JP2002108855A - Document processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a document processor by which a character part in picture data of a document consisting of text data and picture data is recognized with high precision. SOLUTION: Text data is translated (S3) when text data exists in an HTML document (S2: YES), picture data is analyzed (S5) when picture data exists (S4: YES), a character is segmented (S7) when the character part of picture data exists (S6: YES) and, then, the character is recognized (S8). A recognition candidate is compared with the character in text data, the recognition candidate is re-arranged (S9), the re-arranged recognition candidate is translated (S10) and a translation result is synthesized with the graphic part of picture data (S11).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a document processing apparatus for analyzing and translating data having text data and image data.

[0002]

2. Description of the Related Art Conventionally, in a document processing apparatus for analyzing and translating data having text data and image data, when translating a document including both text data and image data, only the text data is translated. A method of displaying the result by combining the image data with image data is known. For example, as shown in FIG. 8, in the A portion of the left half of the document, "Seminar announcement, today, a seminar will be held at the right venue from 1 o'clock" is described in text data, and the right half of the document is described. When a document in which image data including a map and an image of a character “seminar venue” is arranged in part B is translated, as shown in FIG.
In the part, `` Information of seminar Today there is semi
The translation is "nar in right place from 1:00", and the map, which is image data, and the characters "seminar venue" are displayed as they are in the D portion of the right half of the document.

[0003]

However, in the above-mentioned method, since the characters of the "seminar venue" which is the character portion in the image data are not translated, a document in which Japanese and English are mixed as shown in FIG. There was a problem of becoming. Further, as described in Japanese Patent Application Laid-Open No. H10-154145, a technique of extracting a character portion in image data, performing character recognition, and translating the image portion has been proposed.
There was a problem that the accuracy of character recognition was not always satisfactory.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem. In a document processing apparatus which analyzes a document including text data and image data, performs character recognition, and translates the result, An object of the present invention is to provide a document processing apparatus having high character recognition accuracy for character portions in image data.

[0005]

According to a first aspect of the present invention, there is provided a document processing apparatus for storing input text data and image data accompanying the text data. Separating means for separating an image part and a character part from the image data stored in the storage means, character recognition means for performing character recognition on the character part separated by the separating means and converting the character part into a character code; Candidate comparing means for comparing a candidate of the recognition result of the means with a character string in the text data stored in the storage means, and detecting a match between the candidate of the recognition result and the character string in the text data; And a candidate rearranging unit for rearranging the candidates of the recognition result of the character recognizing unit based on the comparison result of the comparing unit.

In the document processing apparatus having this configuration, the storage means stores the input text data and the image data accompanying the text data, and the separation means stores the image part and the character part from the image data stored in the storage means. And separate
The character recognizing means performs character recognition on the character portion separated by the separating means and converts it into a character code.The candidate comparing means compares the character string in the text data stored in the storage means with the candidate of the recognition result of the character recognizing means. Are compared with each other, and when a match between the candidate of the recognition result and the character string in the text data is detected, the candidate sorting unit sorts the candidates of the recognition result of the character recognition unit based on the comparison result of the candidate comparison unit.

According to a second aspect of the present invention, in addition to the configuration of the first aspect,
The candidate comparing unit compares a combination of characters in a plurality of candidates recognized by the character recognizing unit with a character string in text data stored in the storage unit, and the candidate comparing unit compares the plurality of candidates. When it is detected that a combination of characters in the text data matches a plurality of continuous character strings in the text data, the candidate reordering unit sets the plurality of continuous character strings as priority candidates and sets the character recognition unit Are sorted.

[0008] In the document processing apparatus having this configuration, in addition to the operation of the document processing apparatus according to the first aspect, the candidate comparing means includes:
The combination of characters in the plurality of candidates recognized by the character recognition unit is compared with the character string in the text data stored in the storage unit, and the candidate comparison unit compares the combination of characters in the plurality of candidates with the character string in the text data. When it is detected that a plurality of consecutive character strings match, the candidate sorting unit
The candidates of the recognition result of the character recognition means are rearranged with the plurality of continuous character strings as priority candidates.

According to a third aspect of the present invention, in addition to the configuration of the first or second aspect, in addition to the configuration of the document processing apparatus, the candidate character of the recognition result rearranged by the candidate rearranging unit is provided. A translation unit is provided for translating the string into a character string in another language different from the language of the character string.

[0010] In the document processing apparatus of this configuration, in addition to the operation of the document processing apparatus according to claim 1 or 2, the translating means converts the character string of the recognition result candidate rearranged by the candidate rearranging means. Translate to a string in another language different from the language of the string.

According to a fourth aspect of the present invention, in addition to the configuration of the document processing apparatus according to any one of the first to third aspects, the translation result of the translation unit is separated from the translation result by the separation unit. And a document synthesizing means for generating synthesized image data obtained by synthesizing the image portion.

[0012] In the document processing apparatus having this configuration, in addition to the operation of the document processing apparatus according to any one of claims 1 to 3, the document synthesizing means may include a translation result of the translation means and an image portion separated by the separation means. To generate composite image data.

According to a fifth aspect of the present invention, in addition to the configuration of the document processing apparatus according to any one of the first to fourth aspects, the combined image data combined by the document combining means is displayed. The display device is provided with a display means.

In the document processing apparatus having this configuration, in addition to the operation of the document processing apparatus according to any one of claims 1 to 4, the display means displays the combined image data combined by the document combining means.

According to a sixth aspect of the invention, in addition to the configuration of the document processing apparatus according to any one of the first to fifth aspects, the input text data and the text data are attached. The image data is described in HTML.

In the document processing apparatus having this configuration, in addition to the operation of the document processing apparatus according to any one of claims 1 to 5, input text data and image data accompanying the text data are described in HTML. This facilitates the analysis and synthesis of documents.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of the document processing apparatus 1, FIG. 2 is a flowchart of main control of the document processing apparatus 1, and FIG.
5 is a flowchart of a subroutine for comparing a recognition candidate with text data and rearranging candidates.

As shown in FIG. 1, a document processing apparatus 1 according to a first embodiment of the present invention has a CPU 2 for controlling the document processing apparatus 1, a ROM 7 storing programs executed by the CPU 2, and the like. RAM 8 for temporarily storing data
And an I / O interface 3 that mediates input / output of data to / from the CPU 2. The I / O interface 3 has a keyboard 4 for inputting characters,
A mouse 5, a display device 6 composed of an LCD or CRT, a CD-ROM drive 9, a LAN board 10 for connecting to an external network, and a hard disk drive 12 for storing a large amount of data are connected. ing. The LAN board 10 is connected to a router (not shown) via a cable 11, and the router (not shown) is connected to the Internet.

Next, the storage area of the ROM 7 will be described. As shown in FIG. 1, the ROM 7 stores an HTML file analysis program storage area 7a, a layout analysis program storage area 7b, a character cutout program storage area 7c, and a character recognition program storage area 7.
d, character recognition dictionary storage area 7e, candidate comparison program storage area 7f, candidate rearrangement program storage area 7g, translation program storage area 7h, HTML
A file synthesizing program storage area 7i is provided.

The HTML file analysis program storage area 7a stores an HTML file analysis program that causes the CPU 2 to analyze a file described in HTML and extract text data and image data. . The layout analysis program storage area 7b stores a layout analysis program that causes the CPU 2 to perform a layout analysis of the extracted image data and analyze a block of a figure, a block of a character, a block of a table, and the like. .

Further, a character extracting program storage area 7c stores a character extracting program for causing the CPU 2 to perform a process of extracting characters from a character block of the extracted image data. The character recognition program storage area 7d stores a character recognition program for causing the CPU 2 to perform character recognition processing. Also,
The character recognition dictionary storage area 7e stores a character recognition dictionary that is referred to when recognizing Japanese characters. further,
The candidate comparison program storage area 7f stores a candidate comparison program that causes the CPU 2 to perform a candidate comparison process of comparing text data with candidate characters of the character recognition result.

A candidate rearrangement program storage area 7g stores a candidate rearrangement program for causing the CPU 2 to perform a process of rearranging the candidate characters of the character recognition result based on the result of the candidate comparison process. The translation program storage area 7h stores a translation program that causes the CPU 2 to perform a translation process from Japanese to English or from English to Japanese. Further, in the HTML file synthesizing program storage area 7i, the translation result of the text data, the translation result of the character portion in the image data,
The process of combining parts other than the character part in the image data
An HTML file synthesizing program to be executed by PU2 is stored.

Next, the storage area of the RAM 8 will be described. As shown in FIG. 1, the RAM 8 has a text data storage area 8a for storing text data extracted from a file described in HTML, and an image data storage area 8b for storing image data extracted from a file described in HTML. An analysis result storage area 8c for storing analysis results of image data, a translation result storage area 8d for storing translation results, and a character extraction result storage area 8e for storing character extraction results from character portions in image data. A recognition result storage area 8f for storing a result of character recognition, a synthesis result storage area 8g for storing a synthesis result of a translated result of a character portion in the image data and a portion other than the character portion in the image data, A working area 8h for storing data in various processes performed by the CPU 2 is provided.

The operation of the document processing apparatus 1 configured as described above will be described with reference to FIGS. FIG.
FIG. 3 is a flowchart of main control performed by the document processing apparatus 1, and FIG. 3 is a flowchart of a subroutine for comparing recognition candidates and text data and rearranging candidates.
4 and 5 are flowcharts of a modification of the candidate rearrangement process. FIG. 6 is an example of candidate characters recognized as characters. FIG. 7 is a diagram showing the translation result displayed on the browser screen of the display device 6. 8 is a diagram illustrating an example in which an HTML file including text data and image data is displayed on a browser screen of the display device 6. FIG.

Hereinafter, a case where the homepage of the Internet is translated from Japanese to English will be described as an example. On the Internet homepage, as shown in Figure 8,
In the A part of the left half of the document, "Seminar announcement, today,
A seminar will be held at the right venue from 1:00. And an HTML file that displays a screen in which a map and image data composed of images of characters "seminar venue" are arranged in the right half B of the document. In this HTML file, "Seminar announcement, a seminar will be held at the right venue from 1:00 today." Will be described in Japanese text data, and the map and the text "Seminar venue" will be converted from the integrated image data. It is assumed that it is configured.

The document processing apparatus 1 captures an HTML file on the Internet homepage via a router (not shown) via a cable 11 and a LAN board 10 and stores it in a working area 8h of a RAM 8. Then, on the browser screen of the display device 6, as shown in FIG. 8, “Announcement of a seminar, a seminar will be held at the right venue from 1:00 today.”
And text data, and B in the right half of the screen
In the portion, image data including a map and an image of characters “seminar venue” is displayed.

Here, the user is displayed on the display device 6 using the mouse 5. When a translation button (not shown) is clicked, an HTML file, which is the source code of the display screen shown in FIG. 8 and stored in the working area 8h of the RAM 8, is analyzed, and text data and image data are extracted from the HTML file. The text data is stored in the text data storage area 8a of the RAM 8, and the image data is stored in the image data storage area 8b of the RAM 8.
(S1). Here, "Seminar announcement, seminar will be held at the right venue from 1:00 today."
Are extracted as text data, and the map and the characters “seminar venue” are extracted as integrated image data.

Next, in the determination process of S2, it is determined whether or not there is text data extracted from the HTML file. In the display screen shown in Fig. 8, the part of "Seminar announcement, a seminar will be held at the right venue from 1:00 today" is extracted as text data (S2:
YES), the text data is translated into English by a well-known machine translation method (S3). Then, "Information
of seminar Today thereis seminar in right place fr
om 1:00 "and stored in the translation result storage area 8d of the RAM 8 (S3).

Next, it is determined whether there is image data extracted from the HTML file (S4). In the display screen shown in FIG. 8, since the map and the characters “seminar venue” are extracted as one piece of image data, (S
4: YES), the image data is analyzed (S5).
More specifically, a layout analysis is performed on the extracted image data to extract a block of a figure, a block of a character, a block of a table, and the like, and store the extracted block in the analysis result storage area 8c. This analysis process can use a well-known technique of layout analysis. For example, a projection of image data is obtained, and a diagram block, a text block,
Analysis of blocks in the table is performed (S5).

For example, in a block of characters, peaks and valleys having a characteristic of projection are generated, but in the block of the figure, the character block is not easily distinguished from the block of the figure since the projection does not have the characteristic like the above character. it can. In the example shown in FIG. 8, the “seminar venue” which is a character part (character block) is present in the image data, so the “seminar venue” is a character part (character block) and the map is a block of the figure. Is stored in the analysis result storage area 8c.

In the example shown in FIG. 8, since there is a "seminar venue" which is a character portion (a block of characters) in the image data (S6: YES), a character cutout process is performed (S7). In this character extraction processing, first, characters are extracted one by one from the image data of the “seminar venue” which is a block of characters stored in the analysis result storage area 8c. This character can be cut out by using a known technique of OCR. For example, if a block of characters is written horizontally, first scan the image in the horizontal direction, count the number of black dots, and if the total number of black dots periodically repeats peaks and valleys, The part can be determined as a line break. Thereafter, the image is scanned in the vertical direction for one line, black dots are counted in the same manner, and a place where the total number of black dots is reduced is determined to be a character break, and characters are cut out one by one, and the character cutout result storage is performed. It is stored in the area 8e (S7).

Next, in the character recognition process (S8), the character recognition dictionary stored in the character recognition dictionary storage area 7e is referred to for each character of the extracted one character stored in the character extraction result storage area 8e. Then, the character is recognized by a well-known character recognition method such as off-line recognition, and the recognition result is stored in the recognition result storage area 8f (S8).

When the character recognition of the image data of the "seminar venue" which is a block of characters is completed by the above-described character recognition processing (S8), for example, as shown in FIG. ”As the second candidate,
Goyu "and" Zesif @ Zenyaku "are recognized as the third candidate,
It is stored in the recognition result storage area 8f of the RAM 8. However, none of these recognition results is a correct recognition result.

Next, the recognition result storage area 8f of the RAM 8
Candidate and the text data storage area 8
The comparison of the text data stored in a and the rearrangement of the candidates are performed (S9). The process of S9 will be described with reference to a flowchart of a subroutine shown in FIG. In the processing of the subroutine shown in FIG. 3, the combination of the recognition candidates and the character string
When more than characters match, the matching candidates are sorted in priority.

In the following description, it is assumed that the value of the counter i indicating the number of the recognition result of the character and the value of the counter j indicating the number of matching characters are stored in the working area 8h of the RAM 8. .

First, assuming that the counter i = 1 (S21),
It is determined whether or not the first character recognition candidate is stored in the recognition result storage area 8f (S22). Here, the first candidate “Himi + − venue” and the second candidate are stored in the recognition result storage area 8f.
The candidate "Sesana + Goyu" and the third candidate "Zesif @ Zenyaku" are stored, and the first character recognition candidate is "H",
Since "se" and "ze" are present (S22: YES), the first candidate "Himi + -Venue", the second candidate "Sesana + goyu", and the third candidate "Zesif @ Zenyaku" A search is made to determine whether the first character recognition candidates "hi", "se", and "ze" are present in the text data (S23).

The text data "seminar notice, today, 1
We will hold a seminar at the right venue from time to time. , There is "se" which is the same character as the first character "se" of the second candidate "sesana + goyu" (S23: YES), so that one of the rearrangement candidates is placed in the working area 8h. "C" is stored as the character (S24). In addition, the text data "Seminar announcement, seminar will be held at the right venue from 1:00 today."
In the search in, two "se" are found, but the character that first appears is adopted. When it is determined that the answer is NO in S23 (S23: NO), the value of i is increased by "1" (S32), and the process returns to S22.

Next, assuming that the counter j = 1 (S25),
In the text data stored in the text data storage area 8a, "Seminar announcement, today we will hold a seminar at the right venue from 1:00." A certain “mi” exists in the (i + j) character of the first candidate “Himi + —venue”, the second candidate “Sesana + goyu”, and the third candidate “zesif @ zeny”, that is, the second character. It is determined whether or not to perform (S26). here,
Since "mi" is present in the second character of the first candidate "Himi +-venue" (S26: YES), "mi" is stored in the working area 8h as the second character of the sorting candidate (S2).
7).

Next, the value of the counter j is increased by "1" to "j = 2" (S28), and "na", which is the character next to the "semi" character in the text data, is the first candidate "Himi". +
It is determined whether or not the third character is present in the (i + j) -th character of the third candidate “Zesif @ Zen-y”, the second candidate “Sesena + goyu”, and the third candidate (S26). Here, since “na” is present in the third character of the second candidate “sesana + goyu” (S26: YES), “na” is stored in the working area 8h as the third character of the rearrangement candidate. (S27).

Next, the value of the counter j is increased by "1" to "j = 3" (S28), and the next character "-" after the character "Semina" in the text data is the first candidate "Himi". +
It is determined whether or not the second character is present in the (i + j) -th character, ie, the fourth character of the second candidate “Sesana + goyu” and the third candidate “Zesif @ Zen-y” (S26). Here, since “−” exists in the fourth character of the first candidate “Himi + − venue” (S26: YES), “−” is stored as the fourth character of the rearrangement candidate in the working area 8h ( S27).

Next, the value of the counter j is increased by "1" to "j = 4" (S28), and "no" which is the next character of the character of "seminar" in the text data is the first candidate "Himi". + -Venue ”, the second candidate“ Sesuna + goyu ”, and the third candidate“ Zesif @ Zeny ”(i + j), that is, whether or not the fifth character is present (S26). . Here, since it is not found (S26: NO), the processing shifts to the determination processing of S29. In the determination of S29, it is determined whether j is 2 or more. When j is 2 or more, since two or more characters match the characters of the recognition candidate and the characters in the text data stored in the text data storage area 8a (S29: Y
ES), the i-th to (i + j−1) -th characters of the first candidate character stored in the recognition result storage area 8f (in the above example,
Working area 8h from 1st character to 4th character)
(S30).

Thereafter, the counter is set to i = i + j (S3
1). That is, in the above example, "i = 1 + 4 = 5" (S31). Thereafter, returning to S22, it is determined whether or not the fifth character recognition candidate corresponding to i = 5 is stored in the recognition result storage area 8f (S22). Since the fifth character recognition candidate is stored in the recognition result storage area 8f (S22: YES), the fifth to fifth character recognition candidates of the first to third candidates stored in the recognition result storage area 8f are displayed. A search is made to determine whether "", "go", and "all" are in the text data (S23).

The text data “seminar notice, today,
We will hold a seminar at the right venue from time to time. Since the character of the first candidate "kai" is found in the 17th character (without counting the reading points) in "" (S23: YES), "kai" is stored in the working area 8h as the fifth character of the sorting candidate. (S2
4).

Next, assuming that the counter j = 1, in the text data stored in the text data storage area 8a, "Seminar announcement, today, a seminar will be held at the right venue from 1:00.""Place", which is the character next to the character of "Kai", is the first to third candidates
It is determined whether or not the candidate exists at the (i + j) th character, that is, the sixth character (S26). Here, since "place" exists in the sixth character of the first candidate (S26: YES), "place" is stored in the working area 8h as the sixth character of the rearrangement candidate (S27).

Next, the value of the counter j is increased by "1" to "j = 2" (S28), and "de", which is the character next to the character of "venue" in the text data, is the first candidate to the first. It is determined whether or not the third candidate exists in the (i + j) th character, that is, the seventh character (S26). Here, since it is not found (S26: NO), the processing shifts to the determination processing of S29. S2
In the determination of 9, it is determined whether j is 2 or more. When j is 2 or more, the recognition candidate character matches the character in the text data stored in the text data storage area 8a with two or more characters (S29: YES), and thus the recognition result storage area 8f. Are arranged from the i-th character to the (i + j-1) -th character (the fifth to sixth characters in the above example) of the first candidate characters stored in the working area 8h stored in the working area 8h. Replace (S30). Here, the fifth and sixth characters of the first candidate are the “venues”, so they are the same as before sorting.

Thereafter, the counter is set to i = i + j (S3
1). That is, in the above example, "i = 5 + 2 = 7" (S31). Thereafter, returning to S22, it is determined whether or not the recognition candidate of the seventh character corresponding to i = 7 is stored in the recognition result storage area 8f (S22). Since the recognition candidate for the seventh character is not stored in the recognition result storage area 8f (S22: NO), the process of rearranging the recognition candidates is terminated, and the process shifts to S10 in the flowchart shown in FIG. By the above rearrangement processing, the first candidates stored in the recognition result storage area 8f are rearranged as “seminar venue”.

Next, in step S10, the first seminar candidate "seminar venue" which is rearranged is translated into English by a well-known machine translation (S10). Then, "Place of
Since the translation result of "seminar" is obtained, it is stored in the translation result storage area 8d of the RAM 8.

Next, an HTML file synthesizing process is performed (S11). In the process of S11, the translated text data “In” stored in the translation result storage area 8d is read.
formation of seminar Today there is seminar in rig
ht place from 1:00 ", image data other than the character part of the image data analyzed in S5 (in this example, a map part),
The text part “seminar venue” of the image data is translated and “Place of seminar” is synthesized, and as shown in FIG. 7,
"Informatio" is displayed on the left side E of the browser screen of the display device 6.
n of seminar Today there is seminar in right place
from 1:00 "and a map and" Place
An HTML document file in which the letters "of seminar" are displayed is created and stored in the synthesis result storage area 8g of the RAM 8, and the process is terminated. Here, the arrangement of the text data and the arrangement of the images in the synthesis processing of the HTML file are represented by S
The arrangement of the text data after translation and the arrangement of the images are performed by using the information of the arrangement of the text data before translation and the arrangement of the images read at the time of analyzing the HTML file in 1.

Then, the browser screen of the display device 6 displays
As shown in FIG. 7, "Information of semin"
ar Today there is seminar in right place from 1: 0
"0" is displayed, a map is displayed on the right side F, and "Place of seminar" is displayed below it.

As described above, in the document processing apparatus 1 of the above embodiment, the recognition candidates are compared with the character strings in the text data and the recognition candidates are rearranged, so that the accuracy of character recognition is improved. be able to. Further, when two or more characters in the text data match, the candidates are rearranged, so that it is possible to prevent the rearrangement of the candidates by mistake.

Next, the recognition result storage area 8f of the RAM 8
Candidate and the text data storage area 8
A modified example of the comparison of the text data stored in a and the sorting of the candidates (S9) will be described with reference to the flowcharts of the subroutines shown in FIGS. In the above embodiment, when two or more characters match, the candidates are unconditionally rearranged. However, in the processing of the subroutine shown in FIGS. 4 and 5, the recognition score of the recognition candidate is also taken into consideration. Candidates with poor recognition scores do not sort recognition candidates unless more character strings (eg, four or more character strings) match.

In the following description, the value of a counter i indicating the number of characters to which the recognition result is focused, the value of a counter j indicating the number of matching characters, and whether a recognition candidate includes a word having a bad recognition score. The recognition score flag s indicating whether
8 in the working area 8h.
If the recognition score flag s = 0, it means that the recognition candidate includes a word with a bad recognition score, and if the recognition score flag s = 1, the recognition candidate does not include a word with a bad recognition score. means.

First, as shown in FIG.
(S21), and the recognition score flag s = 1 (S4
2). Next, it is determined whether or not the first character recognition candidate corresponding to the counter i = 1 is stored in the recognition result storage area 8f (S43). Here, the first candidate "Himi + -Venue", the second candidate "Sesuna + goyu", and the third candidate "Zesif @ Zenyai" are stored in the recognition result storage area 8f, each of which has one character. Recognition candidates are "hi", "se",
Since "ze" exists (S43: YES), the first candidate "Himi +-venue", the second candidate "Sesana + Goyu", the third candidate
A search is made to determine whether the first character recognition candidates "hi", "se", and "ze" of each of the candidates "zeshifu @ zenyaku" are present in the text data (S44).

The text data “seminar notice, today, 1
We will hold a seminar at the right venue from time to time. , There is "se" which is the same character as the first character "se" of the second candidate "sesana + goyu" (S44: YES), so that one of the rearrangement candidates is placed in the working area 8h. "C" is stored as the character (S45). In addition, the text data "Seminar announcement, seminar will be held at the right venue from 1:00 today."
In the search in, two "se" are found, but the character that first appears is adopted. Further, when it is determined that the result of S44 is NO (S44: NO), the value of i is increased by "1" (S49), and the process returns to S43.

Next, assuming that the counter j = 1, in the text data stored in the text data storage area 8a, "Seminar announcement, today, a seminar will be held at the right venue from 1:00." It is determined whether the recognition score of the first character “se” of the second candidate “sesana + goyu” that matches the found “se” is less than 50 (S47). The recognition score is a value obtained by digitizing a match between the character of the basic shape of the character and the characteristic data when the character composed of the image data is subjected to character recognition by a known character recognition method such as off-line recognition. ] Are normalized numbers. The closer the recognition score is to 100, the better the recognition accuracy, and the closer to 0, the worse the recognition accuracy.

For example, "S" found in the process of S44
When the recognition score of the first character “se” of the second candidate “sesana + goyu” that matches with is 45, the recognition score is 5
If it is less than 0, it is determined that the recognition accuracy of “C” is poor (S47: YES), so that it is stored in the working area 8h as the recognition score flag s = 0 (S48), and S shown in FIG.
The process proceeds to a determination process of 50. If the recognition score is 50 or more (S47: NO), the process shifts to the determination process of S50 shown in FIG.

Next, in the text data stored in the text data storage area 8a, "Seminar announcement, a seminar will be held at the right venue from 1:00 today."
The character "mi" next to the character "se" found in the processing of 44 is the first candidate "Himi +-venue", the second candidate "sesana + goto", and the third candidate "zesif @ all" It is determined whether or not the character exists in the (i + j) -th character of “Easy”, that is, the second character (S5).
0). Here, since "mi" is present in the second character of the first candidate "Himi +-venue" (S50: YES), "mi" is selected as the second character of the rearrangement candidate in the working area 8h.
Is stored (S51).

Next, the value of the counter j is incremented by “1” to “j = 2” (S52), and the two characters of the first candidate “Himi + —venue” that matches “Mi” found in the processing of S50. It is determined whether the recognition score of “mi” is less than 50 (S
53). If the recognition score is less than 50, the recognition score flag s = 0 is stored in the working area 8h (S
54), the process proceeds to the determination process of S50. When the recognition score is 50 or more (S53: NO), the process shifts to the determination process of S50.

In the determination processing of S50, the character "na" which is the character next to the character "semi" in the text data is the first candidate "Himi + -venue", the second candidate "Sesana + goyu", It is determined whether the third candidate "zeshifu @ zenyaku" exists in the (i + j) th character, that is, in the third character (S50). here,
Since "na" is present in the third character of the second candidate "sesana + goyu" (S50: YES), "na" is stored in the working area 8h as the third character of the rearrangement candidate (S5).
1).

Next, the value of the counter j is increased by "1" to "j = 3" (S52), and the recognition score is determined in the same manner as above (S53). If the recognition score is less than 50, the recognition score flag s = 0 (S54), and the next character "-" after the character "Semina" in the text data is the first character.
The (i + j) letters of the candidate "Himi + venue", the second candidate "Sesana + Goyu", and the third candidate "Zesif @ Zenyaku", that is, 4
It is determined whether or not the character exists (S50). Here, since “−” exists in the fourth character of the first candidate “Himi + − venue” (S50: YES), the working area 8
"-" is stored in h as the fourth character of the sorting candidate (S51).

Next, the value of the counter j is increased by "1" to "j = 4" (S52), and the recognition score is determined in the same manner as above (S53). If the recognition score is less than 50, the recognition score flag s = 0 (S54), and the next character "no" in the text data is the first candidate "Himi +-venue", the second candidate "Semina + Goyu",
(I + j) character of the third candidate “Zesif @ Zeny”,
It is determined whether or not the fifth character exists (S50). Here, since it is not found (S50: NO), the processing shifts to the determination processing of S55. In the determination of S55, it is determined whether the recognition score flag s is “0” (S55).

When the recognition score flag s is “0” (S55: YES), recognition candidates having low recognition accuracy are included in the recognition candidates that match the characters in the text data.
When the counter j is 4 or more (S56: YES), that is,
Only when four or more characters of the recognition candidate match characters in the text data stored in the text data storage area 8a (S56: YES), the first candidate character stored in the recognition result storage area 8f From the i character to i +
j-1th character (in the above example, the first to fourth characters)
Are sorted with the sorting seminar "seminar" stored in the working area 8h (S58).

If the recognition score flag s is “1” (S55: NO), since recognition candidates having low recognition accuracy are not included in the recognition candidates that match the characters in the text data,
When the counter j is 2 or more (S57: YES), that is,
If two or more characters of the recognition candidate match the characters in the text data stored in the text data storage area 8a (S57: YES), the recognition result storage area 8
From the i-th character of the first candidate character stored in f to i + j-1
The characters (the first to fourth characters in the above example) are rearranged with the rearrangement candidate “seminar” stored in the working area 8h (S58).

Thereafter, the counter is set to i = i + j (S5
9). That is, in the above example, “i = 1 + 4 = 5” (S59). Thereafter, the process returns to S43, and it is determined whether or not the recognition candidate of the fifth character corresponding to i = 5 is stored in the recognition result storage area 8f (S43). Since the fifth character recognition candidate is stored in the recognition result storage area 8f (S43: YES), the fifth to fifth character recognition candidates of the first to third candidates stored in the recognition result storage area 8f are displayed. A search is made to see if "", "go", and "all" are in the text data (S44).

The text data “seminar notice, today, 1
We will hold a seminar at the right venue from time to time. In the 17th character (without counting the readings) in the 5th
Since the character "kai" is found (S44: YE
S), "kai" is stored in the working area 8h as the fifth character of the sorting candidate (S45).

Next, the counter j = 1 (S46),
In the text data stored in the text data storage area 8a, "Announcement of the seminar, a seminar will be held at the right venue from 1 o'clock today." It is determined whether or not the recognition score of the fifth character of "Himi + -venue" is less than 50 (S47). For example, "kai" found in the processing of S44
When the recognition score of the first character of the first character "kai" of the first character "kai" of the first candidate "Himi + Venue" that matches the recognition score is 48, the recognition score is less than 50, and " It is determined that the recognition accuracy of the event is low (S47: YES), so that the recognition score flag s = 0 is stored in the working area 8h (S48), and the process proceeds to the determination process of S50 shown in FIG.
If the recognition score is 50 or more (S47: N
O), the process proceeds to the determination process of S50 shown in FIG.

Next, in S50, in the text data stored in the text data storage area 8a, "Seminar announcement, a seminar will be held at the right venue from 1:00 today." The first character of the first to third candidates (i +
j) It is determined whether the character exists in the sixth character, that is, in the sixth character (S50). Here, for the sixth character of the first candidate,
Since "place" exists (S50: YES), "place" is stored in the working area 8h as the sixth character of the sorting candidate (S51).

Next, the value of the counter j is increased by "1" to "j = 2" (S52), and the recognition score is determined in the same manner as above (S53). If the recognition score is less than 50, the recognition score flag s = The value is set to 0 (S54), and next, "de", which is the next character of the character of "venue" in the text data, exists in the (i + j) -th character of the first to third candidates, that is, the seventh character. It is determined whether or not it is (S50). Here, since it is not found (S50: NO), the processing shifts to the determination processing of S55. In the determination of S55, the recognition score flag s is “0”
It is determined whether or not (S55).

Here, since the recognition score flag s is “0” (S55: YES), the recognition candidate that matches the character in the text data includes the character of “kai” with low recognition accuracy. Since the counter j is 2 (S56: N
O), the process proceeds to S59 without rearranging S58, and sets the counter i = i + j (S59). That is, in the above example, “i = 5 + 2 = 7” (S59). Then, S
Returning to 43, it is determined whether or not the recognition candidate for the seventh character corresponding to i = 7 is stored in the recognition result storage area 8f (S43). Since the recognition candidate for the seventh character is not stored in the recognition result storage area 8f (S43: NO), the process of rearranging the recognition candidates is terminated, and the process shifts to S10 of the flowchart shown in FIG. By the above sorting process,
In the first candidate stored in the recognition result storage area 8f, only the "seminar" portion is rearranged.

As described above, in the above modification,
Since the recognition candidates are compared with the character strings in the text data and the recognition candidates are rearranged, the accuracy of character recognition can be improved. Also, even when two or more characters in the text data match, if a candidate with a poor recognition score is included in consideration of the recognition score at the time of recognition, sorting is not performed unless four or more characters match the text data. By doing so, it is possible to prevent erroneous sorting of candidates with poor recognition scores.

The document processing apparatus 1 of the above embodiment
It goes without saying that various modifications can be made. For example, in the above-described embodiment, an example of translating from a Japanese homepage to English has been described, but it is also possible to translate from an English homepage to Japanese. Further, the language is not limited to Japanese and English, and it goes without saying that the invention can be applied to various languages such as Chinese, Korean, German, French, Italian, Spanish, and the like. Furthermore, in the first embodiment, the sorting is not performed unless two or more characters match the text data. However, the sorting is not necessarily limited to two characters and a predetermined number of characters such as three, four, five, etc. May not be sorted if does not match the text data. Further, in the above-described modified example, when a candidate having a poor recognition score is included, sorting is not performed unless four or more characters match the text data.
The number of characters is not necessarily limited to four, and the rearrangement may not be performed unless a predetermined number of characters such as three, five, or six characters matches the text data.

[0073]

As described above, in the document processing apparatus according to the first aspect of the present invention, the storage means stores the input text data and the image data accompanying the text data, and the separation means stores the text data. Separating the image portion and the character portion from the image data stored in the storage device, the character recognition device performs character recognition on the character portion separated by the separation device to convert the character portion into a character code, and the candidate comparison device includes the character recognition device. Comparing the candidate of the recognition result with the character string in the text data stored in the storage unit, and detecting a match between the candidate of the recognition result and the character string in the text data, the candidate reordering unit sets the candidate The recognition result candidates of the character recognition means can be rearranged based on the comparison result of the comparison means. Therefore, since the character strings in the text data are referred to for rearranging the recognition candidates, the accuracy of character recognition can be improved.

In the document processing apparatus according to the second aspect of the present invention, in addition to the effects of the document processing apparatus according to the first aspect,
The candidate comparing unit compares a combination of characters in a plurality of candidates recognized by the character recognizing unit with a character string in text data stored in the storage unit, and the candidate comparing unit compares the plurality of candidates. When it is detected that a combination of characters in the text data matches a plurality of continuous character strings in the text data, the candidate reordering unit sets the plurality of continuous character strings as priority candidates and sets the character recognition unit Can be sorted.

According to a third aspect of the present invention, in addition to the effects of the first or second aspect, the translating unit may further include a recognition result rearranged by the candidate rearranging unit. Can be translated into a character string in another language different from the language of the character string.

According to a fourth aspect of the present invention, in addition to the effects of the document processing apparatus according to any one of the first to third aspects, the document synthesizing means includes a translation result of the translation means and the separation result. Combined image data can be created by combining the image portion separated by the means. Therefore, it is possible to obtain composite image data in which different languages are not mixed in the translation result.

According to a fifth aspect of the present invention, in addition to the effects of the document processing apparatus according to any one of the first to fourth aspects, the display means further comprises a synthetic image synthesized by the document synthesizing means. Data can be displayed. Therefore, the user can view the synthesized image in which different languages are not mixed in the translation result by the display means.

According to a sixth aspect of the present invention, in addition to the effects of the document processing apparatus according to any one of the first to fifth aspects, the input text data and the text data are attached. Since the image data is described in HTML, it is possible to easily analyze and synthesize the document.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a document processing apparatus 1;

FIG. 2 is a flowchart of main control of the document processing apparatus 1.

FIG. 3 is a flowchart of a subroutine for comparing a candidate character and text data and reordering candidates in the document processing apparatus 1.

FIG. 4 is a flowchart of a modification of the candidate reordering process.

FIG. 5 is a flowchart of a modification of the candidate reordering process.

FIG. 6 is an example of candidate characters recognized as characters;

FIG. 7 is a diagram illustrating an example in which a translation result is displayed on a browser screen of a display device 6;

FIG. 8 is a diagram illustrating an example in which an HTML file having text data and image data is displayed on a browser screen of a display device 6;

FIG. 9 is a diagram illustrating an example in which a translation result according to a conventional technique is displayed on a browser screen of a display device 6;

[Description of Signs] 1 Document processing device 2 CPU 3 I / O interface 4 Keyboard 5 Mouse 6 Display device 7 ROM 7a HTML file analysis program storage area 7b Layout analysis program storage area 7c Character extraction program storage area 7d Character recognition program storage area 7e Character recognition dictionary storage area 7f Candidate comparison program storage area 7g Candidate sorting program storage area 7h Translation program storage area 7i HTML file synthesis program storage area 8 RAM 8a Text data storage area 8b Image data storage area 8c Analysis result storage area 8d Translation Result storage area 8e Character cutout result storage area 8f Recognition result storage area 8g Synthesis result storage area 8h Working area 9 CD-ROM drive 10 LAN board 11 Cable 12 Hard disk drive

Claims (6)

[Claims] A storage unit configured to store input text data and image data accompanying the text data; a separation unit configured to separate an image portion and a character portion from the image data stored in the storage device; A character recognition unit that performs character recognition on the character portion separated by the separation unit and converts the character portion into a character code; and compares a candidate for a recognition result of the character recognition unit with a character string in text data stored in the storage unit. A candidate comparing unit that detects a match between the candidate for the recognition result and a character string in the text data; and a candidate arrangement for rearranging the candidates for the recognition result of the character recognition unit based on the comparison result of the candidate comparing unit. A document processing apparatus comprising: a replacement unit. 2. The candidate comparing unit compares a combination of characters in a plurality of candidates recognized by the character recognizing unit with a character string in text data stored in the storage unit. Detects that a combination of characters in the plurality of candidates matches a plurality of consecutive character strings in the text data, the candidate rearranging unit sets the plurality of consecutive character strings as priority candidates. 2. The document processing apparatus according to claim 1, wherein the candidates of the recognition result of the character recognition unit are rearranged. 3. A translation unit for translating a character string of a candidate of the recognition result rearranged by the candidate rearranging unit into a character string of another language different from the language of the character string. The document processing device according to claim 1. 4. The apparatus according to claim 1, further comprising document synthesizing means for generating synthetic image data obtained by synthesizing a translation result of said translating means and an image portion separated by said separating means. A document processing device according to claim 1. 5. The document processing apparatus according to claim 1, further comprising display means for displaying the combined image data combined by said document combining means. 6. The document processing apparatus according to claim 1, wherein the input text data and image data accompanying the text data are described in HTML.