JPS63143685A - Recognition result display method in character recognition device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a recognition result display method in a character recognition device.

(Technical Background) As a character recognition device that recognizes character information in a document through computer processing, the character information to be recognized, for example, alphanumeric characters, is photoelectrically converted, and the photoelectrically converted electrical signal is cut out into one character type and recognized. The optical character reader (OCR) recognizes each character one character at a time according to predetermined recognition logic.
)It has been known.

In this type of character recognition device, conventionally, when a recognized character has a low correctness rate and is judged to be suspicious, so-called rejected, the rejected character is displayed on a display unit using a cathode ray tube (CRT), etc. The operator corrects the rejected characters using a correction means such as a keyboard while checking the display and comparing the rejected characters with the original.

However, since the above-mentioned conventional method basically displays the result of recognition processing for one character type, the recognition rate is generally low, that is, the rejection frequency is high, and therefore, a great deal of effort is required for correction. The recognition work efficiency was still unsatisfactory.

(Section M to be solved) The present invention has been made in view of the above problems, and performs spell check and spell correct processing on recognized character strings word by word to effectively reduce the work of correcting recognition results. It is also an object of the present invention to provide a recognition result display method in a character recognition device, which can improve the efficiency of the correction work and make the recognition work highly efficient.

(Structure) In order to achieve the above object, the recognition result display method of the present invention displays the recognized character string from the recognition unit word by word when displaying the result recognized character by character in the recognition unit on the display unit. Spell check and spell correct processing are performed for each extracted word, and when the existence of the corresponding word is denied as a result of the spell check processing, all characters of the word are highlighted and corrected by spell correct processing. This feature is characterized by displaying all characters of the word with an underline.

The present invention will be explained below with reference to the accompanying drawings.

(Embodiment) FIG. 1 shows a block diagram of an optical character reader (OCR) to which the recognition result display method of the present invention can be applied.

In FIG. 1, reference numeral 1 denotes an input section, and this input section 1 has a photoelectric conversion type image caster 2 using, for example, a C0D (charge-coupled device) and an image memory 3. This image caster 2 can scan documents placed on a document table (not shown).
For example, an alphanumeric character image is read and photoelectrically converted, and an electrical signal representing the character image is stored in the image memory 3.

Reference numeral 5 denotes a character recognition section using, for example, a microcomputer, which recognizes character information from the human power section 1 according to a recognition logic formula stored in a ROM (read only memory), not shown.

In the one-line cutting section 6 of the character recognition section 5, the image signal in the image memory 3 is cut out at one-line intervals using a known method, so-called segmentation is performed, and this cut-out character information is stored in the one-line memory 7. be done.

The character information stored in the one-line memory 7 is extracted character by character by a character extraction unit 8 based on the interval information between each character analyzed using, for example, an image density histogram. Each of the extracted characters is then recognized character by character in the character recognition section 9 while referring to a character dictionary 10 storing predetermined alphanumeric characters, and sent to the word memory 11. Further, the character information output from the one-line memory 7 is simultaneously sent to the word segmentation unit 12, and based on the delimiter information between words extracted in the word segmentation unit 2, the character information output from the one-character recognition unit 9 Each character is stored in the word memory 11 in units of words.

Reference numeral 15 denotes a language processing section that performs word spell checking and spell correcting processing, and is composed of a spell checking section 16, a word dictionary 17, and a spell correcting section 18.

Words are sequentially read from the word memory 11 to the spell check section 16 of the language processing section 15, and the spell check section 16 performs spell check processing on the read words. This spell check processing is performed by determining whether or not a word read out from the word memory 11 completely matches a word stored in a word dictionary 17 that stores predetermined English words.

When the first spell check result in the spell check section 16 is determined to be "no", that is, it is rejected, the spell correct section 18 is activated.

This spell correcting unit 18 selects a plurality of word candidates by correcting characters with a low recognition rate among the characters that make up the rejected word based on the recognition rate data of each character in the single character recognition unit 9. do. For example, five word candidates are selected from first to fifth in order of successive success rates. This process of selecting word candidates is referred to as a collect process in this specification.

When all the word candidates selected by the above-mentioned correcting process are rejected by the spell check section 16, the word is rejected as there is no possibility that a word representing the recognition result exists, and the word 20 is rejected based on the character recognition result. This is an output unit that displays and outputs the result data. The spell check section 16 is stored in the output memory 21 of this output section 20.
The processing result, that is, the recognition processing result is stored, and the recognition result is displayed on the display section 22 using, for example, a cathode ray tube (CRT).
and is sent to a device 24 external to the recognition device, such as a translator or word processor. The output memory 21 is connected to a spelling correction means 23 using a keyboard or the like, and the operator spells the word while looking at the character image of the word displayed on the display screen of the display section 22 via the spelling correction means 23. It can be modified at will.

Next, the recognition result display method of the present invention will be explained with reference to the operational flowchart of FIG.

In step 1, when a document to be read, for example, a handwritten English document, is set on the document table (not shown) of the reading device and a recognition operation is started, an image of the document surface is read through the image scanner 2. , a signal representing the read character information is manually input to the image memory 3.

In step 2, the image information in the image memory 3 is read out, and the one-line cutout unit 6 uses a known method to
Extraction or segmentation is performed at administrative reform level,
Further, the character information is cut out character by character. Each character cut out in single character units is recognized by a known method while referring to the character dictionary 10.

In step 3, the character string recognized character by character as described above is segmented word by word by a known method via the word segmentation unit 12, and each segmented word is stored in the word memory 11. .

In step 4, each word is read out from the word memory 11 to the spell check section 16 of the language processing section 15 and spell check processing is performed. This spell check processing is performed by determining whether a word that completely matches all the characters of the read word exists in the word dictionary 17. If the result of this spell check process is "Yes", the word is recognized correctly in step 5, and its image is displayed on the display unit 22 without any spelling correction.

On the other hand, as a result of the spell check process in step 4,
If it is determined to be "absent", a collect process is performed in step 6, and word candidates with high possibility or probability are selected. This collection process is performed by the language processing unit 15.
This is executed in the spell correcting section 18 of.

All word candidates selected by the spell correcting process are spell checked in step 7 in the same manner as described above, and a search for possible words is performed. As a result of this search, if a possible word is found, that is, a word candidate that completely matches any of the words registered in the word dictionary 17, the word is displayed on the display section 22 in step 8. All candidate characters are displayed with an underline. On the other hand, if a possible word cannot be found, in step 9, all the characters of the word before the spell correcting process, that is, the recognition results for each word, are displayed in reverse on the display section 22.

FIG. 3 shows an underlined display example (A) and an inverted display example (B) in steps 8 and 9 above.

In the above embodiment, the characters to be recognized are alphanumeric characters, but by using an appropriate recognition logical formula, the present invention can also be applied to kanji and kana characters.

(Effects) As is clear from the above description, according to the present invention, recognized character strings are spell-checked and spell-corrected word by word, and words containing almost completely misrecognized characters are automatically corrected. Since the operator is able to perform correction work on a word-by-word basis, the correction work can be effectively reduced compared to the conventional method, which corrects on a character-by-character basis. .

In addition, while highlighting words that include almost completely misrecognized words,
Words with a high success rate (in other words, words that are likely to be correct) are automatically corrected and displayed with an underline, allowing the operator to quickly perform the required tasks with just a glance at the display screen. Therefore, the efficiency of recognition work can be effectively increased.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an optical character reading device to which the method of the present invention can be applied, Fig. 2 is an operation flowchart of the device shown in Fig. 1 using the method of the present invention, and Fig. 3 is a block diagram of an optical character reading device to which the method of the present invention is applied. Examples of inverted display and underline (-j) display on the display section of the apparatus shown in FIG.・・1 character cutting part,
9...1 character recognition unit, 10...character dictionary, 12...
・Word extraction unit, 15...Language processing unit, 16...
Spell check section, I7... Word dictionary, 18... Spell correct section, 20... Output section, 22... Display section, 23... Correction means, 24... External device, A...・
Underline display, B...Reverse display.

Claims (1)

[Claims] (1) In order to display the results recognized character by character in the recognition unit on the display unit, the recognized character string from the recognition unit is cut out word by word, and spell check and spell correction processing is performed for each word cut out. and when the existence of the corresponding word is denied as a result of the spell check process, all the letters of the word are displayed in reverse video, and all the letters of the word corrected by the spell correct process are displayed with an underline. A recognition result display method in a character recognition device characterized by: