal data that has detected an abnormality. The added completed journal data is transmitted to the user.

  Then, when the user confirms the final journal data and corrects the final journal data or creates a new journal, the information is fed back to the accounting processing apparatus 10. The fed-back corrected journal data is digitized and vectorized by the digitizing / vectorizing unit 13 and used for further learning of the journal AI by the journal judging unit 14. In addition, in order to maintain the learning accuracy of the journal entry AI by feedback, users who can perform feedback may be limited to tax accountants and accountant professional users. That is, only the specific information terminal 3 has a feedback function, or the accounting processing apparatus 10 is set to accept only the feedback from the specific information terminal 3.

  Next, referring to FIG. 4 and FIG. 5, FIG. 4 shows a flowchart showing an automatic journalizing routine (accounting processing method) in the accounting system configured as described above. (E) shows an example of each data structure, and will be described below with reference to the flowchart of FIG. 4 with reference to FIG.

  First, in step S <b> 1 of the automatic journal entry routine, the user scans a voucher such as a receipt by the reading device 2 to convert it into image data, which is transmitted to the accounting processing device 10 via the information terminal 3.

  In subsequent step S2, the image analysis unit 11 of the accounting processing apparatus 10 generates unformatted data as shown in FIG. 5A in which the journal elements are extracted from the received image data (image analysis step).

  In step S3, the data assigning unit 12 assigns the classification information to the unshaped data based on the information of the classification DB group 16, and generates the shaped data as shown in FIG. 5B (data providing step).

  In step S4, the digitizing / vectorizing unit 13 digitizes and vectorizes (digitizes) the shaped data to generate vector data as shown in FIG. 5C (numericalizing / vectorizing step).

  In step S5, the journal determination unit 14 determines whether or not an appropriate account item can be generated from the vector data of the journal elements by the journal AI, that is, whether or not the journal is possible. If the determination result is true (Yes), that is, if the journal entry by the journal entry AI is possible, the process proceeds to step S6.

  In step S6, the journal determination unit 14 generates journal data as shown in FIG. 5D based on the journal AI (journal determination step).

  In the next step S7, the erroneous determination detection unit 15 determines whether or not there is an erroneous determination in the journal data (error determination detection step). If the determination result is true (Yes), and if the journal data is abnormal, the process proceeds to step S8.

  In step S8, completed journal data to which erroneous determination data indicating that there is an abnormality in the journal data is added is generated, and the process proceeds to step S9. If the determination result in step S7 is false (No), that is, if there is no abnormality in the journal data, the completed journal data is generated as it is, and the process proceeds to step S8. Note that the completed journal data in this case may be completed journal data provided with data indicating that there is no abnormality in the journal data, as shown in FIG. 5E, for example.

  In step S9, the completed journal data is transmitted from the accounting processing apparatus 10 to the user.

  In step S10, the transaction processing apparatus 10 determines whether or not there is correction feedback from the user for the completed journal data transmitted in step S9. If the determination result is false (No), that is, if there is no correction feedback, the routine ends. On the other hand, if the determination result is true (Yes), that is, if the journal is corrected by the user, the process proceeds to step S11.

  In step S11, after the fed-back corrected journal data is converted into vector data by the digitizing / vectorizing unit 13, the journal determining unit 14 learns the journal AI based on the vector data, and the routine ends.

  Further, in the above step S5, when the journal entry by the journal AI is impossible due to an unprecedented journal element, the determination result is false (No), and the process proceeds to step S12.

  In step S12, the journal determination unit 14 generates journal data that does not correspond to the account item and transmits it to the user.

  In subsequent step S13, since the account item is not applicable, manual journaling is performed by the user, and the result is fed back to the accounting processing apparatus 10. In step S10, the manually journaled corrected journal data is learned for the journal AI based on the fed-back new journal information.

  As described above, in the journal processing system according to the present embodiment, the voucher journal elements are extracted by the image analysis unit 11, and the classification information corresponding to the journal elements is added from the classification DB group 16 by the data adding unit 12. After the journalizing element is digitized and vectorized by the conversion / vectorization unit 13, the journal data is generated by selecting the account item by using the journal entry AI in which the journal determination unit 14 has previously learned the journal by machine learning. Yes.

The journal entry AI is an AI specialized for journals learned using highly accurate journal data as teacher data. By using such a journal entry AI, the journal entry elements of the voucher are judged in a complex manner, and an accurate account item is recorded. Will be selected.
In particular, the journal element (unformatted data) extracted by image analysis is given classification information according to the journal element by the data adding unit 12, thereby improving the accuracy of the journal element information, and then the journal determining unit 14. The journal accuracy in can be improved .

  For example, if the company name of the oil company is extracted from the letters and logos of the receipt, the oil company classification is given to the business partner, and the transaction is performed according to the characteristics of the determination of the account item of the vehicle transportation expenses in the journal entry AI. If the destination is set, the account item “vehicle transportation expense” is selected from the name of the supplier.

  In addition, in the case of a specialized shop for stationery stores, if “A stationery store” or the like is listed as the customer of the receipt, the journal entry AI will be immediately available if the customer is set as a feature of the account item judgment. “Office supplies costs” can be selected.

  In addition, there are many cases in which “office supplies costs” are usually used for receipts purchased at home electronics mass merchandise, such as “Gift gifts”. If a description is set for the feature point of the account item for entertainment expenses in the journal entry AI, the account item “Entertainment expenses” can be selected. It is.

  In addition, if the journal AI is the characteristic of the judgment of the account item for the receipt that purchased the computer, the transaction source is a general company and the cost is 100,000 yen or less In contrast to selecting an account item of “office supplies cost” or “consumables cost”, if the transaction source is a general company and is more than 100,000 yen, “construction equipment” is selected.

  In addition, since the journal elements are digitized and vectorized and then journalized by the journal AI, the calculation processing by the journal AI can be simplified and the calculation speed can be improved.

  Furthermore, since the journal data generated by the journal AI is not necessarily correct, the erroneous determination detection unit 15 detects the erroneous determination of the journal data, so that the accuracy can be further improved.

  Further, since the transaction source is included as the journal element, an account item corresponding to the user's business type or the like is selected.

  Thus, for example, if a transaction source is a characteristic feature of the account item judgment in the journal entry AI for a receipt for purchasing a drink at a supermarket, the journal entry AI is “ Whereas the account item “meeting fee” or “welfare expense” is selected, “purchase” is selected when the transaction source is a restaurant that provides drinks to customers.

  In the case of purchasing a book, if the business partner, quantity, and business source are set as feature points, if the business partner is a bookstore and the quantity is as small as 1 or the like, "newspaper book expenses" If there is a large quantity at a book wholesaler and the transaction source is a general company, “advertising expenses”, if this transaction source is a bookstore, “purchase”, etc. Courses can be selected.

  Further, the image analysis unit 11 extracts the voucher logo mark (figure) and the appearance of the voucher, and the data adding unit 12 adds data of journal elements such as business partners based on the logo mark and the appearance of the receipt. Therefore, even when the business partner is not specified as a character in the voucher, it can be extracted as a journal element.

  As a result, for example, when a receipt of a coffee shop is entered, it is possible to extract a business partner from the logo mark at the top of the receipt, and make a coffee shop = meeting decision to select a “meeting fee”.

  In addition, for example, colored handwritten receipts unique to izakayas and special receipts of airlines, railway agencies, etc. are given classification information of the industry from the appearance of the receipt, from which `` entertainment expenses '', `` It is also possible to select “travel expenses”.

Further, by extracting the corporate number, it becomes possible to accurately determine the business partner .

  In addition, when the journal data is corrected on the user side, the accuracy of the journal entry AI can be further improved by feeding back the corrected journal data to the journal entry AI and further learning.

  In particular, the feedback of the corrected journal data is limited to a tax accountant or accountant expert user, so that a reduction in accuracy due to learning an incorrect journal can be prevented.

  For this reason, according to the accounting processing apparatus, accounting processing system, accounting processing method, and accounting processing program in the present embodiment, the accuracy of journal entry in automatic journal entry from a voucher can be improved.

  This is the end of the description of the embodiment of the present invention, but the aspect of the present invention is not limited to this embodiment.

  For example, in the above embodiment, the journal entry AI that has been machine-learned in advance for journal entries is directly used by the journal decision unit 14, but the journal decision unit selects account items corresponding to journal elements in advance through machine learning. The journal elements necessary for selecting the account item are set as judgment criteria for each account item based on the journal entry trend of the journal entry AI that has been learned, and the extracted journal elements are accounted based on the set judgment criteria. Journal data may be generated by selecting a subject.

  In other words, instead of using the journal entry AI directly in the journal entry judgment section, analyze the journal tendency learned by the journal entry AI, and extract the journal elements (feature points) necessary for the judgment of the account item for each account item. Then, a journal determination program is created and the program is applied to the journal determination unit.

  As described above, by using only the journal tendency found from the journal entry AI without directly incorporating the journal entry AI as the accounting processor, it is possible to easily realize an accounting processor with sufficient journal accuracy. it can.

  In the above embodiment, the accounting apparatus 10 assigns the classification data to the journal elements extracted by the image analysis unit 11 by the data adding unit 12, thereby further improving the journal accuracy in the journal determination unit 14. As shown, the journal entry extracted by the image analysis unit without the data adding unit may be directly determined by the journal determination unit. In this case as well, as indicated by the dotted line in FIG. 1, it is preferable to determine the journal by the journal determination unit after digitizing and vectorizing the journal elements extracted by the image analysis unit.

DESCRIPTION OF SYMBOLS 1 Communication network 2 Reading apparatus 3 Information terminal 11 Image analysis part 12 Data provision part 13 Digitization / vectorization part 14 Journal determination part 15 Misjudgment detection part 16 Classification DB group

Claims (22)

An accounting device that performs automatic journaling,
An image analysis unit that extracts journal elements including an appearance including at least a date, a supplier, an amount, a description , and at least a size and a color of the voucher from image data of the voucher;
In advance by machine learning by the learning to select an account corresponding to the journal element journal AI, a journal determination unit to generate a journal data selected accounts for journal element extracted by the image analysis unit,
An accounting processing apparatus comprising: An accounting device that performs automatic journaling,
An image analysis unit that extracts journal elements including at least a date, a supplier, an amount, a description, and a graphic described in the voucher from image data of the voucher;
In advance by machine learning by the learning to select an account corresponding to the journal element journal AI, a journal determination unit to generate a journal data selected accounts for journal element extracted by the image analysis unit,
An accounting processing apparatus comprising: Further, each journal element includes at least one of a date database corresponding to the date of the journal element, a customer database corresponding to the business partner of the journal element, and a summary database corresponding to the description of the journal element. One or more classification databases that store the set classification information;
A data adding unit that adds one or more pieces of classification information corresponding to the journal element from the classification database and sends the journal element extracted by the image analysis unit to the journal determination unit. Or the accounting processing apparatus of 2. The accounting processing apparatus according to claim 3, wherein the data assigning unit assigns a classification classified stepwise according to the amount of money to the amount of the journal element. The accounting process according to any one of claims 1 to 4 , further comprising: a digitizing / vectorizing unit that digitizes and vectorizes the journal elements extracted by the image analysis unit and sends the journalizing elements to the journal determining unit. apparatus. The accounting processing apparatus according to any one of claims 1 to 5 , further comprising a misjudgment detection unit that detects a misjudgment of the journal content for the journal data generated by the journal judgment unit. The journal element also includes information about the transaction source,
The accounting apparatus according to any one of claims 1 to 6 , wherein the journal AI has previously learned a journal for selecting an account item for a journal element including the transaction source by machine learning. The image analysis unit also extracts journal elements from the number set to identify the company described in the voucher,
In the classification database, classification information based on a number set to identify a company is stored,
In order to identify the company by detecting the number set for identifying the company in question from the classification database for the journal element of the number set to identify the company The accounting processing apparatus according to claim 3, wherein classification information corresponding to a set number is given. The journal determination unit determines a journal element necessary for selecting the account item for each account item from the journal tendency of the journal entry AI learned to select the account item corresponding to the journal element by machine learning in advance. The accounting processing apparatus according to claim 1, wherein the journal data is generated by selecting an account item based on the set determination criterion for the journal element. The accounting apparatus according to any one of claims 1 to 9,
A reading device capable of reading the voucher as image data;
An information terminal connected to the accounting processing device via a communication network and capable of transmitting image data read by the reading device to the accounting processing device and receiving journal data generated by the accounting processing device And an accounting system. The information terminal is capable of correcting the received journal data and feeding back the corrected journal data to the accounting processing device,
The accounting processing system according to claim 10, wherein the journal determination unit of the accounting processing device further learns the journal AI based on the corrected journal data fed back. A plurality of information terminals are connected to the accounting processing device via the communication network,
The accounting processing system according to claim 11, wherein an information terminal capable of feeding back the corrected journal data is limited to a specific information terminal. An accounting method for automatic journal entry by a computer ,
An image analysis step of extracting journal elements including an appearance including at least a date, a supplier, an amount, a description , and at least a size and a color of the voucher from image data of the voucher;
In advance by machine learning by the learning to select an account corresponding to the journal element journal AI, a journal determination step of generating a journal data selected accounts for journal element extracted by the image analysis step,
Accounting method. An accounting method for automatic journal entry by a computer ,
An image analysis step of extracting journal elements including at least a date, a supplier, an amount of money, a description, and a graphic described in the voucher from image data of the voucher;
In advance by machine learning by the learning to select an account corresponding to the journal element journal AI, a journal determination step of generating a journal data selected accounts for journal element extracted by the image analysis step,
Accounting method. Furthermore, for the journal elements extracted in the image analysis step, at least a date database corresponding to the date of the journal element, a supplier database corresponding to the business partner of the journal element, and a description of the journal element A data providing step of assigning one or more classification information corresponding to the journal elements from one or more classification databases storing classification information set corresponding to each journal element, including any of the summary databases The accounting method according to claim 13 or 14.   The accounting processing method according to claim 15, wherein in the data adding step, the amount of the journal element is assigned a stepwise classification according to the amount of money. The accounting according to any one of claims 13 to 16 , further comprising a quantification / vectorization step of digitizing and vectorizing the journal elements extracted in the image analysis step and proceeding to the journal determination step. Processing method. The accounting processing method according to any one of claims 13 to 17 , further comprising a misjudgment detection step of detecting a misjudgment of a journal content for the journal data generated in the journal judgment step. The journal element also includes information about the transaction source,
The accounting method according to any one of claims 13 to 18 , wherein the journal AI learns in advance journals for selecting account items for journal elements including the transaction source by machine learning. In the image analysis process, journal elements are extracted from the number set to identify the company described in the voucher,
In the classification database, classification information based on a number set to identify a company is stored,
In the data granting step, for the journal element of the number set for specifying the company, the number set for specifying the company in question is detected from the classification database to specify the company The accounting processing method according to claim 15 , wherein classification information corresponding to a set number is assigned.   In the journal determination step, the journal element necessary for selecting the account item for each account item is determined based on the journal entry tendency of the journal entry AI learned in advance to select the account item corresponding to the journal element by machine learning. 21. The accounting processing method according to claim 13, wherein journal data is generated by selecting an account item based on the set determination criterion for the journal element. An accounting processing program for causing a computer to execute the accounting processing method according to any one of claims 13 to 21.

Images