Character recognition result determination device, character recognition result determination method, character recognition result determination program, and character recognition system
The character recognition result determination device improves OCR accuracy by using multiple related OCR results and database matching to correct and validate the initial OCR result, addressing the challenge of low initial accuracy.
Patent Information
- Application Number
- JP2023083984
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-22
- Publication Date
- 2024-12-04
AI Technical Summary
Existing OCR technologies struggle to improve reading accuracy when the initial OCR result is poor, as the matching rate is insufficient for effective correction.
A character recognition result determination device that acquires OCR results of multiple related items, compares them with matching data from a search database, and corrects the OCR result if a predetermined ratio of matching data exceeds a threshold, ensuring high reliability.
Enhances reading accuracy by leveraging multiple OCR results to correct and validate the initial OCR result, even when initial accuracy is low.
Smart Images

Figure 2024167725000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a character recognition result determination device, a character recognition result determination method, a character recognition result determination program, and a character recognition system. [Background technology]
[0002] Character recognition such as OCR (Optical Character Recognition) is becoming widespread. It analyzes images of paper, detects patterns corresponding to characters contained therein, and recognizes the written content as character data. One method for improving the accuracy of OCR is to compare the correct information with dictionary data that contains the correct information, and if there is a mismatch with the OCR result, correct the OCR result with the value of the dictionary data.
[0003] For example, the technology described in Patent Document 1 discloses a technology that improves the accuracy of reading addresses by searching dictionary data for address information read by OCR and comparing the data, as well as searching for addresses in dictionary data using other OCR results such as telephone numbers and postal codes, and comparing these with the address read by OCR. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Japanese Patent Application Publication No. 6-103404 Summary of the Invention [Problem to be solved by the invention]
[0005] However, the technology described in Patent Document 1 determines whether to correct the OCR results with the matching data based on the ratio of the number of matching characters as a result of the comparison (matching), and in this case, it is assumed that a part of the target OCR results will be corrected. Therefore, if the reading accuracy of the OCR results of the target item is too poor, even if the matching data can be obtained correctly, there is a problem that the matching rate is low and correction is not possible.
[0006] The present invention has been made in consideration of the above circumstances, and has an objective of improving the reading accuracy by using other OCR results even when the reading accuracy of the OCR result for the target item is poor. [Means for solving the problem]
[0007] The above-mentioned object of the present invention can be achieved by the following configuration.
[0008] (1) A character recognition result determination device comprising: a character recognition result acquisition unit that acquires OCR results of three or more related items; a matching data extraction unit that uses the OCR results of other items related to one of the three or more items to extract matching data for the target item by referring to a search database in which the values of each of the items are stored in correspondence with each other; and a reliability determination unit that compares the OCR result of the target item with each of the matching data obtained from the other items and determines whether the number of identical values exceeds a predetermined percentage of the total.
[0009] (2) The character recognition result determination device described in (1) above, wherein the reliability determination unit corrects the OCR result of the target item to the same value of the matching data when multiple matching data are the same value and exceed the specified ratio, and when the OCR result of the target item is different from the same value.
[0010] (3) The character recognition result determination device described in (1) above, further comprising a notification unit that compares the OCR result of the target item with multiple matching data and outputs information that determination is impossible if there are no identical values exceeding the specified percentage.
[0011] (4) A character recognition result determination method for a character recognition result determination device which executes the following steps: a character recognition result acquisition step of acquiring OCR results of three or more related items; a matching data extraction step of extracting matching data of one of the three or more items related to a target item by using the OCR results of other items related to the target item to refer to a search database in which the values of each of the items are stored in correspondence with each other; and a reliability determination step of comparing the OCR result of the target item with each of the matching data obtained from the other items to determine whether the number of identical values exceeds a predetermined percentage of the total.
[0012] (5) A character recognition result determination program for causing a computer to execute the steps of: obtaining OCR results of three or more related items; extracting matching data for one of the three or more items related to the target item by referring to a search database in which the values of each of the items are stored in correspondence with each other, using the OCR results of the target item for other items; and comparing the OCR result of the target item with each of the matching data obtained from the other items to determine whether the number of identical values exceeds a predetermined percentage of the total.
[0013] (6) A character recognition system comprising a character recognition device having a character recognition function and a character recognition result determination device that determines reliability of the character recognition function, wherein the character recognition result determination device comprises: a character recognition result acquisition unit that acquires OCR results of three or more related items from the character recognition device; a matching data extraction unit that uses the OCR results of other items related to one of the three or more items to extract matching data of the target item by referring to a search database in which the values of each of the items are stored in correspondence with each other; and a reliability determination unit that compares the OCR result of the target item with each of the matching data obtained from the other items and determines whether the number of identical values exceeds a predetermined percentage of the total. Effect of the Invention
[0014] According to the present invention, even if the reading accuracy of the OCR result of the target item is poor, the reading accuracy can be improved by using other OCR results. [Brief description of the drawings]
[0015] [Figure 1] 1 is a functional block diagram showing a configuration of a character recognition system including a character recognition result determination device according to an embodiment of the present invention. [Diagram 2] 11 is a diagram for explaining a process of extracting matching data from a search database. FIG. [Diagram 3] 11A and 11B are diagrams for explaining a result of determination regarding the reliability of character recognition by a reliability determination unit according to the embodiment. [Figure 4] FIG. 13 is a diagram showing an example of a case where the OCR result of one target item matches three pieces of two matching data (Case 1). [Diagram 5] FIG. 13 is a diagram showing an example of a case where the OCR result of one target item matches one piece of matching data (Case 2). [Figure 6] FIG. 13 is a diagram showing an example of a case where two pieces of matching data other than the OCR result of one target item match (Case 3). [Figure 7]FIG. 13 is a diagram showing an example of a case where the OCR result of one target item does not match the two pieces of matching data (Case 4). [Figure 8] 4 is a flowchart showing the flow of processing executed by the character recognition result determination device according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0016] Hereinafter, an embodiment of the present invention (hereinafter, referred to as "the present embodiment") will be described with reference to the drawings. However, in all the drawings in this specification, components having corresponding functions are given the same reference numerals, and the description thereof will be omitted as appropriate.
[0017] FIG. 1 is a functional block diagram showing the configuration of a character recognition system 1 including a character recognition result determination device 10 according to this embodiment. The character recognition system 1 includes a character recognition device 20 and a character recognition result determination device 10 that is communicatively connected to the character recognition device 20. The character recognition result determination device 10 is a computer that includes a processor (not shown) and a storage unit 13, and embodies each functional unit by executing a character recognition result determination program (not shown). The character recognition result determination device 10 is communicatively connected to a search database (DB) 30 via a network or the like (not shown).
[0018] The character recognition device 20 is a device equipped with an OCR function that reads handwritten or printed characters using an image scanner or the like, recognizes the text portion of the image data, and converts it into character data. In addition, by using AI-OCR, the character recognition device 20 is provided with a function of not only simply recognizing characters but also automatically extracting reading positions and items using AI (artificial intelligence).
[0019] The character recognition result determination device 10 obtains information on OCR results related to three or more items related to each other from the character recognition device 20 (hereinafter referred to as "OCR results" or "OCR result information"). Then, the character recognition result determination device 10 searches various databases (search databases) for the OCR result of a target item (target item) using two or more other OCR results related to the item as a key, and extracts the hit value as matching data. Then, the character recognition result determination device 10 compares the OCR result with the multiple matching data, and determines that the reliability is high if the number of identical values in the OCR result and the matching data exceeds a predetermined ratio of the whole (number of items), and determines that the reliability of the OCR result is low if the ratio is equal to or less than the predetermined ratio. If the predetermined ratio is exceeded and the OCR result of the target item is different from the identical value, the character recognition result determination device 10 replaces (corrects) the OCR result with the identical value (details will be described later). This allows the character recognition result determination device 10 to improve the reading accuracy of character recognition.
[0020] As shown in FIG. 1, the character recognition result determination device 10 includes a control unit 11, an input / output unit 12, and a storage unit 13.
[0021] The input / output unit 12 inputs and outputs information between other devices (such as the character recognition device 20 and the search database 30). The input / output unit 12 is composed of a communication interface that transmits and receives information via a communication line, and an input / output interface that inputs and outputs information between an input device such as a keyboard (not shown) and an output device such as a monitor (not shown).
[0022] The storage unit 13 is configured with a hard disk, a flash memory, a RAM (Random Access Memory), or the like. The storage unit 13 temporarily stores a program (character recognition result determination program) for executing each functional unit of the control unit 11, and information required for processing by the control unit 22 (for example, matching data, etc.).
[0023] The control unit 11 is responsible for all the processes executed by the character recognition result determination device 10 , and includes a character recognition result acquisition unit 111 , a matching data extraction unit 112 , a reliability determination unit 113 , and a notification unit 114 .
[0024] The character recognition result acquisition unit 111 acquires OCR result information from the character recognition device 20 . In this embodiment, examples of information read by the OCR function of the character recognition device 20 are three items related to the company name written on the invoice: company name, address, and telephone number. In addition to these three items, the corporate number and the name of the president may be included. Furthermore, the information is not limited to character recognition of the invoice, and may include items such as the name of the person in charge of various reports, employee number, extension number, and email address. It may also include items such as the company name, telephone number, and the company's URL (Uniform Resource Locator). The character recognition result acquisition unit 111 acquires OCR result information for three items (company name, address, and telephone number) obtained by the AI-OCR function of the character recognition device 20.
[0025] The matching data extraction unit 112 searches the search database 30 for a certain target item (e.g., "company name") using the OCR results of two or more other items related to the target item (here, the OCR results of "address" and "telephone number") as keys, and extracts the hit value (company name) as matching data. Of the three or more items, one item selected to determine the reliability of character recognition will be referred to as a "target item," and items other than the target item will be referred to as other items.
[0026] Here, the search database (DB) 30 is a database in which values of at least two or more items out of three or more items related to each other are registered in association with each other. This search database 30 is, for example, a telephone number search database that correlates "company names" with "telephone numbers" or an address search database that correlates "company names" with "addresses". The matching data extraction unit 112 acquires matching data by accessing these telephone number search databases and address search databases that exist on the Web or in the cloud via the Internet or the like. The search database 30 may be provided in the storage unit 13 in the character recognition result determination device 10. Alternatively, the character recognition result determination device 10 may be externally connected to a device provided with the search database 30, and the matching data extraction unit 112 may acquire the matching data.
[0027] FIG. 2 is a diagram for explaining the process of extracting matching data from the search database 30. As shown in FIG. Here, an example is shown in which the OCR results include the information "company name" (symbol K1), "address" (symbol J1), and "telephone number" (symbol D1), as shown in the shaded areas in Figure 2.
[0028] <Company name matching data extraction> A case where matching data for the target item "company name" is extracted will be described. The matching data extraction unit 112 searches the address search database using the OCR result (code J1) for "address", which is a target item related to "company name", as a key, and extracts the hit information as matching data for "company name" (code TK1).
[0029] In addition, the matching data extraction unit 112 searches the telephone number search database using the OCR result (code D1) for the target item "telephone number" related to "company name" as a key, and extracts the hit information as matching data for "company name" (code TK2).
[0030] <Extracting address matching data> A case where matching data for the target item "address" is extracted will be described. The matching data extraction unit 112 searches the address search database using the OCR result (code K1) for the target item "company name" related to "address" as a key, and extracts hit information as matching data for "address" (code TJ1).
[0031] In addition, the matching data extraction unit 112 searches the telephone number search database using the OCR result (code D1) for the target item "telephone number" related to "address" as a key, and extracts the hit information as matching data for "address" (code TJ2).
[0032] <Extract phone number matching data> A case where matching data for a "telephone number" is extracted as a target item will be described. The matching data extraction unit 112 searches the telephone number search database using the OCR result (code K1) for the target item "company name" related to "telephone number" as a key, and extracts the hit information as matching data (code TD1) for "telephone number".
[0033] Furthermore, since there is no database that allows direct search of "address" from "phone number" for "address", which is a target item related to "phone number", the matching data extraction unit 112 first searches the address search database using the OCR result for "address" (code J1) as a key to obtain the search result for "company name".Then, the matching data extraction unit 112 searches the telephone number search database using the search result "company name" as a key, and extracts the hit information as matching data for "phone number" (code TD2).
[0034] Returning to Fig. 1, the reliability determination unit 113 compares the OCR result of the target item with each of the matching data obtained from the other items, and determines whether the number of identical values exceeds a predetermined ratio of the total (number of items). Here, the reliability determination unit 113 determines that the reliability of the OCR result is high if the OCR result matches the matching data at a ratio greater than a predetermined ratio, and determines that the reliability of the OCR result is low if the ratio is equal to or less than the predetermined ratio.
[0035] An example of a judgment made when the reliability judgment unit 113 acquires OCR result information for three items (e.g., "company name," "address," and "phone number") will be described. Here, the reliability judgment unit 113 judges that the reliability is high when the majority of the OCR results and the matching data match as a "predetermined ratio," and that the reliability is low when the majority of the data do not match.
[0036] FIG. 3 is a diagram for explaining the determination results of the reliability of character recognition by the reliability determination unit 113 in different cases. (Case 1) If the OCR result for one target item matches three of the two matching data (all three match), the OCR result is determined to be highly reliable, i.e., the OCR result is correct and does not require correction.
[0037] (Case 2) When the OCR result for a certain target item matches one piece of matching data (there is a match between the OCR result and the matching data), the OCR result is judged to be highly reliable because the majority of the data matches the OCR result. In other words, the OCR result is judged to be correct and does not require correction.
[0038] (Case 3) If two pieces of matching data other than the OCR results for a certain target item match (two pieces of data other than the OCR results match), the OCR results are determined to be incorrect and the OCR results are corrected to the value where the two pieces of matching data match.
[0039] (Case 4) If the OCR result of a certain target item does not match each of the two pieces of matching data (all three do not match), it is impossible to determine which information is correct. The reliability determination unit 113 outputs information to the effect that it is impossible to determine to the notification unit 114.
[0040] As described above, the reliability determination unit 113 determines whether the number of identical values between the OCR result of a certain target item and the matching data extracted from the OCR result of another target item exceeds a predetermined ratio of the total (number of items). If the number of matches exceeds the predetermined ratio, the reliability determination unit 113 determines that the reliability of the OCR result is high overall, and if the ratio is equal to or less than the predetermined ratio, the reliability determination unit 113 determines that the reliability of the OCR result is low overall. The predetermined ratio may be a majority, more than half, or may be preset to 60%, 70%, etc. In addition, the reading rate of the character recognition device 20 may be calculated from past data, and when the reading rate is high, the predetermined ratio (predetermined threshold) may be set high. In addition, as described below, when multiple matching data have the same value and exceed a predetermined ratio, and the OCR result of the target item has a different value, the reliability judgment unit 113 executes a process of replacing (modifying) the OCR result of the target item with the same value of the matching data.
[0041] When the result of the determination by the reliability determination unit 133 is equal to or lower than a predetermined ratio and the determination is impossible, the notification unit 114 transmits a notification that the determination is impossible to a system management device (not shown) or the like. In this way, it is possible to encourage the user to visually check the reliability.
[0042] <Example of judgment> We will explain an example of determining whether correction is necessary based on the OCR result and the matching data. Here, we will explain the above judgment examples (Case 1) to (Case 4) specifically, assuming that there are three items, "company name," "address," and "phone number." 4 to 7 show the OCR results. The correct answers for the data read by OCR are "Company name": Konica Minolta, Inc., "Address": 2-7-2 Marunouchi, Chiyoda-ku, Tokyo, "Telephone number": 03-1234-5678. The reliability determination unit 113 determines the predetermined ratio as the majority (more than half). The AI-OCR function of the character recognition device 20 identifies the "company name," "address," and "telephone number" from the content of the document, which are acquired by the character recognition result acquisition unit 111 of the character recognition result determination device 10, and the OCR result information is stored in the memory unit 13. In the following, we will explain how to compare the OCR results of "company name" with the matching data, with "company name" as the target item. The same process is used when "address" and "telephone number" related to "company name" are also target items.
[0043] (Case 1) FIG. 4 is a diagram showing an example of a case where the OCR result of one target item (here, "company name") matches three items of two matching data (Case 1). As shown in Figure 4, the OCR result for "company name" is [Konica Minolta, Inc.]. The OCR result for "address" is "2-7-2 Marunouchi, Chiyoda-ku, Tokyo." The OCR result for "phone number" is "03-1234-5678."
[0044] The matching data extraction unit 112 searches the address search database using the OCR result for "Address", which is an item related to "Company Name", "2-7-2 Marunouchi, Chiyoda-ku, Tokyo", as a key, and extracts the hit value as the matching data for "Company Name" [Konica Minolta, Inc.].
[0045] In addition, the matching data extraction unit 112 searches the telephone number search database using the OCR result "03-1234-5678" for the "telephone number", which is an item related to the "company name", as a key, and extracts the hit value as the matching data for the "company name" [Konica Minolta, Inc.].
[0046] In this case, there are no errors in the OCR results for "Address" and "Telephone Number", so the search results hit the correct values. Therefore, all "Company Name" matches "Konica Minolta, Inc.", so there is no need to correct the OCR results for the company name, and it is determined to be correct.
[0047] (Case 2) FIG. 5 is a diagram showing an example of a case where the OCR result of one target item (here, "company name") matches one piece of matching data (Case 2). As shown in the shaded area in Figure 5, the OCR result for "Company Name" is [Konica Minolta, Inc.]. The OCR result for "Address" is "2-7-2 Marunouchi, Chiyoda-ku, Tokyo." The OCR result for "Telephone Number" is "03-1234-5676," which is incorrect and not the correct last four digits "5678."
[0048] The matching data extraction unit 112 searches the address search database using the OCR result for "Address", which is an item related to "Company Name", "2-7-2 Marunouchi, Chiyoda-ku, Tokyo", as a key, and extracts the hit value as the matching data for "Company Name" [Konica Minolta, Inc.].
[0049] Furthermore, the matching data extraction unit 112 searches the telephone number search database using the OCR result "03-1234-5676" for the "telephone number" that is an item related to the "company name" as a key. However, the OCR result for the telephone number "03-1234-5676" is incorrect, and the telephone number is not registered in the telephone number search database, so the "company name" is not extracted.
[0050] In this case, the value of the "company name" matches both the OCR result and the matching data value based on the "address". Therefore, since it is more than half (the majority), the reliability determination unit 113 determines that there is no need to correct the OCR result of the company name and that it is correct.
[0051] (Case 3) FIG. 6 is a diagram showing an example of a case where two pieces of matching data match each other except for the OCR result of one target item (here, "company name") (Case 3). As shown in the shaded area in Figure 6, the OCR result for "Company Name" is [Konica Minolta, Inc.], which is incorrect and not the correct "Konica Minolta." The OCR result for "Address" is "2-7-2 Marunouchi, Chiyoda-ku, Tokyo." The OCR result for "Telephone Number" is "03-1234-5678."
[0052] The matching data extraction unit 112 searches the address search database using the OCR result for "Address", which is an item related to "Company Name", "2-7-2 Marunouchi, Chiyoda-ku, Tokyo", as a key, and extracts the hit value as the matching data for "Company Name" [Konica Minolta, Inc.].
[0053] In addition, the matching data extraction unit 112 searches the telephone number search database using the OCR result "03-1234-5678" for the "telephone number", which is an item related to the "company name", as a key, and extracts the hit value as the matching data for the "company name" [Konica Minolta, Inc.].
[0054] In this case, the value of "company name" matches the value of the matching data based on "address" and the value of the matching data based on "phone number", but does not match the OCR result of the company name. Since the two matching data match more than half (the majority), the reliability determination unit 113 determines that the OCR result of the company name is incorrect and corrects the OCR result to the value of the matching data.
[0055] (Case 4) FIG. 7 is a diagram showing an example of a case where the OCR result of one target item does not match the two pieces of matching data (Case 4). As shown in the shaded area in Figure 7, the OCR result for "Company Name" is [Konica Minolta, Inc.], which is incorrect and not the correct "Konica Minolta." The OCR result for "Address" is "2-7-2 Marunouchi, Chiyoda-ku, Tokyo." The OCR result for "Telephone Number" is "03-1234-5878," which is incorrect and not the correct last four digits "5678."
[0056] The matching data extraction unit 112 searches the address search database using the OCR result for "Address", which is an item related to "Company Name", "2-7-2 Marunouchi, Chiyoda-ku, Tokyo", as a key, and extracts the hit value as the matching data for "Company Name" [Konica Minolta, Inc.].
[0057] In addition, the matching data extraction unit 112 searches the telephone number search database using the OCR result "03-1234-5878" for "telephone number", which is an item related to "company name", as a key, and extracts the hit value as the matching data for "company name" [XX Co., Ltd.].
[0058] In this case, the value of "company name" is different between the OCR result, the value of the matching data based on "address", and the value of the matching data based on "phone number", and it is impossible to determine which information is correct. Therefore, the reliability determination unit 113 outputs to the notification unit 114 that the reliability of the OCR result and the matching data is low and determination is impossible. The notification unit 114 then transmits a notification that determination is impossible to a system management device (not shown) or the like, thereby prompting the user to visually confirm.
[0059] <Processing flow> Next, the flow of the process executed by the character recognition result determination device 10 will be described. FIG. 8 is a flowchart showing the flow of the process executed by the character recognition result determination device 10 according to this embodiment.
[0060] First, the character recognition result acquisition unit 111 of the character recognition result determination device 10 acquires OCR result information from the character recognition device 20 (step S1).
[0061] Next, the matching data extraction unit 112 searches the search database 30 for a certain target item (e.g., "company name") using the OCR results of two or more other items related to the target item (e.g., "address", "telephone number") as keys, and extracts the hit value (company name) as matching data (step S2).
[0062] Next, the reliability judgment unit 113 compares the OCR result of the target item with each of the matching data obtained from the other items, and judges whether the number of identical values exceeds a predetermined percentage of the total (number of items) (step S3).
[0063] Here, if the number of identical values does not exceed a predetermined percentage of the total (number of items) (step S3 → No), the reliability judgment unit 113 determines that the reliability of the OCR results as a whole is low and cannot be judged, and proceeds to step S4. In step S4, when the notification unit 114 receives information from the reliability determination unit 113 that the ratio is below a predetermined ratio and that the determination is impossible, the notification unit 114 outputs a notification that the determination is impossible to the system management device or the like, and ends the process.
[0064] On the other hand, if the number of the same values exceeds a predetermined percentage of the total (number of items) in step S3 (step S3 → Yes), the reliability judgment unit 113 judges that the reliability of the OCR results is high overall, and proceeds to step S5. In step S5, the reliability determination unit 113 determines whether the value of the OCR result is included in the matching values (identical values) obtained in the comparison between the OCR result of the target item in step S3 and each of the matching data obtained from the other items. If the value of the OCR result of the target item is included (step S5→Yes), the reliability determination unit 113 determines that the OCR result of the target item is correct (high reliability) (step S6), and ends the process. On the other hand, if the OCR result value of the target item is not included (step S5 → No), the reliability judgment unit 113 determines that the OCR result of the target item is incorrect, replaces the OCR result of the target item with a value that matches the matching data (the same value) (step S7), and ends the processing.
[0065] In this way, the character recognition result determination device 10 according to the present embodiment can determine the reliability of the reading accuracy of the OCR result. Even if the reading accuracy of the OCR result of the target item is poor, the character recognition result determination device 10 can correct the data based on the OCR results of other items, thereby further improving the reading accuracy.
[0066] The present invention is not limited to the above-described embodiment, and various changes and modifications can be made without departing from the gist of the present invention. [Explanation of symbols]
[0067] 1. Character Recognition System 10 Character recognition result judgment device 11 Control section 12 Input / output section 13 Storage section 20 Character recognition device 30 Search Database (DB) 111 Character recognition result acquisition unit 112 Matching Data Extraction Unit 113 Reliability Judgment Unit 114 Notification Department
Claims
1. a character recognition result acquisition unit that acquires OCR results of three or more related items; a matching data extraction unit that extracts matching data of one of the three or more items by referring to a search database in which values of the items are stored in correspondence with each other using an OCR result of another item related to the target item; a reliability determination unit that compares the OCR result of the target item with each of the matching data obtained from the other items and determines whether the number of identical values exceeds a predetermined ratio of the total; A character recognition result determination device comprising:
2. When the plurality of matching data are the same value and exceed the predetermined ratio, and the OCR result of the target item is different from the same value, the reliability determination unit corrects the OCR result of the target item to the same value of the matching data. The character recognition result determination device according to claim 1 .
3. The reliability determination unit further includes a notification unit that compares the OCR result of the target item with a plurality of the matching data, and outputs information indicating that the determination is impossible if there are no identical values exceeding the predetermined ratio. The character recognition result determination device according to claim 1 .
4. A character recognition result acquisition step of acquiring OCR results of three or more related items; a matching data extraction step of extracting matching data of one of the three or more items by referring to a search database in which values of the items are stored in correspondence with each other, using an OCR result of another item related to the target item; a reliability determination step of comparing the OCR result of the target item with each of the matching data obtained from the other items and determining whether or not the number of identical values exceeds a predetermined ratio of the total; A character recognition result determination method for a character recognition result determination device that executes the above.
5. On the computer, obtaining OCR results for three or more related items; a step of extracting matching data of any one of the three or more items by referring to a search database in which values of each of the items are stored in correspondence with each other, using an OCR result of another item related to the target item; a step of comparing the OCR result of the target item with each of the matching data obtained from the other items, and determining whether or not the number of identical values exceeds a predetermined ratio of the total; A character recognition result judgment program for executing the above.
6. A character recognition system including a character recognition device having a character recognition function and a character recognition result determination device that determines reliability of the character recognition function, The character recognition result determination device includes: a character recognition result acquisition unit that acquires OCR results of three or more related items from the character recognition device; a matching data extraction unit that extracts matching data of one of the three or more items by referring to a search database in which values of the items are stored in correspondence with each other using an OCR result of another item related to the target item of the three or more items; a reliability determination unit that compares the OCR result of the target item with each of the matching data obtained from the other items and determines whether or not the number of identical values exceeds a predetermined ratio of the total; A character recognition system comprising:
Citation Information
Patent Citations
Business card recognizing device
JP1994103404A
