JPH1185899A - Character reader, its method and record medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce analysis time of a character string. SOLUTION: A user instructs the number of the character strings of the same kinds with the character string written on an original through an input device 17. A CPU 13 counts the number of analysis processing times executed in the analysis of a character recognizing result for every kind, excludes the kind of analysis processing reaching an indicated number of times from the analysis processing of the character string after then and avoids the analysis processing of its kind.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a character reading device for reading a document image on which a character string is described, and analyzing the character strings of the character string, a reading method thereof, and a recording medium.

[0002]

2. Description of the Related Art A character reading device reads a document image on which a character string is described by a scanner, recognizes the read character image, converts the read character image into a character code, and converts the read character image into a character code. Using such a character reading device, it has become possible to read personal related information such as a name and an address written on a health insurance card or a driver's license, and to register the read personal related information in a database. .

[0003] As is well known, a database is one in which one record is composed of data of a plurality of items, and a plurality of records are accumulated and stored. Therefore, it is necessary to classify which item the character code read and converted by the character reading device corresponds to.

The following method is known as a method of indicating the correspondence.

A) The document to be read is limited to a document in which the position of the character string is fixed.

[0006] The user instructs the character reading device to describe the position of the character string on the document, that is, the position of the image reading area and the type of record item corresponding to this position. The character processing device reads characters in a plurality of designated areas, converts them into a character code string, and creates a record for database registration.

B) The description content of a character string on a manuscript, for example, individual information such as a name, a company name, an address, a telephone number, and a zip code such as a business card is a proper noun or a number having a fixed number of digits. The originals to be read are limited to originals.

The character reading device uses a proper noun dictionary, a personal name dictionary, a place name dictionary, and the like to analyze whether the read character string corresponds to an item name such as a name or a company name, and based on the analysis result. Create records to be registered in the database.

[0009]

In the reading condition setting method of the above a). Since the user must specify the reading area and the attribute of the character string described in the reading area, there is a problem that the user's instruction operation becomes more complicated as the reading area increases.

On the other hand, in the above-described reading condition setting method of b), the attribute of the information described in the document is automatically determined, so that the user's instruction operation effort is reduced. However, since the character string is analyzed, the character string is analyzed. There is a problem that attribute analysis takes longer as the number of attributes increases.

Therefore, the object of the present invention is as follows.
An object of the present invention is to provide a character reading device, a reading method, and a recording medium which solve the above-mentioned problem b) and do not complicate a user's instruction operation.

[0012]

In order to achieve the above object, a first aspect of the present invention is to read an original image on which a plurality of character strings are described, and recognize the plurality of read character strings by character recognition. By performing an analysis process on the result of the character recognition, a character reading device that identifies the type of the semantic content of the character string and classifies the result of the character recognition based on the identification result. Instruction means for instructing the type of the semantic content to be targeted, the number of character strings on the document image for each type of the semantic content, and each time the type of the semantic content of the plurality of character strings is determined, the Counting means for counting the number of determinations for each type, and control means for excluding, from the type of analysis processing to be performed, the type in which the result of the counting reaches the number instructed by the instruction means. Was And wherein the door.

[0013] According to a second aspect of the present invention, in the character reading apparatus according to the first aspect, the analysis processing uses a plurality of analysis rules that define characteristics of a character string for each type of the semantic content. It is characterized by being.

According to a third aspect of the present invention, in the character reading apparatus according to the first aspect, the analysis processing is a dictionary in which a description of the character string is described according to the meaning and content of the character string.

According to a fourth aspect of the present invention, in the character reading apparatus according to the first aspect, when a plurality of types of identification results are obtained for one character string recognized as a character, a predetermined selection criterion is set. And selecting means for selecting one of the plurality of types of identification results as a final identification result based on the above.

According to a fifth aspect of the present invention, an original image on which a plurality of character strings are described is read, the read plurality of character strings are recognized as characters, and the result of the character recognition is analyzed. , Identifies the type of semantic content of the string,
In the character reading method of the character reading device that classifies the result of the character recognition based on the identification result, the type of the semantic content to be subjected to the analysis processing and the number of character strings on the document image are defined by the The character reading device instructs the character reading device for each type, and the character reading device counts the number of determinations for each type each time the type of the meaning content of the plurality of character strings is determined. Are excluded from the types of analysis processing to be executed.

According to a sixth aspect of the present invention, in the character reading method of the character reading apparatus according to the fifth aspect, the analysis process uses a plurality of analysis rules that define the characteristics of a character string for each type of the semantic content. Lexical analysis processing.

According to a seventh aspect of the present invention, in the character reading method of the character reading apparatus according to the fifth aspect, the analysis processing is a dictionary in which a description of the character string is described according to the meaning of the character string. And

According to an eighth aspect of the present invention, in the character reading method of the character reading apparatus according to the fifth aspect, the character processing device can obtain a plurality of types of identification results for one character string recognized as a character. In this case, one of the plurality of types of identification results is selected as a final identification result based on a predetermined selection criterion.

According to a ninth aspect of the present invention, an original image on which a plurality of character strings are described is read, the read plurality of character strings are recognized as characters, and the result of the character recognition is analyzed. , Identifies the type of semantic content of the string,
In a recording medium recording the processing program to execute a processing program that defines a series of processing for classifying the result of the character recognition based on the identification result by a computer in a character reading device, the processing program includes: A process of instructing the character reading device for the type of the meaning content to be analyzed and the number of character strings on the document image for each type of the meaning content, and the meaning content of the plurality of character strings Each time the type is determined, a processing procedure of counting the number of determinations for each type, and a type in which the result of the counting reaches the specified number is excluded from the types of analysis processing to be performed. And a processing procedure.

According to a tenth aspect of the present invention, in the recording medium according to the ninth aspect, the analysis processing is a lexical analysis processing using a plurality of analysis rules defining the characteristics of a character string for each type of the semantic content. There is a feature.

According to an eleventh aspect of the present invention, in the recording medium according to the ninth aspect, the analysis processing is a dictionary in which a description of the character string is described according to the meaning of the character string.

According to a twelfth aspect of the present invention, in the recording medium according to the ninth aspect, the processing program is configured such that when a plurality of types of identification results are obtained for one character string whose character is recognized, The method further includes a processing procedure of selecting one of the plurality of types of identification results as a final identification result based on the determined selection criterion.

[0024]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows a system configuration of a character reading apparatus to which the present invention is applied. In FIG. 1, a general-purpose personal computer can be used as the character reading device 10. In the main body of the character reading device 10, the following circuits are connected to a bus. Input / output interface (I /
O) 11 is connected to the scanner 20, inputs a document image read by the scanner 20, and delivers it to the CPU 13.

The system memory 12 stores a system program for controlling the system such as an operating system, data, an image for display, operation data, and the like. CP
U13 executes a character reading program described below,
A record for database registration is created based on the document image input from the scanner 20. Also, the operation of the entire system is controlled according to the system program. The communication interface 14 is connected to a LAN (local area network) or the like, and communicates with another computer.

The hard disk storage (HDD) 15 stores a system program, a character reading program, and a window screen or the like for a user to input information in a character reading process.

Floppy disk drive (FDD) 1
Numeral 6 receives a floppy disk and reads and writes information from and to the floppy disk. The character reading program and related data according to the present invention are installed in the HDD 15 from a floppy disk. The input device 17 has a keyboard and a mouse, and inputs information. The display 18 displays a read original, a graphical interface for inputting read conditions, an attribute analysis result, and the like.

The scanner 20 is a CCD (solid-state image sensor)
And outputs a read image to the character reading device 10 in the form of a digital signal.

Prior to the description of the operation of the character reading apparatus having such a system configuration, a screen (graphical interface) used for setting reading conditions and instructing reading will be described.

In this embodiment, originals having different formats such as a health insurance card as shown in FIG. 2 (abbreviated as an insurance card) and postcards shown in FIG. 4 can be read.
FIG. 3 is a screen for setting reading conditions for the insurance card of FIG. Reference numeral 101 denotes a format name given to the reading condition, and the reading condition is stored and displayed based on the format name.

Reference numeral 102 denotes a column for setting the type of character string described in the document and the number of duplications. The user designates the type and content of the character string described in the item document by attaching the check mark by operating the mouse. In addition, the user inputs a number from the keyboard and indicates the number of character strings of the same type. In the example of FIG. 3, there is one set of character strings for personal names, phonetic characters, genders, and organization names, two sets of identification codes, two sets of character strings for dates, and five sets of character strings for place names and addresses.
It shows that there is a pair.

In the present embodiment, all types of character strings that can be analyzed are displayed as guidance, and the user selects the type of character string to be analyzed on the screen shown in FIG. Further, the number of character strings described in the document is selected for each type on the screen of FIG. The user merely sets the reading conditions, and thereafter, the character reading apparatus automatically analyzes the type of the read character string and creates a record having the specified number of items.

The reading conditions set by the user for reading the postcard of FIG. 4 are shown in FIG. 5 for reference. FIG. 6 shows a screen 201 for setting a mask color when reading a document. An area 205 displays all selectable colors. The color of the sample is displayed in the area 206, and the user specifies the mask color by specifying the sample with the mouse. A setting clear button 202 is a button for instructing to return current settings to initial settings. 20
Reference numeral 3 denotes an OK button, which is a button for confirming the currently set color. A cancel button 204 is a button for instructing to cancel the mask color setting mode in FIG. 6 and delete the window screen. Such a window screen is called a color picker and is often used in image processing software.

FIG. 7 shows a second example of a screen for setting a mask color. In FIG. 7, reference numeral 301 denotes an area for displaying a document image read at a low resolution. By clicking a specific position in this area with a mouse, a color corresponding to the position is set as a mask color. Reference numeral 302 denotes a button for instructing reading of an image, and 303 denotes a button for instructing to clear the set mask color. An area 304 displays the selected mask color. Reference numeral 305 denotes an OK button for instructing confirmation of the current mask color setting, and reference numeral 306 denotes a button for instructing cancellation of the mask color setting.

In this embodiment, the read image is masked with the mask color set in this way, and the color background and characters that are not used for character recognition of a specific color are erased.

FIG. 8 shows a window screen for displaying the result of character recognition and the result of analysis of the recognized character string. 401
Is a column for inputting a format name, and 402 is a button for instructing new creation of a file. A button 403 indicates an editing mode, and a button 404 indicates a reading execution. An area 405 displays the analysis result of the type of character.

When the user operates the read execution button 404 after instructing the new creation button 402 with the mouse, the character reading apparatus instructs the scanner 20 to read an image, and then performs character recognition of the read image.
Analyzes the type contents. The analysis result is displayed in the display area 405 in association with the type content (so-called attribute) of the character string.

This analysis result is stored in a file in a format that can be registered in the database, for example, in various formats such as a text with a comma.

Next, the type contents of the attribute of the character string that can be analyzed in the present embodiment and the analysis rules thereof will be described with reference to FIG. FIG. 9 shows the types of attributes that can be analyzed and the analysis rules.

A) Personal Name A personal name dictionary is used to analyze whether or not the character string to be analyzed is a personal name. Personal name dictionaries include notation of first and last names,
In this case, the dictionary is a dictionary in which a plurality of character code strings are described.
Stored in DD15. If the same character code string (character code string to be analyzed) obtained as a result of character recognition by searching the personal name dictionary is described in the personal name dictionary, the character code string to be analyzed is a person name. Is determined.

B) Furigana Furigana determines whether all character strings to be analyzed are composed of hiragana or katakana. JIS (Japanese Industrial Standards) hiragana character codes and katakana character codes are established so that they fall within a specific code range. Therefore, do the individual character codes of the character strings to be analyzed fall within the above code ranges? By determining whether or not the character string to be analyzed is a phonetic character.

C) Title The title is determined based on whether or not the same character string as the character string to be analyzed is described in a dedicated title dictionary. The title dictionary is a dictionary in which character code strings representing titles such as section chiefs and section managers are described, and is stored in the HDD 15.

D) Age When the character code string is composed of character codes related to numbers, and the value indicated by the character code is within a range corresponding to the age, for example, 0 or more and 120 or less. Is determined to be the age. Since the computer has a function of converting a character code into a numerical code, the character code may be numerically converted using this function. Since an unsuccessful response is returned from the computer when trying to convert a kanji code or a hiragana character code to a numerical value, in this case, it can be determined that the character code string to be analyzed is not age. e) Gender For gender, the character code string to be analyzed is "male",
"Woman", "M", "F", "Male", "Femal"
e ”is determined. If the character code string to be analyzed matches the specific character code string, it is determined that the character code string to be analyzed is gender.

For a place name / address, if the same character string as the character string to be analyzed is described in the place name dictionary, the character string to be analyzed is determined to be a place name / address. The place name dictionary is a dictionary in which character code strings representing place names are described.
5 is stored.

F) Postal code For the postal code, the character code string to be analyzed is composed of only numbers or a number and a "-" sign, and the range of numerical values is the range of values assigned to postal codes ( Or the range of the number of digits), the character string to be analyzed is determined to be a postal code.

G) Group name, business name, organization / department name When the same character string as the character string to be analyzed is described in the group name dictionary, business name dictionary, or organization / department name dictionary, the analysis target Is determined as an organization name, a business name, or an organization / department name according to the dictionary described.

H) Telephone number Character strings to be analyzed are "telephone", "TE"
A character string composed of a character string of numbers accompanied by symbols "L", "("")" and "-" is determined to be a telephone number. In order to perform more accurate analysis, it may be determined whether the number of digits of the number matches the number of digits of the telephone number.

G) Identification code If the character string to be analyzed contains any of the character codes "No.", "No.", and "No", the type is determined to be an identification code. For this purpose, the first character code of the character code string to be analyzed is extracted, and the extracted character code is
Performs a match comparison to see if it is the character code of "go" In the case of a mismatch determination, the first two character codes are extracted from the character code string to be analyzed, and are compared with the character code string of (No) above. In any of the above-described matching determinations,
When a match determination result is obtained, the character code string to be analyzed is identified by an identification code.

H) Date In the character string to be analyzed, "Heisei", "Showa", "/" and "Ja"
n. If a numeric character value is included in the range of the numeric value used in the date, the character string to be analyzed is determined to be a date.

I) Other Numerical Values When the character code to be analyzed is composed of numeric codes, the type is determined to be other numeric values.

J) When two or more types of character strings to be analyzed are obtained, the type is finally determined based on a predetermined selection criterion described later.

In accordance with the above-described analysis rules and the dictionary, the character reading device analyzes the character string obtained as a result of the character recognition and determines the type of the character string.

Next, a character reading process according to the present invention will be described with reference to FIG.
This will be described with reference to FIGS. FIG. 10 shows the contents of the main processing of the character reading processing, and FIGS. 11 to 15 show details of the individual processing in the main processing.

The processing programs shown in FIGS. 10 to 15 are described in a program language that can be executed by the CPU 13.
Stored in DD15. When the activation of the character reading program is instructed, the character reading program is loaded into the system memory 12 and executed by the CPU 13.

K) Setting of reading conditions The user designates a reading condition setting mode with a mouse or the like on the menu screen. As a result, a window screen shown in FIG. 3 is displayed. The user specifies the type of character string described in the document to be read and the number of character strings of the same type. After the instruction, input the format name and instruct to save. The CPU 13 stores the read condition file in the form of a table as shown in FIG. 16B in the HDD 15 (steps S10 → S20 → S30). It goes without saying that the setting relating to the mask processing is also performed during the reading condition setting processing in step S30.

When reading a document, the user sets a document as shown in FIG. The user calls the window screen of FIG. 8 on the menu screen. The name of the read condition file stored earlier is input to the format name entry field 401 from the keyboard, and the read execution button 404 is operated with the mouse. In response to this operation, the execution procedure of the CPU 13 proceeds to steps S10 → S20 → S100 in FIG.

The CPU 13 operates in the same manner as in the prior art.
The scanner driver 20 is controlled by executing driver software for driving to read a document image. The read color document image is input to the character reading device via the I / O 11, and is temporarily stored in the work area in the system memory 12 by the CPU 13. Thereafter, image processing for creating an image suitable for character recognition is performed (step 1).
00).

FIG. 11 shows the detailed procedure of step S100. In the processing procedure of FIG. 11, after a color document image read by the scanner 20 is stored in the system memory 12 (step S1001), mask processing and binarization processing are performed (step S1002).

By the binarization process, each pixel in the character portion of the original image data is converted into a bit 1 value, and each pixel in the background portion of the original image data is converted into a bit 0 value.

The CPU 13 detects the character string area and other blank areas by examining the stroke distribution of the image data after the binarization processing, that is, the distribution of bit 1 (step S1003).

Next, the inclination of the image generated when the document is set obliquely is corrected, and the ruled line image is removed (step S1005).

Returning to FIG. 10, the CPU 13 blocks the character string area detected by the above-described processing by a conventional method. In the present embodiment, blocking means a process of detecting a character string in which characters are continuous and automatically detecting the position of a block (rectangle) circumscribing the character string (step S).
200). The image in this block is the target of character recognition. Block numbers are assigned to the detected blocks in the order of detection, and a table A describing the block numbers and the coordinate positions of the blocks is created in the work area in the system memory 12 as shown in FIG. 12. Step S
2001 to S2003). In addition, a table B for storing the character recognition result corresponding to the block number as shown in FIG. 16B is also created in the work area.

The CPU 13 performs character recognition on each character image in the block area and converts it into a character code (step S300). The character recognition result is stored in the table (see FIG. 15B) in association with the block number. FIG. 13 shows an example of a detailed procedure of the character recognition process. In this detailed procedure, the density histogram of the image in the block is examined, and if there is no change in density (gradation) in the image, the image is determined to be a character image and character recognition is performed (steps S3001 → S3002 → S3003 →). S3004
To S3006). On the other hand, if the image has a density change, the image is dirty or not a character image.
Character recognition is not performed to prevent erroneous recognition, and the image itself is stored in the recognition result storage column corresponding to the block number in table B (steps S3001 → S3002 → S3).
003 → S3004 → S3007)).

When the character recognition of all blocks is performed, the CPU 13 analyzes the attribute of the character recognition result, that is, analyzes the type of the meaning content according to the present invention (step S40).
0). Details of the type analysis are shown in FIGS.

The CPU 13 sets a line pointer on the first line of the table B in FIG. 16 and takes out a character string described as a result of character recognition on the first line into the work area of the system memory 12.

The extracted character string is collated with each of the above-described analysis rules, and a matching attribute name is described in a candidate attribute description column of Table B of FIG. In addition, an analysis program is prepared for each analysis rule, one analysis program is executed, and when it is determined that the analysis rule does not match, the next analysis program is executed.
The lexical analysis of the character string is performed (step S4001 → S
4002 → S4003 → S4004 → S4006).

When the health insurance card shown in FIG. 2 is read, the character recognition result described on the first line of table B is “issued on March 1, 1996”, so that this character string When the analysis program relating to the date is executed, it is determined that the analysis rule matches the date analysis rule, and “date” is given as a candidate attribute (see FIG. 16B).

The CPU 13 checks all the analysis rules shown in the current table C of FIG. 16 even after the determination of the matching of the specific analysis rule is obtained. Is written in the candidate attribute description column of the table B (steps S4002 to S4004 → S4
006 → loop processing of S4002). After completion of the lexical analysis, a dictionary analysis described later is executed (step S401).
0) After the dictionary analysis is completed, the attribute is determined from a plurality of obtained candidate attributes based on a predetermined attribute selection criterion (described later) (S4011).

On the other hand, when the selected lexical analysis program is executed and it is determined that the lexical analysis program does not match the analysis rule (NO in step S4004), another analytic program is selected and lexical analysis is performed. (Step 4004)
→ S4006 → S4002). After the lexical analysis, a dictionary analysis process is performed (step S4002 → S4010).

Since only “date” can be obtained as a candidate attribute relating to the recognized character string on the first line of the original in FIG. 2, “date” is automatically determined as the final attribute, and FIG. As shown, the determination result is entered in the attribute description column. Also,
The set value of the table C relating to the date is updated from the current “3” to “2” (decrement, equivalent to counting the number of identification results obtained for the date) (FIGS. 16C and 17B). reference). Here, it should be noted that when the set value of the attribute is 1 or more, lexical analysis or dictionary analysis is performed on the attribute, and when the set value of the attribute becomes 0 (a specific type of The counting result of the identification result is
(Equivalent to reaching the set value), the attribute is excluded from the analysis. Conventionally, the lexical analysis rules used for analysis have a fixed number of dictionaries, whereas in the present embodiment, the number of analysis rules and dictionaries used is reduced, so the analysis processing time is greatly reduced. .

When the analysis of the recognized character string on the first line of the table C in FIG. 16 is completed, the CPU 13 selects the recognized character string on the second line as an object to be analyzed, and executes the above-described analysis processing. The recognized character strings in the first to fourth lines of the table C can be analyzed in the lexical analysis processing of FIG. 14, but the recognized character strings in the fifth line have the attribute "person name". If no analysis result is obtained in the lexical analysis processing of FIG. 14 (YES determination in step S4002), step S40
The procedure proceeds to ten dictionary analysis processes.

FIG. 15 shows the details of the dictionary analysis process.

The CPU 13 selects the first dictionary (in this case, the personal name dictionary) shown in the table C as the dictionary to be used for analysis (step S5010). The CPU 13 takes out the recognized character string on the fifth line onto the work area of the system memory 12 and performs the following dictionary analysis.

To introduce an example of the personal name analysis, the first character of the extracted character string is extracted (step S503).
0), search the dictionary selected for the first character. For example, the name dictionary is searched, and if the same character as the first character is described as the surname, it is detected that the recognized character string is the name (step S504).
0).

If the same character as the first character is not described in the dictionary, the first character and the second character of the character string extracted next to the work area are combined, and this combined character string Look up the selected dictionary for In this way, the number of combined characters is increased until the written character string is determined (steps S5030 → S504).
0 → S5050 → S5070 → S5030 loop processing).

If there is no description even after searching all combinations of the characters of the extracted recognized character string, the selected dictionary is changed to the second dictionary, and the presence or absence of the description is confirmed (step S5080 →). S5090 → S5030 → S5
040).

As described above, when the type of the dictionary in which the recognized character string is described, that is, the attribute of the recognized character string is detected, the CPU 13 writes the detected attribute in the candidate attribute description column of the table B of FIG. S5060). Less than,
The selected dictionary is changed, and the presence or absence of the recognition character string is confirmed for the remaining dictionaries (step S5060 → S509).
0 to S5095 → S5030 → S5040).

A plurality of candidate attributes, for example, a plurality of candidate attributes such as "person name" and "place name" may be obtained by the lexical analysis and dictionary analysis described above (FIG. 18 (a)).
See the twelfth and thirteenth rows, and the contents of table C at this time are shown in FIG. 18 (b)). In such a case, a selection criterion is provided in advance, and one attribute is determined from a plurality of candidate attributes based on the selection criterion.

An example of selection criteria will be introduced. FIG. 16 (c)
The weights (or priorities) are determined in advance for the types of attributes described in the reference numeral, and the weights of a plurality of candidate attributes are compared. As a result, the candidate attribute having the largest weight can be determined as the final attribute. This weight may be fixed, or a value of the number of attributes set in the reading condition may be used. In this example, since “person name” is initially set to 1 and “place name (address)” is initially set to 5 (see FIG. 16C), a place name / address having a large attribute is determined as an attribute ( Step S4011 in FIG. 14).

As described above, when the lexical analysis and the dictionary analysis are performed on the recognized character strings of all the lines written in the table C, the CPU 13 completes the processing procedures of FIGS. Attribute (determined)
Is displayed as shown by reference numeral 405 in FIG. 8 (step S500 in FIG. 10).

The user looks at the display and, if it is necessary to correct the character, corrects the character in the same manner as the correction of the word processing document to create data for database registration (steps S500 → S600). . Finally, the created data is registered in the database as in the conventional case (step S700), and the processing procedure of FIG. 10 ends.

The following embodiment can be carried out in addition to the embodiment described above.

1) In the present embodiment, the document image unnecessary for character string analysis is deleted by mask processing. However, the area designated by the user is deleted from the document image, and the original image after the deletion processing is deleted. To perform character recognition processing.

2) In the above-described embodiment, the type and number of semantic contents are instructed by a mouse or a keyboard. However, instructions are received from another computer through communication with another computer, or from other application programs. You may receive instructions. Further, data (type, number) to be instructed for each type of document may be registered in the database, and the instruction data may be extracted from the database in accordance with the type of document to be read.

In the above example, another computer, program, or database is the instruction means of the present invention.

[0087]

As described above, claims 1 and 5,
According to the ninth aspect, the types of various types of analysis, such as a person's name and phonetic characters, and the number of types of character strings on the document are specified. For example, when the number of types of character strings of a person's name on a document image is designated as 1 and the identification of the person's name is completed, the analysis of the person's name is not performed in the analysis of other character strings thereafter. Conventionally, all types of analysis have been performed on all character strings. Therefore, the number of types of analysis to be performed is reduced, thereby shortening the analysis processing time. Further, since only the operation of designating the number and type of character strings is required, the user does not need to perform a complicated conventional design operation such as associating the position of the reading area with the designation of the attribute.

According to the second, sixth and tenth aspects of the present invention, by performing lexical analysis, it is possible to identify the type of character string such as phonetic characters, age and gender.

According to the third, seventh and eleventh aspects of the present invention, by performing analysis using a dictionary, a character string including a name, a person's name and an address can be identified.

According to the fourth, eighth, and twelfth aspects of the present invention, analysis is performed among the types of executable analysis processing, a plurality of identification results are obtained, and the final identification result is selected from the candidates. Since it is acquired, for example, erroneous analysis of a character string common to a place name and a person name can be reduced as much as possible.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a system configuration according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram illustrating an example of a document to be read.

FIG. 3 is an explanatory diagram showing setting contents of a graphical interface for setting reading conditions.

FIG. 4 is an explanatory diagram showing another example of a document to be read.

FIG. 5 is an explanatory diagram showing setting contents of a graphical interface for setting reading conditions.

FIG. 6 is an explanatory diagram showing display contents of a graphical interface for setting a mask color.

FIG. 7 is an explanatory diagram showing display contents of a graphical interface for previewing a document image.

FIG. 8 is an explanatory diagram showing display contents of an analysis result.

FIG. 9 is an explanatory diagram showing a dictionary and an analysis rule used for analyzing a character string.

FIG. 10 is a flowchart illustrating a processing procedure executed by a CPU 13;

FIG. 11 is a flowchart illustrating a processing procedure executed by a CPU 13;

FIG. 12 is a flowchart illustrating a processing procedure executed by a CPU 13;

FIG. 13 is a flowchart illustrating a processing procedure executed by a CPU 13;

FIG. 14 is a flowchart illustrating a processing procedure executed by a CPU 13;

FIG. 15 is a flowchart illustrating a processing procedure executed by a CPU 13;

FIG. 16 is an explanatory diagram showing the configuration and description contents of a table used for attribute analysis.

FIG. 17 is an explanatory diagram showing a change in description contents of tables B and C;

FIG. 18 is an explanatory diagram showing a change in description contents of tables B and C.

[Explanation of symbols]

 Reference Signs List 10 Character reading device 11 I / O 12 System memory 13 CPU 14 Communication interface 15 HDD 16 FDD 17 Input device 18 Display

Claims (12)

[Claims] 1. A document image on which a plurality of character strings are described, and the read plurality of character strings are recognized as characters.
By performing an analysis process on the result of the character recognition, a character reading device that identifies the type of the semantic content of the character string and classifies the result of the character recognition based on the identification result, Instruction means for instructing the type of the semantic content and the number of character strings on the document image for each type of the semantic content, and each time the type of the semantic content of the plurality of character strings is determined, Counting means for counting the number of times for each type, and control means for excluding, from the type of analysis processing to be executed, a type in which the result of the counting reaches the number instructed by the instruction means. Character reading device characterized by the above-mentioned. 2. The character reading device according to claim 1, wherein the analysis process is a lexical analysis process using a plurality of analysis rules that define the characteristics of a character string for each type of the semantic content. Character reading device. 3. The character reading apparatus according to claim 1, wherein the analysis processing is a dictionary that describes the notation of the character string for each meaning of the character string. 4. The character reading device according to claim 1, wherein when a plurality of types of identification results are obtained for one character string whose character has been recognized, the plurality of identification results are determined based on a predetermined selection criterion. A character reading device further comprising a selection unit for selecting one of the types of identification results as a final identification result. 5. An original image on which a plurality of character strings are described, and the read plurality of character strings are recognized as characters.
By performing an analysis process on the result of the character recognition, the type of the semantic content of the character string is identified, and the character reading method of the character reading device that classifies the result of the character recognition based on the identification result. The type of the meaning content to be analyzed and the number of character strings on the document image are instructed to the character reading device for each type of the meaning content, and the character reading device includes the plurality of character strings. Each time the type of the meaning content is determined, the number of the determination is counted for each type, and the type in which the result of the counting reaches the specified number is excluded from the types of analysis processing to be performed. A character reading method for a character reading device, comprising: 6. The character reading method according to claim 5, wherein the analysis process is a lexical analysis process using a plurality of analysis rules defining a characteristic of a character string for each type of the semantic content. A character reading method for a character reading device. 7. The character reading method according to claim 5, wherein the analyzing process is a dictionary in which the description of the character string is described for each of the meanings of the character string. How to read characters. 8. The character reading method according to claim 5, wherein the character processing device is configured to output, when a plurality of types of identification results are obtained for one character string recognized as a character, A character reading method for a character reading device, wherein one of the plurality of types of identification results is selected as a final identification result based on a predetermined selection criterion. 9. A document image on which a plurality of character strings are described, and the read plurality of character strings are recognized as characters.
By performing an analysis process on the result of the character recognition, a type of semantic content of the character string is identified, and a processing program that defines a series of processes for classifying the result of the character recognition based on the identification result is defined as a character program. In a recording medium on which the processing program is recorded for execution by a computer in a reading device, the processing program indicates a type of a meaning content to be analyzed and a number of character strings on the document image. A processing procedure for instructing the character reading device for each type of content, and a processing procedure for counting the number of determinations for each type each time the type of the meaning content of the plurality of character strings is determined; A recording procedure for excluding, from the type of analysis processing to be performed, a type whose counting result has reached the specified number. . 10. The recording medium according to claim 9, wherein
The recording medium according to claim 1, wherein the analysis process is a lexical analysis process using a plurality of analysis rules that define the characteristics of a character string for each type of the semantic content. 11. The recording medium according to claim 9, wherein
A recording medium according to claim 1, wherein said analysis processing is a dictionary in which a description of the character string is described for each semantic content of the character string. 12. The recording medium according to claim 9, wherein
When a plurality of types of identification results are obtained for one character string whose character has been recognized, the processing program determines one of the plurality of types of identification results based on a predetermined selection criterion. A recording medium further comprising a processing procedure for selecting as a final identification result.