JP2795003B2 - Character recognition processor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character recognition processing device and, more particularly, to an address recognition processing method for collating an address read result in an optical character recognition processing device (OCR).

[0002]

2. Description of the Related Art Japanese OCR has been rapidly spreading in recent years, but it is difficult to read all characters correctly due to insufficient character reading performance and limited character types to be read.

For this reason, post-processing of character recognition for confirming and correcting the result of character recognition using a group of words registered in advance is indispensable for Japanese OCR.

[0004] In the post-processing of character recognition related to an address, the nearest address is selected from hundreds of thousands of addresses by using a group of address words registered in advance. Can also be easily corrected using a word dictionary, and at present, a great effect is achieved.

[0005] In particular, in the case of an address, the address of an individual or a corporation often relates to an apartment house (apartment or condominium), and the name of the apartment house is written in the address column in addition to the ordinary address. Is becoming frequent.

[0006] Matching a registered word with a registered word for such an apartment house in the same manner as in the prior art involves not only the difficulty of obtaining a database but also the enormous storage capacity and the fact that many apartment houses are described with pseudonyms and numbers. However, simple word matching is difficult due to the fact that many candidates appear and the performance is not sufficient.

[0007] In addition, if a process for exclusive use of an apartment complex is not performed, sufficient performance cannot be obtained, and it is necessary to reduce the number of reference word groups at the same time.

[0008]

SUMMARY OF THE INVENTION It is an object of the present invention to provide a character recognition processing apparatus capable of performing accurate address recognition by minimizing the number of reference words stored in an apartment house in an address recognition process including an apartment house.

[0009]

According to the present invention, there is provided an address having an address collating means for collating a character code group of an inputted character recognition result with a group of address words stored in advance to confirm and correct the character recognition result. A character recognition processing apparatus, wherein an unprocessed part of the character code group, which has not been confirmed and corrected by the address matching means, is compared with an address matching result by the address matching means to search for a place name part. Place name matching means, tail matching means for confirming and correcting the last character code group of the unprocessed portion, and consecutive numbers and alphabetical character strings from the character code groups which were not confirmed and corrected by the processing of the tail matching means Numeral matching means for extracting
And a housing code matching means for checking and correcting the character recognition result of the character code group, which has not been subjected to the correction by the number matching means, with a pre-registered collective housing word group. As a result, a character recognition processing device is obtained.

[0010]

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

First, FIG. 2 shows five representative examples of addresses including the name of an apartment house as A to E. In the figure, the underlined portion is the address portion, and the symbols to are the collective housing portion, respectively.

The collective housing section includes a place name section, a tail section, a room number section, and a collective house word section. In addition, the word section of the apartment complex is composed of general words -A, unique words -B, first name and last name words-
It can be classified into C.

As for each category item of an apartment house, not all of the above items always appear in the apartment house section, and some items do not appear in the apartment house section. In addition, there is a difference in difficulty of matching depending on the item.

Therefore, the collation processing procedure of the apartment house is inevitably determined by these conditions, and FIG. 1 is a functional block diagram of the embodiment of the present invention thus obtained.

The character recognition result is input as an input character code 11. With respect to the input character code 11, the address part is collated, confirmed, and corrected by the address collation part 12, and the address part is determined. Further, when it is determined that the apartment house is described, the apartment house collation unit 13 performs collation, confirmation, and correction of the apartment house unit, and outputs the result as the collation result 14.

In the apartment house collating section 13, first of all, the place name section of the apartment house section of the input character code is collated by the place name collating section 21 using the collation result 12 of the address section. next,
The collation at the end of the unchecked input housing unit is the tail collation unit 2.
2 is made. Next, a character string search of the alphabet and the numbers consecutive from the end of the unmatched character is executed in the number matching unit 23, and the collation between the input character code portion remaining at the end and the group of pre-stored collective housing words is performed by the collective housing word. This is performed in the matching unit 24.

Since the address collating unit 12 can be realized by using a known technique, the details are omitted here.

The character recognition post-processing device having the above-mentioned configuration includes a CPU (Central Processing Unit), a storage medium for storing various processing programs and a group of reference words, and a RAM (random) for temporarily storing various data. Access memory)
And the like.

FIG. 3 shows an example of a code group input as an input character code of this embodiment. The line with the correct answer is the entry, and each line with the input character code is
Is an input character code 11 and is a character recognition result. As an example, here, character recognition ranks first to
Six candidates in the sixth place are to be output.

The address collation unit 12 collates with an address word group registered in advance. FIG. 4 shows an example of the address word group, and FIG. 5 shows the result of collation.

FIG. 4 schematically shows an example of a storage format of address words. One frame corresponds to one word, and an arrow represents a structure in which addresses can be searched hierarchically.
FIG. 5 shows the processing result of the address collation unit.

Hereinafter, in the line of the judgment result, the part judged in each applicable process is underlined, and the unmatched part is represented by a blank. Therefore, in the present operation example of the address collating unit 12, the words "Setagaya-ku" and "Minami Karasuyama" are collated according to the address collation result.

In the place name matching unit 21, the unmatched portion, that is, the apartment house portion in the house matching unit 12 is compared with the word of the address matching result of the address matching unit 12, and the same word exists in this apartment unit. It is searched from the input character code of the housing complex.

In this example, the two sets of "Setagaya-ku" and "Minamikarasuyama"
A search is performed using the word, and “Minami Karasuyama” is detected using the position of the arrow in FIG. 6 as the head position of the word. Next, at the collation position indicated by the arrow, the first collated character and the last collated character in the word are checked from the beginning and end of the word, respectively, and the interval between them is determined as the result of the place name collating unit.

In this example, as shown in FIG. 6, "crow" is the first collated character, "mountain" is the last collated character, and "karasuyama" is collated. Become.

The tail collation unit 22 detects characters such as "No.", "Room No.", "Waka" and "Samakata" which are usually written at the end of the housing unit from the end of the housing unit. FIG. 7 shows the result after this processing.

In the numeral grouping unit 23, a continuous alphabet,-(hyphen) from the end of the part which has not been collated at this time.
Alternatively, a character string of numbers is extracted. The number of extracted characters is limited to a maximum of 5 characters when there is a-(hyphen) and to 4 characters when there is no-(hyphen), and processing is continued until these characters are detected within this range. FIG. 8 shows an example of this processing.

The collective housing word collating unit 24 verifies the unverified portion of the input character code using the collective housing word group as shown in FIG. FIG. 10 shows an example of this processing. Here, the collective house word group is classified in advance into a general, collective house registered word group, and a unique collective house registered word group as shown in FIG.
By selecting one or less words for the general apartment house registered word group and a plurality of words for the unique apartment house registered word group, only one general word such as an apartment / apartment can be determined in the same description. , Character recognition with less contradiction becomes possible.

After the processing of the collective housing word collating unit 24, the collation of the first and last name words is added by using the first and last name processing dictionary that is often used as post-address character recognition processing, so that the first and last names as shown in FIG. It is also possible to check the apartments used.

In the processing of the place name matching unit 21, an address word group (address matching unit 1) is determined based on the word of the address matching result.
If the reverse search is performed on the dictionary used in (2), the related word can be selected from the address word group. As a result, as shown in FIG. 2C, it is possible to recognize a place name (Tokyo), which is a word different from the address part, used in the apartment house part.

Further, in the place name collation by the place name collation unit 21, if the nearest station name is described in the address word group together, the nearest station name is searched from the address collation result, and the station name as shown in FIG. It is also possible to collate apartment houses that seem to have used.

[0032]

As described above, according to the present invention, only a dictionary in which a group of apartment house words is registered is newly added, and the others use existing address word groups, first and last name dictionaries, and the like. There is an effect that character recognition of an address including an apartment house becomes possible.

[Brief description of the drawings]

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

FIG. 2 is a diagram showing a description example of an address including an apartment house.

FIG. 3 is a diagram showing an example of an input character code and an example of a correct answer.

FIG. 4 is a diagram showing a storage example of a house word.

FIG. 5 is an example showing a processing result of a house collation unit.

FIG. 6 is a diagram showing a processing result of a place name matching unit.

FIG. 7 is a diagram illustrating a processing result of a tail collation unit.

FIG. 8 is a diagram showing a processing result of a number collating unit.

FIG. 9 is a diagram showing an example of an apartment house word group.

FIG. 10 is a diagram showing a processing result of an apartment house word matching unit.

[Explanation of symbols]

 11 input character code 12 address collation unit 13 apartment collation unit 21 place name collation unit 22 tail collation unit 23 numeric collation unit 24 apartment housing word collation unit

Continuation of the front page (56) References JP-A-2-77988 (JP, A) JP-A-61-234481 (JP, A) JP-A-63-153689 (JP, A) JP-A-2-181287 (JP) , A) JP-A-2-173885 (JP, A) Proceedings of National Convention of Information Processing Society of Japan, Vol. 38TH (1989), NO. 2, PAGE. 649 Proceedings of the Information Processing Society of Japan, VOL. 41TH (1990), NO. 2, PAGE. 2-142 (58) Field surveyed (Int. Cl. ⁶ , DB name) G06K 9/72

Claims (5)

(57) [Claims] 1. An address character recognition processing device having an address collation unit for collating a character code group of an inputted character recognition result with an address word group stored in advance and confirming and correcting the character recognition result. A place name matching unit that searches for a place name part by comparing an unprocessed part of the character code group that has not been confirmed and corrected by the address matching unit with an address matching result obtained by the address matching unit; Tail collating means for confirming and correcting the character code group at the end of the unprocessed portion, and numeral collating means for extracting continuous numeric characters and alphabetical character strings from the character code group which has not been confirmed and corrected by the processing of the tail collating means. And a character code group for which confirmation correction has not been performed by the processing of the numeral matching means,
A character recognition processing device, comprising: a collective housing word collating means for collating with a pre-registered collective housing word group and confirming and correcting the character recognition result. 2. A multi-unit housing word collating unit for performing verification and collation of a character recognition result by a general multi-unit housing registration word group consisting of pre-registered general multi-unit housing words, 2. The character recognition processing device according to claim 1, further comprising a second apartment house word collating unit for confirming and collating the character recognition result by the unique apartment house word registration group composed of the unique apartment house words. 3. The place name matching unit, wherein the place name match unit searches the address word group based on the address matching result by the address match unit and selects a place name match word, and the selected place name word. 3. The character recognition processing device according to claim 1, further comprising: a place name collating unit that performs confirmation correction of the character code group. 4. The place name matching means, wherein: an address dictionary to which station name information of a nearest station corresponding to each address is added; and a nearest station corresponding to a related address by searching the address dictionary based on the address matching result. 4. The character recognition processing device according to claim 1, further comprising a place name word generation unit for selecting a word. 5. A method according to claim 1, further comprising adding a first name and a second name checking means for confirming and correcting a character recognition result by a first and last name dictionary after the processing of the apartment house word checking means.
Or the character recognition processing device according to 4.