JPS6027938A - Character string comparing device - Google Patents

Abstract

PURPOSE:To prevent an omission of data to be retrieved by applying categories of similar, coincident, and anticoincident character strings while comparing a character string with a reference character string, character by character, by a character comparing device. CONSTITUTION:The reference character string and character string to be compared are stored in shift registers 5 and 6, and compared with each other, character by character, by comparators 1-4. A similarity counter 8 decides on complete coincidence when the counted value is 0 or semicoincidence when 1 when all objective characters are compared or when the cumulative value exceeds a preset value. For example, an evaluation bit 0 is added to a completely coincident character string and an evaluation bit 1 is added to a semicoincident string; and the character string of the register on the side where the compared character string is stored is stored in a memory for storing a character string after retrieval.

Description

[Detailed Description of the Invention] Field of the Invention The present invention relates to a character string comparison device, and in particular, to a device for comparing character strings, which is effective in a system that attempts to search documents organized by adding keywords using keywords as a medium. This invention relates to a method for extracting similar character strings.

<Prior art and background> Rather than using a high-level judgment method such as meaning judgment, a match judgment method is used to input a search string from data already filed in a storage device into a specified register,
A search system that compares character strings extracted from sequentially filed data and extracts matching strings is often used, for example, in literature searches, access to conversion dictionaries in kana-kanji conversion, and the like.

However, in such a system, if we look at the interface between man and machine that is somewhat unnecessary, even if unnecessary things are allocated in such a system, humans can easily decide and exclude unnecessary things. The idea is that the disadvantage of not being able to allocate similar items is greater than the disadvantage of having something mixed in.

Conventionally, methods for detecting similar character strings that are similar to the above idea include ■Comparison by forward match, ■Comparison by backward match, and comparison by 0-middle match, etc., which always match within the entire string on the comparison standard side. In addition to the method of detection by specifying and comparing the parts that need to be changed, and the idea of similar groups, changes in expressions are memorized in a dictionary format in advance and a match search is performed using character strings as criteria. In this case, a match search for similar groups is also performed, but in the former method, a large amount of unnecessary data will be generated unless you consider in advance what kind of similar character strings will be included and specify the matching parts. The latter is a substring specification standard character string assignment that is different from the data source that stores in advance the differences in expressions as possible similar groups corresponding to the character string you want to search, synonyms, conjugations, phonetic expressions, etc. However, the disadvantage of this dictionary is that the bjk of this dictionary is too large to be used as a device, and for the search purpose I, it is necessary to prepare a dictionary for substrings. In view of these problems, the present invention has the disadvantage of having to select one character string and use it as a standard character string for comparison. This is intended to provide a method to prevent the omission of data that should be used for searching, and in the case of searches, the differences in word notation are based on the fact that in kana expressions, most of the characters are omitted and replaced. While comparing the strings to be compared character by character against the columns, those included in these categories are semi-matching string lengths with differences of 2 or more characters and characters with the same length - 2 characters in the strings.
Of course, by providing a character string comparison device that applies the category of mismatching two or more characters and determining that there are two or more pairs of consecutive transposed characters as mismatching, it is possible to solve the problem of exhaustive laziness to some extent. Based on the above categories seen as a means of achieving this, it is possible to obtain all valid data fairly efficiently.
The means for realizing the determination is extremely simple and simple compared to the conventional method described above.

Therefore, the purpose of the present invention is to provide a practical character string comparison device that can determine whether or not character strings match in the above categories. first and second shift registers for storing character strings and compared character strings; A comparator that compares the characters stored in the second shift register with characters in the same order, a comparator that compares characters that are out of order, and the results of these comparisons also include matches and mismatches for each character. , a determination circuit for determining similarity, a similarity counter for accumulating the number of similarity determinations of the determination results, and a shift register (7), and a character string is compared by comparing characters while sequentially shifting them. As a result, those in the same order and the same length are determined to be matching character strings, and those whose accumulated similarity is less than or equal to a predetermined value are determined to be semi-matching character strings.

In addition to the comparator, there is also a pair of character type detectors for detecting the character types of the characters in the same order. The determined matching character strings and quasi-matching character strings are stored in a form that is at least distinguishable. 2
, 3.4 are comparators, and 5 and 6 are first and second comparators, respectively.
This is a shift register, and the delimiter in the shift register means a block that stores the code for characters.The comparison string on the comparison standard side and the compared string on the side to be compared are written one character at a time, starting from the left side in the figure. It is assumed that each character string is stored in each shift register, and each character string can be shifted leftward one character at a time by respective shift pulses from a control circuit (not shown).

Numeral 7 is a judgment circuit, which performs a suki evaluation judgment on the match/mismatch judgment results compared by the comparators 1 to 4; Count the matching ratings and the cumulative discrepancy rating during one character string comparison is 0.
It is a similarity counter that counts 1 as an exact match when there are 1, a semi-match when 1 is a match, and a mismatch when 2 or more. The code for each shift register is placed at the beginning (
If the letters on the left side in the figure are matched and transmitted, then comparator 1 is accommodated at the beginning, and character-to-character comparator 2 is 2.
Comparator 3 compares the characters stored at the beginning of shift register 5 with the characters stored in the second row of shift register 6, comparator 4 compares the characters stored in shift register 5 with the characters stored in the second row of shift register 6.
The second character stored in the shift register 6 is compared with the first character stored in the shift register 6, and if they match, the result is 0, for example.
] If there is a mismatch, then ■] is output, and the outputs are indicated by a l b l e l d, respectively. Also, the judgment circuit 7
In order to compare the next characters in the string, the character string stored in the shift register is shifted to the left in character units using shift pulses e and f respectively.Evaluation judgment for each comparison regarding the degree of binding similarity. The operation of the determination circuit 7 is shown in a truth table based on one or more definitions where the resultant output is g.

Table 1 In other words, if both a and b match (α), it is clear that there is a complete match of the same characters and the same order for the two characters, so the character strings in the upper and lower shift registers 5 and 6 are as shown in e and f. Send two of each and compare the strings in the following order.

Also, the complete mismatch [all (9)] with a r b HC + d + is outside the category of similarity, so at that time, the character strings are mismatched and the judgment 24+ is raised. The comparison is discontinued.

What, if either a or b is (1), and a
, b are both ■], but one of c and d is α], assuming that the characters being compared on the register are one character ahead and behind the other, and if C is (1), then e is 1. If d is [1], set f to 1 and send one character of each shift register, and the similarity counter 8
0] is sent as the signal output g to be output, and if [1] is obtained for both a and b as the comparison output after the above shift, then shift by two characters and perform the next comparison between characters. Execute until the end of the string.

In addition, either a or b is [υ and c and d are both Φ]
In the case of , it is determined that one character is different and the same, and 1 is added to the similarity counter 8, and both e and f are set to 1, and the shift registers 5 and 6 are
The next comparison is performed by sending the characters of -character by character.

Also, both a and b are ■], and both c and d are [1]
In the case of , it is determined that the order of characters for one consecutive pair has been changed, and 1 is added to the similarity counter 8, and e is executed.

The next comparison is performed by setting both f to 1 and moving the shift registers 5 and 6 ic forward by one character.

The above procedure for comparing characters is repeated until the character strings end or the accumulated count value of the similarity value exceeds a preset number (in this example, if the cumulative value is 2 or more, there is a mismatch). ) Comparator for one character string +11
6 is completed (completed or discontinued) If the count value of similarity counter 8 is 0, it is determined to be a complete match, and if it is 1, it is determined to be a semi-match.For example, a complete match is determined as Add the evaluation bit ■], and if there is a near match, add the evaluation bit α] to the string in the register that contained the compared string, and store the convex string. At the same time, the next character string to be compared is stored in a shift register for the next comparison. Although we set the cut-off limit of the similarity counter to 2, depending on the purpose of use, it can be set to 3 or more, and there is no distinction between semi-matches such as matches with similarity 1 and matches with similarity 2. Similarly, if the identification bit is known, the determination criterion can be set to 7$ as the tracking standard.

Further, in order to shorten the evaluation procedure, in this example, a similarity counter is used to determine whether or not a determination has been made and the process is immediately terminated, but it is of course possible to compare even the text of the character strings.

FIG. 2 is an explanatory diagram of another embodiment of the present invention, in which numerals 11 and 4 are comparators, 15 and 16 respectively, a shift register 17 is a judgment circuit, and 17m is provided in the judgment circuit 17. 1-bit memory 18 has a similar comparator omitted, and the comparison result input to the judgment circuit 17 is 2 a and d.
The purpose is basically the same as in Figure 1, except that the criteria for producing the evaluation output g or h of the judgment circuit and the control criteria for sending the shift register 5.16 to the result are different. , just by counting and annotating the number 10, it was number 1.
According to the diagram.

The truth table of the determination circuit is shown below as Table 2.

The notation that indicates a match is [1] and the mismatch is (9), and the notation that indicates the shift amount shall conform to the donation in Table 1.

Table 2 As is clear from Table 2, comparators 11 and 1 are located in shift registers 5 and 16, respectively, with their heads aligned on the left side in the figure and containing the comparison string and the compared string. As in the case of the embodiment shown in the figure, the comparison will start from this state.

From this state, the comparison begins and if the current output a of the comparator 11 is [1] (match), the output d of the comparator 1η
Since the characters in the same row match regardless of [1] or ■], the judgment circuit 17 judges that there is a match, and both c and d are 1.
That is, both shift registers are sent one by one and the next character is compared. If the output signal a is [■] and d is [1], there is a possibility that the order is out of order, that they have been switched, or that they are similar patterns, so e = 1 f =
2 That is, the shift register 15 performs 1 shift and the shift register 6 performs 2 shifts, and the next comparison is performed, and the output g is set to 1, and 1 is added to the similarity counter. Furthermore, if the output signals a and d are both (9), there is a possibility of similar patterns as well as a possibility of mismatch, so e = f
= 1 In other words, the upper and lower shift registers are advanced one by one to perform the next comparison, and in this case, 1 is set in the 1-bit memory M provided in the judgment circuit 17', and 1 is also added to the similarity counter 18. do.

If a=(1) in the next comparison, the 1-bit memory M is reset.

Note that after the above state a=d-■] appears once and the 1-bit memo IJM is set to 1, the above a='d-(6) appears during the comparison outside the string. In this case, it is assumed that there are two mismatches in one character string.
The method +1A is to reset the bit memory M to 0 and add a value larger than the count value that is determined to be similar to the similarity counter based on the output signal to determine a mismatch, and then compare the character strings. Abort #) End.

Note that the similarity tolerance limit is preset for the value counted by the similarity counter 18, and if the tolerance value is exceeded during the above comparison procedure, a mismatch is determined. In the case of a match, a=α], so the value of the similarity counter 18 is zero, and if one character is missing or missing, the similarity is 1 or 2. Similarity is 1 or 2 for character strings of the same length but in the same order with 1 character difference
In this way, it is not the same as the first embodiment in terms of classification ability, but when it comes to matching character strings and similar character strings, the same function can be realized by a simpler comparison means.

It should be noted that in such character string comparisons, it is known that if you try to match similar character strings as mentioned above, some unnecessary data (data noise) will be added. It is also true that there are character strings whose meanings differ greatly just by changing the order, and although these similar groups are allowed to some extent for purposes such as searches, they are not very obvious. If something can be easily removed, it is better to exclude it in advance.

One of the typical character strings that causes this [case where the onion flavor differs greatly] is a number, and when expressed in terms of scale, no deviation in the order is allowed.

Another typical example is ``aunty'' with a negative word.
, "Fanti", "Itchi", [Twitchj.

They are called ``Touki'' and ``Hitouki''.

If it is possible to remove even some of the unnecessary data noise from character strings that are susceptible to data noise using a relatively simple determination means, then that method can be said to be a practically effective means.

FIG. 3 is an explanatory diagram of a third embodiment in which some additional attribute determination is added to the embodiment of FIG. 2 in the above sense. The configuration is similar to the one in Figure 2, and the numbers 100 and 20 are displayed in correspondence, but the characters are packed into shift registers 25 and 26 with the beginnings lined up on the left side compared by the character string comparator 21. The first difference is that the character type detectors 29 and 30 are provided for the characters at the beginning of the registers in the character string, and the respective character type determination outputs are set to jk, and the determination logic of the determination circuit 27 is slightly different. This is different from the configuration in Figure 2.

The operation truth table of FIG. 3 is shown in Table 3.

Table 3 The notation in the above table is basically the same as Table 1 and Table 2, but the outputs of the character type determiner 29 and 30 are expressed as ]Other 34
1 go will be explained as ■].

To explain the comparison in the case of Fig. 3 based on the above settings, when the output a of the comparator 21 is [1], the output d&L
Regardless, it is a match, so as in the case of FIG. 2, one /7 register is sent as a match and the next comparison is made. Note that in this case, the result J of character type discrimination is also irrelevant.

Furthermore, even if the output a is (9), if the output d is 0], the first character of the shift register 25 and the next character of the shift register 26 match, so the signal K (the first character of the shift register 26 If the character type) is ■] (not a number), e=1. When f=2, one extra signal is sent from the shift register 26, and the comparator 21 compares the second character of the shift register 25 with the third character of the shift register 26, determines the character type, and uses a similarity counter. 1 in 28
Add.

Note that signal a is (9) and signals J and K are both [1]
In the case of (number), since there was a discrepancy between the character types called numbers, a discrepancy signal was sent to the similarity counter regardless of other conditions, and subsequent comparisons were eliminated and the discrepancy evaluation was confirmed. The comparison between the strings ends.

Note that signal a is [■], signal K is [1], and signal J is (9
), the second shift register 25 and the first shift register 26! Since there is a possibility that the characters located at the eyes match, e=1. f = o Send one character only to the stingy shift register 25 side, perform the next comparison, add 1 to the similarity counter 28, and set 1 bit memory M to α]
Set to .

Furthermore, if the signal a continues to be ■ in the next comparison, the mismatch evaluation is determined to be confirmed due to two consecutive mismatches, and the subsequent comparisons are discontinued, as in the example in Figure 2, and the next comparison signal a is In the case of (1), the wrinkles in the 1-bit memory M are reset, as in FIG.

When the signal a r d r j + k are both ■] (
(It is not a number and does not match), consider the possibility that there is a difference in one character and use e = 1f = 0, that is, shift register 25.
A procedure is performed in which only one is sent, the next comparison is made, the 1-bit memory M is set to [1], and the similarity counter 28 is incremented by 1.

In addition, if the signals a and d are both [■] and the signal J is [1] (number), it is clear that the characters as numerical data do not match, so in this case, the mismatch signal is also sent to the similarity counter 28. It is determined that the feed does not match.

To summarize the above, when the signal a is other than α], the similarity counter 28 always receives the addition value from the signal g or h, and in the case of a decision, the inconsistency signal alone calculates the similarity by the cumulative value of the similarity counter. If you set the value to be larger than the evaluation level of the cumulative value to be evaluated, matching character strings will have zero similarity, semi-matching character strings can be determined as those whose similarity is less than the evaluation level value, and the letter a element will be added to the comparison standard category. In addition, by changing the weight depending on the type of character to be compared,
It becomes possible to prevent data noise that is too obvious.

In the example of Figure 3, ````'' is used as a character type discrimination category, and cases of numbers and cases of non-numbers are assigned to [1] and ■], respectively, to distinguish them, but negative words, etc., which create opposite words with one character, are classified into the same group as numbers. It is clear that there is no harm in including it.

Effects> As explained above, according to the present invention, in a character string comparison device configured to compare matches and mismatches character by character, without making extensive evaluations such as grammatical relationships, it is possible to achieve To provide a device that can relatively flexibly extract similar word character strings that appear in response to differences in kana characters selected as they are written without introducing too much data noise. It is.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram of one embodiment of the present invention, Fig. 2 is an explanatory diagram of another embodiment of the present invention, and Fig. 3 is an explanatory diagram of 15.25 and 6 of the present invention.
．． 16.26 are upper and lower shift registers, 7, 17.
27 is a judgment circuit, 8, 18.28 is a similarity counter, 2
9 and 30 are character type detectors, M is 1-bit memory, a +
b, C, d + e r f + g + b +
J + k respectively added to the signal output (the g code is −==knee) and the fth /6 rhinoceros?廚

Claims (1)

[Scope of Claims] 1) First and second shift registers that store a comparison string and compared character string character by character, and characters stored in the first and second shift registers in the same order. (1) A comparator that compares characters that are out of order, and a judgment circuit that judges whether each character matches, mismatches, or resembles each other. As a result of comparing the character strings by comparing the characters while sequentially shifting the similarity counter that accumulates the number of similarity judgments of the result and the shift means of the shift register, the result is that the character strings are in the same order and the same length. In addition to determining the matching string as a matching string, the accumulated force 1
Is the similarity a predetermined 1+? A character string comparison device that determines the following as quasi-matching character strings. 2) In addition to the above comparator, a pair of character type detectors that detect the character types of the characters in the same order as above are activated, and if the comparison results of the above comparable comparators do not match, the unmatched characters are of a specific character type. In the case of , the character string comparison device according to claim 1, wherein a number of discrepancies larger than the number of similarity determinations is registered in the similarity counter. 3) The character string comparison device according to claim 1 or 2, wherein the determined matching character string and semi-matching character string are stored in at least a distinguishable form.