JPS6386089A - Character recognizing device - Google Patents

Abstract

PURPOSE:To decrease reading errors by comparing a designating character group with the prescribed character group of a knowledge dictionary and executing recognizing processing of the designating character group. CONSTITUTION:The character group of a slip is read by a reading part 1 and the character of the read character group is recognized part 2. A segmenting part 3 segments the designating character group of the same character species designated out of the character groups of a slip in which a character recognition is completed. A semantically unfair character group is stored into a knowledge dictionary part 4. The designating character group and the prescribed character group are compared by a recognizing processing part 5 and the recognizing processing of the designating character group is executed.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a character recognition device that recognizes characters on a form.

(Prior Art) Character recognition devices that read and recognize characters on forms have been widely used in various office processing fields.

The following types of forms are read by such a character recognition device.

FIG. 3 shows a list-type form P1, where a indicates an address entry frame. Also, Figure 4 shows the form P in manuscript paper format.
2, A is a sentence, B1.82.83.84.
85 indicates a kanji compound word, and b indicates a character entry frame.

By the way, when reading a form such as form P1 in which the character entry frames are set in a certain degree of grouping, it is known in advance what kind of classification content the characters are written in the character entry frames. , the validity of the recognition results can be checked.

For example, when recognizing the Kanji proper noun "Tokyo" in the address entry box a of form P1, the recognition result shown in FIG. 5 may be obtained.

Here, as a recognition result, as shown in the figure, there are cases in which multiple candidates are obtained with ranking, and cases in which only a single candidate is obtained, but in both cases, address entry frame a is Since it is known in advance that the written kanji is a kanji that indicates an address, the recognition result is compared with a predetermined word in the knowledge dictionary in which the address word is stored to determine whether it is valid as an address. Accordingly, it is possible to select the third place "Tokyo" shown in FIG. Therefore, in the former case, a valid one is selected and output, and in the latter case, the occurrence of reading errors can be minimized by outputting only when a correct recognition result has been obtained.

However, in such a conventional character recognition device, when a form is used in which the character entry frames of form P2 are consecutive, as shown in FIG. It is not possible to know in advance what will happen. Therefore, when a document in the form of a manuscript paper is used, the above-mentioned validity check of the recognition result cannot be carried out, and there is a problem in that reading errors increase.

(Problems to be Solved by the Invention) The present invention is intended to solve the above conventional problems, and recognizes sentences such as sentences containing kanji written in a form with continuous character entry frames. It is an object of the present invention to provide a character recognition device that can satisfactorily perform recognition processing for word-based character groups such as kanji and compound words, and can significantly reduce the occurrence of reading errors.

[Configuration of the Invention (Means for Solving Problems)] In order to achieve the above object, the present invention provides a reading means for reading a group of characters on a form, and a character for recognizing each character of the read character group. a recognition means, an extraction means for cutting out a specified character group of the same character type from among the character group of the form for which character recognition has been completed, a knowledge dictionary storing a semantically valid character group, and the specified character. and recognition processing means for performing recognition processing on the designated character group by comparing the group with a predetermined character group in the knowledge dictionary.

(Operation) The present invention uses the above means to recognize character groups in units of words such as kanji compound words, even when recognizing sentences such as sentences containing kanji written in a form with continuous character entry frames. Processing can be performed satisfactorily, and the occurrence of reading errors can be significantly reduced.

(Example) Hereinafter, an example of the present invention will be described in detail based on the drawings.

FIG. 1 is a block diagram showing the configuration of a character recognition device according to an embodiment of the present invention.

In the figure, 1 is a reading unit that reads a group of characters, such as a sentence, on a form, and 2 is a reading unit that reads each character of the read character group.
Character recognition unit that performs HLIJ at position p, 3 is an extraction unit that extracts a specified character group of the same character type from the character group for which character recognition has been completed, 4 is a semantically valid kanji idiom, hiragana, and katakana. The knowledge dictionary section 5 stores the characters recognized by the character recognition section 1 and the recognition result of the specified character group (word) extracted by the extraction section 3, and the knowledge dictionary section 5 stores the words such as a recognition processing unit that performs recognition processing by comparing the words with
is an input section for inputting and specifying information on words of a character group to be specified in the cutting section 4, such as kanji compound words, hiragana, and katakana.

The operation of the character recognition device configured as described above will be explained with reference to the flowchart shown in FIG.

First, before processing the form P, information on words to be specified in the cutting section 3, such as kanji compound words, hiragana, and katakana, is input through the input section 6 (STEP 1). Next, the form P is conveyed to the reading section 1, and the characters on the form P are read by the reading section 1 (STEP 2). The image of the read characters is then recognized character by character by the character H'92 unit 2 (S
TEP3).

Thereafter, a specified character group of the same specified character type is cut out by the cutting unit 3 from the character group of the text on the form P (S
TEP4). For example, when cutting out kanji phrases, it is determined whether the recognized characters are kanji or not based on the recognition results of the character recognition unit 2, and only idioms containing two or more kanji characters are cut out.

Thereafter, the recognition processing unit 5 compares the recognition result of the word that is the specified character group with the words in the knowledge dictionary unit 4. If the word in the knowledge dictionary matches the recognition result, this is taken as the final recognition result. On the other hand, in disagreement,
If there are multiple candidates in the recognition result, the next candidate is compared with the word in the knowledge dictionary. In this way, the comparison is performed until the words match the words in the knowledge dictionary (STEP 5).

For J3 kanji compound words with three or more words, if all the candidates in the recognition result do not match the words in the knowledge dictionary, the last word of the character string that makes up the compound word is deleted and the words in the knowledge dictionary are replaced with the words in the knowledge dictionary. Compare with. Then, delete each word from the end and compare until a match is found.

If there is a match, compare the remaining strings in the same way. For example, in the case of the kanji idiom "operation manual", "
``Operation Manual'' → ``Operation Instructions'' → ``Operation Instructions'' → ``Operation'',
"Instruction" → "Explanation" are compared with the words in the knowledge dictionary in this order, and "operation" and "explanation" match the words in the knowledge dictionary.

Therefore, when recognizing form P1 shown in FIG. 4, for example, when recognizing the "structure" of word B2 F composition of sentence A in form P1, "ko" is the first candidate and the correct answer is " 81 "Proposal" from text A, B2 "Composition J, B3, B4" is recognized as the second candidate.
, B5 are extracted by the extraction unit 3.

In this case, the input unit 6 specifies the word to be extracted as a kanji compound word. Then, the recognition processing unit 5 compares the kanji compound words B+, 82, etc. with the words in the knowledge dictionary unit 4. Here, it is recognized that the word B1 "proposal" is correct. Furthermore, since the word B2 "composition" does not exist in the knowledge dictionary 8 part 4, the second candidate "composition" is compared with the word in the knowledge dictionary 2 part 4. Then, it is recognized that "composition" of word B2 is correct, and "composition" is taken as the final recognition result. B3
, B 4 , and B s are similarly recognized.

In the character recognition device configured in this manner, words 81, B;! of the same character type (kanji compound word) designated by the input unit 6 are selected from the text A of the form P1 for which character recognition has been completed. , B
3, B4, BS (character group) is extracted by the extraction unit 3, and the recognition processing unit 5 compares it with the words in the knowledge dictionary unit 4 and performs recognition processing, so that sentences containing kanji written on a form, etc. Even when recognizing sentences, the recognition process can be performed well on a word-by-word basis such as kanji and phrases, and reading errors can be significantly reduced.

[Effects of the Invention] As explained above, the character recognition device of the present invention recognizes sentences such as kanji compound words written in a form with continuous character entry frames. Recognition processing for character groups in word units can be performed satisfactorily, and the occurrence of reading errors can be significantly reduced.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a character recognition device according to an embodiment of the present invention, FIG. 2 is a flowchart for explaining the operation of the character recognition device shown in FIG. 1, and FIG. 3 is a list-type transfer form. FIG. 4 is a plan view showing a form of manuscript paper, and FIG. 5 is a diagram showing proper nouns that are considered to be candidates for recognition results in a general character recognition device. 1...Reading unit 2...Character recognition unit 3...Cutout unit 4...Knowledge dictionary Part 5: Recognition Processing Part 6: Input Part Applicant: Toshiba Corporation, Patent Attorney, Satoshi Suyama - Figure 1 Figure 3 Figure 5

Claims (1)

[Claims] (1) A reading means for reading a group of characters on a form, a character recognition means for recognizing each character in the read character group, and designation of the same character type from among the group of characters on the form for which character recognition has been completed. A cutting means for cutting out a character group, a knowledge dictionary storing semantically valid character groups, and a recognition process for the specified character group by comparing the specified character group with a predetermined character group in the knowledge dictionary. 1. A character recognition device characterized by comprising a recognition processing means for performing recognition processing.