JPH04215183A - Key word retrieving method - Google Patents

Abstract

PURPOSE:To realize a retrieving method for improving retrieval efficiency by rationally recovering even when some original characters in a documents are incorrectly read out. CONSTITUTION:This retrieving device is composed of an image scanner 1, a read-out part 2, a memory 3 for candidate character, a input part 4 for registration key word, a memory 5 for registration key word, a collation part 6 and a CRT 7. The read-out part selects based on the image if the original character, the prescribed number (e.g. three characters) of candidate character and stores them in the memory 3. When at least one of each candidate character of each original character is coincident with the character of the order for which the registration key corresponds, the original character is decided to be the character for which the key word to be retrieved corresponds.

Description

[Detailed description of the invention]

0001

[Industrial Application Field] This invention is a method for searching a document using a single character or string of characters that is the same as a registered keyword as a keyword. Concerning an efficient keyword search method.

0002

2. Description of the Related Art Generally, it is effective to use keywords to quickly and accurately understand the contents of a document. For example, documents related to global environmental protection issues use keywords such as "radiation,""ozonelayer," and "global pollution." Now, when converting a document into a database, the characters of the document, that is, the original characters, are sequentially read by a character reading device, and the character with the highest degree of similarity to the standard character is selected as the reading character, and a text expressed by a character code is created. This text is searched for the same single character or character string as the registered keyword as a keyword.

0003

In the conventional method, if there is an error in the result read by a character reading device, that is, even one character is misread, the keyword is excluded from the search target even though the keyword exists. That is, when search efficiency is defined as the ratio of the number of search keywords to the total number of keywords in a document, the search efficiency decreases significantly.

[0004] An object of the present invention is to provide a keyword search method that solves the above-mentioned problems of the conventional technology and has good search efficiency even if there is some misreading of each character in a document. be.

[0005]

[Means for Solving the Problem] In order to solve this problem, the keyword search method according to claim 1 is a method for searching a document using the same single character or character string as a registered keyword as a keyword. Each character is sequentially read as a manuscript character by a character reading device, and up to a predetermined number of candidate characters are selected for each manuscript character based on the degree of similarity to a standard character; When one of the characters matches the first character of the registered keyword, set the manuscript character as the first character of the keyword to be searched;
Select up to the same predetermined number of candidate characters as above for each character following the manuscript character corresponding to this first character; at least one of the candidate characters up to the predetermined number corresponding to each subsequent manuscript character When a character in the corresponding succeeding rank of the registered keyword matches, each document character in the succeeding order is set as each character in the corresponding succeeding rank of the keyword to be searched. In a keyword search method according to a second aspect of the present invention, in the method according to the first aspect, the predetermined number of keywords is three.

[0006]

[Operation] In the keyword search method according to claim 1 or 2, even if there is some misreading of each manuscript character of a document, up to a predetermined number of read characters, for example, 3 as in claim 2. If you search for the first readable character in which at least one of the candidate characters matches the first character of the registered keyword, there is a high probability that the matching character is the correct character. Thereafter, each successive manuscript character is read in turn, the same predetermined number of candidate characters, for example, up to three, are selected for each character, and at least one of them is searched for a character that matches each subsequent character corresponding to the registered keyword.
The match is with high probability each corresponding subsequent character of the keyword.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A search device to which the keyword search method according to the present invention is applied will be described below with reference to the drawings. FIG. 3 is a correspondence diagram of registered keywords, manuscript characters, and candidate characters related to the search device. In FIG. 3, the first line is the registered keyword, the second line is the manuscript character to be searched, the third line is the first candidate character of the reading result of each manuscript character, the fourth line is the second candidate character, and the fifth The line is also its third candidate character. The registered keyword is "joint development", the first candidate character in the reading results for the manuscript character "Kyo" is "Kyo", the second candidate character is "I", and the third candidate character is "Don". Hereinafter, the manuscript character ``same'' corresponds to the character ``same,'' and the manuscript character ``open'' corresponds to the character ``open''.
Close and Seki are selected as candidate characters for the manuscript character ``Hatsu'', and ``Hatsu'', 溌, and HAI are respectively selected as candidate characters. For each of the first, second, and third candidate characters, up to three characters are selected in order of similarity to the standard character of each manuscript character, or by relaxing the threshold related to similarity in order. be done. If the degree of similarity is extremely low, it cannot be said to be a candidate character, so there is no need to select all three characters.

FIG. 4 is a correspondence diagram when each character in FIG. 3 is encoded, and shows the constituent characters of the registered keyword K: Ki,
These correspond to the constituent characters of the original character W: Wi, and the first, second, and third candidate characters: Ai, Bi, and Ci, respectively. Here, i=1, 2, 3, and 4, which are common character ranking codes for the registered keyword and manuscript characters.

FIG. 2 is a block diagram showing the configuration of the search device. In FIG. 2, reference numeral 1 denotes an image scanner that obtains an image of original characters of a document, and 2 a reading unit that selects up to three candidate characters based on the image of original characters. Note that the first, second, and third candidate characters are as already described. Reference numeral 3 denotes a memory for candidate characters related to read characters, in which each of the first, second, and third candidate characters to be read is stored as a character code. 4 is an input section for registered keywords,
5 is a memory for registered keywords. 6 is the collation part,
Corresponding character codes from each memory 3 and 5 are compared to determine whether they match or do not match. 7 is a CRT, which displays the verification results on the screen. Note that this CRT 7 may also be provided with a printer for printing and outputting the verification results.

FIG. 1 is a flowchart showing the operation of the search device. In FIG. 1, in step S1, each constituent character Ki of a registered keyword consisting of four characters, the original character W
The common character order code i of i is initialized to i=1. 3 as a result of reading the original characters Wi in step S2.
The first, second, and third candidate characters Ai, Bi, and Ci are selected. In step S3, it is determined whether the first candidate character Ai matches the keyword constituent characters Ki, and YE
If S, the process moves to step S6; if NO, the process moves to step S4. After it is determined in step S6 that Wi and Ki match, the above process is repeated for all four characters through steps S7 and S8. Returning to step S4, it is determined whether the second candidate character Bi matches Ki. If YES, the process moves to step S6; if NO, the process moves to step S5. In step S5, it is determined whether the third candidate character Ci matches Ki, and YES is determined.
If so, proceed to step S6; if NO, proceed to step S9. In step S9, Wi is made unreadable, and therefore, in the next step S10, Wi is made unsearchable.

[0011] Here, to add a little bit, even if there is some misreading of the first manuscript character W1, which should be a keyword, up to three candidate characters A1, B1,
If C1 is increased and at least one of them matches the first character K1 of the registered keyword, it can be considered that there is a high probability that the matched character is the correct character. Hereafter, each subsequent manuscript character Wi is read in order,
If you select up to three candidate characters Ai, Bi, and Ci for each, and search for a character in which at least one of them matches each subsequent character Ki of the registered keyword, the matched character has a high probability of being the keyword's character. It can be said that each corresponding subsequent character. Note that the greater the number of candidate characters, the higher the reading accuracy, but the longer the processing time, so it is necessary to find the harmonious points. The choice of up to 3 candidate characters is based on experience, and the processing time is moderate, approximately 9.
This is based on the fact that it has been proven that characters can be read accurately with a probability of 9%.

0012

Effects of the Invention In the keyword search method according to claim 1 or 2, even if there is some misreading of each manuscript character of a document, a predetermined number of read characters, for example, 3 as in claim 2, are read. If you search for the first readable character in which at least one of the candidate characters matches the first character of the registered keyword, there is a high probability that the matching character is the correct character. are read in order, and the same predetermined number of candidate characters, for example, up to 3, are read for each character, and if at least one of them matches the corresponding subsequent character of the registered keyword, then the matched character has a high probability. Each corresponding subsequent character of the keyword. Therefore, even if there is some misreading of each character in the document, the effect of keyword search with good search efficiency can be obtained as a result. In particular, by limiting the number of candidate characters to three,
It has been demonstrated that accurate character reading can be achieved with approximately 99% probability of processing time at a reasonable level, and that a balance between processing time and retrieval accuracy can be achieved.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing the operation of a search device to which the method according to the present invention is applied.

[Figure 2] Block diagram showing the configuration of this search device

[Figure 3]
Correspondence diagram of registered keywords, manuscript characters, and candidate characters related to this search device

[Figure 4] Correspondence diagram when each character in Figure 3 is encoded

[Explanation of symbols]

1 Image sensor 2 Reading section 3. Memory 4 Input section 5. Memory 6　　　　　　Verification section 7 CRT

Claims (2)

[Claims] Claim 1. A method of searching a document using a single character or string of characters as a keyword as a registered keyword, in which each character of the document is sequentially read as a manuscript character by a character reading device, and the similarity of each manuscript character to a standard character is determined. select up to a predetermined number of candidate characters based on the degree of occurrence; the search should be performed for the manuscript character when at least one of the candidate characters of the most preceding character in each of the manuscript characters matches the first character of the registered keyword; Select the first character of the keyword; select up to the same predetermined number of candidate characters as above for each character following the manuscript character corresponding to this first character; select each candidate character up to the predetermined number corresponding to each subsequent manuscript character; When at least one of the characters matches a character in a corresponding succeeding rank of the registered keyword, each manuscript character in the succeeding order is set as a character in a corresponding succeeding rank of the keyword to be searched; Keyword search method. 2. The method according to claim 1, wherein the predetermined number is three.