JP2003223608A - Method and device for correcting recognized character string, and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To increase the accuracy of character correction while enabling the batch correction of the results of recognition. <P>SOLUTION: Any character based on an operation instruction from an operator is selected from among the text data of the results of recognition, and the selected character C1 is substituted for a substitute character C2 based on the operation instruction from the operator. Next the processed character is moved on the text data from the selected character C1 backward to determine whether character is the same as or similar to the selected character C1 (S270, S275). When the character is determined to be the same as or similar to the selected character, the character as an automatically corrected character Cm is temporarily substituted for the substitute character C2 (S290). Then the syntax at the front and rear of the substitute character C2 is analyzed (S300) and, when the syntax is discriminated to be correct, the substitution for the substitute character C2 is defined to be reasonable (S310 to 330). <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for correcting a character string obtained by a character recognition process in which an original is input as image information and characters are recognized.

[0002]

2. Description of the Related Art In a character recognition apparatus for performing character recognition processing, each character in a character dictionary prepared in advance is compared with an input character, and the character with the highest degree of coincidence is taken as the recognition result. However, since the recognition result is not always correct, the character string as the recognition result is then displayed on the display, and the erroneously recognized character can be corrected on the screen. In other words, when the character to be corrected is specified on the screen, multiple characters that have a relatively high degree of matching at the time of recognition are displayed as candidate characters in the next rank, and if there is a correct character among them, the character is selected. If not, for example, the correct character is input from the keyboard by the kana-kanji conversion method or the like.

By the way, in the above character recognition device, the same character is often erroneously recognized at a plurality of places. In such a case, if the above-mentioned correction work is performed character by character, the workability is poor and the time required for error correction extends for a long time. For this reason, when correcting one misrecognized character, the image information that matches the misrecognized character is also searched for,
A character recognition device equipped with a batch correction function capable of correcting a plurality of points at once has been proposed.

[0004]

However, in the above conventional technique, when the character read as the image information is a character similar to the correction target character, the similar character is also collectively corrected. Sometimes, even correctly recognized characters were changed accidentally. As a result,
There was a problem that the accuracy of character correction was reduced.

The present invention has been made in view of the above problems, and an object thereof is to improve the accuracy of character correction while enabling batch correction of recognition results.

[0006]

[Means for Solving the Problem and Its Action / Effect] As a means for solving at least a part of the above-mentioned problems, the following constitution is adopted.

The recognized character string correction method of the present invention is a recognized character string correction method for correcting a character string obtained by a character recognition process of inputting a document as image information and recognizing characters. A step of selecting an arbitrary character from the character string based on an operator's operation command;
(B) a step of replacing the selected arbitrary character with a predetermined character according to an operator's operation command, and (c) an arbitrary character selected from the character string in the step (a). A step of sequentially selecting substantially matching characters as a character to be automatically corrected, and (d) the character to be automatically corrected,
Based on the result of the syntax analysis by the step (e), the step of substituting with the predetermined character; (e) the step of analyzing the syntax before and after the character replaced by the step (d); And the step of determining the result of the replacement in step (d) as valid.

The character recognition processing may be performed by driving the image information reading device to directly read the image information of the original as long as the image information representing the original is input. The image information of the document obtained in advance by using the information reading device may be read from the storage device.
In addition, the above-mentioned “substantial match” may be a complete match, a similar match, or a match and a similar match.

According to the method for correcting a recognized character string having the above configuration (hereinafter referred to as a basic configuration), the recognized character string (hereinafter, referred to as the character string based on the operation instruction of the operator is executed in steps (a) and (b). An arbitrary character is selected from among the recognition character strings) and is replaced with a predetermined character. In step (c), the characters that substantially match the selected arbitrary character are sequentially selected from the recognized character strings as automatic correction target characters, and in step (d), they are once replaced with predetermined characters. . The syntax before and after the replaced character is analyzed in step (e), and based on the result of the analysis, in step (f), the result of the upper replacement is determined to be valid.

Therefore, according to the recognized character string correction method of the present invention, the correction is confirmed as valid only when the automatically corrected character has the correct syntax. Therefore, there is an effect that the accuracy of character correction can be improved.

In the recognition character string correction method having the above structure,
(G) a step of returning the automatic correction target character to the state before the replacement in step (d) when the result of the syntax analysis in step (f) indicates that the syntax is not correct It can be configured.

According to this structure, when the automatically made correction of the character is not the correct syntax, the correction is undone. Therefore, the accuracy of character correction can be further improved.

In the recognition character string correction method having the above structure,
(H) It may be configured to include a step of displaying the replacement character determined to be valid in the step (f) so as to be distinguishable from other characters.

According to this structure, a character whose substitution result is determined to be valid is displayed in a distinguishable manner with respect to other characters. Therefore, the operator can know from the display that the characters are automatically corrected.

In the recognition character string correction method having the above structure,
The step (f) includes (f-1) the step (f)
When the result of parsing the syntax by means of correct syntax is inquired to the operator whether to accept the result of the replacement by the step (d),
(F-2) When it is determined that the result of the replacement is accepted based on the operation command from the operator in response to the inquiry in the step (f-1), the result of the replacement in the step (d) is validated. And a step of confirming.

According to this structure, since it is possible to inquire of the operator whether or not the result of the correction can be accepted, it is possible to further improve the accuracy of the character correction.

According to this structure, it is possible to select candidate character string data having an excellent matching result from a plurality of candidate character string data and output the selected candidate character string data as a conversion result.

In the recognition character string correction method having the above structure,
When the arbitrary character selected in the step (a) is composed of a plurality of continuous characters, the replacement in the step (b) replaces the plurality of continuous characters with one predetermined character. Can be

According to this structure, when a plurality of characters are recognized from the image data representing one character, for example, "hour"
When the characters of "day" and the characters of "temple" are recognized separately from the image data of, it is possible to correct them to the original one character.

In the recognition character string correction method having the above structure,
The replacement in step (b) can be configured to replace one character with a plurality of characters.

According to this structure, when one character is recognized from the image data representing a plurality of characters, it can be corrected to the original plurality of characters.

The recognized character string correction device of the present invention is a recognized character string correction device for correcting a character string obtained by a character recognition process for inputting a document as image information and recognizing characters. Manual correction target character selection means for selecting an arbitrary character from the character string based on an operation command, and manual correction means for replacing the selected arbitrary character with a predetermined character according to an operation command of an operator. And an automatic correction target character selecting unit that sequentially selects, as the automatic correction target character, a character that substantially matches an arbitrary character selected by the manual correction target character selecting unit from the character string, and the automatic correction target character Is replaced by the predetermined character, an automatic correction means, a syntax analysis means for analyzing the syntax before and after the character replaced by the automatic correction means, and a syntax by the syntax analysis means Based on the results of the analysis, it is characterized by comprising a modification result determination means for determining a valid result of substitution by the automatic correction means.

The recognized character string correcting method having the above-described structure has the same operation and effect as the recognized character string correcting apparatus of the present invention, and can improve the accuracy of character correction.

A computer program according to the present invention is a computer program for executing a process of correcting a character string obtained by a character recognition process of inputting a document as image information and recognizing a character. A function of selecting an arbitrary character from the character string based on the operation command of (b), and a function of (b) replacing the selected arbitrary character with a predetermined character according to an operation command of the operator,
(C) a function of sequentially selecting, from the character strings, characters that substantially match an arbitrary character selected by the function (a) as a character to be automatically corrected; and (d) the character to be automatically corrected, Based on the function of replacing with a predetermined character, (e) the function of analyzing the syntax before and after the character replaced by the function (d), and (f) the result of the syntax analysis by the function (e). , A function of determining the result of the replacement by the function (d) as a valid one is realized by a computer.

The computer program having the above configuration has the same operation and effect as the recognized character string correction device and the recognized character string correction method of the present invention, and can improve the accuracy of character correction.

The recording medium of the present invention is characterized by a computer-readable recording medium recording the computer program of the present invention. This recording medium has the same operations and effects as the computer programs of the present invention.

[0027]

Other Embodiments of the Invention The present invention also includes the following other embodiments. A first aspect thereof is an aspect as a program supply device for supplying the computer program of the present invention via a communication path. In the first aspect, the computer program is placed in a server or the like on a computer network, the necessary program is downloaded to a computer via a communication path, and the computer is executed to realize the above-described device or method. You can

[0028]

BEST MODE FOR CARRYING OUT THE INVENTION In order to further clarify the structure and operation of the present invention described above, the embodiments of the present invention will be described below based on Examples.

1. Overall Hardware Configuration FIG. 1 is a block diagram showing the schematic hardware configuration of a computer system to which the first embodiment of the present invention is applied. The computer system mainly includes a so-called personal computer (hereinafter, simply referred to as a computer), and a CRT display 12 and an image scanner 14 are provided around the personal computer. The computer includes a computer main body 16, a keyboard 18, and a mouse 20. It should be noted that the computer main body 16 includes a CD-RO.
A CD drive 24 for reading the contents of M22 is mounted.

The computer main body 16 has a CPU 31, which is a central processing unit, as a center, and a ROM 31, a RAM 32, a display image memory 33, a mouse interface 34, and a keyboard interface 3 which are interconnected by a bus.
5, CDC 36, HDC 37, CRTC 38, I / O interface 40 and I / O port 41. The ROM 31 is a read-only memory that stores therein various programs and the like. RAM32 is
It is a readable / writable memory that stores various data. The display image memory 33 is the CRT display 1
2 is a memory for storing the image data of the image to be displayed in 2.

The mouse interface 34 is used for the mouse 20.
It is an interface that controls the exchange of data and so on.
The keyboard interface 35 is an interface that controls key input from the keyboard 18. CDC36
Is a CD controller that controls the CD drive (CDD) 24. The HDC 37 is a hard disk controller that controls a hard disk drive (HDD) 42. The HDD 42 stores in advance a computer program described later, a character dictionary used for character recognition described later, a word dictionary and a rule dictionary used for syntax analysis described later, and the like.

The CRTC 38 is a CRT controller that controls the display of an image on the CRT display 12 based on the display image data stored in the display image memory 33. The input / output device interface 40 is an input / output device connected to the outside, which is the image scanner 1 in this embodiment.
4 is an interface for controlling input / output of data to / from. The I / O port 41 has a serial output port and is connected to the modem 44.
Is connected to the public telephone line 46 via. The computer main body 16 is connected to an external network via a modem 44 and can be connected to a specific server 47.

In this computer system, the operating system is stored in the HDD 42, and when the computer main body 16 is powered on, the RAM 32 follows the loader written in the boot block of the HDD 42.
Is loaded into a predetermined area of. Further, character recognition software (computer program) for converting an image captured by the image scanner 14 (hereinafter referred to as a scan image) into text data is stored in the CD-ROM 22 in advance and starts a predetermined installation program. From the CD drive 24 to the computer body 16
Installed on. The installed computer program is stored in the HDD 42 and is loaded into a predetermined area of the RAM 32 when a predetermined start command is received.

This computer program is executed by the CPU 30.
By executing the above, various constituent requirements of the present invention are realized. As described above, this computer program is stored in the CD-ROM 22, but instead of this, a floppy disk, a magneto-optical disk, an IC
It may be configured to be stored in another portable recording medium (portable recording medium) such as a card. Further, the computer program described above can be obtained by downloading program data provided via a network from a specific server 47 connected to an external network and transferring the program data to the RAM 32 or the HDD 42. . The network may be the Internet or a computer program downloaded from a specific home page.
Alternatively, it may be a computer program supplied in the form of an email attachment.

The state of control processing according to the character recognition software by the computer system having the above-described hardware configuration will be described below. FIG. 2 is a block diagram showing a state of processing according to the character recognition software 50 executed by the computer main body 16.

As shown in FIG. 2, the computer main body 16
According to the character recognizing software 50 operating inside, first, the scanner driver 60 is operated by the scan image capturing unit 51 so that the image (scan image) of the document P in which the document is written is obtained from the image scanner 14. Perform the import process. Then, the character string recognition module 52 performs character recognition on the scanned image captured by the scanned image capturing unit 51. By the character string recognition module 52, the data (text data) of the character string written on the document can be obtained.

Next, the recognized character string correction module 53 corrects the character string data obtained by the character string recognition module 52. Recognition character string correction module 53
The function of is, in detail, as follows. First, based on the operation command of the mouse 20 or the keyboard 18 by the operator, the manual correction target character selection unit 53a
An arbitrary character is selected from the text data. Then, the selected arbitrary character is replaced with a predetermined character according to an operation command of the keyboard 18 or the mouse 20 by the operator by the manual correction unit 53b.

Next, from the character string data obtained by the character string recognition module 52, a character that substantially matches an arbitrary character selected by the manual correction target character selection section 53a is selected as an automatic correction target character selection section. The characters 53c are sequentially selected as the characters to be automatically corrected. After that, the automatic correction target character is replaced by the predetermined character by the automatic correction unit 53d. Next, the syntax before and after the character replaced by the automatic correction unit 53d is analyzed by the syntax analysis unit 53d.
Analyze by e. Then, based on the result of the syntax analysis, the result of replacement by the automatic correction unit 53d is confirmed by the correction result confirmation unit 53f as valid.

Manual correction unit 53b and correction result confirmation unit 5
The character string data determined by 3f is sent to the CRT display 12 via the display driver 70 and displayed.

By executing the character recognition software 50 by the CPU 30 of the computer main body 16, the above-mentioned character string recognition module 52 and the recognized character string correction module 53 are realized. The control processing according to the character recognition software 50 will be described in detail below. Figure 3
It is a flow chart which shows the routine of this control processing.
This routine is repeatedly executed at predetermined time intervals after the instruction to execute the character recognition software 50 is issued.

As shown, when the process is started, C
The PU 30 first determines whether or not it is the first time after the activation of the character recognition software 50 (step S10).
0). Here, when it is determined that it is the first time, a process of displaying the application window WD on the CRT display 12 in the initial state is performed (step S11).
0). FIG. 4 is an explanatory diagram showing the application window WD in the initial state. As shown in the figure, the application window WD is provided with a menu bar BR1 and toolbars BR2, BR3. The toolbar BR2 includes a batch correction button BT11, a combination button BT12, a separation button BT13, and a re-recognition button BT, which will be described later.
14 are provided. On the toolbar BR3, the buttons BT for "scan", "area extraction", "recognition", and "save" for executing the commands in order according to the work procedure.
1, BT2, BT3, BT4 are provided. In addition,
Returning to FIG. 3, when it is determined in step S100 that it is not the first, the process of step S110 is not executed.

Next, the CPU 30 causes the toolbar BR3.
It is determined whether or not the "scan" button BT1 above is clicked by the mouse 20 (step S1).
20). Here, when the determination is affirmative, the CPU 30
Performs scan processing for scanning the document P that is the target of character recognition (step S130). The manuscript P is, for example, a document in which a document is described. Step S130
Then, the CPU 30 operates the scanner driver 60 to take in the scan image data indicating the image of the document P from the image scanner 14 on which the document P is set. After the execution of step S130, the process returns to "return" to end the process once. It should be noted that, instead of the configuration of step S130, the scan image of the document P captured by using the image scanner is stored in the HDD 42 in advance, and this HDD is stored.
It is also possible to adopt a configuration in which the scan image in 42 is read out. That is, as long as the document P can be input as image information, the image scanner may be used to read each time, or it may be prepared in advance in a storage unit such as the HDD 42 and read from the storage unit. You can also Further, the image information of the document may be taken in from the outside via a network.

On the other hand, when a negative decision is made in step S120, the CPU 30 advances the processing to step S140, and decides whether or not the "area extraction" button BT2 has been clicked by the mouse 20. Here, when the determination is affirmative, the CPU 30 executes a region extraction process (step S150). This area extraction process
From the scan image data taken in by the scan processing of step S130, an arbitrary image area that is a target of character string recognition is extracted. Using the mouse 20, the operator performs an operation of designating the range of the image area to be the target of character string recognition for the scan image of the document P displayed in the application window WD. CPU
In response to the mouse operation by the operator, the reference numeral 30 extracts the data of the image area designated by the mouse 20 from the scan image data. When this area extraction processing is completed, the process returns to "return" and this processing is once completed.

On the other hand, if the determination in step S140 is negative, the process proceeds to step S160. Although not described in detail in the flow chart, even if a positive determination is made in step S140, step S130
If it is not after the execution of the scan processing in step S1, step S1
The process proceeds to 60.

By the processing up to this point, the extracted image data obtained by the area extraction processing in step S150 is prepared in the HDD 42 as image data to be subjected to character recognition (hereinafter referred to as recognition target image data). It will be. If only the scan process of step S130 is executed but the region extraction process of step S150 is not executed, the scan image data obtained by the scan process of step S130 becomes the recognition target image data as it is. .

When the process proceeds to step S160, it is determined whether or not the "recognition" button BT3 on the toolbar BR3 has been clicked by the mouse 20. Here, when the determination is affirmative, the CPU 30 executes a character string recognition process for converting the recognition target image data into text data (step S170). This character string recognition processing is performed by using each character in a character dictionary (prepared by this character recognition software 50) prepared in advance in the HDD 42,
This is a well-known technique in which an image of each character represented by the recognition target image data is compared and the character having the highest degree of coincidence is used as the recognition result, and detailed description thereof will be omitted here.

FIG. 5 is an explanatory diagram showing an example of the application window WD after completion of the character string recognition processing. As shown, application window WD
Below the toolbar BR3, a window F1 for displaying an original image and a window F2 for displaying a recognition result are prepared. After the character string recognition processing in step S170 is completed, the recognition target image data TD is displayed in the original image window F1, and the recognition result window F2 displays the character string (text data) which is the recognition result of this character string recognition processing. ) Is displayed. A candidate character field F3 is prepared below the recognition result window F2, and the recognition result window F3 is provided in this candidate character field F3.
The next candidate for recognition of the character at the position of the cursor in the text data displayed in 2 is displayed. In the illustrated example, the cursor is located at the beginning of the sentence in the text data (the cursor is designed to be located at the beginning of the sentence immediately after the character string recognition processing). In this example, the next candidate for recognition of spaces (actually all spaces as shown) is displayed.

Returning to FIG. 3, if a negative decision is made in step S160, the process proceeds to step S180. Although not described in detail in the flow chart, even if a positive determination is made in step S160,
If it is not after the execution of the scanning process of 130, the process proceeds to step S180. In step S180, it is determined whether or not the "save" button BT4 on the toolbar BR3 has been clicked by the mouse 20, and if an affirmative determination is made here, the CPU 30 obtains the character string recognition processing routine. The saved text data is saved (step S190). After execution of step S190, or when a negative determination is made in step S190, the process returns to "return" and the control processing routine is temporarily terminated.

After execution of the character string recognition processing in step S170, the application window WD shown in FIG. 5 is displayed on the CRT display 12, as described above. By comparing the text data displayed in the recognition result window F2 of the application window WD with the image data displayed in the window F1 for displaying the original image, the operator can correctly perform the character string recognition process. It can be confirmed whether or not. When an operator finds an error in the text data as a recognition result, he / she corrects the erroneous character. This correction work is performed according to a recognition character string correction routine executed by the CPU 30.
Hereinafter, this recognition character string correction routine will be described in detail.

FIGS. 6 and 7 are flowcharts showing a recognized character string correction routine executed by the CPU 30. This recognition character string correction routine is repeatedly executed at predetermined time intervals. As shown in FIG. 6, when the process is started, the CPU 30 first receives an operation signal from the mouse 20 or the keyboard 18 and performs an operation of selecting a desired character from the text data as a recognition result. It is determined whether or not (step S200). This character selection operation is an operation performed by the operator using the mouse 20 or the keyboard 18, and when the mouse 20 is used, a desired position of the text data displayed on the screen of the CRT display 12 is clicked. An operation such as selecting one character is performed. When it is determined in step S200 that the character selection operation has not been performed, the CPU 30 advances the process to return, and once ends the process of the recognized character string correction routine.

On the other hand, if it is determined in step S200 that the character selection operation has been performed, the following processing is executed. First, the CPU 30 operates to select the character,
That is, a process of selecting the character C1 at a position determined by the operation of the mouse 20 or the keyboard 18 (a position in the text data, hereinafter referred to as a predetermined position P0) is performed (step S210). FIG. 8 shows this step S
It is explanatory drawing which shows an example of the application window WD after the process of 210. FIG. 9 is an enlarged view around the recognition result window F2 in FIG. By the processing of the CPU 30, as shown in FIG. 9, in the text data displayed in the recognition result window F2, the second line, “It is an unusual row of grass”.
The character of "grass" in the character is a character (hereinafter referred to as the selected character)
Selected as C1. At this time, in the candidate character field F3, “car”, “desk”, which are the next candidates for recognition,
“Skewer”,… are displayed in order.

Returning to FIG. 6, the CPU 30 then temporarily stores the selected character C1 in the RAM 32 as the memory selected character C1X (step S215). Then, CPU3
At 0, the selected character C1 in the text data is replaced with a predetermined character C2 in accordance with an operation command from the mouse 20 or the keyboard 18 by the operator (step S220). In the example of FIG. 9, the operator uses the mouse 20 to click the correct character “car” in place of the “grass” that was erroneously recognized from the next candidate group displayed in the candidate character field F3. In step S220,
The operation signal from this mouse 20 is taken in and the mouse 2
Candidate character determined by 0 operation (in this example,
“Vehicle”) is set as the predetermined character C2, and the selected character C1 in the text data is overwritten.

Returning to FIG. 6, after the execution of step S220,
The CPU 30 determines whether or not an instruction to activate the batch correction function is set in advance (step S230). This instruction is given by a mouse operation by the operator on the application window WD. A batch correction button BT11 (see FIG. 8) is provided on the toolbar BR2 of the application window WD. When the batch correction button BT11 is clicked with the mouse, an instruction to activate the batch correction function is given to the computer main body 16. Can be set. Step S2
At 30, it is determined whether or not an instruction to operate this collective correction function is set in advance. If it is determined that the instruction has not been set, the process returns to "return" and the process ends.

On the other hand, when it is determined in step S230 that the instruction to activate the collective correction function is preset, the CPU 30 performs the following process. First, from the recognition target image data, image data (hereinafter referred to as character image data) P in a character range corresponding to the predetermined position P0.
D is extracted and stored in the RAM 32 (step S24)
0). The character image data PD is character image data corresponding to the selected character C1. Next, the CPU 30 sets the value 1 to the variable m (step S250).

After that, the CPU 30 determines whether or not there is a recognized character in the text data which is the recognition result, behind the predetermined position P0 by the value of the variable m (step S260). Here, if it is determined that there is not, the process returns to "return" and the process ends. On the other hand, if it is determined in step S260 that the recognized character exists, the CPU 30 advances the process to step S270 in FIG.

In step S270, the CPU 30 determines the predetermined position P determined to exist in step S260.
It is determined whether or not a character (hereinafter, referred to as the m-th character) Cm that is located behind 0 by the value of the variable m is the same as the selected storage character C1X obtained in step S215. Here, when it is determined that they are the same, the m-th character Cm
Is temporarily stored in the RAM 32 as the uncorrected character CmX (step S280), and the m-th character (corresponding to the automatic correction target character in claims) Cm on the text data of the recognition result is used in step S220. The specified character (hereinafter referred to as a replacement character) C2 is replaced (step S290).

On the other hand, in step S270, the m-th character C
When it is determined that m is not the same as the selected character C1X for storage, the CPU 30 advances the process to step S275, and the character image data corresponding to the m-th character Cm is the character image data PD stored in step S240. And is similar to. This similarity judgment is
The feature of the image represented by the character image data is extracted, the degree of coincidence of the feature is quantitatively calculated, and the determination is made from the calculation result. If it is determined in step S275 that they are similar, the process proceeds to step S280. That is, in step 270, the mth character Cm is the same as the selected character C1 selected in step 210, or the character image data corresponding to the mth character Cm is the character corresponding to the selected character C1. If it is similar to the image data, the process proceeds to step 280.

Note that this step S270 and step S
Instead of the configuration including 275, step S275 may be deleted and only step S270 may be included.
In this case, when a negative determination is made in step S270, the process proceeds to step S350. Alternatively, step S270 may be deleted and only step S275 may be provided. In this case, step S260 is followed by step S275.

After the replacement with the character C2 is performed in step S290, the CPU 30 advances the process to step S300 to analyze the syntax before and after the replacement character C2 in the text data as the recognition result. . The processing of parsing is to divide an input text into minimum linguistic units called morphemes, collect sentences divided into morphemes into units called bunsetsu, and analyze the syntactic structure of the sentence. The division into the above morphemes is performed based on a word dictionary containing all parts of speech. The analysis of the syntactic structure attempts to analyze the dependency structure of each clause, and is performed based on a rule dictionary that stores knowledge necessary for the syntactic analysis. The word dictionary and the rule dictionary are stored in the HDD 42 in advance as described above.

The dependency structure of a bunsetsu classifies the type of partner's bunsetsu that the bunsetsu can modify, and the type of the other bunsetsu that the bunsetsu can receive,
It has a structure that is related to and receives each. In the analysis of the syntactic structure, the dependency structure of the bunsetsu is analyzed to evaluate the degree of the dependency of the bunsetsu, that is, the degree of the semantic connection of the bunsetsu. Since a specific syntax analysis method is well known, its description is omitted here. In step S300, a character string of a predetermined number of characters centering on the replacement character C2 is extracted from the text data that is the recognition result, and the parsing process is performed by using this character string as the input text. Note that the syntax analysis method is not limited to the above description,
Anything can be used as long as it can semantically parse the syntax. Further, the ranges before and after the above need not necessarily be determined from the predetermined number of characters, and may be extracted with appropriate clauses or sentences.

When the syntax analysis in step S300 is completed,
Next, the CPU 30 uses this syntax (the replacement character C2 in the text data as the replacement character C2 based on the evaluation result of the syntax analysis.
It is determined whether the syntax before and after is correct (step S3)
10). That is, if the evaluation result shows that the degree of the semantic connection between the clauses is large (equal to or more than a predetermined value), the syntax is determined to be correct in step S310, and the CPU 30 advances the process to step S320. hand,
C on the dialog box DB1 of "Confirm advanced correction"
Processing for displaying on the RT display 12 is performed.

In FIG. 10, the dialog box DB1 for "Confirm advanced correction" is displayed in the application window WD.
It is an explanatory view shown with. FIG. 11 is an enlarged view of the dialog box DB1. As shown in FIG. 11, this dialog box DB1 asks "Do you want to modify this part?" Along with the character data before modification (character CmX before modification) and the character data after modification (replacement character C2). Message is described. The operator looks at the dialog box DB1 together with the text data displayed in the recognition result display window F2 to determine whether the correction of the recognition result is correct (whether the correction can be accepted). When it is determined that the worker accepts, the operator clicks the "Yes" button BT21 of the dialog box DB1. On the other hand, when the worker determines that the correction cannot be accepted, the “No” button BT22 in the dialog box DB1.
Click.

Returning to FIG. 7, when the dialog box DB1 of "Confirm advanced correction" is displayed in step S320, the CPU 30 then reads the output signal from the mouse 20 and the operation of the operator is " It is determined which of the "Yes" button BT21 and the "No" button BT22 has been clicked (step S330). Here, if it is determined that the “No” button BT22 is clicked, the process proceeds to step S340,
The replacement character C2 replaced in step S290 is returned to the character before replacement, that is, the uncorrected character CmX stored in step S280. After execution of step S340, CPU3
0 increments the variable m by 1 (step S350).

On the other hand, if it is determined in step S330 that the "Yes" button BT21 has been clicked, step S340 is executed without executing step S340.
Proceed to. That is, when the operator determines that the correction of the recognition result is correct, the CPU 30 determines that the replacement with the character C2 performed in step S290 is valid, and immediately proceeds to step S350. Proceed.

After execution of step S350, the CPU 30
Returns the processing to step S260 of FIG. 6 and performs the processing of steps S260 to S350 on the automatic correction target character (mth character) Cm determined from the new variable m incremented by the value 1. Note that a negative determination is made in step S275, that is, the automatic correction target character Cm
When it is determined that the image data corresponding to is not similar to the character image data PD, the CPU 30 advances the process directly to step S350 and increments the variable m by the value 1.

According to the recognized character string correction routine having the above-described structure, first, by the process of step S210, an arbitrary character based on the operation command of the operator is selected from the text data of the recognition result, for example, in the example of FIG. , The character C1 of "grass" in "the unusual row of grass" on the second line is selected. Then, by the process of step S220, the selected character C1 is replaced with the replacement character C2 based on the operator's operation command. Then, steps S250 to S2
By the processing of 75, "da" following the letter C1 of "grass" above
Character C1 'of the selected character C1 (selected character for storage C
It is determined whether or not it is the same as or similar to 1X). Since it is determined that the character C1 'of "Da" is not the same as the character C1X of "Grass" and that the respective character image data are not similar, special processing is performed for the character C1' of "Da". The processing target is shifted to the subsequent character without performing the processing. By repeating the processing of steps S260 to S350, it is determined whether or not the characters match the character C1X of "grass" toward the rear of the text data or the respective character image data are similar.

In the example of FIG. 9, when the object to be processed moves to the character C1 ″ of “grass” in the “11th grass” on the 14th line in the text data, a match with the character C1X of “grass” occurs. Can be seen. The character is once replaced with the replacement character C2 as the automatic correction target character Cm. Then, the syntax before and after the replacement character C2 is analyzed, and when it is determined that the syntax is incorrect, the automatic correction target character Cm replaced by the replacement character C2 is returned to the state before the replacement. . On the other hand, when the syntax analysis determines that the syntax is correct, as shown in FIG.
Dialog box DB1 is displayed to prompt the operator to decide whether or not to accept the result of the replacement, and then the replacement of the character Cm with the replacement character C2 is confirmed as valid.

Therefore, according to the first embodiment, when batch correction is performed on the text data as the recognition result, the correction is determined to be valid only when the correction has the correct syntax. . Therefore, there is an effect that the accuracy of character correction can be improved.
Also, according to this first embodiment, if the modification is not the correct syntax, the modification is automatically reverted. Therefore, the accuracy of character correction can be further improved.

Further, according to the first embodiment, if the syntax is determined to be correct by the syntax analysis, the dialog box DB1 of "Confirm advanced correction" is displayed, and the operator replaces it. You can ask if you can accept the results of. Therefore, the accuracy of character correction can be further improved.

Further description of the first embodiment will be continued. In the first embodiment, it is explained that the arbitrary character selected in step S210 is one character exemplified by "grass", but this arbitrary character does not necessarily have to be one character. It is possible to select multiple consecutive characters. On the second line in the text data displayed in the recognition result window F2 shown in FIG. 9,
It contains two consecutive characters such as "day" and "temple",
In step S210, the two characters "day" and "temple" can be selected. After this, the worker selects the toolbar BR of the application window WD.
The combining button BT12 (see FIG. 8) provided in No. 2 is clicked using the mouse 20. In this case, in step S220, the operation signal from the mouse 20 is fetched and the continuous plural The selected character C1 is replaced with the replacement character C2 with the single character in which the characters are combined, that is, the character "hour" in the above example as the replacement character C2.

According to this configuration, when a plurality of characters are erroneously recognized from the image data representing one character, for example,
When the characters of "day" and "temple" are recognized separately from the image data of "hour", they can be corrected to the original one character. In addition, as an example of the character separated into a plurality of characters, "uta", "bayashi", "nephew", "luster",
There are various things such as "mosquitoes".

Further, as a result of character recognition, two characters may be recognized as one character. For example, two half-width characters such as 00 may be erroneously recognized as one full-width character such as}. In this case, the arbitrary character selected in step S210 is set to the character of}. After this, the operator performs an operation of clicking the separation button BT13 (see FIG. 8) provided on the toolbar BR2 of the application window WD with the mouse 20, but in this case, in step S220, the mouse 20 is clicked. By taking in the operation signal from, the selected character C1 is replaced with the replacement character C2 with a plurality of characters, that is, "00" in the above example as the replacement character C2.

According to this structure, when one character is erroneously recognized from the image data representing a plurality of characters, it can be corrected to the original plurality of characters.

The second embodiment of the present invention will be described below. The second embodiment has the same hardware configuration as that of the first embodiment. The software configuration is almost the same, and only a part of the recognition character string correction routine is different. FIG. 12 is a flow chart showing the latter half of the recognized character string correction routine in the second embodiment. The first half of the recognized character string correction routine in the second embodiment is the same as that in the first embodiment and is shown in FIG. As shown in FIG. 12, the latter half of the recognized character string correction routine in the second embodiment includes steps S270 to S310 and step S340 which have the same processing contents as those in the first embodiment.

If it is determined in step S310 that the syntax is correct based on the evaluation result of the syntax analysis,
The CPU 30 highlights the replacement character C2 replaced in step S290 on the text data of the recognition result (step S400). The highlight display here means that the background portion of the replacement character C2 is black,
The replacement character C2 is displayed in white. As a result of step S400, the replacement character C2 is confirmed as valid. After executing step S400, the CPU 30 advances the process to step S350.

According to the recognition character string correction routine having the above structure, all the characters Cm to be automatically corrected are automatically replaced by the replacement character C2.
When the syntax before and after the replacement character C2 is analyzed to be correct, the replacement result is highlighted and confirmed as a valid one. Figure 13
It is explanatory drawing which shows an example of the application window WD after completion | finish of a recognition character string correction routine. FIG. 14 is an enlarged view around the recognition result window of FIG. As shown in these figures, all replacement characters C2 that have been automatically replaced and parsed as syntactically correct are highlighted.

According to the second embodiment constructed as described above, the accuracy of character correction can be improved, as in the first embodiment. Particularly, in the second embodiment, as in the first embodiment, a dialog box DB1 of "confirmation of advance correction" is displayed every time the correction is performed, and the operator can accept the result of the replacement. There is no such thing as inquiring about what can be done. The correction target character is corrected immediately,
The corrected portion can be easily known by highlighting it on the screen of the CRT display 12.

The above highlighted display shows the replacement character C.
2 may be any display as long as it can be distinguished from other characters, and may be underlined or bold.

Another embodiment of the present invention will be described below. In the first and second embodiments, the operator uses the replacement character C
As an operation of inputting 2, a correct "car" character, which replaces the "erroneously recognized" grass ", is clicked using the mouse 20 from the next candidate group displayed in the candidate character field F3. In addition to this, when there is no correct character in the next candidate group displayed in the candidate character field F3, the character dictionary used for character recognition is a character dictionary with a large number of records (or biased to a specialized field). A character dictionary) can be converted to a character re-recognition process. The worker recognizes the re-recognition button BT14 provided on the toolbar BR2 of the application window WD.
By performing an operation of clicking (see FIG. 8) with the mouse 20, it is possible to instruct the execution of the re-recognition process. When inputting the replacement character C2, the CPU 30 takes in an operation signal for clicking the re-recognition button BT14, executes the re-recognition process, and recognizes the candidate obtained by the re-recognition as the first recognition process. Similarly, the candidate character field F3 is displayed, and the candidate character designated by the candidate group displayed in the candidate character field F3 is determined as the replacement character C2. According to this embodiment, the accuracy of character correction can be further improved.

In the first and second embodiments, the automatic correction target character Cm is replaced with the replacement character C2 on the text data which is the recognition result, and then the replaced character is before and after (before and after on the text data). Was parsed, and based on the result of parsing the syntax, the result of the substitution was confirmed as valid. Instead of this, character string data in a predetermined range before and after the automatic correction target character Cm is extracted from the text data as the recognition result, and the character string data is once transferred to another area of the RAM, and this transfer is performed. When the above-mentioned automatic correction target character Cm in the generated character string data is replaced with the replacement character C2, and then the syntax of the character string data is analyzed, and the result of the analysis of this syntax is determined to be the correct syntax. In addition, it can be configured such that the character string data is returned to the text data on the assumption that the replacement result is valid. With this configuration, the accuracy of character correction can be improved as in the first and second embodiments.

Although one embodiment of the present invention has been described in detail above, the present invention is not limited to such an embodiment and can be carried out in various modes without departing from the gist of the present invention. Of course, you can.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of hardware of a computer system to which a first embodiment of the present invention is applied.

FIG. 2 is a block diagram showing a state of processing according to the character recognition software 50 executed by the computer main body 16.

FIG. 3 is a flowchart showing a control processing routine according to the character recognition software 50.

FIG. 4 is an explanatory diagram showing an application window WD in an initial state.

FIG. 5 is an explanatory diagram showing an example of an application window WD after completion of character string recognition processing.

FIG. 6 is a flowchart showing a first half portion of a recognized character string correction routine executed by CPU 30.

FIG. 7 is a flowchart showing the latter half of the recognized character string correction routine executed by the CPU 30.

FIG. 8 is an explanatory diagram showing an example of an application window WD after designating a selected character C1 to be modified in step S210.

9 is an enlarged view around the recognition result window F2 in FIG.

FIG. 10 is an explanatory diagram showing a dialog box DB1 for “confirm advance correction” together with an application window WD.

FIG. 11 is an enlarged view of a dialog box DB1.

FIG. 12 is a flow chart showing the latter half of the recognized character string correction routine in the second embodiment of the present invention.

FIG. 13 is an explanatory diagram showing an example of an application window WD after completion of a recognized character string correction routine.

FIG. 14 is an enlarged view around the recognition result window in FIG.

[Explanation of symbols]

12 ... CRT display 14 ... Image scanner 16 ... Computer body 18 ... Keyboard 20 ... Mouse 30 ... CPU 31 ... ROM 32 ... RAM 33 ... Display image memory 34 ... Mouse interface 35 ... Keyboard interface 36 ... CDC 37 ... HDC 38 ... CRTC 40 ... I / O interface 41 ... I / O port 42 ... Hard disk drive 44 ... Modem 46 ... Public telephone line 47 ... server 50 ... Character recognition software 51 ... Scan image capturing unit 52 ... Character string recognition module 53 ... Recognized character string correction module 53a ... Character selection part for manual correction 53b ... Manual correction unit 53c ... Character selection part for automatic correction 53d ... Automatic correction unit 53e ... syntax analysis unit 53f ... Correction result confirmation section P ... manuscript WD ... application window F1 ... Original image window F2 ... Recognition result window F3 ... Candidate character field BT11 ... Batch correction button BT12 ... Combine button BT13 ... Separation button BT14 ... Re-recognition button C1 ... Selected character C1X ... Selection character for storage C2 ... Substitution character Cm ... Characters for automatic correction CmX ... Character before correction DB1 ... Dialog box for "Confirmation of advanced correction"

Claims (16)

[Claims] 1. A recognition character string correction method for correcting a character string obtained by a character recognition process for inputting a document as image information and recognizing characters, comprising: (a) an operator's operation command. , A step of selecting an arbitrary character from the character string, (b) a step of replacing the selected arbitrary character with a predetermined character in accordance with an operation command of an operator, and (c) the character string From among these, a step of sequentially selecting a character that substantially matches the arbitrary character selected in step (a) as an automatic correction target character, and (d) a step of replacing the automatic correction target character with the predetermined character And (e) a step of analyzing the syntax before and after the character replaced in the step (d), and (f) a step of (d) based on the result of the syntax analysis of the step (e). Yo A method for correcting a recognized character string, which comprises the step of confirming the result of the replacement as valid. 2. The recognition character string correction method according to claim 1, wherein when (g) the result of the syntax analysis in step (f) indicates that the syntax is not correct, the automatic correction target A method for correcting a recognized character string, which comprises a step of returning a character to a state before the replacement in the step (d). 3. The method for correcting a recognized character string according to claim 1 or 2, wherein (h) a replacement character determined to be valid in step (f) is identified with respect to another character. A method for correcting a recognized character string, which comprises a step of displaying the character string. 4. The recognition character string correction method according to claim 1, wherein in step (f), the result of the syntax analysis by (f-1) step (f) is a correct syntax. If it indicates that the step (d)
It is determined that the result of the replacement is accepted based on the step of inquiring the operator whether to accept the result of the replacement by (f-2), and (f-2) the operation command from the operator in response to the inquiry in the step (f-1). At this time, the recognition character string correction method comprising the step of determining the result of the replacement in step (d) as valid. 5. When the arbitrary character selected in step (a) is composed of a plurality of consecutive characters,
The replacement according to the step (b) is a structure for replacing the plurality of consecutive characters with one predetermined character.
5. The method for correcting a recognized character string according to any one of 1 to 4. 6. The replacement according to the step (b) is a structure in which one character is replaced with a plurality of characters.
A method for correcting a recognized character string according to any one of 1. 7. A recognition character string correction device for correcting a character string obtained by a character recognition process for recognizing a character by inputting an original as image information, wherein the character is corrected based on an operation command from an operator. Manual correction target character selection means for selecting an arbitrary character from a string, manual correction means for replacing the selected arbitrary character with a predetermined character according to an operator's operation command, and among the character strings , An automatic correction target character selection unit that sequentially selects, as an automatic correction target character, a character that substantially matches an arbitrary character selected by the manual correction target character selection unit, and replaces the automatic correction target character with the predetermined character Based on the result of the syntax analysis by the syntactic analysis unit, the automatic correction unit for analyzing the syntax before and after the character replaced by the automatic correction unit, A recognition character string correction device comprising correction result confirmation means for confirming that the result of replacement by the correction means is valid. 8. The recognition character string correction device according to claim 7, wherein when the result of the syntax analysis by the correction result determination means indicates that the syntax is not correct, the automatic correction target character is A recognition character string correction device comprising means for returning to a state before replacement by automatic correction means. 9. The recognition character string correction device according to claim 7 or 8, wherein the correction result confirmation means indicates that the result of the syntax analysis by the replacement result confirmation means is a correct syntax. At some time, whether to accept the result of the replacement by the automatic correction means, when it is determined to accept the result of the replacement based on an operation command from the operator and a means for inquiring to the worker,
A recognition character string correction device, comprising: means for deciding that the result of replacement by the automatic correction means is valid. 10. A computer program for executing a process of correcting a character string obtained by a character recognition process for recognizing a character by inputting a document as image information, comprising: (a) an operation command from an operator. A function of selecting an arbitrary character from the character string, (b) a function of replacing the selected arbitrary character with a predetermined character according to an operator's operation command, and (c) the character string From among the above, a function of sequentially selecting, as an automatic correction target character, a character that substantially matches an arbitrary character selected by the function (a), and (d) replacing the automatic correction target character with the predetermined character. A function, (e) a function of analyzing the syntax before and after the character replaced by the function (d), and (f) the function (d) based on the result of the syntax analysis by the function (e). Positive result of replacement by A computer program that causes a computer to realize a function that is determined to be appropriate. 11. The computer program according to claim 10, wherein when (g) the result of the syntax analysis by the function (f) indicates that the syntax is not correct, the automatic correction target character is A computer program for causing a computer to realize a function of returning to a state before replacement by the function (d). 12. The computer program according to claim 10, wherein (h) the replacement character that has been determined to be valid by the function (f) is displayed in a distinguishable manner from other characters. A computer program that causes a computer to realize the function to perform. 13. The computer program according to claim 10, wherein the function (f) indicates that (f-1) the result of the syntax analysis by the function (f) is a correct syntax. If it is, the function (d) is based on the function of inquiring the operator whether to accept the replacement result by the function (d), and (f-2) the operation command from the operator for the inquiry of the function (f-1). A computer program having a function of determining the result of the replacement by the function (d) as valid when it is determined to accept the result of the replacement. 14. When the arbitrary character selected by the function (a) is composed of a plurality of consecutive characters, the replacement by the function (b) replaces the plurality of consecutive characters by one predetermined character. The computer program according to any one of claims 10 to 13, wherein the computer program is configured to substitute characters. 15. The replacement according to the function (b) is configured to replace one character with a plurality of characters.
The computer program according to any one of 3 above. 16. A computer-readable recording medium in which the computer program according to claim 10 is recorded.