JPS5839378A - Post processing system for character recognition - Google Patents

Abstract

PURPOSE:To perform accurate post processing by weighting candidated characters according to their order and finding the matching with a word dictionary, and finding a word matched best. CONSTITUTION:Candidated characters recognized and outputted by a recognition part 1 are outputted to a character matrix register 5 in the order of the recognition. A matching circuit collates words read out of a word dictionary 8 with the 1st-5th order recognized candidated characters and when one of the candidated character matches with one of words held in the register 5, the matching level which corresponds to the order of recognition is outputted. A matching result output register 9 holds matching levels, the degree of coincidence, outputted from the circuit 7, corresponding to the words. A result decision circuit 10 outputs a word matched best to an output register 11 on the basis of the result of the matching by the circuit 7.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a character recognition post-processing method, and in particular, the present invention relates to a character recognition post-processing method.
After performing character recognition processing with a kanji dictionary (for example, a kanji dictionary), the recognition results are weighted according to their rankings and matched with a word dictionary, so that input words can be accurately recognized. This concerns the second processing method.

In conventional character recognition methods, for example, as shown in Figure 1, I
In the I iron section 1, the features of the input characters are extracted, compared with the file, and the one with the highest M fiber rank is output to the output register 2, and then output to this output register 2 as the character w1w & post-processing. If it is known in advance that the three characters shown represent the prefecture base, these output characters are sequentially compared with the prefecture dictionary 3 in the matching circuit 4 to accurately match the input characters. I'm trying to recognize it.

In other words, in Figure 1, 3 is entered in the prefecture-based entry column.
A data input form (not shown) on which characters have been written is read using OCR (not shown), and based on the nine data obtained, the recognition unit 1 extracts features for each character and selects the highest recognition rank. ”, “Sai”,
This book outputs "prefecture" to the output register 2, sequentially compares them with the prefecture base set in the prefecture dictionary 3 in the matching circuit 4, and outputs the prefecture base with the highest degree of matching as a read output. It is.

However, in such a post-processing method, as shown in FIG. 1, when the recognition unit 1 outputs "Miya", "Sai", and "Prefecture" and matches them with the prefecture based on the nine,
Since ``Miyazaki-kei'' and ``Miyagi Prefecture'' exist with the same priority, it is not possible to automatically select one of them.

To this end, as shown in Figure 2, it may be proposed to output candidate characters of multiple ranks when the recognition unit recognizes an input character, and when the recognition unit recognizes the 3* character Tohenfuken name, the first
Regarding the #i character, as shown in Figure 2, the first rank is "Kana", the second rank is "Chichi", the third rank is "Autumn", and the fourth rank is "Ka".
The rank is "chi", the fifth rank is "bayashi", and for the second character, the first to fifth ranks are "ta", "uchi", "
For the third character, the 1st to 5th positions are ``gu'', ``prefecture'', ``me'', ``and'', and ``dan''. In this case, each of these candidate characters is sequentially compared with the prefecture base. That is, "Hokkaido" is read out gt-th from the prefectural dictionary 3, and the first character "Kita" is compared with the above-mentioned "Ke, Chichi, Autumn, Timber, Forest" to perform matching. Then, the second character ``Umi'' is compared with the above-mentioned ``田, 内, 口, EN, ゆ'', and the third character ``道'' is compared with the above-mentioned ``gu, prefecture, eye, and tan''. A match is made for each, but none of them match. First, the same matching is performed as for the second word "Aomori Prefecture", but the third character "Prefecture" is
When compared with ``gu, prefecture, eye, and dan'', the second choice is ``prefecture''.
A match is obtained. When matching the third word ``Akita-kei'', a match is obtained by matching the first character ``Autumn'' with ``Kana, Chichi, Autumn, Wood, Forest'', and similarly the second word ``Akita-kei''. A match was obtained by matching the character ``田'' with ``田, 内, 口, EN, ゆ'', and by matching the third character ``#'' with ``gu, prefecture, eye, and dan''. Again, a match is obtained.

In this way, all the characters in "Akita-type" match one of the candidate characters, so this "Akita-type" with the highest degree of "Rry chi/g" is output as the read character.

However, when simply comparing candidate characters in multiple ranks, the reading output is reduced to one in the example shown in Figure 3.
It may not be possible to be certain. In other words, for the characters in the 1st book of 1g, the first to fifth recognition rankings are "Miya,"
Government, Wealth, Lu, Space'', and the same for the second character <M#J&) - the position is ``Sai, Saki, Pass, Castle, Earth'', and the third # character Similarly, “prefecture, ingredients, eyes, and,
When this is output from the character matrix register 5 to the rank register 6 for each rank, three characters at a time, and compared with the words in the prefectural dictionary 3 in the matching circuit 4, "
There are cases where "Miyazaki-kei" and "Miyagi Prefecture" have the same matching degree and cannot be classified.

Therefore, in order to improve this problem, the present invention assigns weights to the candidate output from the recognition unit for each rank and matches them with the word dictionary, and then finds the word that is matched with the highest rank. The object of the present invention is to provide a character recognition post-processing method that is easy to use. For this purpose, in the character recognition post-processing method of the present invention,
It is equipped with a character recognition means for recognizing read characters, a word holding means for holding words, and a matching means for detecting that the characters recognized in the character W literacy stage match the words held in the word holding means. In the character recognition post-processing method, the character recognition means outputs recognized character candidates of a plurality of ranks, and the matching means matches the word held in the word holding means with the recognition character candidates of the plurality of ranks, and also outputs the recognized character candidates of the plurality of ranks. When one character of the candidate is matched with one character of the word, a weighting output means is provided which gives a weight according to the utl and face position and outputs it, thereby obtaining a matching ranking according to the recognition ranking. The present invention is characterized in that words having the highest matching ranking are selectively output.

An embodiment of the present invention will be described below based on FIG.

In FIG. 4, the same reference numerals as in other figures indicate the same parts, 7 is a matching circuit, 8 is a word dictionary, 9Fi matching result output register, 10Fi result determination circuit, and 11 is an output register.

The matching circuit 7 matches the word read from the word dictionary 8 with the first word outputted from the recognition unit 1 to the character matrix register 5.
Matching is performed by comparing the recognition candidate characters from the face rank to the 5th iWs rank, and if a matched character exists, it is determined according to its recognition rank. "1" when it matches the recognition candidate character of the second rank, "2" when it matches the recognition candidate character of the third rank, "3" when it matches the recognition candidate character of the fourth rank.
, "4" is output when it matches with the sixth rank, and "5" is output when it does not match with any of the first JIa to fifth ranks.

The word dictionary 8 is a file containing a plurality of word collections necessary for post-processing, such as word collections for prefecture names, and classified word collections for each region and prefecture, such as the names of municipalities, towns, and villages in Akita Prefecture. , words of a predetermined classification are sequentially output according to classification and in a fixed order by a control signal C8 for each word from the matching circuit 7.

The matching result output register 9 is a register that holds the degree of matching between the candidate characters set in the character matrix register 5 and the words output from the word dictionary @8 in word correspondence.

The result determination circuit 10 selects and outputs the word with the highest degree of matching as a result of the matching performed by the matching circuit 7.

Next, the operation shown in FIG. 4 will be explained.

(1) The recognition candidate characters output from the recognition unit 1 are output to the IJ contest register 5 according to their recognition order. For example, for the 1st character of fI, the 1st to 5th ranks of ``Kana, Chichi, Autumn, Wood, Hayashi'' are output,
For the second character, ``田, 内, 口, EN, ゆう'' is output, and for the third character, ``gu, prefecture, eye, and,
"Dan" is output.

Since it is known in advance that the output of the recognition unit 1 is a prefecture name, word collection file portions for prefecture names are sequentially read from the word dictionary 41'8. In this case, the control signal C for each word output from the matching circuit 7
(Yo) First, ``Hokkaido'' is mentioned one after another. Then, "Shidagu" is first set in the order register 6 by the order control signal C1- from the matching circuit 7, and compared with "Hokkaido", but at this time, there is a match only because there is no character in the fourth position. "Chichiuchi Prefecture", which is the second priority, is set in the order register 6 by the order control signal Cs-1 from the matching circuit 7, and is similarly matched with "Hokkaido". In this way, the order control signals Cl-2 to C1-4 from the matching circuit 7 cause the order register 6 to select the third
The ranking ``Akiguchime'' and the 4th ranking ``Hayashi Yudan'' with 5% of J% votes are set sequentially and matched with ``Hokkaido'', but these characters do not match.夛、
As a result, there is only a match because the fourth character does not exist, so "0" is written in (4) K of section 1 of the 1 matching result output register 90, and (1) to (
“5” is entered in 3).

(2) When the matching with the first single @ "Hokkaido" is completed in this way, the matching circuit 7 sends the control signal C! is output, and the second word ``Aohama Prefecture'' is output. Then, the sequence control signal C1-.~C1/%-4 is outputted to sequentially set the 1st) V4 rank "Shidagu" to the 5th face rank "Yudan Hayashi" in the rank register 6, and the above-mentioned "Semori" Matching with "Prefecture".

At this time, there is a match in two points: "Prefecture" in "Chichiuchi Prefecture" in the second face position and the absence of the fourth character, so the matching circuit 7 outputs the i-matching result output register 90 to section 2 (4). "0", "1" in (3), (2) and (1) 4
"5" will be entered for each C.

(3) Next, the third word "Akita-kei" is outputted by the matching circuit 7Fi control signal Cm, and then the (
11, (Same as 21, enter "Shidagu" in rank register 6.
~ "Yutan Hayashi JfJ@Next set and check with this "Akita j. Since a match is obtained for "prefecture" in "prefecture" and "autumn" in the third rank "early autumn" and the absence of the fourth character, (2) in category 3 of matching result output register 9, ( roJ will be added to 4), ``1'' will be added to (3), and ``2'' will be added to (1).

(4) When the matching with all prefecture names is completed in this way, the result judgment circuit 10Fi determines the category with the lowest total score of each category in the matching result output register 9, and selects the category with the highest degree of matching. The selected output will be as follows.

Therefore, in this case, category 3 is the minimum due to the total score of 3, thus fs 3rd prefecture name "Akita-kei"
is output from the output register IIK as the final read output. In this way, it is possible to accurately extract the [Akita type] by post-processing.

In addition, as shown in figure t7J5, the result judgment circuit 10'K
A first manual register 12, a second manual register 13, and a comparison control section 14 are provided, and the matching state of each category from the first manual register 12//c matching circuit 7 is inputted first. When the matching degree of the newly transmitted first human-powered register 12 is large compared to the matching state of the category held in the second human-powered register 13 (in the state shown in FIG. 4, when the total score is small) ) is entered in the second manual register 13, and if the matching degree for the next word is entered directly into the first manual register 12 when it is small, the matching result output register 9Fi in FIG. 4 becomes unnecessary. When the number of matching words read from the word dictionary is large, this can be handled with a simple configuration.

Thus, according to the present invention, as shown in FIG. 6, even when "Miyazaki Prefecture" to "Chujidan" appear in the character matrix register 5 as the first to fifth rankings, the matching circuit In 7, the total score is "1" and 61" people are entered in category 9-0 of the prefecture stationary star, but for "Miyagi-kei", the total score of category 9-xaK is also entered as "3". Therefore, ``Miyazaki Prefecture'' will be output as the post-processing result.

To explain the present invention generally, as shown in Figure μ8, an input word f Lt yllg ,...L% (a word is composed of duck characters) is assumed, and a recognition candidate for Ll is Lt(11eLt( L +++ L>(y (when mR candidate and L"r 1st JQi position to 5th face position are adopted). Also, weight is set to Wi for 1st application position to 5th application position)
ll ~ Vdb+ and for those below 6th place ■・
), Wllll (Wd... Express the degree of difference between 81,811.8% and the input word LxeL-...Ln as the sum of the degrees of difference for each character.The difference for each character Ll, Ls... and Ls. For example, when considering sl, the recognition candidate L 1 (when matching is obtained with 41, the difference f
d(Sx, L is N4+.If there is no matching 11! compensation, the degree of dissimilarity will be Wsu.Therefore, the degree of dissimilarity of the word is D-, L' d(sii, Ll) 11, and the word in the word dictionary with the minimum D is determined to be correct.

In addition, in Kamiki Eiaki, an example was explained in which candidate characters were selected up to the fifth rank, but this is not limited to this, and an example was explained in which a small number was given to a candidate character with a high rank, but the opposite The same applies to cases.

As described above, according to the present invention, clear post-processing can be performed by weighting candidate characters according to their candidate ranks and calculating the sum of the weights for each word.

[Brief explanation of drawings]

Figures 1 to 3 are schematic diagrams of conventional post-processing, Figure 4 is a configuration diagram of one embodiment of the present invention, Figure 5 is another embodiment of the result judgment circuit, and Figures 6 to 8. FIG. 2 is a detailed explanatory diagram of the present invention. In the figure, 1 is the g fiber section, 2 is the output register, 3 is the prefectural dictionary, 4 is the matching circuit, sFi character matrix, register 2, 6 is the order register, 7 is the matching circuit, 8il
t word dictionary, 9ii matching result output register, 10
11 indicates a result determination circuit, and 11 indicates an output register. Patent Applicant Fujitsu Limited Representative Patent Attorney Akira Yamatani

Claims (1)

[Claims] (1) Read the characters I! The character recognition means includes a word holding means in which a word is held, and a matching means for detecting that nine characters recognized by the character recognition means match a word held in the word holding means. In the character recognition post-processing method, the character recognition means outputs the gw & character candidates of the numeric rank, and the matching means matches the word held in the word holding means with the m-character candidate of the plurality of ranks, and the recognition When one character of the character candidate is matched with one character of the word, a weighting output means is provided which assigns and outputs a weight according to the recognition rank, thereby obtaining a matching order position according to the recognition rank. A character recognition post-processing method is characterized in that the word with the highest matching order position is selected and outputted.