JP2890241B2 - Optical character recognition device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character recognition apparatus and, more particularly, to an optical character recognition system for performing a knowledge process on an optically read result.

[0002]

2. Description of the Related Art Conventionally, in this type of optical character recognition system, as shown in FIG. 10, a kanji character "Ugajin" and its corresponding pseudonym "Ugasin" are written one by one in one box. In addition, a clouded point “゛” and a semi-voiced mark “゜” given to a pseudonym are also entered in one box.

As shown in FIG. 11, an optical character reading device outputs a recognition result of a kanji character or a hiragana character as a similar character ranked as a candidate character. In other words, the recognition result of the kanji is that the first candidate character is “Uojin”, the second candidate character is “Shoga detail”, the third candidate character is “Kotorakaki”, the fourth candidate character is “Yuuoi Naka”, The 5 candidate characters are “seasonal sleeve”, the 6th candidate character is “Evening Insane Insertion”, the 7th candidate character is “9 Min Shin”, the 8th candidate character is “Soushu Zen”, and the 9th candidate character is “ It is output in a similar character to which a ranking is given in advance, such as "light frame".

In addition, the recognition result of the phonetic kana is a similar character that has been given a ranking in advance, such as a first candidate character of "Ukahashin", a second candidate character of "Katabuni", and a third candidate character of "Tausoso". Is output.

In the optical character recognition method, a word is created by combining kanji candidate characters output from an optical character reader, and the word is stored in a database in which kanji and shakugana are stored in association with each other in advance. Compare and determine kanji candidate words. By determining the kanji candidate word, the furigana word is determined from the contents of the database.

[0006] This kana word is compared with the candidate character of the recognition result of the kana character, and the score of the kana word is calculated according to the candidate order. The score and the score of the kanji candidate word determined by the above method are compared with the score. Is the total score of the candidate word that pairs the kanji and the Chinese kana.

[0007] The above operation is repeatedly executed, and finally, words having a good total score are output as candidate words. This optical character recognition method is disclosed in Japanese Patent Laid-Open No. 4-328892.
The details are described in Japanese Patent Publication No.

[0008]

In the above-mentioned conventional optical character recognition method, since the syllabic point and the syllabary point are entered in one box as in the case of other characters (kanji and syllabic kana), misreading or unreading is performed. Occurs, the character order between the pseudonym data in the database and the pseudonym recognition result is shifted by the amount of voiced or semi-voiced dots, and the data does not match, and a normal candidate word cannot be determined.

Accordingly, an object of the present invention is to solve the above-mentioned problems and to provide an optical character which can normally determine a candidate word even if the character order is shifted by the amount of the voiced or semi-voiced characters assigned to the pen-kana. An object of the present invention is to provide a recognition device.

[0010]

SUMMARY OF THE INVENTION An optical character recognition apparatus according to the present invention is an optical character recognition apparatus for optically reading and recognizing a kanji character and a katakana character corresponding to the kanji character. A kanji word selecting means for extracting a kanji word by comparing the result with a pre-stored kanji; and whether or not there is a clouded point and a semi-voiced point in a shakana corresponding to the kanji word extracted by the kanji word selecting means. Detecting means for detecting whether
When the detection means detects the voiced and semi-voiced points, the pseudonym word selection means for extracting a pseudonym word by comparing the pseudonym and the pseudonym recognition result excluding the voiced and semi-voiced points, and the kanji word Means for determining the order of the candidate words based on the candidate order of the kanji words extracted by the selecting means and the candidate order of the pseudonym words extracted by the shaking kana word selecting means.

According to another aspect of the present invention, there is provided an optical character recognizing apparatus for storing, in addition to the above-described configuration, the kanji word and the syllabary including the syllabary to which the syllabic point and the syllabic point are added. Is provided.

Another optical character recognition apparatus of the present invention further comprises, in addition to the above-described configuration, holding means for holding the positions of the voiced and semi-voiced points in the kurikana detected by the detecting means. It is configured to remove the cloudy point and the semi-voiced point from the Chinese kana based on the contents of the holding means.

[0013]

A kanji word is extracted by comparing a word obtained by combining the characters of the kanji recognition result with a word previously stored in the database unit, and the score of the kanji word is calculated.

Further, the phonetic character corresponding to the kanji word is read out from the database unit, and the position of the voiced or semi-voiced voice in the character is detected. Based on the positions of the detected voiced or semi-voiced voices, the voiced or semi-voiced voice data in the database unit 6 is excluded from the pseudonym data, and the pseudonym word is extracted by comparing the pseudonym data with the pseudonym recognition result. Calculate word scores.

The total score of the kanji word and the kana word is summed to calculate a total score of the kanji word, and the highest score of the total kanji word is sequentially output as a candidate word.

In this way, even when the character order in the kana and the kana recognition results in the database unit is out of alignment by the amount of the voiced or semi-voiced voice and the data does not match, the candidate word can be determined normally.

[0017]

Next, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention. In FIG. 1, when a kanji word selection unit 1 receives a kanji recognition result output from an optical character reading device (not shown), the kanji word selection unit 1 combines a word obtained by combining characters of the kanji recognition result with a word stored in the database unit 6 in advance. Compare with

As a result of the comparison, if a kanji word exists in the database unit 6, the kanji word selection unit 1 calculates the score of the kanji word based on the ranking of the kanji recognition result, and determines the kanji word data and the score as candidate. Pass to word output unit 5.

At the same time, the kanji word selection unit 1 passes a pointer in the database 6 for the kana character corresponding to the kanji word to the kana word selection unit 2. Furigana word selection part 2
Handovers the pointer of the furigana from the kanji word selection unit 1 to the voiced dot search unit 3.

When receiving the pointer of the pseudonym from the pseudonym word selecting section 2, the voiced point search section 3 searches the database for the pseudonym data in the database section 6 based on the pointer, and searches for the voiced or semi-voiced point in the pseudonym data. Is stored in the turbid point position holding unit 4.

The kana kana word selection unit 2 includes a voiced dot position holding unit 4
Is compared with the phonetic character recognition result obtained by removing the phonetic character or semi-voiced character from the phonetic character data of the database unit 6 while referring to the position information held in the database.

As a result of the comparison, if a kanji word exists in the database unit 6, the kana-kana word selection unit 2 calculates the score of the kana-kana word based on the ranking of the kana-kana recognition result, and And the score are passed to the candidate word output unit 5.

The candidate word output unit 5 repeats the above operation until there are no more combinations of words in the kanji recognition result.
The kanji word data and its score from the kanji word selection unit 1 and the kana character data and its score from the furigana word selection unit 2 are summed to calculate the total score of the kanji word.

After calculating the total score of all the kanji words, the candidate word output section 5 sequentially outputs the candidate words having the highest total score as candidate words having the highest rank. In this case, the candidate word output unit 5 outputs the first candidate word,
The second candidate word is sequentially output.

FIG. 2 is a diagram showing a form entry example according to one embodiment of the present invention. In the figure, the form A has an entry column A1 for the pseudonym “Ugasin” and an entry column A2 for the kanji character “Ugajin”.
Are provided. Here, a case is shown in which a pseudonym including a voiced point is entered in one box, such as “ga”.

FIG. 3 is a diagram showing the configuration of the database unit 6 of FIG. In the figure, when one kanji word has a plurality of kana characters, for example, a kanji word “Ukajin” in the database unit 6, the plurality of kana words “Ukakami”, “Ugasin”, “Ugakami”, “ “Ugajin” and “Ugami” are stored in advance in association with the kanji word “Ugajin”.

FIG. 4 is a diagram showing an example of a kanji recognition result according to one embodiment of the present invention. In the figure, the optical character reading device outputs the kanji recognition result B as a similar character with ranking. FIG. 4 shows similar characters that are output when the optical character reader reads and recognizes “Ugajin”.

In this case, in the kanji recognition result B, the first candidate character is “Uhuangjin”, the second candidate character is “Shoga detail”, the third candidate character is “Kotorakaki”, and the fourth candidate character is “Evening evening”. Naka, the 5th candidate character is “seasonal sleeve”, the 6th candidate character is “Evening Insect”, the 7th candidate character is “9 Min Shin”, the 8th candidate character is “Souzen Zen”, the ninth
The candidate character is output as a similar character to which a ranking is given in advance such as “character thin frame”.

FIG. 5 is a diagram showing an example of a word created by combining the kanji of the kanji recognition result B of FIG. In the figure,
The kanji in the kanji recognition result B are combined, and "Uhuangjin"
An example is shown in which words such as “Shinhuangjin”, “Kohuangjin”,..., “Ugajin”, “Utorajin”, “Uoijin”,.

FIGS. 6 and 7 are diagrams showing an example of the comparison operation between the pseudonym recognition result and the contents of the database according to one embodiment of the present invention. FIG. 6 shows the comparison operation between the pseudonym recognition result C and the pseudonym data 1 when no voiced or semi-voiced point is detected in the pseudonym data in the database unit 6.

At this time, the characters of the first candidate character “U? Shin”, the second candidate character “Kanni”, and the third candidate character “Tasso” of the kana kana recognition result C are stored in the kana kana data of the database unit 6. 1 The character is compared with the character at the same position of "Ukakami".

FIG. 7 shows a comparison operation between the pseudonym recognition result C and the pseudonym data 1 and 2 when a voiced or semi-voiced voice is detected in the pseudonym data in the database unit 6.

At this time, the first and second characters of the first candidate character “U? Shin”, the second candidate character “Kanni”, and the third candidate character “Tasso” of the kana-kana recognition result C are stored in the database. The first character and the second character of the pseudonym data 1 “Uka @ shin” and the hiragana data 2 “Uka @ kami” of the part 6 are compared.

The third and fourth characters of the first candidate character "U? Shin", the second candidate character "Kani", and the third candidate character "Tasoso" of the kana-kana recognition result C are stored in the database unit 6. Are compared with the fourth and fifth characters of the pseudonym data 1 "Uka @ shin" and the hiragana data 2 "Uka @ kami", which are obtained by skipping the cloudy point "@".

FIGS. 8 and 9 are flowcharts showing the operation of one embodiment of the present invention. The operation of optical character recognition according to one embodiment of the present invention will be described with reference to FIGS.

When an optical character reader reads a form A (see FIG. 2) in which a phonetic character with voiced dots is entered in one box, a kanji recognition result B (see FIG. 4) and a character recognition result C (see FIG. 4). The similar characters ranked as candidate characters composed of the candidate characters (see FIGS. 6 and 7) are output from the optical character reading device.

In this case, the recognition result C of the pseudonym with respect to the pseudonym with a voiced dot entered in one box of the form A is an unreadable character (indicated by "?" In the figure) or misread.

Upon receiving the kanji recognition result B from the optical character reading device, the kanji word selection section 1 creates a kanji word by combining the characters of the kanji recognition result B (step S1 in FIG. 8).
(See FIG. 5).

The kanji word selection unit 1 compares the kanji word created with a word stored in the database unit 6 in advance, and searches whether or not the kanji word is registered in the database unit 6 (step in FIG. 8). S2, S3).

The kanji word determined by this search is scored according to the order of candidate characters for each character constituting the word, and the total score of each word character is the kanji word score. .

When the kanji word is determined, the kanji word selection section 1 determines the phonetic kana word corresponding to the kanji word and passes the pointer in the database section 6 of the phonetic kana word to the phonetic kana word selection section 2.

The furigana word selection unit 2 is a kanji word selection unit 1
When the pointer is received from the database, it reads the pseudonym data from the database unit 6 based on the pointer, and performs a pseudonym word selection process.

That is, the hiragana word selecting section 2 compares the hiragana data read out from the database section 6 with the hiragana recognition result C based on the pointer from the kanji word selecting section 1, and determines the word in the hiragana recognition result C. Is searched in the database unit 6 (steps S4 and S5 in FIG. 8).

In this case, since the kana with the syllabary point is entered in one box, the total number of characters of the kana recognition result C becomes 4 characters, and the kana kana word selection section 2 refers to the database section 6 from the database section 6. There is no word because there is a deviation from the pseudonym data.

When there is no word, the furigana word selection unit 2 passes the pointer of the furigana word from the kanji word selection unit 1 to the voice mark search unit 3. When receiving the pointer of the pseudonym word from the pseudonym word selection unit 2, the voiced dot search unit 3 reads out the pseudonym data from the database unit 6 based on the pointer and performs the voiced word search process (step S6 in FIG. 8).

That is, the dakuten search unit 3 searches whether or not there is a dakuten or a semi-dakuten in the kana data read from the database unit 6 based on the pointer from the kana word selection unit 2 (step S6 in FIG. 9). 1, S6-2).

If there is a dakuten or semi-dakuten in the shakugana data, the dakuten search unit 3 holds the position information indicating the position of the dakuten or semi-dakuten in the shagana data in the dakuten position holding unit 4 (step S6 in FIG. 9). -3).

When the position information of the voiced and semi-voiced voices is stored in the voiced voice position storage unit 4, the pseudonym word selection unit 2 performs a pseudonym word selection process (step S6-4 in FIG. 9). That is, while referring to the position information stored in the voiced character position storage unit 4, the pseudonym word selection unit 2 recognizes the pseudonym recognition result C and the pseudonym obtained by removing the voiced and semi-voiced characters from the pseudonym data in the database unit 6. Compare one character at a time.

The furigana word selection unit 2 determines whether a kanji word exists in the database unit 6 based on the result of the comparison (step S7 in FIG. 8). If a kanji word exists in the database unit 6, the kana-kana word selection unit 2 calculates the score of the kana-kana word based on the order of the kana-kana recognition result, and outputs the kana-kana word data and the score to the candidate word output unit. Pass to 5.

The candidate word output unit 5 sums up the kanji word data and its score from the kanji word selection unit 1 and the kana character data and its score from the kana word selection unit 2 to obtain the total score of the kanji word. The calculation is performed (step S8 in FIG. 8).

The candidate word output unit 5 repeats the above operation until there are no more combinations of words in the kanji recognition result (FIG. 8).
Steps S1 to S9), and sequentially output the candidate words having higher total scores as candidate words having higher ranks (step S10 in FIG. 8). In this case, the candidate word output unit 5 sequentially outputs the first candidate word and the second candidate word in order from the candidate word having the highest rank.

As described above, the pen-kana to which the voiced point and the semi-voiced point are added is entered in one box. By comparing the result with the result C, the candidate word can be determined normally even when the character order between the phonetic kana data in the database unit 6 and the phonetic kana recognition result C is different from each other due to the shift of the voiced or semi-voiced point and the data does not match. be able to.

[0054]

As described above, according to the present invention, a kanji word is extracted by comparing a kanji recognition result with a previously stored kanji character, and a kanji word corresponding to the extracted kanji word is extracted. When a voiced point or semi-voiced point is detected, the phonetic alphabet excluding the voiced point or semi-voiced point is compared with the character recognition result to extract the word pseudonym. By determining the order of the candidate words based on, there is an effect that the candidate words can be normally determined even if the character order is shifted by the amount of the voiced or semi-voiced characters assigned to the pen-kana.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.

FIG. 2 is a diagram showing a form entry example according to an embodiment of the present invention.

FIG. 3 is a diagram illustrating a configuration of a database unit in FIG. 1;

FIG. 4 is a diagram illustrating an example of a kanji recognition result according to an embodiment of the present invention.

5 is a diagram showing an example of a word created by combining the kanji of the kanji recognition result of FIG. 4;

FIG. 6 is a diagram illustrating an example of a comparison operation between a pseudonym recognition result and the contents of a database according to an embodiment of the present invention.

FIG. 7 is a diagram showing an example of a comparison operation between a pseudonym recognition result and the contents of a database according to an embodiment of the present invention.

FIG. 8 is a flowchart showing the operation of one embodiment of the present invention.

FIG. 9 is a flowchart showing the operation of one embodiment of the present invention.

FIG. 10 is a diagram showing a form entry example according to a conventional example.

FIG. 11 is a diagram showing an example of a comparison operation between a recognition result and the contents of a database according to a conventional example.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 Kanji word selection unit 2 Shakukana word selection unit 3 Dakuten search unit 4 Dakuten position holding unit 5 Candidate word output unit 6 Database unit

Claims (2)

(57) [Claims] (1) A kanji and a kana corresponding to the kanji.
Optical character recognition means for reading and recognizing each optically
And a candidate word consisting of a kanji word and the corresponding kana
Candidate word storing means for storing the read kanji recognition result and the previously stored kanji
Candidate kanji to extract candidate kanji words by comparing with kanji words
A word selecting means, and a kana character in a kana word corresponding to the candidate kanji word
Is a cloud point / semi-voice point detection means for detecting whether there is a semi-voice point
And, to respond to the candidate Kanji word and the read the furigana
If the total number of characters does not match
From the kana word corresponding to the candidate kanji word
Alternatively, remove the semi-voiced point and re-
Candidate kana to extract candidate kana words by comparing with the name
A word selecting means, and a candidate kanji word extracted by the candidate kanji word selecting means.
Extracted by the candidate rank and the candidate pseudonym word selecting means.
Based on the candidate rank of candidate kana words
Optical sentence having means for final determination
Character recognition device. 2. The turbidity detected by the turbidity / semi-turbidity point detecting means.
Point / semi-voice point position
2. The optical statement according to claim 1, wherein the optical statement has holding means.
Character recognition device.