JP5145287B2 - Input error warning device - Google Patents

Description

  The present invention relates to an input error warning device that warns an input error according to an input word input by a user.

  There has been proposed a technique for providing input support for preventing erroneous input when a computer apparatus receives character input from a user. For example, Patent Literature 1 stores a combination of an input character and a correction character corresponding to the input character as a correction rule based on a character arrangement of a keyboard that accepts character input and a typo error for each user. A technique for correcting an input character based on a stored correction rule has been proposed. In Patent Document 2, every time a stored correction rule is used, the weight of each correction rule is updated to learn the user's error habits, and the input characters input from the user according to the updated weight are stored. Techniques for correction are described.

JP-A-4-245319 JP-A-5-252799

  However, the above-described technique corrects an erroneous input by using a correction rule generated based on a typo at the time of keyboard operation. Here, for example, when an operator receives an inquiry about a company product by telephone and inputs it to the computer device, or reads a sentence printed or written on a paper medium and inputs it to the computer device, an error due to a typo is made. In addition to the input, it is considered that an erroneous input based on a user's listening error or reading error may occur. Depending on the correction rule based on the typo as described above, such an erroneous input cannot be detected appropriately. Here, when the computer device accepts a character input from the operator, it is desirable to warn the operator of an erroneous input based on a hearing error or a reading error.

  The present invention has been made in view of such a situation, and an object of the present invention is to provide an input error warning device that detects an erroneous input based on a mistake in reading or a reading error with higher accuracy and warns the user.

  In order to solve the above-described problem, the present invention provides a correct word that is predetermined character information and a number of correct words that are included in processing target data that is processing target for detecting character information by pattern recognition processing. An appearance number storage unit in which the number of occurrences of 1 and the second number of appearances indicating the number of correct words included in the character information input by the user are associated with each other, the correct word, and the correct word are misread A first error number indicating the number of error words that have been misread as error words by the pattern recognition process, and a second error indicating the number of error words that have been misread by the user as error words A number of errors stored in association with the number, a weighting coefficient storage for storing a weighting factor indicating the degree of reliability of the second number of errors relative to the first number of errors, and a correct answer First occurrence and second occurrence corresponding to the word An error rate calculation unit that calculates an error rate in which a correct word is misread as an error word based on the first error count, the second error count, and a weighting factor, and character information input from a user And the character information input to the input unit is stored as an error word in the error number storage unit, and the error rate calculated by the error rate calculation unit according to the error word is predetermined. And a warning unit that outputs a correct word corresponding to the error word when it is determined that the threshold value is exceeded.

  Further, the present invention calculates the number of correct words included in the character information input to the correction input unit, the correction input unit that receives the correction input of the character information input to the input unit, and the calculated number appears. The appearance number calculation unit for adding and storing the second appearance number stored in the number storage unit, the character information input to the input unit, and the character information input to the correction input unit are compared, and the difference is calculated. Detects and calculates the number of characters that have a difference between the character input to the input unit as an error word and the character input to the correction input unit as a correct word, and stores the calculated number in the error number storage unit An error number calculation unit to be added to and stored in the second error number, a first appearance number corresponding to the correct word, a second appearance number, a first error number, and a second error number And a weighting factor calculation unit that calculates a weighting factor and stores the weighting factor in a weighting factor storage unit; Characterized in that it comprises a.

  Further, according to the present invention, the pattern recognition processing is performed by any one of character recognition processing for detecting character information from image data that is processing target data and voice recognition processing for detecting character information from voice data that is processing target data. It is characterized by being.

  In addition, the present invention includes a user attribute information storage unit that stores user information indicating user attributes, and the appearance number storage unit stores a second appearance number for each user attribute, and an error number storage unit Stores the second error count for each user attribute, the weight coefficient storage unit stores the weight coefficient for each user attribute, and the error rate calculation unit calculates the error rate for each user attribute. It is characterized by doing.

In the present invention, the error rate calculation unit sets the first number of appearances corresponding to the correct word to n ₀ , the second number of appearances to n ₁ , the first number of errors to e ₀ , When the number of errors is e ₁ and the weighting factor is T, the error rate P in which the correct word is misread as an error word is expressed by the following equation (a)

  It is characterized by calculating by.

In the present invention, the correction input unit accepts correction input over a plurality of stages, and the weighting factor calculation unit sets P as the error rate calculated by the error rate calculation unit in the last stage among the plurality of stages. The first occurrence number corresponding to the correct word is n ₀ , the second occurrence number is n ₁ , the first error number is e ₀ , the second error number is e ₁ , and the weight coefficient T is Formula (b) below

  It is characterized by calculating by.

  As described above, according to the present invention, the input error warning device includes a correct word that is predetermined character information and a correct word that is included in processing target data that is a processing target for detecting character information by pattern recognition processing. A first appearance number indicating the number of words, a second appearance number indicating the number of correct words included in the character information input by the user, an error word that is a result of misreading the correct word, and pattern recognition processing The first error number indicating the number of correct words misread as an error word, the second error number indicating the number of correct word misreads as an error word by the user, and the second error number with respect to the first error number Based on the weighting factor indicating the degree of reliability of the number of errors, the error rate at which the correct word is misread as an error word is calculated, the character information input from the user is stored as the error word, and the error Error rate calculated for each word If it is determined that a predetermined threshold value is exceeded, the correct word corresponding to the error word is output, so the number of errors for the character information detected by the pattern recognition process and the character information input by the user It is possible to calculate the error rate of the character information input from the user based on the correlation existing between the number of errors about and the error input based on the user's listening error and reading error more accurately. It becomes possible to warn the user.

It is a block diagram which shows the structural example of the input system by one Embodiment of this invention. It is a figure which shows the example of a display of the input item by one Embodiment of this invention. It is a figure which shows the example of a display of the input item by one Embodiment of this invention. It is a figure which shows the example of data of the input information memorize | stored by one Embodiment of this invention. It is a figure which shows the example of data of the number of appearance memorize | stored by one Embodiment of this invention. It is a figure which shows the example of data of the number of errors memorize | stored by one Embodiment of this invention. It is a figure which shows the example of data of the weighting coefficient memorize | stored by one Embodiment of this invention. It is a figure which shows the operation example which memorize | stores the number of appearances and the number of errors by one Embodiment of this invention. It is a figure which shows the operation example which outputs warning information by one Embodiment of this invention.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing the configuration of the input system 1 according to the present embodiment. The input system 1 includes computer devices such as an operator input terminal 100, an input device 200, a correction input terminal 300, and an input error warning device 400, and the respective computer devices are connected via a network.

  The operator input terminal 100 is a computer terminal that includes a display unit such as a display and an input unit such as a keyboard and a mouse, and receives input information input from an operator. The operator input terminal 100 displays, for example, an input screen including input items corresponding to the inquiry contents from a customer to a company product or the like on the display unit, and input information input from the operator according to the displayed input items Accept. In the present embodiment, the operator reads characters such as a customer's name, address, telephone number, and inquiry content written by hand on the paper medium by the customer, and inputs the read characters to the operator input terminal 100 as input information. The operator input terminal 100 transmits input information input from the operator to the input device 200.

  Here, the input information that is input to the operator input terminal 100 is information that is input by the operator reading characters that are handwritten on a paper medium by the customer. There may be shaking. Thereby, it is considered that the same information may be input as different information. For example, it may be considered that the character “Aoda” indicating the name of the customer written by hand on a paper medium by the customer may be misread as “Aoi” by the operator and input.

The input device 200 is a computer device in which input information input to the operator input terminal 100 or the correction input terminal 300 is transmitted and stored, and includes an input control unit 210, an operator information storage unit 220, and an input information storage unit 230. And.
The input control unit 210 transmits an inquiry content input screen to the operator input terminal 100 or the correction input terminal 300, and input information transmitted to the operator input terminal 100 or the correction input terminal 300 according to the transmitted input screen. Receive. Here, the input device 200 has a web server function, and communicates with the operator input terminal 100 or the correction input terminal 300 through HTTP (HyperText Transfer Protocol) communication or the like. The input screen that the input control unit 210 transmits to the operator input terminal 100 or the correction input terminal 300 is a web page described in, for example, HTML (HyperText Markup Language).

  Here, when the input control unit 210 transmits an inquiry content input screen, an operator information input screen for login processing is transmitted in advance, and when the login processing is successful, the inquiry content input screen is displayed as an operator input terminal. To 100. Here, the input control unit 210 inputs operator information in which the operator input terminal 100 or the correction input terminal 300 includes input items of an operator ID (Identifier) for identifying the operator and a defined password corresponding to the operator ID. The screen is transmitted, and operator information transmitted from the operator input terminal 100 or the correction input terminal 300 is accepted according to the transmitted operator information input screen. When the input control unit 210 determines whether operator information that matches the combination of the operator ID and password received from the operator input terminal 100 is stored in the operator information storage unit 220, and determines that it is stored Is determined to be successful login, and if it is determined that it is not stored, it is determined that login failed. When it is determined that the login is successful by such login processing, the input control unit 210 stores the operator ID of the logged-in operator in its own storage area and transmits an inquiry content input screen to the operator input terminal 100. . On the other hand, if it is determined that the login has failed, the inquiry content input screen is not transmitted to the operator input terminal 100, and the process is terminated without accepting transmission of input information. In the input system 1 according to the present embodiment, the history information of the operator who has input the input information can be stored by storing the input information in association with such operator information.

  Further, the input control unit 210 receives input information and transmits it to the input error warning device 400 every time a character is input to the operator input terminal 100 in accordance with the transmitted inquiry content input screen. The input control unit 210 displays the correct word included in the warning information transmitted from the input error warning device 400 when the characters included in the input information are characters that are erroneously input at a certain ratio or more. 100 to display. For example, FIGS. 2 and 3 are diagrams illustrating display examples when the correct word is transmitted to the operator input terminal 100 and the transmitted correct word is displayed on the input screen of the operator input terminal 100. FIG. For example, FIG. 2A shows a correct word corresponding to a character input in a drop-down format as indicated by reference symbol a2 when the character “dog” is input from the user to a text box indicated by reference symbol a1. This is an example in which “Large” is displayed. Here, such real-time information display can be performed by applying, for example, the Ajax technique. FIG. 2B is a diagram illustrating a display example in which “Large” which is a correct word displayed in a drop-down format is selected, as indicated by reference numeral 2. For example, when the enter key of the keyboard is pressed in this state, the selected correct word “Large” is input to the text box b1. FIG. 3 shows an example in which the character “Aota” is input in the text box a1 corresponding to the input item “name” and the correct word “Aoi” corresponding to the input character is displayed.

  Returning to FIG. 1, operator information corresponding to an operator registered in advance is stored in the operator information storage unit 220. The operator information includes, for example, attribute information indicating a name, age, sex, years of work experience, password for login, and the like in association with an operator ID for identifying the operator. The work experience years are information indicating the years of experience in which the corresponding operator performs the work of reading the characters written on the paper medium and inputting them to the operator input terminal 100. By storing such work experience years, it is possible to calculate a statistical value of the number of input errors for each work experience years. Further, the operator information stored in the operator information storage unit 220 is read when the operator login process is performed by the input control unit 210 as described above.

  The input information storage unit 230 stores input information transmitted from the operator input terminal 100. FIG. 4 is a diagram illustrating an example of input information data stored in the input information storage unit 230. The input information storage unit 230 stores input information in association with the input information ID, the input number, and the operator ID. The input information ID is identification information that identifies input information. The input number is information given when input information is corrected and input by a corrector. Here, if the input number is “1”, it indicates that the corresponding input information is the first input information. If the input number is “2”, the corresponding input information is input information that has already been input. Indicates that the input information has been corrected by the corrector and input to the correction input terminal 300. The operator ID is information for identifying an operator who has input corresponding input information. For example, in the data example of FIG. 4, the input information ID “1” and the input number “1” are input and stored by the operator with the operator ID “1”. Also, for the input information whose input information ID is “1”, the name “Aoda” is corrected to “Aoi” by the operator with the operator ID “2” and stored in association with the input number “2”. . As described above, the input information storage unit 230 stores the input information transmitted from the operator input terminal 100 and the corrected input information transmitted from the corrected input terminal 300 after the input information is stored.

  Returning to FIG. 1, the correction input terminal 300 has the same configuration as the operator input terminal 100, reads and displays the input information stored in the input device 200, and the operator who is the corrector according to the displayed input information. Accepts corrected input information to be input. The operator who inputs the corrected input information to the corrected input terminal 300 compares the input information read and displayed from the input device 200 by the corrected input terminal 300 with other input information or customer information registered in advance. Then, the corrected input information is input to the correction input terminal 300 by correcting the error word as a correct word or correcting the notation fluctuation of the same customer information. The corrected input terminal 300 transmits the input corrected input information to the input device 200. The input information transmitted by the modified input terminal 300 is associated with the input information ID of the read input information, and “2” is assigned to the input number and stored in the input information storage unit 230.

  Here, with regard to input information related to the contents of inquiries from customers, for example, when trying to manage input information for each customer and for each customer attribute, if there is an error or fluctuation in the input information, the information is centrally managed by computer processing However, it is difficult to calculate useful statistics. For this reason, when the information included in the input information is substantially the same as the information included in the other input information, it is desirable that the information is stored in the same notation. For example, as described above, when different inquiry contents are stored in association with the names of “Aoda” and “Aoi”, the statistics are calculated as different persons. Also, it is not possible to determine mechanically which name is correct. Therefore, the operator who is the corrector refers to the input information displayed on the correction input terminal 300 and corrects the input information corresponding to the same person in the input information corresponding to the same person among the plurality of input information. The corrected input information is input to the correction input terminal 300.

  The input error warning device 400 is a computer device that detects and outputs the correct word corresponding to the case where the input information is an error word in accordance with the input information input from the operator. The input error warning device 400 includes a processing target data input unit 410, a pattern recognition processing unit 420, a character information input unit 430, an error pattern storage unit 440, an appearance number calculation unit 450, an error number calculation unit 460, An error rate calculation unit 470, a weight coefficient calculation unit 480, and a warning unit 490 are provided.

  The processing target data input unit 410 receives processing target data that is a processing target for which character information is detected by the pattern recognition processing unit 420. In the present embodiment, the processing target data input to the processing target data input unit 410 will be described by applying image data including characters. The image data is used as a sample for calculating a character input error rate by character recognition (OCR (Optical Character Recognition)). For example, as image data, information obtained by reading a paper medium on which a character code table including all characters defined in a specific character code is printed by a scanner or the like can be applied.

  The pattern recognition processing unit 420 performs pattern recognition processing on the processing target data input to the processing target data input unit 410, and detects character information included in the processing target data. In the present embodiment, the pattern recognition processing unit 420 reads the image data input to the processing target data input unit 410, performs pattern recognition processing, and includes character information analyzed based on the form of characters included in the image data. Pattern recognition processing result is generated and stored in its own storage area. Here, it is considered that the character information included in the pattern recognition processing result may include character information that is erroneously analyzed due to character distortion or dirt in the image data. For example, there may be a case where a “large” character is misread by the pattern recognition process and output as a “dog” character.

  In the character information input unit 430, character information corresponding to the image data input to the processing target data input unit 410 is input and stored. Here, for example, the user inputs character information corresponding to the image data into a text editor of a computer device to generate a document file, and converts the generated document file into an image, thereby converting the processing target data input unit 410. The processing target data to be input to is generated. As a result, the pattern recognition processing result obtained by analyzing and storing the processing target data by the pattern recognition processing unit 420 is compared with the character information stored in the character information input unit 430, thereby detecting a reading error due to the character recognition processing. Is possible.

The error pattern storage unit 440 includes an appearance number storage unit 441, an error number storage unit 442, and a weight coefficient storage unit 443, and stores information indicating an error pattern of input information.
The number-of-appearance storage unit 441 includes, for each predetermined correct word, characters included in the character information input to the character information input unit 430 and characters included in the input information stored in the input information storage unit 230. The correct word is stored, and for each correct word, the number of correct words detected from the character information input to the character information input unit 430 or the input information stored in the input information storage unit 230 is associated as the number of appearances. Remembered.

  FIG. 5 is a diagram illustrating a data example of the number of appearances stored in the appearance number storage unit 441. In the appearance number storage unit 441, the correct word, the OCR appearance number, the total number of manual appearances, the number of work experience years less than one year, and the number of work experience years more than one year are stored in association with each other. The The correct word includes all of the characters included in the character information input to the character information input unit 430 and the characters included in the input information stored in the input information storage unit 230. The OCR appearance number is the number of corresponding correct words detected from the character information input to the character information input unit 430. The total number of manual appearances is the number of corresponding correct words included in the input information stored in the input information storage unit 230. The number of appearances of less than one year of work experience is the number of corresponding correct words included in the input information input by an operator who has less than one year of work experience among the total number of appearances by hand. The number of appearances of one year or more of work experience is a number in which the corresponding correct word is included in input information input by an operator who has one or more years of work experience among the total number of appearances by hand.

  For example, in this example, the character information input to the character information input unit 430 includes 20000 characters for “Large”, 15000 characters for “Small”, and 60000 characters for “H”. Is detected, and it is detected that the input information stored in the input information storage unit 230 includes 5000 characters of “Large”, 4000 characters of “Small”, and 10,000 characters of “H”. It has been shown.

  The appearance number storage unit 441 stores the number of appearances calculated by the appearance number calculation unit 450. The appearance number calculation unit 450 calculates the number of appearances for each of the characters included in the input information stored in the input information storage unit 230 and the characters included in the character information input to the character information input unit 430. The number is stored in the number storage unit 441. At this time, the appearance number calculation unit 450 reads the work experience years associated with the operator ID stored in the input information storage unit 230 from the operator information storage unit 220. The appearance number calculation unit 450 determines whether the work experience year is less than one year or more than one year, calculates the appearance number based on the determination result, and stores the appearance number in the appearance number storage unit 441.

  In the error number storage unit 442, the correct word stored in the appearance number storage unit 441, the error word in which the correct word is misread, and the number of errors indicating the number of misreads are stored in association with each other. . FIG. 6 is a diagram illustrating an example of data on the number of errors stored in the error number storage unit 442. In the error number storage unit 442, the correct word, the error word, the number of OCR errors, the total number of manual errors, the number of errors of less than one year of work experience, and the number of errors of one year or more of work experience are associated with each other. Is remembered.

  The number of OCR errors indicates the number of correct words that have been misread as corresponding error words by the pattern recognition processing unit 420 among the OCR appearance numbers stored in the appearance number storage unit 441. The total number of manual errors indicates the number of correct words that are erroneously read by the operator as the corresponding error words out of the total number of manual appearances stored in the appearance number storage unit 441. The number of errors of less than one year of work experience indicates the number included in the input information misread by the operator having less than one year of work experience out of the total number of appearances by hand. The number of errors for one year or more of work experience indicates the number included in the input information misread and input by an operator having a work experience of one year or more out of the total number of appearances by hand.

  In the error number storage unit 442, the error number calculated by the error number calculation unit 460 is stored. For example, the error count calculation unit 460 compares the pattern recognition processing result stored in the pattern recognition processing unit 420 with the character information stored in the character information input unit 430, and the correct word is misread as an error word by OCR. The number of OCR errors which is the calculated number is calculated and stored in the error number storage unit 442. Further, the error number calculation unit 460 compares the input information stored as the input number “1” and the input information stored as the input number “2” among the input information stored in the input information storage unit 230. Then, the difference between the character included in the input information whose input number is “1” and the character included in the input information whose input number is “2” is detected. Then, for the character from which the difference is detected, the character included in the input information whose input number is “1” is set as an error word, and the character included in the input information whose input number is “2” is set as a correct word. The number of errors is calculated and stored in the error number storage unit 442. At this time, the error number calculation unit 460 reads the work experience years associated with the operator ID stored in the input information storage unit 230 from the operator information storage unit 220. The error number calculation unit 460 determines whether the work experience year is less than one year or more than one year, calculates the error number based on the determination result, and stores the error number in the error number storage unit 442.

  The weighting coefficient storage unit 443 stores a weighting coefficient indicating the degree of reliability of the number of manual errors with respect to the number of OCR errors. The weight coefficient is referred to when the error rate of the correct word is calculated. The weighting factor is a predetermined numerical value of 1 or more, and a predetermined value is stored as an initial value, and is updated and stored with the weighting factor calculated by the weighting factor calculation unit 480.

Based on the number of appearances stored in the number of appearances storage unit 441, the number of errors stored in the number of errors storage unit 442, and the weighting factor stored in the weighting factor storage unit 443, the error rate calculation unit 470 An error rate indicating the probability that the correct word stored in the number storage unit 442 is misread as an error word is calculated. Here, the error rate calculation unit 470 sets the number of OCR appearances corresponding to the correct word to n ₀ , sets the total number of manual appearances to n ₁ , sets the number of OCR errors to e _0, and sets the total number of manual errors to e _1. The error rate P at which the correct word is misread as an error word is calculated by the following equation (1), where T is a weighting factor.

Here, input information is input to the operator input terminal 100 and the correction input terminal 300 over a plurality of stages. More specifically, the calculation formula of the error rate at each stage is the input information in which the OCR appearance number stored in the appearance number storage unit 441 is n ₀ and the input number stored in the input information storage unit 230 is 1. Overall the number of occurrences and n ₁ manual stored in the appearance count storage unit 441 based on, is stored in the occurrence count storage unit 441 based on the input information stored input number is 2 in the input information storage unit 230 Let n _{2 be the} total number of appearances made manually. Hereinafter, if the input number stored in the input information storage unit 230 is i, the total number of manual appearances stored in the appearance number storage unit 441 is n _i .

Further, the number of OCR errors stored in the error number storage unit 442 is set to e _0, and the input number stored in the input information storage unit 230 is manually stored in the error number storage unit 442 based on the input information of 1 The total number of errors is e _1, and the total number of manual appearances stored in the error number storage unit 442 based on the input information whose input number is 2 stored in the input information storage unit 230 is e ₂ . Hereinafter, if the input number stored in the input information storage unit 230 is i, the total number of manual errors stored in the error number storage unit 442 is set to e _i . The weighting factor stored in the weighting factor storage unit 443 is T.

  Here, regarding the number of errors based on the pattern recognition process and the number of errors based on the manual correction process, it is assumed that the number of errors based on the manual correction process is more reliable, and the error rate calculation unit 470. Calculates the error rate by multiplying the number of errors based on manual correction processing by one or more weighting factors. For example, the error rate calculation unit 470 calculates the error rate P (Step 2) in stage 2 by the following equation (2).

  Similarly, the error rate P (Step 3) in stage 3 is calculated by the following equation (3).

  Here, the error rate calculation unit 470 may calculate the error rate by applying a different weighting factor for each attribute corresponding to the operator ID of the operator who has input the input information. For example, a large value is calculated as the weighting coefficient when the work experience years are small, and a small value is calculated when the work experience years is large. Further, for example, a small value is calculated as the weighting factor when the age corresponding to the operator ID of the operator who has input the input information is high, and a large value is calculated when the age is low.

ここでは、入力情報記憶部２３０に同一の入力情報ＩＤに対応付けられた複数の入力番号に応じた入力情報が記憶されている場合には、より後の段階における入力情報に基づいて算出された誤り率の方がより信頼性が高いと考えられる。そこで、最後の段階での誤り率がより正しいとして、誤り率算出部４７０は、一つ前の段階において算出した誤り率が、最後の段階において算出した誤り率となるような重み係数Ｔを算出し、算出した重み係数を、重み係数記憶部４４３に記憶させる。 The weighting factor calculation unit 480 calculates a weighting factor based on the appearance number stored in the appearance number storage unit 441 and the error number stored in the error number storage unit 442. Here, the weighting factor calculation unit 480 sets P as the error rate calculated by the error number storage unit 442 in the last step among the plurality of steps, sets n _{0 as the} number of OCR appearances corresponding to the correct word, and makes the total appearance by hand. The number is n ₁ , the number of OCR errors is e ₀ , the total number of manual errors is e ₁ , and the weight coefficient T is calculated by the following equation (4).

Here, when input information corresponding to a plurality of input numbers associated with the same input information ID is stored in the input information storage unit 230, it is calculated based on the input information at a later stage. The error rate is considered to be more reliable. Therefore, assuming that the error rate at the last stage is more correct, the error rate calculation unit 470 calculates a weighting factor T such that the error rate calculated at the previous stage becomes the error rate calculated at the last stage. Then, the calculated weighting coefficient is stored in the weighting coefficient storage unit 443.

For example, assuming that stage 3 is the last stage when the weighting coefficient calculation unit 480 calculates the weighting coefficient, the value of T is calculated such that the error rate in stage 2 matches the error rate in stage 3. . When such a condition is satisfied T and _{T new,} relationship between _{T new} error rate P (Step3) can be expressed as: Equation (5).

When this is solved for T _new , it can be expressed as in equation (6) below.

Weight coefficient calculation unit 480, a weighting factor _{T new} calculated by the equation (6), is stored in the weight coefficient storage unit 443.

Similarly, if stage 4 is the last stage, the value of T is calculated such that the error rate in stage 4 becomes the error rate in stage 3. When such a condition is satisfied T and _{T new,} relationship between _{T new} error rate P (Step4) can be expressed as: Equation (7).

When this is solved for T _new , it can be expressed as shown in equation (8) below.

  Here, the weight coefficient calculation unit 480 stores the number of appearances stored in the appearance number storage unit 441 and the number of errors stored in the error number storage unit 442 for each attribute such as the number of years of work experience. Calculates a weighting factor for each attribute.

  The warning unit 490 receives input information transmitted from the input device 200 and input to the operator input terminal 100. The warning unit 490 determines whether the character information included in the received input information is stored in the error count storage unit 442 as an error word. If the warning unit 490 determines that the character information included in the received input information is not stored in the error count storage unit 442 as an error word, the warning unit 490 does not transmit the warning information. On the other hand, if the warning unit 490 determines that the character information included in the received input information is stored in the error count storage unit 442 as an error word, the error rate calculation unit 470 calculates the error word according to the error word. It is determined whether the error rate exceeds a predetermined threshold. When the warning unit 490 determines that the error rate calculated by the error rate calculation unit 470 according to the error word exceeds a predetermined threshold, the correct word corresponding to the error word is input as warning information to the input device. 200. Here, the warning unit 490 detects a plurality of correct words corresponding to the character information included in the input information from the error number storage unit 442, and among the detected plurality of correct words, the corresponding error rate exceeds a threshold value. If there are a plurality of correct words, a plurality of correct words are arranged in descending order of error rates, for example, and transmitted to the input device 200. On the other hand, if the warning unit 490 determines that the error rate calculated by the error rate calculation unit 470 according to the error word does not exceed a predetermined threshold, the warning unit 490 does not transmit the warning information.

Next, an operation example of the input system 1 according to the present embodiment will be described. FIG. 8 is a flowchart illustrating an operation example in which the input system 1 stores information in the error pattern storage unit 440 of the input error warning device 400.
The appearance number calculation unit 450 determines whether or not the correct word input to the correction input terminal 300 and transmitted to the input control unit 210 is stored in the appearance number storage unit 441 (step S1). If the appearance number calculation unit 450 determines that the correct word is stored in the appearance number storage unit 441 (step S1: YES), the appearance number calculation unit 450 adds 1 to the appearance number corresponding to the correct word stored in the appearance number storage unit 441. The stored value is stored (step S2). Here, the value of both the total number of appearances manually and the number of appearances according to the years of experience according to the attribute information of the operator is increased.

  The error number calculation unit 460 determines whether or not the combination of the corrected correct word and the error word is stored in the error number storage unit 442 (step S4). If it is determined by the error number calculation unit 460 that the combination of the correct word and the error word is stored in the error number storage unit 442 (step S4: YES), the correct word stored in the error number storage unit 442 A value obtained by incrementing the number of errors corresponding to the combination with the error word by 1 is stored (step S5). The weighting factor calculation unit 480 calculates a weighting factor based on the number of appearances stored in the error pattern storage unit 440 and the number of errors stored in the error number storage unit 442, and the calculated weighting factor is used as the weighting factor storage unit. It memorize | stores in 443 (step S7), and complete | finishes a process.

  On the other hand, when the appearance number calculation unit 450 determines in step S1 that the correct word is not stored in the appearance number storage unit 441 (step S1: NO), the appearance number storage unit 441 stores data regarding the correct word. Newly added and the number of appearances “1” is stored in association with each other (step S3). In step S4, if the error number calculation unit 460 determines that the combination of the correct word and the error word is stored in the error number storage unit 442 (step S4: NO), the error number storage unit 442 stores Data of a combination of a correct word and an error word is newly added, the number of errors “1” is stored in association with each other (step S6), and the process proceeds to step S7.

Next, an operation example in which the input system 1 outputs warning information in accordance with input information input from an operator will be described with reference to FIG.
Here, when the login process according to the operator information input to the operator input terminal 100 is successful, the input control unit 210 reads out the tail operator attribute information corresponding to the logged-in operator ID from the operator information storage unit 220 ( Step S11). When the input screen for the inquiry content transmitted from the input device 200 is displayed on the operator input terminal 100, the operator inputs input information for each input item. The operator input terminal 100 transmits the input information input from the operator to the input error warning device 400. The warning unit 490 determines whether or not the number of errors having the character (input word) included in the input information transmitted from the operator input terminal 100 as an error word is stored in the error number storage unit 442 (step S12). ).

  When the warning unit 490 determines that the number of errors having the input word transmitted from the operator input terminal 100 as an error word is stored in the error number storage unit 442 (step S12: YES), the error rate calculation unit 470 The work experience years included in the operator attribute information read in step S11 is read from the operator information storage unit 220 of the input device 200, and the number of errors corresponding to the read experience years is read from the error number storage unit 442. Further, the error rate calculation unit 470 reads a weighting factor corresponding to the years of experience corresponding to the operator ID from the weighting factor storage unit 443 (step S13).

  The error rate calculation unit 470 calculates an error rate based on the read appearance number, error number, and weighting factor (step S14). When the warning unit 490 compares the calculated error rate with a predetermined threshold and determines that the error rate exceeds the threshold, the warning unit 490 sends warning information including a correct word corresponding to the input information to the operator input terminal 100. Transmit (step S15). Here, when the error rate does not exceed the threshold, the warning unit 490 does not transmit warning information. When the operator selects the correct word that has been warned in accordance with the warning information displayed on the operator input terminal 100, the operator input terminal 100 transmits the selected correct word to the input error warning device 400 (step S16). ).

  When the error count calculation unit 460 receives the correct word transmitted from the operator input terminal 100 (step S16), the error count calculation unit 460 updates and stores the error count stored in the error count storage unit 442 (step S17). On the other hand, when the warning unit 490 determines in step S12 that the number of errors having the character included in the input information transmitted from the operator input terminal 100 as an error word is not stored in the error number storage unit 442 (step S12: NO), the warning unit 490 ends the warning process. In step S16, when the correction input corresponding to the warning information is not transmitted (step S16: NO), the warning unit 490 ends the warning process.

  In the present embodiment, sample image data input to the processing target data input unit 410 is a paper medium on which a character code table including all characters defined in a specific character code is printed using a scanner or the like. Although it is the read information, for example, image data obtained by digitizing a paper medium on which inquiry contents are written by a customer by a scanner or the like may be applied.

Further, in the present embodiment, the operator input terminal 100 and the correction input terminal 300 are described as different computer terminals. However, the same operator uses the same computer terminal to read input from a paper medium and input information. You may perform both correction input.
In the present embodiment, the input device 200 has a web server function and receives input information transmitted from the operator input terminal 100 by HTTP communication or the like. However, the operator input terminal 100 and the input device 200 are different from each other. Information may be transmitted and received by other communication methods. Further, for example, a single computer device may be configured to include a functional unit included in the operator input terminal 100 and the input device 200, or a single computer device may include a functional unit included in the input device 200 and the correction input terminal 300. It may be configured to include a functional unit included in all of the operator input terminal 100, the input device 200, and the correction input terminal 300, or may be configured to include a single computer device.

  Further, in the present embodiment, an example in which the error rate and the like are calculated for each year of work among the operator attribute information stored in the operator information storage unit 220 has been shown. However, the error rate and the like are calculated for each age. Alternatively, an error rate or the like may be calculated for each gender. The administrator of the input system 1 may calculate an error rate or the like for each attribute useful for calculating the error rate according to the content of the input information.

  In this embodiment, image data is input as processing target data to the processing target data input unit 410, and the pattern recognition processing unit 420 performs character recognition processing to detect character information included in the processing target data. As shown, voice data is input to the processing target data input unit 410 as processing target data, and the pattern recognition processing unit 420 performs voice recognition processing to detect character information included in the processing target data and performs pattern recognition processing results. May be output.

  In this embodiment, for example, the error rate and the weighting factor are calculated for each attribute of the operator such as years of work experience. For example, for each input item (for example, name, address, telephone number, etc.) Different error rates and weighting factors may be calculated, and different error rates and weighting factors for each character type (for example, kanji / hirakana mixed, kanji only, hiragana only, katakana only, numbers / symbols, etc.) May be calculated.

  As described above, according to the input system 1 of the present embodiment, an input error is detected by utilizing the fact that there is a correlation between misreading by pattern recognition processing and misreading when a character is read by an operator. In order to warn, it is possible to warn of an input error based on a combination of a correct word and an error word with higher accuracy before an actual input error by the operator is accumulated. In addition, by recalculating the error rate of the input word based on the input information actually input by the operator and the corrected input information, and learning the input error of the operator, the input error warning is more accurately performed. Can be done.

  Furthermore, since it is possible to calculate an error rate based on attributes such as years of work experience, age, and sex of an operator who inputs input information, more accurate error estimation can be performed. In addition, since the error rate is calculated based on the ratio between the number of occurrences and the number of errors, input errors can occur even when input words with a small number of stored words or errors occur simultaneously in multiple input items. Can be estimated.

  Note that a program for realizing the function of the processing unit in the present invention is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into a computer system and executed, thereby executing an input error warning. May be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer system” includes a WWW system having a homepage providing environment (or display environment). The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

DESCRIPTION OF SYMBOLS 1 Input system 100 Operator input terminal 200 Input device 210 Input control part 220 Operator information storage part 230 Input information storage part 300 Correction input terminal 400 Input error warning device 410 Processing target data input part 420 Pattern recognition processing part 430 Character information input part 440 Error pattern storage unit 441 Appearance number storage unit 442 Error number storage unit 443 Weight coefficient storage unit 450 Appearance number calculation unit 460 Error number calculation unit 470 Error rate calculation unit 480 Weight coefficient calculation unit 490 Warning unit

Claims (6)

A correct word that is defined character information, a first appearance number that indicates the number of correct words included in the processing target data that is a processing target for detecting character information by pattern recognition processing, and a character input by the user An appearance number storage unit that stores the second appearance number indicating the number of correct words included in the information in association with each other;
The correct word, an error word that is a result of misreading the correct word, a first error number that indicates the number of misread words as the error word by the pattern recognition process, and the correct answer by the user An error number storage unit in which a second error number indicating the number of words misread as the error word is associated and stored;
A weighting coefficient storage unit that stores a weighting coefficient indicating a degree of reliability of the second error number with respect to the first error number;
Based on the first number of occurrences, the second number of occurrences, the first number of errors, the second number of errors, and the weighting factor corresponding to the correct word, An error rate calculation unit for calculating an error rate misread as the error word;
An input unit for receiving character information input by the user;
The character information input to the input unit is stored as the error word in the error number storage unit, and the error rate calculated by the error rate calculation unit is determined in advance according to the error word. A warning unit that outputs the correct word corresponding to the error word;
An input error warning device comprising: A correction input unit that receives a correction input for the character information input to the input unit;
The number of occurrences of calculating the number of correct words included in the character information input to the correction input unit, and adding the calculated number to the second number of appearances stored in the appearance number storage unit to be stored A calculation unit;
The character information input to the input unit and the character information input to the correction input unit are compared to detect a difference, and the character input to the input unit is used as an error word for a character having a difference. An error number calculation unit that calculates a number that uses the character input to the correction input unit as a correct word and adds the calculated number to the second error number stored in the error number storage unit; ,
Based on the first number of appearances, the second number of appearances, the first number of errors, and the second number of errors corresponding to the correct word, the weight coefficient is calculated, and the weight A weight coefficient calculation unit to be stored in the coefficient storage unit;
The input error warning device according to claim 1, further comprising: The pattern recognition process is one of a character recognition process for detecting character information from image data that is the processing target data and a voice recognition process for detecting character information from voice data that is the processing target data. The input error warning device according to claim 1, wherein the input error warning device is a warning. A user attribute information storage unit for storing user information indicating the attribute of the user;
In the appearance number storage unit, the second appearance number is stored for each attribute of the user,
The error number storage unit stores the second error number for each attribute of the user,
The weighting coefficient storage unit stores the weighting coefficient for each attribute of the user,
The input error warning device according to any one of claims 1 to 3, wherein the error rate calculation unit calculates the error rate for each attribute of the user. The error rate calculation unit sets the first number of appearances corresponding to the correct word as n ₀ , sets the second number of appearances as n ₁ , sets the first number of errors as e ₀ , When the number of errors is e ₁ and the weighting factor is T, the error rate P in which the correct word is misread as the error word is expressed by the following equation (a)

The input error warning device according to claim 1, wherein the input error warning device is calculated by: The correction input unit receives the correction input over a plurality of stages;
The weighting factor calculation unit sets the error rate calculated by the error rate calculation unit in the last step among the plurality of steps as P, sets the first appearance number corresponding to the correct word as n _0, and The second occurrence number is n ₁ , the first error number is e ₀ , the second error number is e ₁ , and the weight coefficient T is expressed by the following equation (b)

The input error warning device according to claim 2, wherein the input error warning device is calculated by:

Images