JP2003050968A - Device and method for document filing and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately retrieve a document image affected adversely by external disturbance, which has been impossible before. SOLUTION: When image information in a process area extracted from an inputted document image is recognized and converted into a character code and conversion results are classified by specific document items, respective process results in the processing stage of the input document images are compared with previously set specific decision conditions, and it is judged whether there is a process error according to the comparison results. When it is decided that there is a process error, correction indication information for the corresponding process result is presented. The process result is corrected according to correction information inputted in response to the presentation, and document filing data regarding the input image area generated by using the process result and/or correction result and stored.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document filing apparatus for classifying an input document image into preset document items for filing, and particularly, when filing a document image, an error that interferes with document retrieval. The present invention relates to a document filing apparatus and a document filing method for determining whether or not the occurrence has occurred and appropriately correcting the recognition result when the above error occurs, and a document filing program for causing a computer to execute the same.

[0002]

2. Description of the Related Art A document filing device is a device for automatically recognizing characters in a document, classifying document items, and then accumulating (filing) the document.
There is one disclosed in Japanese Patent No. 40901.

FIG. 23 is a block diagram schematically showing the configuration of a conventional document filing apparatus disclosed in Japanese Patent Laid-Open No. 10-240901. In the figure, reference numeral 100 denotes an image file that holds a document image read by a scanner or the like, and outputs the document image to the layout analysis unit 101. Reference numeral 101 denotes a layout analysis means for performing a document layout analysis on the document image from the image file 100, and a document layout knowledge file 107.
With reference to, information about the document layout such as character strings and coordinate information of characters is extracted from the document image. Reference numeral 102 is a character recognition means for inputting information about the document layout from the layout analysis means 101, and converts the character image cut out for each character into a character code based on the information about the document layout. Reference numeral 103 denotes a document item classification unit that compares the information about the document layout from the layout analysis unit 101 with the document layout knowledge file 107, and the character recognition result obtained by the character recognition unit 102 and the document content knowledge file 108. And the character recognition results for each document item are automatically classified. A storage unit 104 stores a character recognition result for each document item obtained from the document item classification unit 103 in a character recognition result file 109.

Reference numeral 105 denotes a search means for searching the character recognition result file 109 in response to a search request from the outside.
A document satisfying the search request is identified from the character recognition result file 109. An output unit 106 outputs the document recognition result and the corresponding document image to the outside, and the retrieval unit 10
The character recognition result identified by No. 5 is extracted from the character recognition result file 109 and output, or the document image corresponding to the character recognition result extracted from the character recognition result file 109 is extracted from the image file 100 and output. To do. Reference numeral 107 denotes a document layout knowledge file that stores knowledge (document layout knowledge) about the layout of a document to be read, which is created in advance, and includes a storage device that can be appropriately referred to by the layout analysis unit 101 and the document item classification unit 103. Reference numeral 108 denotes a document content knowledge file that stores knowledge (document content knowledge) about the description content of a document, and is composed of a storage device that can be appropriately referred to by the document item classification unit 103. A character recognition result file 109 stores the character recognition result for each document item obtained from the document item classification means 103, and outputs the document searched and identified by the search means 105 to the image file 100 or the output means 106.

Next, the operation will be described. FIG. 24 shows FIG.
3 is a flow chart showing the document file creation operation by the document filing apparatus in FIG. 3, and FIG. 25 is a flow chart showing the document search operation by the document filing apparatus in FIG. 23. The document file creation will be described with reference to these flow charts. First, an image input device such as a scanner (not shown) is used to read a document as document image information into a document filing device (step ST100) and store it in an internal image file 100 (step ST101). Figure 2
6 is a diagram showing an example of the document image stored in the image file in FIG. FIG. 26 shows an example in which the minutes of a certain regular progress meeting are read as a document image, which will be described below using the document image.

As a document file creation process, first, a document image is read from the image file 100 and the layout analysis means 101 analyzes the document layout (step ST102). More specifically, the layout analysis unit 101 uses the document layout knowledge file 1
By using the contents described in 07, the character string and the character in the document image are cut out. FIG. 27 is a diagram showing the contents of the document layout knowledge file relating to the document image of FIG.
In the figure, 110 is the "meeting name" item in the document image of FIG. 26, and 111 is the "venue" in the document image of FIG.
Items 112 and 112 represent "held date and time" items in the document image of FIG. 26, 113 represents layout information related to "attendees" item in the document image of FIG. 26, and 114 represents "held date and time" item 112. The coordinate information is shown. The layout analysis means 101 uses the ratio of width and height in the document layout knowledge shown in FIG. 27, for example, “meeting name” item 110, “holding place” item 111, “holding date” item 112, and “attendee” item. It is determined that both 113 are written horizontally, and the character string is cut out on the assumption that the document is written horizontally, and then the characters are cut out.

The layout analysis means 101 forms a rectangular area in which a plurality of continuous character strings are combined from the character string cutout result, and extracts the coordinates and size information for each area to create layout information. FIG. 28
FIG. 6 is a diagram showing layout information created by a layout analysis unit. In the figure, 115 is a layout analysis unit 1.
01 is a start block indicating the order of the character strings at the beginning of the rectangular area, and 116 is an end block indicating the order of the character strings at the end of the rectangular area. Reference numeral 117 is area coordinate information indicating coordinate / size information of a rectangular area from the start block 115 to the end block 116, and 118 is coordinate information for an area where the start block 115 is “5” and the end block 116 is “6”. is there. As shown in FIG. 28, the layout information includes a start block 115, an end block 116, and coordinate / size information associated with the rectangular area in which a plurality of consecutive character strings are combined.

The character recognition means 102 cuts out the character string obtained by the layout analysis means 101, and obtains a character recognition result by converting a character image into a character code based on the character cutout result (step ST103). FIG. 29 is a diagram showing the result of character recognition by the character recognition means. In the figure, reference numeral 119 is the character recognition result for the area in which the start block is "5" and the end block is "6" in FIG.

Next, the document item classification means 103 classifies the document items using both the document layout and the document contents acquired from the character recognition means 102 (step S).
T104). First, the classification using the document layout information is realized by collating the document layout information shown in FIG. 28 with the document layout knowledge shown in FIG. This matching is performed using the coordinate information of the rectangular area from the start block 115 to the end block 116 shown in FIG. 28 and the coordinate information of each item of the document layout knowledge shown in FIG. The matching degree Pij is calculated for each item. Pij = 100 × (area of overlapping area of item i and block j) × 2 / (area of item i + area of block j) (1)

FIG. 30 is a diagram showing a result of collation of document layout information and document layout knowledge by the document item classification means. In the figure, 120 is the degree of matching between the coordinate information 118 and the holding date and time for the region in which the start block 115 is “5” and the end block 116 is “6” in FIG. 28, which is “46”. That is, the area of the region formed by the coordinate information 118 is 9 for the region where the start block 115 is “5” and the end block 116 is “6” in FIG.
0 (width) × 10 (height) = 900, the area of the region formed by the coordinate information 118 with respect to the date and time of holding in FIG. 27 is 90 (width) × 3 (height) = 270, and the region where both regions overlap Since the area of 270 is 270, the matching degree is 100 × 270
X2 / (900 + 270) = 46. By the method described above, the document item classification unit 103 calculates the matching degree based on the document layout information for all areas and document items.

On the other hand, the classification using the document content knowledge is realized by collating the character recognition result from the character recognition means 102 with the document content knowledge from the document content knowledge file 108. That is, when a character group composed of consecutive characters in the character recognition result is collated with the document content knowledge and the character string registered in the document content knowledge as a keyword is matched, the collation by the following formula (2) is performed. The degree Pc is added. On the other hand, in the case of matching with a keyword that should not exist, the following formula (3)
The negative matching degree Pe due to is added. Finally, the total matching degree Pa is calculated from the matching degrees Pc and Pe according to the following equation (4). Pc = 20 × (number of matching keyword characters) (2) Pe = −50 × (number of matching keywords) (3) Pa = Pc + Pe (4)

A specific example will be described. FIG. 31 is a diagram showing an example of the contents of the document content knowledge file, FIG. 32 is a diagram showing the result of collation between the character recognition result by the document item classification means and the document content knowledge, and FIG. 33 is integrated by the document item classification means. FIG. 34 is a diagram showing a collation result between the character recognition result and the document content knowledge, and FIG. 34 is a diagram showing a correspondence relationship between the document items classified by the document item classification means and the character strings. In the figure, 121 is the keyword information corresponding to the item "holding place", 122 is the keyword "holding place", 123 is the keyword "place", 124 is the keyword "meeting room", 125 is the keyword "room", 126
Is the keyword “meeting”, 127 is the keyword “meeting name”, 128 is the keyword “date and time of event”, 12
9 is the degree of matching between the area where the starting block is “2” and the ending block is “3” and the item of the venue, and 130 is the area where the starting block is “1” and the ending block is “1” and the meeting name. The degree of matching with the item, the start block of 131 is “2”
The degree of matching between the area whose end block is "3" and the item of the venue is 132, and the degree of matching 132 between the area whose start block is "4" and the end block is "5" and the item of the holding date and time, 1
33 is the degree of matching between the area in which the start block is “6” and the end block is “7” and the items of the attendees, 134 is the conference name, the place of the event, the date and time of the event, the correspondence relationship between the attendees and the block, and 135 is The degree of matching 136 in the correspondence relation 134 indicates a combination of blocks that gives the maximum degree of matching.

Here, a case will be described in which the degree of matching 129 between the area where the starting block is "2" and the ending block is "3" shown in FIG. First, the document item classification means 103 has the keyword information 121 corresponding to the character recognition result 119 shown in FIG. 29 corresponding to the area where the start block is “2” and the end block is “3” and the “holding place” shown in FIG. 31. And match. Specifically, the document item classification means 103 displays the keywords "holding place" 122 and "place" 123 located at the beginning of the holding place item and the FIG.
The character recognition result 119 shown in FIG. At this time,
The keyword located at the beginning of the venue item is matched only with the character string starting from the beginning of the area. In this case, since the first character string of the character recognition result 119 does not match any of the keywords “holding place” 122 and “place” 123, the matching degree is not added.

Next, the document item classification means 103 collates the keywords "meeting room" 124 and "room" 125 located at the end of the holding place item with the character recognition result 119. At this time, the keyword located at the end of the holding place item is matched only with the character string ending at the end of the area. In this case, since the character string at the end of the character recognition result 119 matches the keyword “meeting room” 124, 20 × 3 (the number of characters) = 60 is added to the matching degree according to the above equation (2).

The keyword "meeting" whose position is arbitrary
The collation between 126 and the character recognition result 119 is performed on all the character recognition results. In this case, character recognition result 1
Since the keyword "meeting" 126 never appears in 19, the matching degree is not added.

The matching with the prohibited keywords "meeting name" 127 and "holding date and time" 128 is carried out for all character recognition results. At this time, the document item classification means 1
When the collation is successful, 03 deducts the collation degree because there is a keyword that should not be entered. In this case, since the leading character string of the character recognition result 119 matches the keyword “meeting name” 127, the matching degree is decremented by 50 according to the above equation (3).

As described above, the document item classification means 10
The addition amount and the subtraction amount calculated by 3 are added, and as shown in FIG. 31, as the final matching degree 129 of the area where the starting block is “2” and the ending block is “3” and the venue, “10 Get

In the same manner as described above, the document item classification means 1
03 calculates the matching degree based on the document content information for all the areas and the document items.

Next, the document item classification means 103 obtains the degree of matching with respect to the document layout information and the document content knowledge, and then adds up the degree of matching of both, to obtain a continuous character string area and document items shown in FIG. Compute the overall matching degree of.

After calculating the total degree of collation between the continuous character string area and the document item, the document item classification means 103 calculates the coordinate value of the item of the document layout knowledge shown in FIG.
It is required that the documents to be read are arranged in the order of "meeting name", "venue", "date and time", and "attendees" from the top. Based on this, document item classification means 1
03 extracts a combination of continuous character string regions that may correspond to each item of the document. In the case of the document shown in FIG. 26, 4 areas are extracted from the 7-row character string, and at least 1 character string is assigned to each area, and the "meeting name", "venue", and "holding" Extract the combinations arranged in the order of "date and time" and "attendees". As a result, the combination shown in FIG. 34 is obtained.

Next, the degree of collation of the entire document when the respective areas and the document items are combined is obtained. In the case of the meeting name, the place of the meeting, the date and time of the meeting, and the correspondence relationship 134 between the attendees and the block shown in FIG. 34, the matching degree for each item is obtained from FIG. That is, the area corresponding to the conference name has the start block “1” and the end block “1”, and the matching degree 130 is “−50” from FIG. 33. In addition, an area in which the starting block is “2” and the ending block is “3” corresponds to the holding place, and the matching degree 131 is “10” from FIG. 33. Furthermore, an area in which the start block is “4” and the end block is “5” corresponds to the holding date and time, and the matching degree 132 is “136” from FIG. 33. Finally, the attendee corresponds to the area in which the starting block is “6” and the ending block is “7”, and the matching degree 133 is “3” from FIG. 33.
8 ". By adding these matching degrees, FIG.
“134” is obtained as the collation degree 136 for the meeting name, the venue, the date and time of the conference, and the correspondence relationship 134 between the attendees and the block shown in FIG.

According to the method described above, the document item classification means 1
03 calculates the matching degree for all combinations of continuous character string areas that have the possibility of corresponding to each item of the document, and sets the combination having the largest matching degree as the item classification result. In the example shown in FIG. 34, the character string on the second line for the conference name, the character string on the third line for the venue,
When the character string on the fourth line is associated with the date and time of the event and the character string on the fifth line is associated with the attendees, the maximum matching degree is shown. As a result, the correct item classification result is obtained.

The accumulating means 104 accumulates the character recognition results classified for each document item obtained as described above in the character recognition result file 109 (step ST105).

Next, the document search process will be described. A search request is input to the search means 105 using an input means (not shown) (step ST100a). When the search means 105 receives the search request, the character recognition result file 1
09 is searched to identify a document that matches the search request (step ST101a). Output means 106
Extracts the document image corresponding to the character recognition result identified by the search means 105 from the character recognition result file 109 or the document image corresponding to the character recognition result extracted from the character recognition result file 109 to the image file 100. (Step ST102a).
The output unit 106 outputs the character recognition result and / or the document image extracted as described above to the outside (step S).
T103a).

[0025]

Since the conventional document filing apparatus is configured as described above, there is a problem that it is not possible to accurately retrieve a document image with disturbance.

To explain the above problem in detail, in the conventional document filing apparatus, a desired character string is not cut out from a document image, an error occurs in the character recognition itself, or a document item is classified. Even if an error occurs, the character recognition result after document item classification is forcibly output. In other words, even if a form image with disturbance such as shading or remarkable image rotation is input, a series of processing such as character string cutout, character recognition, and document item classification is performed as in the normal form image. Is stored as a character recognition result file. Therefore, an accurate search cannot be performed for the character recognition result in which an error due to disturbance has occurred.

Further, since it is impossible to distinguish between a normal form and a document to which disturbance has been added, in order to confirm that the document image has been converted into the correct character code, the document image and the document data for all the document data can be confirmed. There is a problem that it is necessary to compare the reading result, and it takes a lot of time and labor to input data.

The present invention has been made to solve the above problems. When filing a document image for each preset document item, it is determined whether or not an error that hinders document retrieval has occurred. A document filing apparatus capable of accurately retrieving a disturbed document image without storing a character recognition result containing an error by appropriately correcting the recognition result when the above error occurs, and A document filing method and a document filing program that causes a computer to execute the method are provided.

[0029]

A document filing apparatus according to the present invention recognizes image information in a processing area extracted from an input document image, converts it into a character code, and converts the conversion result into a predetermined document item. A document filing unit that classifies each of the input document images and a predetermined filing condition that is set in advance when the document filing unit processes the input document image are compared with each processing result in the process of processing the input document image, and based on the comparison result. Error determining means for determining whether or not there is an error in the processing of the document filing means,
When the error determining unit determines that the process has an error, the correcting unit presents the correction instruction information for the corresponding process result, and corrects the process result in accordance with the correction information input in accordance with the correction instruction information. Using the processing result of the document filing means and / or the correction result of the correction means,
Document data storage means for generating and storing document filing data relating to an input image.

The document filing apparatus according to the present invention is
The error determining means uses the arrangement direction of the character strings in the processing area extracted from the input document image by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The error determining means uses the number of lines of the character string in the processing area extracted from the input document image by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The error determination means uses the position of the character string in the processing area extracted from the input document image by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The error determining means uses the character string size in the processing area extracted from the input document image by the document filing means and / or the character size forming the character string as the determination condition.

The document filing apparatus according to the present invention is
The error determination means uses the likelihood of character recognition in the processing area extracted from the input document image by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The error determination means uses the number of characters in the processing area extracted from the input document image by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The error determining means uses the character code conversion result of the image information in the processing area extracted from the input document image by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The error determination means uses the satisfaction degree of the document items classified by the document filing means as the determination condition.

The document filing apparatus according to the present invention is
The correction means presents only the portion having a difference from the determination condition of the processing result as the correction instruction information, and corrects the processing result according to the correction information input to the portion having the difference.

The document filing apparatus according to the present invention is
The correction means presents correction instruction information that can specify a partial area in the document image as the correction information.

The document filing apparatus according to the present invention is
The correction means presents correction instruction information capable of externally correcting a portion that differs from the determination condition of the processing result.

The document filing apparatus according to the present invention is
The correction means presents correction instruction information capable of externally determining the determination result that the processing result is incorrect.

The document filing apparatus according to the present invention is
The document data storage means stores, as document filing data, data obtained by combining the processing result of the document filing means and the correction result of the correction means.

The document filing method according to the present invention is
A document filing step of recognizing the image information in the processing area extracted from the input document image and converting it into a character code, and classifying the conversion result for each predetermined document item, and an input document image in the document filing step In processing the above, a predetermined judgment condition set in advance is compared with each processing result in the processing process for the input document image, and an error is judged based on the comparison result whether or not there is an error in the processing in the document filing step. When it is determined that the processing has an error in the determination step and the error determination step, the correction instruction information for the corresponding processing result is presented, and the processing result is corrected according to the correction information input accordingly. Correction step to be performed, the processing result in the document filing step and / or the correction in the correction step With fruit, but with a filing data generation step of generating a document filing data about the input image.

A document filing program according to the present invention recognizes image information in a processing area extracted from an input document image, converts the image information into a character code, and classifies the conversion result into predetermined document items. In processing the input document image in the step and the document filing step, a predetermined determination condition set in advance is compared with each processing result in the process of processing the input document image, and based on the comparison result, the document filing step is performed. An error determination step of determining whether or not there is an error in the process, and when it is determined that the process has an error in the error determination step, the correction instruction information for the process result corresponding to this is presented and The correction step to correct the processing result according to the correction information input according to Tsu using a modified result of processing results and / or modification steps in the flop, is intended to execute the filing data generating step to the computer to produce a document filing data about the input image.

[0045]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below. Embodiment 1. 1 is a diagram showing the configuration of a document filing apparatus according to Embodiment 1 of the present invention. In the figure, 1 is an image file for storing a document image, which is a hard disk device of a computer or C for storing a document image input using an image input device (not shown) such as a scanner.
It is realized by a recording medium such as a D-ROM. Reference numeral 2 denotes a layout analysis means (document filing means) for performing a layout analysis of the document image stored in the image file 1, and extracts an image area for character recognition from the document image according to the document layout knowledge stored in the document layout knowledge file 8. To do. Reference numeral 3 denotes a character recognition means (document filing means) for converting the image information in the processing area extracted by the layout analysis means 2 into a character code. For example, a character image cut out for each character by the layout analysis means 2 is converted into a character code. Convert. Reference numeral 4 denotes a character item classification means (document filing means) for classifying the conversion result of the character recognition means 2 for each preset character item, and character strings / characters outputted from the layout analysis means 2 and the character recognition means 3. The cutout result and the character recognition result are collated with the constraint conditions relating to the document items stored in the document layout knowledge file 8 and the document content knowledge file 9, and the character recognition result is classified into partial areas for each document item.

Reference numeral 6 is a retrieval means for retrieving the document data in the keyword file 13, which retrieves according to the retrieval information (keyword) for each document item inputted from the outside. 7
Is an output means for outputting the document obtained by the search means 6 in the form of a document image and a character code, and comprises a display device of a computer or the like. Reference numeral 8 denotes a document layout knowledge file that describes the rules on the document layout for each document item, which is appropriately read by the layout analysis means 2 and used for layout analysis. Reference numeral 9 is a document content knowledge file holding information of words described for each document item, which is appropriately read by the document item classification means 4 and used for classification of document items. Reference numeral 10 is an automatic keyword attaching determination means (error determination means), which is a character string / character cutout result, character recognition result, and document item classification result output by the layout analysis means 2, the character recognition means 3, and the document item classification means 4. It is determined whether or not there is a problem in performing the search.

Reference numeral 11 is a keyword inputting means (correcting means) for adding a keyword to a document item for which the automatic keyword attaching judging means 10 has determined that the automatic keyword attaching result is impossible, and is executed by the CPU of the computer. In addition to the software relating to the correction processing, the automatic keyword assignment determination means 10 includes a display device of a computer for displaying the processing result determined to be impossible, an input means for inputting the correction information via the display device. 12
Is a keyword storage unit (document data storage unit) that outputs and stores the keyword by combining the automatically created keyword and the keyword manually input through the keyword input unit 11. 13 is a keyword storage means 1
2 is a keyword file (document data storage means) for storing the document data generated in 2.

FIG. 2 is a diagram showing a document image stored in the image file in FIG. 1, FIG. 3 is a diagram showing a character recognition result for each document item obtained from the document image, and FIG. 4 is a document item shown in FIG. It is a figure which shows the content with respect to "meeting content" in detail,
FIG. 5 is a diagram showing the automatic keyword attaching determination conditions set in the automatic keyword attaching determining means in FIG. 6 is a diagram showing the result of automatic keyword assignment determination by the automatic keyword assignment determination means in FIG. 1, FIG. 7 is a diagram showing the initial state of the correction screen for the document item "meeting name", and FIG. 8 is the document item "conference name". FIG. 9 is a diagram showing a state after the correction screen is corrected, FIG. 9 is a diagram showing an initial state of the correction screen for the document item “meeting date and time”, and FIG. 11 is a diagram showing a state in which a keyword is input on the correction screen for the document item "meeting date and time".
2 is a diagram showing the initial state of the correction screen for the document item "contents of proceedings", and FIG. 13 is a diagram showing the contents of the keyword file in FIG.

Next, the operation will be described. First, the layout analysis unit 2 reads the document image 14 shown in FIG. 2 from the image file 1 and analyzes the document layout.
More specifically, the layout analysis unit 2 uses the contents described in the document layout knowledge file 8 to
Cut out character strings and characters in a document image (document filing step).

Next, the character recognizing means 3 detects a character string, which is a processing area extracted from the document image by the layout analyzing means 2.
A character recognition result is obtained by converting a character image into a character code based on the character cutout result. Further, the document item classification unit 4 appropriately refers to the document content information in the document content knowledge file 9 as in the conventional technique, and the character recognition result acquired from the character recognition unit 3 is set for each preset document item. Classify. By performing these operations on the document image 14, a character string cutout, a character cutout result, a character recognition result, and a document item classification result as shown in FIG. 3 are obtained. In FIG. 3, there are five document items, and the results for the document items “meeting name” 15 to “meeting content” 19 are stored. "Number of lines of document item" for each item 15 to 2
0, “character number of document item” 21 is stored, “character string direction” 22 for each line, “character string coordinate (coordinate of character string cutout)” 23, “number of characters (in line)” 24, “each Character recognition result (cutout coordinates, character code, similarity) "
25 is stored. Figure 4 shows the document item "Agenda" in Figure 3.
Is described in detail.

The automatic keyword attaching judging means 10 is shown in FIG.
The results of the character string / character cutout, character recognition, and document item classification shown in are compared with the automatic keyword assignment determination conditions shown in FIG. 5, and automatic keyword assignment is possible for each document item based on the comparison result. Determine if there is. The automatic keyword addition determination condition shown in FIG. 5 is given to each item of, for example, a conference name, a venue, a date and time of attendance, attendees, and agenda items. Allowable character string direction 28 in FIG.
Indicates that the allowable character string direction is stored in each item, and only the horizontal (write) character string is allowed in each item. The allowable number of lines 29 defines the number of lines of a character string that is allowed for each item, 1 line to 4 lines for the meeting name, 1 line or more for the proceedings, and 1 line to 2 lines for other items. Allowed under the following conditions. The allowable character string position 30 defines the positional relationship allowed for each item, and is allowed under the condition that the items such as the meeting name, the place of the meeting, the date and time of the meeting, the attendees, and the agenda item are arranged in order from the top.

The allowable character string size 31 is defined with respect to the size of the character string allowed for each item, and is allowed under the condition that the height of the character string is less than 23 for all items. The allowable character similarity 32 represents the evaluation value obtained as a result of the character recognition allowed for each item by the similarity, and all the items are allowed under the condition that the evaluation value is 85 or more with the maximum value of the similarity being 100. . Here, the similarity is the likelihood of character recognition obtained from the evaluation value obtained as a result of character recognition. The allowable number of characters 33 defines the maximum number of characters allowed in each item and stores the maximum number of characters. The condition is that the meeting name and the venue are 30 characters or less, and the date and time and attendees are 50 characters or less. Tolerate. In addition, the terms of the proceedings are not specified. The allowable character size 34 defines the size of the character image belonging to each item, and is allowed under the condition that the character width or height is less than 23 with all items. The essential character string 35 defines words that are essential for each item, and in the case of FIG. 5, “month”, “day”, and
It is allowed on condition that it contains either "/". No conditions are specified for other document items. The allowable item satisfaction number 36 defines the number of document items in which a character string exists, and is allowed under the condition that a character string exists in five items, that is, all items.

The automatic keyword attaching determination condition shown in FIG. 5 may be selected from the outside through an input means such as the keyword input means 11. Furthermore, the automatic keyword addition determination condition shown in FIG. 5 is set by the following two methods, for example. First, there is a method of determining a condition value according to the characteristics of a form that is input in advance, that is, based on a sample of the form, and setting the condition value externally through an input unit such as the keyword input unit 11. In addition to this, for items such as the "allowable number of characters", the number of characters in the document item is automatically counted during the character recognition process, and automatic keyword addition is not possible for items that do not meet the document content knowledge. May be determined.

The automatic keyword attaching judging means 10 collates the automatic keyword attaching judging condition shown in FIG. 5 with each result of the character string / character cutout, character recognition, and document item classification for each item (error judging step). ). This will be described in detail. First, the document item “meeting name” is collated. Regarding the allowable character string direction 28, the character string direction of each line stored in the “meeting name” 15 shown in FIG. The character string direction of each line of the conference name 15 shown in FIG. 3 is “horizontal”, which coincides with the allowable character string direction “horizontal” of the automatic keyword attaching determination condition. The condition is judged to be clear.

With the allowable number of lines 29, the number of lines stored in the "meeting name" 15 shown in FIG. 3 is "4", which matches the allowable number of lines "1 or more and 4 or less" of the automatic keyword attaching determination condition. The automatic keyword attaching determination means 10 determines that the condition of the allowable number of lines is clear. Regarding the allowable character string position 30, the "meeting name" 15 and the "conference" are displayed based on the character string coordinates of each line stored in the "meeting name" 15 and the character string coordinates of each line stored in the "venue" 16 shown in FIG. The positional relationship of the area with "place 16" is obtained. Specifically, the coordinates of the “meeting name” 15 are obtained by a circumscribing rectangle including the character strings in the first to fourth lines, and SX = 30, SY = 40, W = 310, H = 12.
5, and the coordinates of "holding place" 16 are SX = 45, SY
Since = 160, W = 180, and H = 20, it can be seen that the “meeting name” 15 is located above the “holding place” 16.

On the other hand, since the allowable character string position 30 of the automatic keyword attaching determination condition is "above the holding place", the condition of the allowable character string position is determined to be clear. Regarding the allowable character string size 31, the maximum value of the character string height is obtained from the character string coordinates of each line stored in the "meeting name" 15 shown in FIG. In the case of FIG. 3, the heights of the first, second, and fourth lines are 25, 26, and 25, and the maximum value of the character string height is 2.
It becomes 6. Further, since the allowable character string size 31 of the automatic keyword attaching determination condition is "character string height <23", the condition of the allowable character string size is rejected. As for the allowable character similarity 32, the minimum value of the similarity values of the characters in each line stored in the “meeting name” 15 of FIG. 3 is obtained and “86” is obtained.

Further, the allowable character similarity 32 of the automatic keyword attaching determination condition is "85 or more", and it is judged that the condition of the allowable character similarity is clear. For the allowable number of characters 33, "39" is obtained by referring to the number of characters stored in the "meeting name" 15 of FIG. The allowable number of characters 33 of the automatic keyword attaching determination condition is “30”, which exceeds the maximum allowable number of characters, so the condition of the allowable number of characters is rejected.

Next, regarding the allowable character size 34,
The coordinates of each character image are extracted from the recognition result of each character in each line stored in the "meeting name" 15 in FIG. 3 to obtain the maximum width / height values. In the case of FIG. 3, the maximum width is “26” and the maximum height is “25”. The allowable character size 34 of the automatic keyword attaching determination condition is "character width or height <23", which exceeds the allowable character size. Therefore, the allowable character size condition is rejected. Further, with respect to the essential character string 35, since the condition of the allowable character size 35 of the automatic keyword attaching determination condition is not defined, it is determined that the condition of the essential character string 35 is clear. Finally, the number of acceptable items is 36
As for the above, since the five items shown in FIG. 3 are associated with character strings of one line or more, and the allowable item satisfaction number “5” of the automatic keyword attaching determination condition is satisfied, the condition of the allowable item satisfaction rate is Judge as clear.

As described above, the “meeting name” 15 in FIG. 3 is determined to be rejected under the three conditions of the allowable character string size 31, the allowable number of characters 33, and the allowable character size 34. Similarly to the above, "holding place" 16, "holding date and time" 17 of FIG.
The “attendees” 18 and the “meeting details” 19 are collated with the automatic keyword addition determination conditions. As a result, the "holding place" 16 and the "attendee" 18 are all cleared. On the other hand, regarding the “date and time of event” 17, the “date and time of event” 17 of FIG. 3 is indicated by the hatching of the characters (in FIG. 2, it is indicated by the shaded area on the background of the characters for the sake of electronic application).
Since the similarity 26 of the 20th character in the first line is "52", which is less than the allowable character similarity "85" of the automatic keyword attaching determination condition, the allowable character similarity is rejected. Also,
Regarding the “meeting content” 19, there is a “vertical (writing)” 27 in the character string direction on the 4th line in FIG. 4, which does not match the allowable character string direction “horizontal” in the automatic keyword attaching determination condition, so an allowable character Rejected in the row direction.

The items rejected by the automatic keyword assigning determining means 10 are displayed on a screen by a screen displaying means (not shown) to prompt correction, together with the keyword assigning determining conditions that do not match as shown in FIG. 6 (correction step). The operator designates a document item to be corrected by an instruction means (not shown). Specifically, a document item is designated using an instruction means such as a mouse along the automatic keyword addition determination result screen shown in FIG. For example, a GUI in which the correction area can be designated by placing the mouse cursor on the "meeting name" portion in the automatic keyword attaching determination result screen appearing in front of the document image 14 shown in FIG. 6 and clicking.
Software for realizing the above is incorporated in the document filing program of the present invention. According to such an operator's instruction, the keyword input means 11 displays the correction screens shown in FIGS. 7 to 12.

FIG. 7 shows the initial state of the correction screen for the document item "meeting name". A character recognition result 39 of automatic keyword addition is displayed in the upper part of the correction window 38, and an input field 40 for inputting a keyword is displayed in the lower part. In the initial state, the keyword input field 40 displays the automatic keyword addition result as it is. The automatic keyword assignment determination means 10 rejects the document item "meeting name" under three conditions of the character string size, the number of characters, and the character size, but cancels the rejected state on the screen shown in FIG. For example, in order to clear the condition of the number of characters, it is necessary to delete the number of characters to 30 or less. In this case, it is realized by moving a cursor (not shown) to the keyword input field and deleting extra characters. The screen after the deletion is shown in FIG. 8 (keyword input field 40a in FIG. 8).
In, "○ month ○ date" has been deleted). By finally pressing the output button 41, the description content of the item is confirmed. At this time, the data in the character recognition result shown in FIG. 3 is edited. That is, the character code, coordinates, and similarity of the recognition result corresponding to the deleted character are all deleted.

FIG. 9 shows the initial state of the correction screen for the document item "meeting date and time". In the example of FIG. 9, the dated area is shaded (in FIG. 9, it is shown by a hatched portion on the background of characters for the sake of electronic application).
Since the character with noise was recognized as it is (the incorrect character code due to the character with noise is input in the keyword input field 40b in FIG. 9), the automatic keyword assigning result 39a does not contain the correct code. In this case, as shown in FIG. 10, the characters in the portion where the character recognition did not work well are temporarily deleted (the inaccurate character code due to the noisy characters is deleted in the keyword input box 40c in FIG. 10), The corrector inputs correct characters as shown in FIG. 11 (keyword input box 40d in FIG. 11) via the keyword input means 11. By finally pressing the output button 41, the description content of the item is confirmed. At this time, the data in the character recognition result shown in FIG. 3 is edited. That is, the character code, coordinates, and similarity of the recognition result corresponding to the deleted character are all deleted, and then the character additionally input through the keyword input unit 11 is added to the character recognition result. At this time, since the coordinates and the similarity do not exist for the characters manually input, the coordinates are NU.
LL and the degree of similarity are set to 100 by judging that they are 100% accurate by manual input.

FIG. 12 shows the initial state of the correction screen for the document item "contents of proceedings". In the example of FIG. 12, the keyword input means 11 is displaying the screen by rejecting a vertically written character string mixed in regarding the agenda item. At this time, if the corrector browses all the recognition results using, for example, the scroll bars 42a and 42b, judges the rejection conditions and determines that they are appropriate, the automatic keyword addition result can be output as it is. Allow the document filing program to have the functions that can be performed. Specifically, FIG.
By pressing the output button 41 without touching the keyword input field 40e on the initial screen, the description content of the item is confirmed. In this case, since the result of automatic keyword addition has not been modified, the result of character recognition for the agenda item in FIG. 3 is not changed at all.

When the correction of all rejected items is completed by the keyword input means 11, the character string cutout, the character cutout, the character recognition, and the item classification result reflecting these corrections are transferred to the keyword storage means 12. The keyword accumulating unit 12 generates a keyword file from the character string cutout, the character cutout, the character recognition, and the item classification result in which both the automatic keyword addition result and the contents input by the keyword are stored. As a result, the keyword file 13 shown in FIG. 13 is obtained (filing data generation step).

As described above, according to the first embodiment, the image information in the processing area extracted from the input document image is recognized and converted into the character code, and the conversion result is determined for each predetermined document item. In classifying into, the predetermined determination condition set in advance and each processing result in the processing process for the input document image (automatic addition keyword obtained as a character recognition result) are compared, and there is an error in the processing based on the comparison result. If it is determined that there is an error in the processing, the automatic keyword assigning results 39a and 39b and the keyword input field 40a for the processing result corresponding to this are determined.
For example, while presenting correction instruction information such as 40e, the processing result is corrected according to the keyword input in accordance with this, and the processing result and / or the correction result is used to generate and store the document filing data regarding the input image. ,
For documents that are not disturbed, the results of character string cutout, character cutout, character recognition, and document item classification are automatically accumulated, and forms with disturbance such as shading and remarkable image rotation are input. In this case, it is possible to prompt the user to manually input the keyword. This allows
It is not possible to distinguish between a normal form and a document with disturbance, and it is possible to reduce the effort of comparing the document image with the document reading result for all document data in order to confirm that the document image has been converted correctly. it can.

Embodiment 2. FIG. 14 is a diagram showing the structure of a document filing apparatus according to the second embodiment of the present invention. In the figure, 11a is a keyword input means (correction means), which performs the same operation as that of the above-described first embodiment, and adds a function capable of designating a partial area in the document image as a keyword to the software that performs this operation. is doing. In addition, the same components as those in FIG. 1 are designated by the same reference numerals, and duplicate description will be omitted.

FIG. 15 is a diagram showing a document image stored in the image file shown in FIG. 14, FIG. 16 is a diagram showing a character recognition result for each document item obtained from the document image, and FIG. 17 is an automatic diagram shown in FIG. FIG. 18 is a diagram showing a result of automatic keyword assigning determination by the keyword assigning determining means, and FIG. 18 is a diagram showing a state after the correction screen of the document item “holding place” is corrected.
9 is a diagram showing a state in which the automatic keyword is deleted on the correction screen for the document item "date and time", and FIG. 20 is a diagram showing a state in which a partial area of the document image is input as a keyword on the correction screen for the document item "date and time". FIG. 21 is a diagram showing a state in which a partial area of the document image is input as a keyword on the correction screen for the document item “held date / time”, and FIG. 22 is a diagram showing the contents of the keyword file in FIG. 14.

Next, the operation will be described. First, the layout analysis unit 2 reads the document image 14a shown in FIG. 15 from the image file 1, and the layout analysis unit 2, the character recognition unit 3, and the document item classification unit 4 perform the character string reading, as in the first embodiment. The results of cutout, character cutout, character recognition result, and document item classification are obtained (document filing step). In FIG. 16, there are five document items, and the results for the document items “meeting name” 43 to “meeting content” 47 are stored. "Number of lines of document item" 48 and "number of characters of document item" 49 are stored for each item, and "character string direction" 50, "character string coordinates (cutting coordinates of character string)" for each line. 51, "Number of characters (in line)" 52,
The “recognition result of each character (cutout coordinate, character code, similarity)” 53 is stored. In FIG. 16, the document image 14a shown in FIG. 15 has the items of “holding place” and “holding date and time” in the same area, and the items are arranged side by side. It shows an example when it is not omitted. As a result, the item extraction of the holding date 45 fails, and the number of lines 54 and the number of characters 55 are set to zero.

The automatic keyword assigning determination means 10 applies the automatic keyword assigning determination condition shown in FIG. 5 to each result of the character string / character cutout, character recognition, and document item classification, and automatically applies to each document item. It is determined whether keywords can be added (error determination step). When the automatic keyword attaching determination means is applied in the same manner as in the first embodiment, the "meeting name" 56 and the "venue place" shown in FIG.
Rejection occurs with respect to the three items 57, "holding date" 58. The rejection for the "meeting name" 56 is due to the fact that there are 5 lines and the number of characters exceeds 30 characters, and the rejection for the "venue" 57 is the character string of the "conference name" 56. It is a rejection due to the position of the character string because it is not located above the character string of "holding place" 57. In addition, the rejection of the "held date and time" 58 was due to the fact that the "held date and time" 58 item was not extracted, and the "held date and time".
This is because none of the characters "month", "day", and "/", which are 58 essential character strings, do not exist.

Next, as in the first embodiment, the rejected document item is corrected using the keyword input means 11a (correction step). As for the modification of the “meeting name” 56, the modification regarding the number of lines and the number of characters is performed as in the first embodiment. Since this can be realized by the same operation as in the first embodiment, the description thereof will be omitted. "Place" 57
In the modification, the automatic keyword addition result 60 and the keyword input field 61 are displayed as the initial screen 59. In this case, since the part where the date is output is unnecessary, the latter two lines are deleted. As a result, the state shown in FIG. 18 is entered, and by pressing the output button 62, the character recognition result corresponding to the "holding place" 57 becomes only the previous two lines. Since the character string for the "held date and time" 58 does not exist in the correction for the "held date and time" 58, the initial screen as shown in FIG. 19 is displayed.

"(Not applicable)" is displayed as the automatic keyword addition result 60a in FIG. Here, in the second embodiment, the relevant area of the item can be input from the document image 14a via the keyword input means 11a. For example, a rectangular area 63 as shown in FIG. 20 can be set. Next, the keyword input means 11a again executes layout analysis and character recognition for the area designated by the rectangle. As a result, the character recognition result is automatically input to the keyword input field 61a as shown in FIG. Finally, when the output button 62 is pressed, the content of the keyword input field 61a is output as the result of character string / character cutout, character recognition, and item classification. At this time, not only the character code but also the coordinate value and the similarity are output together.

When the correction of all rejected items is completed by the keyword input means 11a, the character string / character cutout, the character recognition, and the item classification result reflecting these corrections are transferred to the keyword storage means 12. The keyword storage unit 12 generates a keyword file from the character string / character segmentation storing both the automatic keyword addition result and the keyword input content, the character recognition, and the item classification result. As a result, the keyword file 13 shown in FIG. 22 is obtained (filing data generation step).

As described above, according to the second embodiment, since the correction instruction information capable of designating the partial area in the document image is presented, not only the character code but also the coordinate information on the document image is input. Therefore, even when the keyword is manually set, the search result can be displayed by being superimposed on the document image at the time of the search.

[0074]

As described above, according to the present invention, the image information in the processing area extracted from the input document image is recognized and converted into the character code, and the conversion result is obtained for each predetermined document item. In classifying, a predetermined determination condition set in advance is compared with each processing result in the processing process for the input document image, it is determined whether there is an error in the processing based on the comparison result, and the processing has an error. When it is determined, the correction instruction information for the corresponding processing result is presented, and the processing result is corrected according to the correction information input accordingly, and the processing result and / or the correction result is used to input. Document filing data related to images is generated and stored, so the results of automatically character string / character segmentation, character recognition, and document item classification are stored for documents without disturbance. , When the form of applied disturbance such as a rotating shaded and marked image is input it can be modified. This makes it possible to accurately search even for forms that have disturbances such as shading and marked image rotation, and it is impossible to distinguish between ordinary forms and documents that have disturbances, and the document images are converted correctly. There is an effect that it is possible to reduce the effort for comparing the document image and the document reading result for all the document data in order to confirm that the document data is present.

According to the present invention, the arrangement direction of the character string in the processing area extracted from the input document image, the number of lines of the character string,
Position of character string, size of character string in processing area and / or character size of character string, likelihood of character recognition, number of characters in processing area, character code conversion result of image information in processing area, document item Since the degree of sufficiency is used as the judgment condition, it is impossible to distinguish between a normal form and a document with disturbance, and in order to confirm that the document image has been converted correctly, the document image and document reading results for all document data There is an effect that it is possible to reduce the labor for comparing.

According to the present invention, only the portion having a difference from the determination result of the processing result is presented as the correction instruction information, and the processing result is corrected according to the correction information input to the portion having the difference. There is an effect that the process regarding correction can be simplified.

According to the present invention, since the correction instruction information capable of designating the partial area in the document image is presented as the correction information, not only the character code but also the coordinate information on the document image can be input. Even when the correction information is set, it is possible to display the search result by superimposing it on the document image at the time of search.

According to the present invention, since the correction instruction information capable of externally correcting the portion that is different from the determination condition of the processing result is presented, the processing relating to the correction can be simplified.

According to the present invention, since the correction instruction information by which the determination result indicating that the processing result has an error can be externally determined is presented, the processing relating to the correction can be simplified.

According to the present invention, the data in which the processing result and the correction result of the correction means are combined is stored as the document filing data, so that accurate retrieval is also possible for a form to which disturbance such as shading or remarkable image rotation is added. The effect is that document data that can be stored can be stored.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a document filing device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a document image stored in an image file in FIG.

FIG. 3 is a diagram showing a character recognition result for each document item obtained from a document image.

FIG. 4 is a diagram showing in detail the contents for the document item “meeting contents” of FIG. 3;

FIG. 5 is a diagram showing automatic keyword assignment determination conditions set in the automatic keyword assignment determination means in FIG.

6 is a diagram showing an automatic keyword assignment determination result by the automatic keyword assignment determination means in FIG.

FIG. 7 is a diagram showing an initial state of a correction screen for a document item “meeting name”.

FIG. 8 is a diagram showing a state in which the correction screen for the document item “meeting name” has been corrected.

FIG. 9 is a diagram showing an initial state of a correction screen for a document item “meeting date and time”.

FIG. 10 is a diagram showing a state in which an automatic keyword is deleted on the correction screen for the document item “meeting date and time”.

FIG. 11 is a diagram showing a state in which a keyword is input on the correction screen for the document item “meeting date and time”.

FIG. 12 is a diagram showing an initial state of a correction screen for the document item “contents of proceedings”.

FIG. 13 is a diagram showing the contents of a keyword file in FIG.

FIG. 14 is a diagram showing a configuration of a document filing device according to a second embodiment of the present invention.

15 is a diagram showing a document image stored in the image file in FIG.

FIG. 16 is a diagram showing a character recognition result for each document item obtained from a document image.

FIG. 17 is a diagram showing an automatic keyword assignment determination result by the automatic keyword assignment determination means in FIG.

FIG. 18 is a diagram showing a state after the correction screen for the document item “holding place” is corrected.

FIG. 19 is a diagram showing a state in which an automatic keyword is deleted on the correction screen for the document item “held date”.

FIG. 20 is a diagram showing a state in which a partial area of a document image is input as a keyword on the correction screen for the document item “held date / time”.

FIG. 21 is a diagram showing a state where a partial area of a document image is input as a keyword on a correction screen for the document item “held date”.

22 is a diagram showing the contents of a keyword file in FIG.

FIG. 23 is a block diagram schematically showing a configuration of a conventional document filing device.

FIG. 24 is a flowchart showing a document file creating operation by the document filing device in FIG. 23.

FIG. 25 is a flowchart showing a document search operation by the document filing device in FIG. 23.

FIG. 26 is a diagram showing an example of a document image stored in the image file in FIG. 23.

FIG. 27 is a diagram showing the contents of a document layout knowledge file relating to the document image of FIG. 26.

FIG. 28 is a diagram showing layout information created by a layout analysis unit.

FIG. 29 is a diagram showing the result of character recognition by the character recognition means.

FIG. 30 is a diagram showing a result of collation of document layout information and document layout knowledge by the document item classification unit.

FIG. 31 is a diagram showing an example of the content of a document content knowledge file.

FIG. 32 is a diagram showing a result of collating the character recognition result and the document content knowledge by the document item classification unit.

FIG. 33 is a diagram showing a collation result of the character recognition result and the document content knowledge integrated by the document item classification unit.

FIG. 34 is a diagram showing a correspondence relationship between the document items classified by the document item classification means and the character strings.

[Explanation of symbols]

1 image file, 2 layout analysis means (document filing means), 3 character recognition means (document filing means), 4 character item classification means (document filing means), 6 retrieval means, 7 output means, 8 document layout knowledge file, 9 Document content knowledge file,
10 automatic keyword attaching determining means (error determining means),
11, 11a Keyword input means (correction means), 12
Keyword storage means (document data storage means), 13
Keyword file (document data storage means), 100
Image file, 101 layout analysis means, 10
2 character recognition means, 103 document item classification means, 104
Storage means, 105 search means, 106 output means, 10
7 Document layout knowledge file, 108 Document content knowledge file, 109 Character recognition result file, 1
10 "Meeting name" item, 111 "Venue place" item,
112 "Date and time" item, 113 "Attendee" item,
114 coordinate information, 115 start block, 116 end block, 117 area coordinate information, 118 coordinate information, 119 character recognition result, 120 matching degree, 121
Keyword information corresponding to the item "venue", 122
Keyword "holding place", 123 keyword "place", 124 keyword "meeting room", 125 keyword "room", 126 keyword "meeting", 127 keyword "meeting name", 128 keyword "holding date", 129, 130, 131 , 13
2, 133 matching degree, 134 correspondence relationship, 135 matching degree, 136 combinations.

Front page continuation (51) Int.Cl. ⁷ identification code FI theme code (reference) G06T 7/40 100 G06T 7/40 100A 7/60 150 7/60 150B F term (reference) 5B050 AA10 BA06 BA10 CA07 EA06 EA07 FA03 FA19 5B064 AA07 EA10 5B075 ND07 ND24 NK02 NK25 NK37 UU06 5L096 BA17 FA06 FA44 FA67 FA69

Claims (16)

[Claims] 1. The image information in the processing area extracted from the input document image is recognized and converted into a character code, and
A document filing unit that classifies the conversion result for each predetermined document item, a predetermined determination condition set in advance when the document filing unit processes an input document image, and each processing result in a processing process for the input document image. Error determination means for determining whether or not there is an error in the processing of the document filing means based on the comparison result, and if the error determination means determines that there is an error in the processing, it corresponds to this. A correction means for presenting correction instruction information for the processing result and correcting the processing result according to the correction information input in response to the processing result, the processing result of the document filing means and / or the correction result of the correction means. And a document data storage means for generating and storing document filing data relating to the input image. Grayed apparatus. 2. The document filing apparatus according to claim 1, wherein the error determination means uses the arrangement direction of the character string in the processing area extracted from the input document image by the document filing means as a determination condition. 3. The document filing apparatus according to claim 1, wherein the error determination means uses the number of lines of a character string in the processing area extracted from the input document image by the document filing means as a determination condition. 4. The document filing apparatus according to claim 1, wherein the error determination means uses the position of the character string in the processing area extracted from the input document image by the document filing means as the determination condition. 5. The error determination means uses the size of a character string in the processing area extracted from the input document image by the document filing means and / or the size of a character forming the character string as a determination condition. 1. The document filing apparatus described in 1. 6. The error determination means uses the likelihood of character recognition in the processing area extracted from the input document image by the document filing means as a determination condition.
Document filing device described. 7. The document filing apparatus according to claim 1, wherein the error determination unit uses the number of characters in the processing area extracted from the input document image by the document filing unit as a determination condition. 8. The document filing apparatus according to claim 1, wherein the error determination unit uses a result of character code conversion of image information in the processing area extracted from the input document image by the document filing unit as a determination condition. . 9. The document filing apparatus according to claim 1, wherein the error determination means uses the degree of satisfaction of the document items classified by the document filing means as a determination condition. 10. The correction means presents only a portion having a difference from the determination condition of the processing result as the correction instruction information, and corrects the processing result according to the correction information input to the portion having the difference. The document filing apparatus according to claim 1, wherein: 11. The document filing apparatus according to claim 1, wherein the correction means presents correction instruction information capable of designating a partial area in the document image as the correction information. 12. The document filing apparatus according to claim 1, wherein the correction means presents correction instruction information capable of externally correcting a portion that differs from the determination condition of the processing result. 13. The document filing apparatus according to claim 1, wherein the correction means presents correction instruction information capable of externally determining a determination result indicating that the processing result has an error. 14. The document filing apparatus according to claim 1, wherein the document data accumulating unit stores, as document filing data, data obtained by combining the processing result of the document filing unit and the correction result of the correction unit. 15. A document filing step of recognizing image information in a processing area extracted from an input document image and converting it into a character code, and classifying the conversion result for each predetermined document item, and the document filing step. In processing the input document image, the predetermined determination condition set in advance is compared with each processing result in the processing process for the input document image, and there is an error in the processing in the document filing step based on the comparison result. An error determination step for determining whether or not there is an error, and when it is determined that there is an error in the processing in the error determination step, the correction instruction information for the processing result corresponding to this is presented and input in accordance with this. The correction step that corrects the processing result according to the correction information that has been corrected, and the processing result in the document filing step. Fruit and /
Or a filing data generation step of generating document filing data regarding an input image using the correction result in the correction step. 16. A document filing step of recognizing image information in a processing area extracted from an input document image and converting it into a character code, and classifying the conversion result for each predetermined document item, and the document filing step. In processing the input document image, the predetermined determination condition set in advance is compared with each processing result in the processing process for the input document image, and there is an error in the processing in the document filing step based on the comparison result. An error determination step for determining whether or not there is an error, and when it is determined that there is an error in the processing in the error determination step, the correction instruction information for the processing result corresponding to this is presented and input in accordance with this. The correction step that corrects the processing result according to the correction information that has been corrected, and the processing result in the document filing step. Fruit and /
Alternatively, a document filing program that causes a computer to execute a filing data generation step of generating document filing data regarding an input image using the correction result in the correction step.