JP3750406B2 - Document filing device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a document filing apparatus for correcting an error included in search information for searching a document image.
[0002]
[Prior art]
In a document filing apparatus for filing document images, when registering a document image, the character recognition result of the document image is stored in the database as search information, and when searching for the document image, the keyword input by the user is included in the search information. Some present document images to the user. An example of such a document filing method is disclosed in Japanese Patent Laid-Open No. 6-28588. Hereinafter, this conventional document filing method will be described.
[0003]
FIG. 17 is a block diagram of a document filing apparatus showing a conventional document filing method. In the figure, 1 is a scanner, 2 is a character recognition device, 3 is a document image database, 4 is a search information database, 5 is a search engine, and 6 is a search terminal.
[0004]
FIG. 18 is a diagram for explaining the contents of the search information. In the figure, 50 is character code data obtained by character recognition of a document image, and 51 is character position data indicating where each recognized character is on the document image.
[0005]
FIG. 19 is a diagram illustrating a display state of search results. In the figure, 52 is a document image hit by the search, and 53 is a frame line indicating an area on the document image hit by the keyword.
[0006]
The operation of the conventional document filing method will be described below with reference to the drawings. This description will be divided into document image registration processing and search processing.
First, in the document image registration process, a paper document to be registered is read by the scanner 1 to create a document image. The document image is stored in the document image database 3 which is a database for storing the document image and is sent to the character recognition device 2.
[0007]
The character recognition device 2 recognizes characters in the document image read by the scanner 1 and creates search information for the document image. As shown in FIG. 18, this search information includes character code data 50 obtained by recognizing characters in the document image and character position data 51 indicating where each recognized character is located on the document image. The character position data 51 includes the upper left point coordinates (X, Y) of the circumscribed rectangle of each character and the width W and height H of the circumscribed rectangle.
Finally, the created search information is stored in the search information database 4 in association with the document image stored in the document image database 3. The document image registration process has been described above.
[0008]
Next, document image search processing will be described. When searching, the user inputs a desired keyword from the search terminal 6 and inputs it to the search engine 5.
The search engine 5 refers to the search information for each document stored in the search information database 4 and searches for the keyword from the character code data 50. If a keyword is found, the position on the document image is checked. Specifically, an area on the document image that each character code that matches the keyword has is obtained from the character position data 51, and one rectangular area that includes these character codes is determined as an area on the document image that hits the keyword. judge. Thereafter, the document image of the document including the keyword is taken out from the document image database 3 and sent to the search terminal 6 together with the keyword and the area on the document image that has been hit.
[0009]
The search terminal 6 draws a frame line surrounding the area on the document image hit with the keyword on the document image sent from the search engine 5 and displays it on the monitor. FIG. 19 shows an example of a search using the keyword “document filing”. As shown in the figure, a region including a keyword is displayed surrounded by a frame line 53 on a document image 52 hit by the search. This makes it easy to see where the keyword is included in the document image. The above is the document image search process.
[0010]
[Problems to be solved by the invention]
As described above, in the conventional document filing method, search information can be automatically generated when search information is generated from a character recognition result. However, if the character recognition result is incorrect, there is a problem that the search performance is degraded. For errors in character recognition results,
(1) A character extraction error that divides one character into a plurality of characters or recognizes a plurality of characters as one character,
(2) misidentification of another similar character,
(3) Character string extraction error that does not recognize a character as a figure or the like,
(4) There is a misprinting of characters that recognizes a part of a figure or photograph as a character. When searching, the character code of the recognition result is incorrect in (1) and (2), so an extra document hits or a desired document does not hit, and in (3) the correct character recognition result is missing. The search performance deteriorates such that the desired document does not hit, and in (4), the search information includes an extra character code, so that the extra document hits.
[0011]
In addition, regarding an error in the character recognition result, a plurality of candidate characters for one character pattern, a feature indicating a rough shape of the character pattern, and the like are added to the search information as additional information, and character code data and keywords Although there is a method in which the desired document can be hit by matching the keyword with the additional information even if the IDs do not match, there is a problem that searching for unnecessary documents is reduced by having additional information, but unnecessary documents are searched incorrectly There is.
[0012]
The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a document filing apparatus that can correct or delete search information so that search performance does not deteriorate. To do.
[0013]
The correction of the recognition result is generally performed by a conventional character recognition device. However, this is to edit the character code of the recognition result in order to create an accurate text, and does not have a function of correcting the character position data. For this reason, the recognition result correcting means used in the conventional character recognition apparatus cannot be used to correct the search information.
[Means for Solving the Problems]
A first invention is a document image database in which document images read by a scanner are stored;
Search information for selecting a correction target character area in the document image, comprising character code data of characters in the document image read by the scanner and character position data indicating the position of a circumscribed rectangle of the characters. A character recognition means to create;
A search information database for storing the search information created by the character recognition means;
The document image stored in the document image database and the search information stored in the search information database are displayed in a display area, and the document image is corrected or deleted based on the displayed document image and the search information. When a correction area that requires correction is selected and a correction character is input to the correction area, the position of the circumscribed rectangle of the input correction character is obtained, and based on the position of the correction character and the circumscribed rectangle of the correction character The character code data of the correction target character in the correction area stored in the search information database is replaced with the character code data of the correction character, and the character position data of the correction target character is replaced with the character position data of the correction character. When the correction target character in the correction area is deleted, the correction target character stored in the search information database is overwritten. And search information modification means for deleting the character code data and the character position data, and with a.
[0014]
According to a second aspect of the present invention, when a correction character is input to the correction area, a rectangular area obtained by equally dividing the correction area by the number of the correction characters is obtained as a position of a circumscribed rectangle of the correction character, and the obtained correction character is obtained. The character code data and character position data of the correction target character stored in the search information database are replaced with the character code data and character position data of the correction character based on the position of the circumscribed rectangle of the character and the correction character. Search information correction means is provided.
[0015]
3rd invention is equipped with the table which saved the average character width for every character,
When a corrected character is input to the correction area, the search information correction means obtains a rectangular area obtained by dividing the correction area by the average character width for each character stored in the table as a position of a circumscribed rectangle of the correction character. Based on the position of the circumscribed rectangle of the corrected character obtained and the corrected character, the character code data and the character position data of the correction target character stored in the search information database are Replace with character position data.
[0016]
According to a fourth aspect of the present invention, when the number of correction characters input to the correction area is equal to the number of correction target characters in the correction area, the position of the circumscribed rectangle of the correction target character is changed to the position of the circumscribed rectangle of the correction character. And search information correction means for replacing the character code data of the correction target character stored in the search information database with the character code data of the correction character.
[0017]
According to a fifth aspect of the present invention, when the circumscribed rectangle position of the corrected character thus obtained is required to be corrected, the position of the circumscribed rectangle of the corrected character is newly corrected by an operator instructing rewriting of the search information, and the search information database Search information correction means for rewriting the character position data of the corrected character stored in the file to the newly corrected character position data of the corrected character.
[0018]
According to a sixth aspect of the present invention, when a keyword for searching for a predetermined character is input from the document image, the character code data and character position data of the character constituting the keyword are searched from the search information database. Search means for obtaining the keyword area in the document image based on the retrieved character code data and character position data, and sending the keyword area together with the document image stored in the document image database;
The search information correction means displays the document image in which the keyword area is surrounded by a frame line in the second display area based on the keyword area sent by the search means, and the frame is surrounded by the frame line. The selected area is selected as a correction area that requires correction or deletion.
[0019]
7th invention contains the character code data of the character in the said document image read with the said scanner, the character position data which shows the position of the circumscribed rectangle of a character, the candidate character for every character, and the rough shape of a character Additional information for selecting the region of the character to be corrected in the document image, and character recognition means for creating search information constituted by:
Search information correction for selecting the correction target character based on the additional information stored in the search information database, replacing or deleting the character code data and the character position data of the search information, and then deleting the additional information Means.
[0020]
DETAILED DESCRIPTION OF THE INVENTION
Embodiment 1 FIG.
Embodiment 1 of the present invention will be described below with reference to the drawings.
FIG. 1 is a configuration diagram of the document filing apparatus according to the first embodiment. In the figure, parts that are the same as or correspond to those in the prior art are assigned the same reference numerals, and descriptions thereof are omitted. In the figure, reference numeral 7 denotes a search information correction terminal.
[0021]
FIG. 2 is a flowchart showing the operation procedure of the search information correction method.
FIG. 3 is a diagram for explaining a method for correcting a character segmentation error included in search information. 11 is a document image, 12 is a correction area, and 13 is a character code recognized from within the correction area.
[0022]
FIG. 4 is a diagram for explaining a method of correcting the character position data in the search information, and 15 is the position of each character obtained by equally dividing the correction area.
[0023]
FIG. 5 is a diagram showing the contents of search information before and after correction, 16 is search information before correction, and 17 is search information after correction.
[0024]
FIG. 6 is a diagram for explaining a method for correcting a character string extraction error included in search information. 18 is a document image, and 19 is a title.
[0025]
FIG. 7 is a diagram for explaining a method of correcting a misreading of characters included in search information (character recognition of a part of a figure, a photograph, etc.), 20 is a document image, and 21 is an extra character. It is a correction area including the recognition result.
[0026]
Next, the operation will be described.
As shown in FIG. 1, the configuration of the present embodiment is the same as the conventional configuration except for the search information correction terminal 7, and is the conventional configuration except for the search information correction process using the search information correction terminal 7. Performs the same operation as. Hereinafter, a method for correcting the search information will be described with reference to the flowchart of FIG.
[0027]
First, in the document selection process in step S1, an operator who corrects or deletes search information refers to the document image database 3 and the search information database 4 from the search information correction terminal 7 and selects a document to be corrected. The document image and search information of the selected document are displayed on the monitor of the search information correction terminal 7.
[0028]
Next, in the correction area designating process in step S2, the operator surrounds the area on the document image where the character recognition result is incorrect based on the document image and its search information with the mouse. That is, an area on the document image that is different from the document image in the search information composed of the character code data and the character position data obtained by recognizing the character of the document image is surrounded by the mouse. This enclosed area is called a correction area. For example, an area 12 surrounded by a rectangular frame on the document image 11 shown in FIG. 3 is a correction area. In the correction area 12, the character code 13 “L year wholesale” which is recognized by dividing the 4-character string “1. control” into 5 characters due to a character cutout error is stored in the search information.
[0029]
Next, in the correction location extraction process in step S3, the character position data included in the search information is examined, and the character whose character position is included in the correction area is picked up as an edit target. Hereinafter, these characters are called correction target characters.
[0030]
Next, in the correction content selection processing in step S4, the operator selects whether to delete or correct the search information for the correction target character. If “delete search information” is selected in step S5, the correction target character data is removed from the search information stored in the search information database 4 in the search information correction process 27 in step S8, and the process ends. If “correction of search information” is selected in step S5, the operator inputs a correct character string in the correction area from the keyboard or the like in the correct character code input process in step S6. For example, the correct character string “1. control” of 4 characters is input to the correction area 12 shown in FIG.
[0031]
Next, when the correct character string is composed of N correct characters in the character position calculation process in step S7, the position on the document image of each correct character is obtained. As the position of each correct character on the document image, N rectangular areas obtained by dividing the width of the correction area into N equal parts are obtained as the positions of the respective characters. For example, when a correct character string “1. control” of 4 characters is designated for the correction area 12, the correction area 12 is divided into four rectangular areas 15 shown in FIG. Here, the first rectangular area from the left is the position of the correct character “1”, the second is the position of the correct character “.”, The third is the “control” position, and the fourth is the “go” position.
[0032]
Finally, for the search information stored in the search information database 4 in the search information correction process in step S8, the character code data of the correction target character obtained in the correction part extraction process in step S3 is used as the correct answer in step S6. Replace with the correct character string entered in the character code input process. Similarly, the character position data of the correction target character obtained in the correction part extraction process in step S3 is replaced with the character position of each correct character string obtained in the character position calculation process in step S7. As a result, as shown in FIG. 5, the search information 16 before correction for the correction area 12 is corrected as search information 17. The above is the procedure for correcting the search information.
[0033]
Here, a method for correcting search information including character segmentation errors has been described as an example. For example, a character string made up of misidentified characters (character segmentation position is correct but segmented character recognition result is incorrect). The search information including erroneous identification can also be corrected by specifying the correction area surrounding and performing the above processing.
[0034]
Also, search information including character string extraction errors (characters are not recognized by being determined as a figure or the like) can be corrected by the same processing. For example, in the document image 18 shown in FIG. 6, since the title 19 is filled with a background, it is determined as a figure and character recognition is not performed. Therefore, even if a word included in the title 19 is searched as a keyword, this document cannot be searched. When the above processing is performed by designating a correction area surrounding the title 19, the correct character string and the position of each character are added to the search information. As a result, this document can be searched using the words included in the title 19 as keywords.
[0035]
Furthermore, with regard to the appearance of characters (an error in recognizing a part of a figure or photo as a character), for example, a correction that surrounds a figure or photo that does not contain characters on a document image as shown in FIG. The region 21 is designated and the above processing is executed. However, in the correction content selection process 24, “delete correction target character from search information” is designated. As a result, the correction target character data included in the correction area 21 is collectively deleted from the search information.
[0036]
As described above, according to the present embodiment, when an operator designates a correction area on a document image and inputs a correct character string of N characters, a rectangular area obtained by equally dividing the correction area into N pieces is obtained. Since the character code data and the character position data constituting the search information are corrected as the position of the correct character, the search information including character cutout error, misidentification, and character string extraction error can be corrected. This also makes it possible to obtain accurate search information.
[0037]
In this embodiment, the user designates a correction area on the document image, and the character data included in the correction area can be deleted from the search information. Can be easily deleted from search information.
[0038]
Embodiment 2. FIG.
Embodiment 2 of the present invention will be described below with reference to the drawings. FIG. 8 is a diagram for explaining a method for obtaining the position of each correct character based on the average character width of the correct characters. In the figure, 22 is a table storing the average character width for each character, 23 is the position of each character obtained by dividing the correction area, and 24 is search information after correction.
[0039]
Next, the operation will be described.
The search information correction method follows the same operation procedure as in the flowchart of FIG. Here, since the processes other than the character position calculation process in step S7 are the same as those in the first embodiment, a description thereof will be omitted.
[0040]
First, a table 22 storing an average character width of each character is prepared in advance. In the character position calculation process in step S7, the average character width of correct characters is checked with reference to this table 22. A rectangular area obtained by dividing the width of the correction area into N according to the average character width ratio is set as the position of each character.
For example, as shown in FIG. 8, when a correct character string “1. control” of 4 characters is designated for the correction area 12, the average character width of the character “1” can be obtained by referring to the table 22. Is 20, "." Is 5, "System" is 40, and "Go" is 42. Therefore, a rectangular area obtained by dividing the width of the correction area at a ratio of 20: 5: 40: 42 is set as the position of each character. As a result, the search information 16 before correction is changed to search information 24 after correction.
[0041]
As described above, according to the present embodiment, N rectangular areas that can be obtained by dividing the correction area into non-equal parts based on the average character width of correct characters are used as the positions of correct characters. Since the character position data of the information is corrected, it is possible to obtain search information that can accurately determine the character position even when the character width is different.
[0042]
Embodiment 3 FIG.
Embodiment 3 of the present invention will be described below with reference to the drawings. FIG. 9 is a diagram illustrating a method for obtaining the position of the correct character in the present embodiment. In the figure, 25 is a correction area, 26 is a character position of a correction target character, 27 is search information before correction, and 28 is search information after correction.
[0043]
Next, the operation will be described.
The search information correction method follows the same operation procedure as in the flowchart of FIG. Here, in the correction of the search information, the processes other than the character position calculation process in step S7 in FIG.
[0044]
First, in the character position calculation process in step S7, as in the first embodiment, a rectangular area obtained by dividing the width of the correction area into N pieces is set as the position of each correct character. However, when the number of correction target characters is equal to the number N of correct characters, it is considered that each character is cut out correctly, and the value of the character position data of the correction target character is used as the position of each correct character.
For example, as shown in FIG. 9, if the character code in the search information for the correction area 25 is “Horiko” and the character position is the position indicated by the rectangle 26, the correct character string If “in” is given, the number of characters to be corrected and the correct character are both 4, so that the original value shown in the rectangle 26 is used as it is for the character position of each correct character, and the character code data is replaced with “in”. . As a result, the corrected search result 28 is obtained from the search result 27 before correction.
[0045]
As described above, according to the present embodiment, when the number of correction target characters is equal to the number N of correct answer characters, the value of the character position data held by the correction target characters is used as the position of each correct character as it is. Since the character position data is corrected, if the character cutout result at the time of character recognition is correct, the character position data becomes an accurate value without error.
[0046]
Embodiment 4 FIG.
Embodiment 4 of the present invention will be described below with reference to the drawings. FIG. 10 is a diagram for explaining a method for correcting a character position in the present embodiment. In the figure, 29 is an automatically obtained initial character position, 30 is a result of the operator correcting the character position, and 31 is search information created.
[0047]
Next, the operation will be described.
The search information correction method follows the same operation procedure as in the flowchart of FIG. Here, in the correction of the search information, the processes other than the character position calculation process in step S7 in FIG.
[0048]
First, in the character position calculation process in step S7, as described in the previous embodiment, if the correction area is divided by the number N of correct characters or the number of correction target characters is equal to N, the correction target character The position of each correct character is automatically obtained by using the character position data value of. The position of the correct character thus obtained is displayed on the monitor of the search information correction terminal 7 so as to overlap with the image of the correction area.
[0049]
For example, as shown in FIG. 10, the position of each correct character is displayed in a rectangle 29 on the correction area 12. If the position of each correct character is incorrect, the operator designates an arbitrary rectangle with the mouse and corrects the position and size correctly. As a result, a rectangle 30 indicating the correct character position is obtained. Search information is created using the position of the rectangle thus corrected as the character position of each correct character. As a result, for example, the search information 31 obtained in this embodiment has correct character position data as compared with the search information 17 obtained by the method of dividing the correction area shown in Embodiment 1 into N equal parts.
[0050]
As described above, according to the present embodiment, since the operator can correct the position of the automatically obtained correct character, the correct character position is correct even if the automatically obtained correct character position is incorrect. And accurate search information can be obtained.
[0051]
Embodiment 5. FIG.
Embodiment 5 of the present invention will be described below with reference to the drawings. FIG. 11 is a diagram for explaining the correction area designation process (refer to the correction area designation process in step S2 in FIG. 2) in the present embodiment. In the figure, 32 is the character code data displayed on the monitor of the search information correction terminal 7, 33 is the character code selected by the operator, 34 is the document image, and 35 is the correction area.
[0052]
Next, the operation will be described.
The search information correction method follows the same operation procedure as in the flowchart of FIG. Here, the processes other than the correction area designation process in step S2 of FIG. 2 are the same as those in the first embodiment, and thus description thereof is omitted.
First, in the correction area designation process in step S2, the operator selects a plurality of characters to be corrected from the character code data displayed on the monitor. At this time, the position of each selected character on the document image is read from the character position data of the search information, and a rectangular area on the document image including all the selected characters is obtained.
[0053]
For example, as shown in FIG. 11, when seven characters “m, a, c, h, i, n, e” indicated by 33 are selected from the character code data 32 displayed on the monitor, the document image 34 is displayed. A rectangular area 35 containing the character selected above is obtained. Thereafter, the same processing as the search information correction processing described in the first embodiment is performed using this rectangular region as a correction region.
[0054]
As described above, according to the present embodiment, the operator selects a plurality of characters to be corrected from the character code data displayed on the monitor, and sets the position of each selected character on the document image as the character of the search information. A rectangular area on the document image that is read from the position data and contains the data is used as a correction area. As a result, the correction area can be specified from the text editor interface displaying the character code data, and the search information can be corrected based on the correction of the characters in the correction area.
[0055]
Embodiment 6 FIG.
Embodiment 6 of the present invention will be described below with reference to the drawings. FIG. 12 is a configuration diagram of the document filing apparatus according to the sixth embodiment. In the figure, reference numeral 36 denotes a search terminal with a correction function. The search terminal with correction function 36 has a function of correcting search information based on the search result in addition to the search function of the search terminal 6 described in FIG.
[0056]
FIG. 13 is a diagram illustrating an example of a search result. In the figure, 37 is a document image, 38 is a correctly searched area, and 39 and 40 are erroneously searched areas.
[0057]
Next, the operation will be described.
When the user inputs a keyword from the search terminal 36 with a correction function, the search engine 5 refers to the search information database 4 and searches for a document that includes the keyword in the search information. The found document image is displayed on the monitor of the search terminal 36 with the correction function by enclosing the keyword and the area on the hit document image with a frame line. However, if there is an error in the search information, an area that does not include a keyword may be accidentally hit.
[0058]
For example, FIG. 13 shows the result of a search using the keyword “communication means”. The hit area on the document image 37 is displayed with frame lines 38, 39, and 40. Here, the frame line 38 correctly surrounds the area including the keyword, but the frame line 39 erroneously surrounds the character string “communication procedure” different from the keyword. Further, the frame line 40 erroneously surrounds a part of the figure.
[0059]
Similar to the operation of the search information correction terminal 7, the search terminal with correction function 36 corrects the search information stored in the search information database 4 according to the flowchart of FIG. However, in the document selection process in step S1, a document hit by the search displayed on the monitor of the search terminal with correction function 36 is set as a correction target. Further, in the correction area designating process of step S2, the user selects an incorrect frame line in the search terminal with correction function 36, and the area surrounded by the frame line is set as the correction area.
[0060]
The subsequent processing is the same as in the first embodiment. For the region 40 in which a part of the figure is hit by mistake, the processing content “delete search information” is selected in step S5, and the search included in the region 40 is performed. Delete information. For the region 39 that has been hit with a character string that is different from the keyword, the processing content “correction of search information” is selected in step S5, and “communication means” is input as a correct character string in step S5. Appropriate search information is added to.
[0061]
With the above operation, the search information included in the erroneously hit area is correctly corrected or deleted. Further, when the search is performed using the keyword “communication means” after correction, the erroneous frame lines 39 and 40 shown in FIG. 13 are not displayed.
[0062]
As described above, according to the present embodiment, an erroneous frame line is designated as a correction area from among the frame lines indicating the keyword and the area on the document image that has been hit when searching, and the correction area is corrected or When deleted, the character code data and character position data of the search information in this correction area are corrected or deleted, so that the search information included in the erroneously hit area is corrected correctly, and the search information Become accurate.
[0063]
Embodiment 7 FIG.
In the sixth embodiment, an area on a document image that has been accidentally hit with a keyword is set as a correction area, and search information included in the area is corrected. However, on the contrary, the region on the document image including the keyword may be lost without being searched. In the present embodiment, a case will be described in which an area on a document image that has been retrieved is set as a correction area to correct search information.
Embodiment 7 of the present invention will be described below with reference to the drawings. FIG. 14 is a diagram illustrating an example of a search result. In the figure, 41 is a document image, 42 is a correctly searched area, and 43 is a searched area.
[0064]
Next, the operation will be described.
The user corrects the contents of the search information database 4 from the search terminal with correction function 36 based on the search result. This correction operation follows an operation procedure similar to the flowchart of FIG.
[0065]
First, in the document selection process in step S1, a document hit by the search displayed on the monitor of the search terminal with correction function 36 is set as a correction target.
[0066]
Next, in the correction area designating process in step S2, the user refers to the search result displayed on the monitor of the search terminal with correction function 36, and designates a rectangle surrounding the retrieved area with the mouse. The inside of this rectangle is set as a correction area. For example, FIG. 14 shows the result of a search using the keyword “communication means”. There is a correctly searched area 42 on the document image 41, but there is also a character string 43 that has been searched. A rectangle surrounding the character string 43 is set as a correction area.
[0067]
Next, in the correction location extraction process in step S3, the character included in the correction area is picked up from the search information as in the first embodiment, and this is used as the correction target character.
[0068]
Next, in the correction content selection processing in step S4, “correct search information” of the correction target character is selected as the subsequent processing. In the correct character code input process in step S6, the keyword used for the search is set as a correct character string.
[0069]
The subsequent processing is the same as in the first embodiment. The character position of each correct character is obtained in the character position calculation processing in step S7, and the correct character string and the character position of each correct character are used in the search information correction processing in step S8. To correct the search information.
[0070]
With the above operation, the search information included in the area that has been searched is correctly corrected. When the search is performed using the same keyword “communication means” as the search result shown in FIG. 14 after the correction, the character string 43 is also searched in addition to the area 42 that has been correctly searched before the correction.
[0071]
As described above, according to the present embodiment, when an area surrounding a character string that has been retrieved is designated as a correction area, and a keyword is input to the correction area as a correct character string, the character code of the search information in the correction area Since the data and the character position data are replaced with the character code data and the character position data of the correct character string, the search information can be corrected based on the search result, and the search information becomes accurate as the search is continued. Moreover, since the keyword used for the search is automatically set as the correct character string, the operation performed by the user only needs to enclose the character string that has been searched in the search terminal with a correction function.
[0072]
Embodiment 8 FIG.
In the first to seventh embodiments, the correction method in the case where the search information includes character code data and character position data has been described. However, in order to enable the search even if the character recognition result is incorrect, the search information as additional information includes a plurality of candidate characters obtained when the character pattern is recognized, the characteristics indicating the rough shape of the character pattern, etc. May be added. When searching, even if the character code data does not match the keyword, if the similarity between the keyword and the additional information is high, it is determined that a hit has occurred. As a result, it is possible to search for documents in which the character code data is incorrect. However, since the keyword is checked against additional information with high ambiguity, there is a problem that an unnecessary document is erroneously searched. Therefore, the search information including the additional information may be corrected from the search information correction terminal 7.
[0073]
Embodiment 8 of the present invention will be described below with reference to the drawings. FIG. 15 is a diagram showing search information having additional information. FIG. 16 is a flowchart showing an operation procedure of a method for correcting search information having additional information.
[0074]
Here, the search information including the additional information includes character code data, character position data, and additional information. FIG. 15 shows search information when there are both three candidate characters for each character and features indicating the rough shape of the character as additional information. The feature indicating the rough shape of the character is a feature used to identify the character in a general character recognition device. In this example, the black pixel density (a value indicating the percentage of black pixels included in the character pattern as a percentage) is used as a feature.
[0075]
Next, the operation will be described.
First, as in the first embodiment, in the document selection process in step S11, the operator selects a document to be corrected from the search information correction terminal 7, and in the correction area designation process in step S12, the operator corrects an arbitrary area on the document image. Designated as an area, the character position data included in the search information is checked in the correction location extraction process in step S13, and the character whose character position is included in the correction area is picked up, and this is used as the correction target character.
[0076]
Next, in the correction content selection process of step S14, the process of deleting the correction target character from the search information “deletion of search information”, the process of correcting the correction target character “correction of search information”, or the correction target The operator selects a process “deletion of additional information” for deleting additional information of characters from the search information.
[0077]
Next, when “delete search information” is selected in step S15, the character code data and character position data of the correction target character from the search information stored in the search information database 4 in the search information correction processing in step S18. Remove. Thereafter, in the additional information deletion process in step S19, the additional information of the correction target character is deleted from the search information.
[0078]
If “correction of search information” is selected in step S15, the correct character code input process in step S16 and the character position calculation process in step S17 are performed as in the first embodiment, and the correct character and each correct character are The position is obtained, and the character code data and the character position data of the search information are corrected by the search information correction process in step S18.
When the character code data is corrected and becomes the correct character code data, additional information for searching the character of this character code data is no longer necessary, and finally, in the additional information deletion process in step S19, the correction target character is deleted. Delete additional information from search information.
[0079]
Next, when “deletion of additional information” is selected in the correction content selection process in step S15, the additional information deletion process 161 deletes the additional information held by the correction target character from the search information.
The above is the procedure for correcting search information having additional information.
[0080]
As described above, according to the present embodiment, the character code data and the character position data of the search information of the correction target character stored in the search information database are corrected correctly, or the character code data and the character position data are deleted. In this case, the additional information of the correction target character was deleted. Thereby, it is possible to obtain accurate search information in which extra documents are not erroneously searched during a search.
Moreover, since the additional information is deleted after correcting or deleting the character code data and the character position data of the search information correctly, the highly ambiguous additional information is not used in the search for a new keyword, Accurate search information can be obtained, and the problem that an excessive document is erroneously searched during a search can be suppressed.
[0081]
Embodiment 9 FIG.
In the present embodiment, a case will be described in which the user corrects search information having additional information from the search terminal with correction function 36 based on the search result.
Embodiment 9 of the present invention will be described below with reference to the drawings.
[0082]
Next, the operation will be described.
The search information correction method follows the same operation procedure as the flowchart of FIG.
First, in the document selection process in step S11, the searched document displayed on the monitor of the search terminal with correction function 36 is selected as a correction target.
Next, in the correction area designating process in step S12, the user selects a correctly searched area from the areas hit by the search, and sets this as a correction area. For example, FIG. 14 shows the result of a search using the keyword “communication means”. An area 42 correctly searched on the document image 41 is selected and set as a correction area.
[0083]
Next, in the correction location extraction process in step S13, the characters included in the correction area are picked up from the search information as in the first embodiment, and are used as correction target characters.
[0084]
Next, in step S14, the correction content selection process, and in step S15, the process of deleting the correction target character from the search information, that is, "deletion of search information", or the process of correcting the correction target character, that is, The operator selects “correction” or processing for deleting additional information of the correction target character from the search information, that is, “deletion of additional information”.
[0085]
Next, in the correct character code input process of step S16, the keyword used for the search is used as it is for the correct character string. Thereafter, the character position calculation process in step S17 determines the position of each correct character as in the first embodiment.
[0086]
Next, in the search information correction process in step S18, as in the first embodiment, the correction is performed using the correct character obtained in the correct character code input process in step S16 and the character position obtained in the character position calculation process in step S17. Correct the character code data and character position data of the target character.
[0087]
Finally, in the additional information deletion process in step S19, the additional information held by the correction target character is deleted from the search information.
[0088]
As described above, according to the present embodiment, when the user specifies a correctly searched area when searching, the character code data and the character position data in the area are automatically corrected, and an extra addition is made. Information is deleted. Thereby, the search information having the additional information can be corrected based on the search result, and the search information becomes accurate as the search is continued.
[0089]
In this embodiment, a method has been described in which a correct hit area is designated as a correction area and search information including additional information is corrected from the search terminal with correction function 36. In addition to this, an erroneous hit area is corrected. The search information including the additional information included in the correction area that has been erroneously hit may be corrected according to the method for correcting the search information including the additional information described in the eighth embodiment.
[0090]
In the present invention, the processing content has been described by taking a horizontally written document as an example, but may be applied to a vertically written document. Further, when a character string extending over two lines is corrected, the correction area may be divided and designated for each line.
[0091]
【The invention's effect】
Since the present invention is configured as described above, the following effects can be obtained.
[0092]
In the first invention, a correction area that requires correction or deletion of the document image is selected based on the document image stored in the document image database and the search information stored in the search information database, and a correction character is stored in the correction area. When input, the character code data and character position data of the correction target character stored in the search information database are replaced with the character code data and character position data of the correction character, and when the correction target character in the correction area is deleted, By deleting the character code data and character position data of the correction target character stored in the search information database, the search information of the correction target character that needs correction can be corrected, and unnecessary search information can be deleted. it can.
[0093]
In the second invention, when a correction character is input to the correction area, a rectangular area obtained by equally dividing the correction area by the number of correction characters is obtained as the position of the circumscribed rectangle of the correction character, and the circumscribed rectangle of the obtained correction character is obtained. Based on the position and the corrected character, the character position of the search information is replaced by replacing the character code data and the character position data of the correction target character stored in the search information database with the character code data and the character position data of the corrected character. Data can be corrected accurately.
[0094]
In the third invention, when a correction character is input to the correction area, a rectangular area obtained by dividing the correction area by the average character width for each character is obtained as the position of the circumscribed rectangle of the correction character, and the circumscribed rectangle of the obtained correction character By replacing the character code data and character position data of the correction target character stored in the search information database with the character code data and character position data of the correction character based on the position of the Character position data can be corrected accurately.
[0095]
In the fourth invention, when the number of correction characters input in the correction area is equal to the number of correction target characters in the correction area, the circumscribed rectangle position of the correction target character is set as the circumscribed rectangle position of the correction character, and the above search is performed. By replacing the character code data of the correction target character stored in the information database with the character code data of the correction character, the character position data of the search information can be corrected accurately.
[0096]
In the fifth invention, when correction is required for the circumscribed rectangle position of the corrected character obtained previously, the position of the circumscribed rectangle of the corrected character is newly corrected by the operator, and the corrected character character stored in the search information database By rewriting the position data to the newly corrected character position data, the position of the corrected character obtained earlier can be further corrected, so the character position data in the search information is more accurate.
Can be corrected.
[0097]
In the sixth invention, when a keyword for searching for a predetermined character in a document image is input, the keyword area is selected as a correction area from the document image, and the correction target character in the selected correction area is corrected. Or, by rewriting or deleting the character code data and character position data of the correction target character that has been deleted and saved in the search information database, the search information of the correction target character that has been corrected or deleted by the character search by the keyword is corrected or deleted can do.
[0098]
In the seventh invention, additional information for selecting a correction target character region in the document image is added to the search information, and the correction target character in the correction region selected based on this additional information is corrected or deleted. By rewriting or deleting the character code data and character position data of the correction target character stored in the search information database, and deleting the additional information, the search information of the correction target character that needs correction can be corrected. In addition, unnecessary search information can be deleted.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of a document filing apparatus according to a first embodiment.
FIG. 2 is a flowchart showing an operation procedure of a search information correction method according to the first embodiment.
3 is a diagram for explaining a method for correcting a character segmentation error included in search information according to Embodiment 1. FIG.
4 is a diagram for explaining a method for correcting character position data in search information according to Embodiment 1. FIG.
5 is a diagram showing the contents of search information before and after correction in Embodiment 1. FIG.
6 is a diagram for explaining a method for correcting a character string extraction error included in search information according to Embodiment 1. FIG.
FIG. 7 is a diagram for explaining a method of correcting a misreading of characters included in search information in the first embodiment.
FIG. 8 is a diagram illustrating a method for obtaining the position of a correct character in the second embodiment.
FIG. 9 is a diagram for explaining a method for obtaining the position of a correct character in the third embodiment.
FIG. 10 is a diagram for explaining a character position correction method according to the fourth embodiment;
FIG. 11 is a diagram for explaining correction area designation processing according to the fifth embodiment.
FIG. 12 is a configuration diagram of a document filing apparatus according to a sixth embodiment.
FIG. 13 is a diagram showing an example of a search result in the sixth embodiment.
14 is a diagram showing an example of search results in Embodiment 7. FIG.
FIG. 15 is a diagram showing search information having additional information according to the eighth embodiment.
FIG. 16 is a flowchart showing an operation procedure of a method for correcting search information having additional information according to the eighth embodiment.
FIG. 17 is a configuration diagram of a conventional document filing apparatus.
FIG. 18 is a view for explaining the contents of search information of a conventional document filing apparatus.
FIG. 19 is a diagram showing a display state of a search result of a conventional document filing apparatus.
[Explanation of symbols]
1 scanner, 2 character recognition device, 3 document image database, 4 search information database, 5 search engine, 6 search terminal, 7 search information correction terminal.

Claims (7)

A document image database in which document images read by the scanner are stored;
Search information for selecting a correction target character area in the document image, comprising character code data of characters in the document image read by the scanner and character position data indicating the position of a circumscribed rectangle of the characters. A character recognition means to create;
A search information database for storing the search information created by the character recognition means;
The document image stored in the document image database and the search information stored in the search information database are displayed in a display area, and the document image is corrected or deleted based on the displayed document image and the search information. When a correction area that requires correction is selected and a correction character is input to the correction area, the position of the circumscribed rectangle of the input correction character is obtained, and based on the position of the correction character and the circumscribed rectangle of the correction character The character code data of the correction target character in the correction area stored in the search information database is replaced with the character code data of the correction character, and the character position data of the correction target character is replaced with the character position data of the correction character. When the correction target character in the correction area is deleted, the correction target character stored in the search information database is overwritten. And search information modification means for deleting the character code data and the character position data, document filing apparatus comprising the. When the correction character is input to the correction area, the search information correction means obtains a rectangular area obtained by equally dividing the correction area by the number of the correction characters as the position of the circumscribed rectangle of the correction character, and the obtained correction Based on the position of the circumscribed rectangle of the character and the corrected character, the character code data and the character position data of the correction target character stored in the search information database are converted into the character code data and the character position data of the corrected character. 2. The document filing apparatus according to claim 1, wherein the document filing apparatus is replaced. A table that stores the average character width for each character
When a corrected character is input to the correction area, the search information correction means obtains a rectangular area obtained by dividing the correction area by an average character width for each character stored in the table as a position of a circumscribed rectangle of the correction character. Based on the position of the circumscribed rectangle of the corrected character obtained and the corrected character, the character code data and the character position data of the correction target character stored in the search information database are 2. The document filing device according to claim 1, wherein the document filing device is replaced with character position data. When the number of correction characters input in the correction area is equal to the number of correction target characters in the correction area, the search information correction means determines the position of the circumscribed rectangle of the correction target character in the circumscribed rectangle of the correction character. 2. The document filing apparatus according to claim 1, wherein the character filing data of the correction target character stored in the search information database is replaced with character code data of the correction character. When the search information correction means needs to correct the circumscribed rectangle position of the corrected character, the operator instructing rewriting of the search information newly corrects the position of the circumscribed rectangle of the corrected character, and the search information 5. The document filing apparatus according to claim 2, wherein character position data of the corrected character stored in the database is rewritten with character position data of the corrected character newly corrected. When a keyword for searching for a predetermined character is input from the document image, the character code data and character position data of the characters constituting the keyword are searched from the search information database, and the searched character code data Search means for obtaining the keyword area in the document image based on the character position data and sending the keyword area together with the document image stored in the document image database;
The search information correction means displays the document image in which the keyword area is surrounded by a frame line in the second display area based on the keyword area sent by the search means, and the frame is surrounded by the frame line. 2. The document filing apparatus according to claim 1, wherein the selected area is selected as a correction area that requires correction or deletion. The character recognition means includes character code data of a character in the document image read by the scanner, character position data indicating a position of a circumscribed rectangle of the character, a candidate character for each character, and a rough shape of the character. Creating search information composed of additional information for selecting a correction target character area in the document image, and
The search information correcting means selects a correction target character based on additional information stored in the search information database, replaces or deletes the character code data and the character position data of the search information, and then adds the additional information. 2. The document filing apparatus according to claim 1, wherein the information is deleted.

Images