JP4982587B2 - Data entry system and data entry method - Google Patents

Description

  Embodiments described herein relate generally to a data entry system and a data entry method.

  For example, there is a data entry system for business in which an image of a form such as an application form is displayed on a screen, and an entry worker inputs information on the form with a keyboard (key input) while viewing the screen.

  In general, entry is made by key input when converting the contents of a paper form received by FAX or the like into data.

  However, this work tends to cause problems such as typographical errors and conversion errors when an unskilled person works. For example, when an obfuscated character is included in an address or name, the key entry itself may be difficult.

  As a technique corresponding to the input of such an obfuscated character, for example, there is a technique for recognizing a handwritten character image. In the case of this technique, character input can be performed without entering all characters by presenting not only handwritten characters but also candidate character strings or candidate words that follow (see, for example, Patent Document 1).

JP-A-8-96085

  The above-described conventional technique is based on the premise that a character or word written by a person is correct, and has a problem that it is not suitable for a data entry operation that may cause a typo such as key input.

  Moreover, it is not 100% certain to recognize a character from an image obtained by optically (mechanically) reading a character written on a form, and misreading occurs in some cases. For this reason, in a system in which no error is allowed, there is a problem that it takes a lot of time and labor to inspect the recognized character.

  In addition, when optical address reading processing recognizes characters in the address entry field of a form image, for example, the character string “Kanagawa” is changed to a string of consecutive characters such as “Kyodogawa”. When an intermediate character is misread, there is a problem that if only the character at the corresponding position is to be corrected, the number of times of input conversion increases, and the work takes much time.

  The present invention has been made to solve such a problem. When a worker performs data entry work while looking at a form image, the data entry work can be performed easily and efficiently while covering input errors. It is an object of the present invention to provide a data entry system and a data entry method that can support the above.

  In order to solve the above-described problem, an image obtained by optically reading a form on which a key entry target character is described in a data entry device including an input device and a display device is provided in the data entry device of the embodiment. Is stored in the image storage unit, a structure storage unit capable of storing a candidate structure of a character string connecting the key entry target characters, and an image stored in the image storage unit. A character recognition unit that recognizes each character from the image of the cut region, generates a character string structure in which the character is associated with a position and a score, and stores the structure in the structure storage unit; A screen including an image of the form and an input box for inputting characters described in the form is displayed on the display device, and a screen system for accepting key input to the input box. And characters or character string candidates related to the received key are read from the structure of the structure storage unit and displayed as conversion candidates in the input box, or displayed as reference information in the vicinity of the input box An entry support unit.

  Further, the data entry method of the embodiment is an image obtained by optically reading a form on which a key entry target character is described in the data entry method in a data entry device including an input device, a storage device, and a display device. Are stored in the storage device, the image stored in the storage device is read out, the read image is cut out along the character string, each character is recognized from the image of the cut out region, and the character is linked by the position and the score. A screen having a step of generating a structure of attached character strings and storing the structure in the storage device, an image of the form, and an input box for inputting characters described in the form is displayed on the display device. Displaying a key input to the input box, and characters or character string candidates related to the received key. And a step of displaying the display as a conversion candidate to read the input box from the body, or in the vicinity of the input box as the information for reference.

It is a figure which shows the structure of the data entry system of one Embodiment of this invention. It is a figure which shows the data of a recognition result memory | storage part. It is a flowchart which shows a character recognition process. It is a flowchart which shows creation operation of a recognition result structure. It is a figure for demonstrating the process which produces a recognition result structure. It is a figure which shows an example of the preparation procedure of a character candidate structure. It is a figure which shows an example of a data entry screen. It is a flowchart which shows key entry assistance operation | movement. It is a figure for demonstrating the operation | movement which produces | generates a complement character string from the most likely character string from the recognition result structure. It is a figure which shows the example when a part of recognition result structure is missing and produced. It is a figure for demonstrating becoming undecided in the recognition result structure of FIG. It is a flowchart which shows the operation | movement which complements the missing character. It is a figure for demonstrating the operation | movement which complements the missing character. It is a figure for demonstrating the operation | movement which complements the missing character. It is a figure which shows a mode that each element of a recognition result structure hold | maintains a candidate character and each similarity. It is a figure which shows the example which displayed the character string of both the case where the input character is utilized, and the case where the recognition result 1st place is utilized.

  Hereinafter, a data entry system according to an embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing the configuration of the data entry system of the first embodiment.

(First embodiment)
As shown in FIG. 1, the data entry system according to the first embodiment includes an image scanner 1 (hereinafter referred to as “scanner 1”) as an image reading device or an image input device, a text input, a data entry start operation, A text input device such as a keyboard 2 for performing an instruction operation such as an end operation, a computer main body 3 as a data entry device, and a display device such as a monitor 4 are provided. The text input device includes, for example, a device that is specified by clicking and specifying kana or roman text displayed on the screen by a combination of a mouse, a trackball, and a GUI.

  The scanner 1 includes, for example, a CCD line sensor in which imaging elements such as CCDs are arranged in a line, and a paper feed mechanism for continuously supplying a form to the CCD line sensor. In addition, the scanner 1 may be provided with a flat bed for reading a form one by one.

  The scanner 1 scans (reads) an object to be read (form such as an application form) by a CCD line sensor and inputs it to the computer main body 3 as a form image (digital image). The form has a predetermined character entry position, such as an order form (see FIG. 3), for example, and the form image (digital image) is character-recognized based on a preset order form format. It is assumed that text data of characters described in the purchase order is obtained.

  In this embodiment, the scanner 1 is exemplified as the image input device. However, if an “image obtained by optically reading a form” is input, for example, a server on a network, a network line, a communication interface thereof, and the like can be input. Included in the device. In addition, various types of image input devices such as a digital camera are allowed.

  The monitor 4 displays a data input screen (data entry screen 50 in FIG. 7) output from the computer main body 3. That is, the monitor 4 functions as a display unit that displays a form image (digital image) obtained by optically reading a form.

  The computer main body 3 includes an input receiving unit 31, a memory 32, a character recognition unit 33, a character recognition dictionary 34, a basic dictionary 36 as a Japanese conversion dictionary, an entry unit 35, and the like. The memory 32 includes an image storage unit 32a and a recognition result storage unit 32b.

  The character recognition dictionary 34 stores a large number of general dictionary data for character recognition (character images (or character feature vectors) associated with text data). That is, the character recognition dictionary 34 is a dictionary used for character recognition. In addition to the character recognition dictionary 34, the memory 32 is provided with a basic dictionary 36 for Japanese conversion that is referred to when a key is input using, for example, “Romaji” or “Kana”. The basic dictionary 36 associates the character code input by the key with text data.

  Japanese characters or character string candidates corresponding to the keys input from the keyboard 2 in a state where the cursor is displayed in the input boxes 52 to 55 of the data entry screen 50 of FIG. It is displayed in the input boxes 52-55. The basic dictionary 36 is a Japanese conversion dictionary in which conversion candidate characters corresponding to keys input from the keyboard 2 are stored. For example, by inputting “Nihon”, a kanji character “Japan” is output. Is for.

  The image storage unit 32a stores a form image, a character image area, and the like input from the scanner 1. That is, the image storage unit 32a stores an image obtained by the scanner 1 optically reading a form on which a key entry target character is written.

  The recognition result storage unit 32b stores the character recognition result recognized by the character recognition unit 33. For example, as shown in FIG. 2, the character recognition result is stored as a structure (table) of character elements (nodes) in which each recognized character is associated with a score added to each candidate as a character candidate. A structure (table) of character elements (nodes) is called a recognition result structure.

  This recognition result structure is a structure in which a form image is cut out for each character line, a character string is recognized from an image of the cut-out area, and the relationship between the characters and characters constituting the character string is linked with a score. A specific method of creating the recognition result structure will be described in detail in the operation description to be described later.

  The memory 32 holds text data input from the keyboard 2. In addition, the memory 32 stores form image format information (hereinafter referred to as “form format”) for the form to be read, input field definition information for configuring the data entry screen 50, and the like.

  The form format is information indicating what information is written in which position of the form image corresponding to the form of the reading source. In other words, the memory 32 stores a form format including the arrangement of character description portions in the form.

  That is, the memory 32 temporarily stores input data and functions as a work area (temporary data storage area) for the character recognition unit 33 and the entry unit 35 to perform respective processes. Elements other than the memory 32 and the character recognition dictionary 34 (input receiving unit 31, character recognition unit 33, entry unit 35, etc.) are executed by a CPU provided in the computer main body 3 by executing a program installed in the hard disk device. Realized. Each of the above elements may be realized not only by software but also by hardware.

  The input receiving unit 31 receives a form image (digital image) input from the scanner 1 and stores it in the image storage unit 32 a of the memory 32. The input receiving unit 31 receives input data such as text data input from the keyboard 2 and passes it to the entry unit 35.

  In accordance with the form format of the memory 32, the character recognition unit 33 cuts out a character string image from the form image for each character line, performs character recognition processing with reference to the character recognition dictionary 34 for the cut out character line, and extracts the extracted character line. A structure in which the character strings are connected is constructed (created) using the character recognition result of each character line, and is stored in the recognition result storage unit 32b. Note that the form format may be input by a key operation by an entry worker during data entry work.

  That is, the character recognizing unit 33 cuts out the image stored in the image storage unit 32a in character string region units, cuts out the cut out character string region in a recognizable minimum unit, and extracts the cut out minimum unit image. The character recognition dictionary 34 is used to recognize characters or character elements, and adjacent character or character elements of the recognition result are combined to form a character string candidate structure configured for each row region of the character string (this is referred to as “recognition result”). Structure ”) and is stored in the recognition result storage unit 32b. In other words, this recognition result structure is a structure that recognizes each character from the region of the character line cut out from the form image, associates the character with the position and the score, and indicates the connection between the characters. The connection between letters is called a path.

  The form format is arrangement data such as which item's character image area exists at which position in the form image. For example, in the case of an order form, the orderer's name, delivery address, order number, amount , Coordinate information (start point (upper left corner) and end point (lower right corner)) indicating the cutout range of the character image area such as a price. The coordinate information is, for example, coordinates on a two-dimensional plane with the upper left corner of the form image as a reference point (x0, y0, etc.).

  The input field definition information includes coordinate information (start point) indicating the position where the text input box is displayed on the data entry screen 50, the type of characters to be input (kanji / alphanumeric, half-width / full-width, etc.), the number of characters, prohibited characters, color (background , Characters), input order, and the like. The input order is the order of activating the target text input box to which text should be input among multiple text input boxes, and the character image area (frame) corresponding to the activated text input box on the form image. Information for pointing.

  The entry unit 35 displays a data entry screen 50 (see FIG. 7) having a form image 51 and a text input box 52 in accordance with the input field definition information, and the keys entered in the input box 52 of the data entry screen 50 The character or character string (text data) read out from the basic dictionary 36 and the recognition result storage unit 32b and stored in the memory 32 or the hard disk device as key entry data.

  That is, the entry unit 35 outputs (displays) a data entry screen 50 having a form image 51 and an input box 52 for inputting characters described in the form to the monitor 4. It functions as a screen control unit that accepts key input to the input box 52.

  The entry unit 35 searches the basic dictionary 36 for a conversion candidate character corresponding to the received key, displays it in the input box 52, and displays the character or character string associated with the key in the recognition result storage unit 32b. It functions as an entry support unit that displays as a conversion candidate equivalent to the basic dictionary 36 or information for reference on the screen read out from the structure, for example, directly below the input box 52.

  The operation of this data entry system will be described below with reference to FIGS.

  As shown in FIG. 3, an example in which data entry is performed while a form image 42 obtained by reading with the scanner 1 is displayed on the data entry screen 50 using a form such as an order form 41 as an original document will be described. In addition to the order form 41, the same processing can be performed for other types of forms such as an application form for insurance or a card. Further, not only the form in this example but also a general document such as a book or a notebook may be set as a data entry processing target.

  In this data entry system, when the form code of the order form 41 to be processed is designated by key input and a data entry start operation is performed, the order form 41 is scanned by the scanner 1 and the form image 42 is transferred to the computer main body 3. Entered. The form code may be detected from the form image at the time of scanning.

  In the computer main body 3, the input receiving unit 31 receives the form image 42 input from the scanner 1, and stores it in the image storage unit 32a. The form image is, for example, in JPEG format or PNG format.

  Next, the character recognition unit 33 reads out the form image stored in the image storage unit 32a, performs character recognition on the read out form image, creates a recognition result structure of the entry contents, and recognizes the recognition result storage unit 32b. Remember it.

  Here, the process in which the character recognition unit 33 creates the recognition result structure will be described in detail with reference to FIGS. The character recognition unit 33 reads a form image from the image storage unit 32a, and extracts a character line area from the read form image (step S101 in FIG. 4).

  Then, the character recognition unit 33 binarizes the extracted image of the region of the character line (step S102).

  The character recognition unit 33 cuts out the positions of the character candidates from each of the binarized character line images (step S103), and extracts all the positions of the character image areas from the image. As a method for character extraction, for example, a method of referring to a projection histogram and cutting out at a blank portion is used. In addition to this, as a means for extracting character candidates, for example, a model of a form to be entered may be defined, and the position of the character string may be determined in accordance with the model.

  Next, the character recognition unit 33 cuts out an image area of the character according to the extracted position, performs an integration process of the cut out image area (step S104), generates a character candidate (step S105), and generates a character candidate. Character recognition is performed (step S106), and a recognition result structure is constructed (created) (step S107).

  The change of the image in the above process will be described in order with reference to FIG. As shown in FIG. 5, a character line image 61 to be processed is cut out according to the position of the extracted character line, and binarization processing is performed on the cut out character line image 61 by Otsu's method or the like (step S102 in FIG. 5). ). In addition, for the binarization process, a known binarization method such as the Kittler method can also be used.

  Next, character component candidates are extracted from the binary image 62. Specifically, for example, a histogram in which the number of black pixels is projected in a direction perpendicular to the line is created, all areas having a value of 0 on the histogram are detected, and the center of each area is defined as a character component candidate delimiter. .

  In addition to this, for example, a so-called peripheral feature may be acquired, and a candidate may be acquired by using another method, such as a place where the value is large or a place where the change is large.

  Next, the character recognition unit 33 performs integration determination (inspection) of the character component candidates, integrates those that can be integrated, and creates a character candidate structure (step S104 in FIG. 5). An example of the procedure for creating the character candidate structure is shown in FIG.

  As shown in FIG. 6, each element of the character candidate structure is a character component candidate itself or a plurality of character component candidates connected to each other. First, a character candidate structure (path between characters) is created assuming that all character component candidates are character candidates.

  Next, in the character candidate structure, it is determined whether or not adjacent character candidates can be connected. If the character candidates can be connected, the combination is searched. As a condition for determining that connection is possible, it is assumed that (aspect ratio) <(threshold) is satisfied on the basis of the aspect ratio (width) / (height) of the character components when connected.

  In this example, the threshold value is set to 1.2, for example, but this value may be another value. For example, character recognition is performed on a character candidate generated when concatenated and a certain degree of similarity is obtained. Different criteria may be employed, such as

  And about the combination which couple | bonds and produces | generates a new character candidate, the element is inserted in parallel with what treats each as a character independently. A character candidate structure can be constructed by repeating this procedure until there are no more pairs of character components to be connected.

  Finally, character recognition is performed on each character candidate area of the character candidate structure created by the above processing, and finally recognition is performed by changing the combination of character candidates and connecting them with different paths as shown in FIG. A result structure 65 is created.

  In this example, for example, a subspace method is used as a character recognition method. In addition, other known different character recognition methods such as the nearest neighbor method may be used. In this example, each element of the recognition result structure 65 includes character information acquired as a result of performing character recognition on the element of the corresponding character candidate structure.

  Next, the operation during the key entry work will be described with reference to FIGS. In this system, a recognition result structure stored in the recognition result storage unit 32b as a character recognition result is used as a dictionary for presenting entry candidates for key entry work to the entry worker.

  First, the data entry screen 50 will be described with reference to FIG. In the case of this system, when a data entry start operation is performed by a key operation by an entry operator, the entry unit 35 outputs a data entry screen 50 to the monitor 4, and the data entry as shown in FIG. A screen 50 is displayed.

  On the data entry screen 50, a purchase order image 51 is displayed on the left side of the screen as a form image, and a plurality of input boxes 52 to 55 are vertically arranged on the right side of the screen. Further, in the vicinity of each of the input boxes 52 to 55, a character string candidate 56 as a result of searching the recognition result structure in the recognition result storage unit 32b is displayed for reference. The character string candidates 56 may be displayed in the input boxes 52 to 55 as conversion candidates.

  The input boxes 52 to 55 displayed on the data entry screen 50 are activated in order from the top, and when the entry worker inputs text, the corresponding character image area of the purchase order image is surrounded by a frame. Displayed. In this example, the input box 52 activated first is a text input box for inputting the name of the orderer.

  By inputting the text in the input box 52 and then confirming it by operating the enter key or the like, the text data (input data) input in the input box 52 is stored in the memory 32 by the entry unit 35, and at the same time The input box 53 is switched to active so that an address can be input.

  The entry unit 35 receives a key input from the entry worker. The entry worker inputs the data described in the form as it is, and performs so-called data entry processing for converting the image into character data.

  The entry unit 35 first reads out the form image acquired by the scanner 1 from the image storage unit 32 a and displays it on the left side in the data entry screen 50. In this example, the form image data is displayed on the screen of the computer main body 3 that has taken in. However, it may be displayed on another computer dedicated to data entry that is connected to the computer main body 3 via a LAN or the like.

  When the entry worker inputs a key into the input box 52 on the right side in the data entry screen 50, the input receiving unit 31 receives the key input from the entry worker and passes it to the entry unit 35.

  Then, the entry unit 35 acquires the recognition result structure from the recognition result storage unit 32b using the received key as a keyword (step S201 in FIG. 8), and the text data corresponding to the input key from the basic dictionary 36. Obtain (step S202).

  When the data acquisition ends (step S203), the entry unit 35 collates the acquired text data with the character string of the recognition result structure (step S204), and matches the acquired text data string (character string) ( A character string to be matched is searched (searched) on the recognition result structure (step S205).

  As a result of the search, if there is a matching part on the recognition result structure (Yes in step S205), the entry unit 35 determines the most likely character for the subsequent part of the matching part of the recognition result structure. A column is determined, and the character string is complemented as an input candidate (step S206) and displayed in the input box 52 (step S207). Note that the most likely character string for the continuation of the matching part may be displayed below the input box 52 separately from the input and presented to the entry worker.

  Note that the most probable method for determining a character string in this example is to use an aspect ratio as an image of a character candidate, and use a character string that connects characters on a path whose average aspect ratio is closest to 1 to the character string. It is assumed that the character string is highly likely.

  Hereinafter, in this example, an example in which the entry worker inputs a character string up to “Kanagawa” in the “Destination” input box 53 (input field) of the data entry screen 50 will be described.

  First, the entry unit 35 searches the recognition result structure in the recognition result storage unit 32b for “Kanagawa”, which is a character string key-inputted this time. When searching for a character string, it is assumed that the input character string is from the beginning of the character string image and that the search result is continuous in the recognition result structure. Search as the head of the recognition result structure.

  In this example, the positions indicated by the thick black frames in FIG. 9 match. Paths parallel to this, that is, all paths in which the character component candidates included in the character string “Kanagawa” are used are not likely to be correct, and are deleted from the candidates. In this example, the portion that is displayed in light color in FIG.

  However, for example, when a portion that can be confirmed such as an address written together with a zip code is included, the search start point may be processed in the same manner as the head of the portion that cannot be confirmed. In the above example, since the address can be determined up to the municipality without the need for entry, the recognition result may be adopted as it is, and entry may be made from the head of the block.

  Further, even in the above case, the above-mentioned two places may be set as the search starting points, and it may be started from either of the areas with a high degree of certainty up to the block and with a low degree of certainty below the block. In this case, when the input is started from a portion with a low certainty factor, the recognition result may be automatically complemented with a portion with a high certainty factor.

  Finally, a complementary character string is generated by concatenating the most likely candidates for the remaining portions. In this example, the path with the average aspect ratio closest to 1 is set as the most likely path, and “prefecture Kawasaki city” obtained by concatenating the character strings on the path is set as the supplementary character string.

  Finally, “Kawasaki City, Kanagawa Prefecture”, which is obtained by concatenating the input character string “Kanagawa” with the complementary character string “Kawasaki City, Kanagawa”, is presented as an input candidate character string. If this input candidate character string is correct, the entry operator can end the entry only by confirming the above, and data entry can be performed easily and efficiently.

  As described above, according to the data entry system of the first embodiment, the recognition result structure configured by recognizing the form to be entered in advance is used as a prediction dictionary when the entry worker performs key entry. Thus, even an unskilled person can easily and efficiently perform entry work. Even if the entry operator makes a mistake in key input, the corresponding character candidate is displayed from the recognition result structure created using the form image, so that input mistakes can be covered.

(Second Embodiment)
Hereinafter, the second embodiment will be described with reference to FIGS. 10 to 14. As the second embodiment, a process will be described in the case where a character that is not included in the recognition result structure of the recognition result storage unit 32b exists in the character string that is key-input.

  In the recognition result structure shown in FIG. 10, for example, a case where the “river” portion of “Kanagawa Prefecture” is missing will be described as an example. Since the processing up to the entry unit 35 is the same as described above, the description is omitted.

  First, it is assumed that the entry worker has entered keys in the input box 53 up to “Kanagawa” as shown in FIG. As in the previous examples, even if “Kanagawa” is searched from within the recognition result structure, the search results need to be continuous within the recognition result structure as described above. Has only “1” and “I” paths, and no path following the word “river”. For this reason, among the “Kanagawa” key input, the last character “river” does not match the recognition result structure, and the information in the input box 53 is not fixed.

  In this case, the entry unit 35 performs alignment with the recognition result structure for each character, and checks whether there is a matching portion.

  In this example, since the “KANA” part matches the recognition result structure, the part corresponding to the matching “KANA” part is first determined in the same manner as in the above embodiment.

  The lack of the “river” part does not match the data in the recognition result structure, so the “river” is not fixed. There are two methods of processing in such an example, a method of giving priority to key input and a method of filling a hole from a character input thereafter, which will be described below.

  First, a method for giving priority to key input will be described. In this case, the entry unit 35 does not perform the complement processing, and reads the text data corresponding to the key input up to “Kanagawa” from the basic dictionary 36 and inputs “Kanagawa” as the key is input, as shown in FIG. Displayed in box 53. The entry worker looks at this display content to determine whether conversion is correct or not, and when it is determined to be correct, confirms “Kanagawa” by performing a confirmation operation.

  Next, a method of filling a missing character by complementing will be described. In this case, when “Kanagawa” is input to the input box 53 by the entry worker and then “Kawasaki-shi ...” is key-inputted to the input box 53, the entry unit 35 first displays each character. Alignment is performed (step S301 in FIG. 12), and it is determined whether or not there is a matching portion between the text data read from the basic dictionary 36 by key input and the data of the recognition result structure among the unconfirmed portions. If there is no matching part (No in step S302), the check is NG (step S303) and the process is terminated.

  If there is a matching part (Yes in step S302), it is further determined whether or not the matching part is completely matched (step S304). If there is no perfect matching (No in step S304), the character string of the text data It is determined whether or not there is a missing character from the arrangement order and the width of the character string area (No in step S305). If it is a missing character (Yes in step S305), the missing part is cut out and re-executed. Recognize (step S306).

  That is, the entry unit 35 searches for a matching character candidate that exists near the most determined part “Kana” among the unconfirmed parts.

  In the case of this example, as shown in FIG. Assuming that “prefecture” has been matched, by performing the same processing as in the first embodiment, the range in which the missing character (“river”) exists is determined, the range is cut out, By re-recognizing the part, the missing character “river” was derived, and this “river” was matched with the “river” of “Kanagawa”, which was the keyed-in character string (step) As shown in FIG. 14, a complementary processing result of “Kawasaki City, Kanagawa Pref.” Is obtained as matching OK (Yes in S307), and even if the recognition result structure has an error, the key entry is correct. It can be carried out.

  As described above, according to the data entry system of the second embodiment, when the missing part matches the data read from the character recognition result structure, the character determined in the vicinity of the missing part and the text data obtained thereafter are obtained. Since that part is complemented from the relationship, even if there is an error in the character recognition result structure configured by recognizing the form to be entered in advance, the key entry can be performed correctly.

(Third embodiment)
Hereinafter, the third embodiment will be described with reference to FIGS. 15 and 16. However, in the third embodiment, when the character recognizing unit 33 creates the character recognition result structure, the character recognizing unit 33 returns the recognition result candidate and the similarity between each candidate instead of the recognition result character string. It is assumed that each element of the recognition result structure holds a plurality of candidate characters and respective similarities.

  The processing procedure up to character recognition performed by the character recognition unit 33 is executed in the same manner as in the first embodiment. In the third embodiment, as shown in FIG. 15, each element of the recognition result structure holds a candidate character and each similarity.

  Next, the entry unit 35 accepts a data entry by an entry worker. In this example, it is assumed that the entry worker has entered, for example, “Kawasaki City, Kanagawa Prefecture” in the destination box 53 of the data entry screen 50.

  In this case, the entry unit 35 first searches the input character string in the recognition result structure. However, since each character candidate has a plurality of candidates in this example, a search is performed for the second and lower ranks, and the portion with the lowest average rank of the matched portions is determined as the matched portion. Further, the present invention is not limited to this, and for example, a portion that matches on different criteria, such as a portion having the highest average similarity, may be determined.

  In this example, the portions indicated by thick black frames in FIG. Next, the input is searched for an example using a character that is not the first candidate. In this example, “倚” corresponds. If it exists, as shown at the bottom of FIG. 16, the character strings of both “when using the input character” and “when using the first recognition result” are displayed so that the correct one can be selected. To.

  In this example, the character strings are simply arranged and displayed. However, only the corresponding part may be presented in a double manner. In this case, for example, the color may be changed for easy understanding.

  In addition, when a character that is not the first candidate is used, when the similarity of the character input in the character candidate is sufficiently close to the first candidate of the recognition candidate, the above display is not performed. You may make it employ | adopt the character input unconditionally. As described above, the entry worker can surely input the correct answer even if it is an error-prone character that causes misreading.

  In addition, since the character recognition result and the entry work result are compared, the input work is performed to confirm the character recognition result at the same time. As a result, high-accuracy input can be performed.

  In the above-described embodiment, the supplement processing is started when input is started. However, even when input is not started, the supplement processing is performed and displayed, and the result of the supplement is displayed to the entry worker from the beginning. You may make it show. The character string displayed when this processing is performed may be a character string on the most likely path obtained from the entire recognition result structure.

  If the complemented result is correct, it is only necessary to confirm it as it is, and if there is an error in the complemented result, it is only necessary to continue input, so that a simple entry operation can be realized.

  In the example described so far, the starting point of the character string matching search is limited, but this may be started from all elements of the recognition result structure. In this way, it is possible to perform correction by entering only a character string including a portion that needs to be corrected when a complementary result is presented before input as described above, which is easier than inputting a single character. It becomes possible to make corrections.

  As described above, according to the data entry system of the third embodiment, the character strings of “when the input character is used” and “when the first recognition result is used” are displayed, and the correct one is selected. By making it possible, for example, correction when a character between character strings is wrong can be performed only by a selection operation, so that entry work can be performed efficiently. In addition to using the input character itself, for example, an input character string converted by a conversion dictionary may be used.

  As a result of the above-described embodiments, it is possible to assist the entry worker so that the data entry operation can be performed efficiently and reliably while looking at the image of the form while covering his / her key input mistake.

  In addition, this invention is not limited only to the said embodiment, You may deform | transform a component in the range which does not deviate from the summary in an implementation stage. For example, each component may be realized by a program installed in a storage such as a hard disk device of a computer.

  The program may be stored in a computer-readable electronic medium, and the computer may realize the functions of the present invention by causing the computer to read the program from the electronic medium. Examples of the electronic medium include a recording medium such as a CD-ROM, flash memory, and removable media.

  Further, the configuration may be realized by distributing and storing components in different computers connected via a network, and communicating between computers in which the components are functioning.

  DESCRIPTION OF SYMBOLS 1 ... Scanner, 2 ... Keyboard, 3 ... Computer main body, 4 ... Monitor, 50 ... Data entry screen, 31 ... Input reception part, 32 ... Memory, 32a ... Image storage part, 32b ... Recognition result storage part, 33 ... Character recognition 34, character recognition dictionary, 35, entry portion, 36, basic dictionary.

Claims (4)

In a data entry device comprising an input device and a display device,
An image storage unit that stores an image obtained by optically reading a form in which characters for key entry are described;
A structure storage unit capable of storing a candidate structure of a character string connecting the key entry target characters;
Reads out the image stored in the image storage unit, cuts out the read image along the character string, recognizes each character from the image of the cut-out area, and generates a character string structure in which the character is linked by position and score A character recognition unit stored in the structure storage unit;
A screen control unit for displaying a screen having an image of the form and an input box for inputting characters described in the form on the display device, and receiving a key input to the input box;
Entry support for reading a character or character string candidate associated with the received key from the structure in the structure storage unit and displaying it as a conversion candidate in the input box or as reference information in the vicinity of the input box A data entry device. The entry support unit
The key conversion candidate character or character string obtained from the basic dictionary by the search is modified to match the character or character string candidate read from the character string candidate structure in the structure storage unit. 2. The data entry device according to claim 1, wherein the corrected character string is displayed in or near the input box. The entry support unit
When reading the character or character string related to the key from the character string candidate structure in the structure storage unit, the character or character string corresponding to the input key is complemented, 2. The data entry device according to claim 1, wherein the complemented character or character string is displayed in or near the input box. In a data entry method in a data entry device including an input device, a storage device, and a display device,
Storing in the storage device an image obtained by optically reading a form in which characters for key entry are described;
Reads out the image stored in the storage device, cuts out the read image along the character string, recognizes each character from the image of the cut out region, and generates a character string structure in which the character is linked by the position and the score. Storing in the storage device;
Displaying a screen having an image of the form and an input box for inputting characters described in the form on the display device, and receiving a key input to the input box;
A step of reading a candidate of a character or a character string related to the accepted key from the structure and displaying it as a conversion candidate in the input box, or displaying it as information for reference in the vicinity of the input box. Data entry method.

Images