JP2007034870A - Language processor based on concept of distance - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To basically provide a character evaluation system allowing rapid and easy retrieval of a similar character string. <P>SOLUTION: In this system, similarity is evaluated by basically finding a 'distance' between two compared character strings, setting an evaluation value based on the distance as the similarity, and comparing the evaluation value. Specifically, the evaluation system or the like for evaluating the similarity between a stored character string and an input character string has: a character string input means 2; a character number calculation means 3; a numerical value allocation means 4 for allocating a numerical value of n/N to each character of the input character string such that the characters between the head and the end have equal intervals; a storage means 5 for allocating a numerical value of m/M to the m-th character of the storage character string and storing it; a character consistency decision means 6; a difference means 7; a square means 8; and an evaluation value calculation means 9. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a device for correcting fluctuation of a word by calculating a distance between words using a geometric technique, a device for calculating a similarity, and the like.

  In an information search apparatus using keyword input, search omission may occur due to fluctuations in keyword notation. In order to eliminate such a search omission, various similar keyword / search devices have been devised.

  For example, if a similar keyword / search device does not have a character string that matches the input keyword in the document to be searched (that is, does not “hit”), it automatically determines a character string similar to the input keyword. Extract and perform a new search using this similar character string as a keyword. Even if there is a “hit”, if there is another keyword similar to that keyword, it is also included in the search.

  In such an apparatus, a method for determining various similar keywords has been devised. For example, there is a method in which replacement of a character string that is likely to occur is made regular and applied to an input keyword. In this method, various character strings obtained by replacement are compared with a database dictionary (word index) or the like, and those in the vocabulary are regarded as similar character strings. On the other hand, an input character string and a word index entry are compared in association with each character, and similar ones are used as similar character strings.

  In the case of determining similar keywords, the replacement rule maintenance is troublesome in the method of regularizing character string replacement as described above. If the replacement rule is not appropriate, no hit will occur and similar keywords cannot be output. Actually, there are many reasons why the input keyword does not hit a new word, coined word, proper noun, etc., and there are cases where the method using the replacement rule cannot cope with it.

  Therefore, a similar character string search device capable of accurately determining and outputting a keyword similar to the input keyword has been studied so that a search omission is not caused by fluctuation of the keyword notation. As such an example, Japanese Patent No. 353122 (the following Patent Document 1) discloses the following invention. That is, in a similar character string search device having a function of outputting a character string similar to an input character string, a character replacement table that associates a character with a replaceable character, a list of characters that constitute the input character string, When there is a character that can be replaced with the input character string by referring to the character replacement table in the character component table of the input character string that holds information indicating the appearance position of each character in the input character string A character component table creation unit that creates a character component table in which a replacement character and information indicating the appearance position are added, with the appearance position of the character to be replaced in the input character string as the appearance position of the replacement character, and various character strings The character component table is searched for each character of the character string registered in the dictionary and the character string registered in the dictionary, and when the searched character exists, the appearance position of the character is extracted and registered in the dictionary. String and input A character string correspondence table creating unit that creates a character string correspondence table representing a one-to-one correspondence between each character of the character string, and an evaluation value of the similarity between the input character string and the registered character string for each registered character string in the dictionary Is calculated by subtracting the uncorresponding characters based on the character string correspondence table from the values when all the characters in the input character string and the registered character string correspond as the optimum corresponding characters, and the replacement character is replaced. Similarity determination that determines the similarity between the input character string and each registered character string in a form that reflects easy-to-occur character replacement by making the deduction value when the character is inconsistent less than the deduction value A similar character string search device (see claim 1 of Patent Document 1 below).

  However, in such a similar character string search device, a replacement character group must be created by referring to a character replacement table for replacing the replacement character, and many calculations are required to evaluate the similarity. Therefore, there is a problem that a character string having a high similarity cannot be searched quickly.

Japanese Patent No. 3531222

  An object of the present invention is basically to provide a character evaluation system capable of quickly and easily searching for similar character strings.

  The present invention provides an evaluation system capable of correcting input words to pre-registered words and easily counting questionnaire data in which various words are input for words having the same concept. Is for another purpose.

  Another object of the present invention is to provide a system that can correct an input character string to a pre-registered character string and has an automatic word correction function.

  The above problem is basically solved by a system that evaluates the similarity by calculating the "distance" between the two strings to be compared, using the evaluation value based on the distance as the similarity, and comparing the evaluation values. Is done.

  According to the system of the present invention, the similarity is evaluated based on the concept of “distance” of characters without requiring a replacement character table. Therefore, the similarity of a plurality of replacement character groups created by the replacement character table is evaluated. It is no longer necessary to evaluate and similarity can be evaluated by a simple algorithm. Therefore, the system of the present invention is easy to implement, can evaluate the degree of similarity quickly and properly, and can search and extract a character string having a high degree of similarity.

  According to the system of the present invention, the input character string can be corrected to the registered character string registered in advance by the replacement means for replacing the input character string with the character string having the lowest evaluation degree among the recorded character strings. The present invention can be used as a system for easily counting questionnaire data in which various words are input for words having the same concept.

  In addition, according to the system of the present invention, an input character string can be corrected to a pre-registered word by a replacement means for replacing the input character string with a character string having the smallest evaluation degree among recorded character strings. It can be used as word processing software. In the system of the present invention, the similarity between each word stored in a database including various words and the input word can be obtained, and the similarity of the selected word can be set. It can be used as a search system in a search engine.

  Hereinafter, an evaluation system according to a first aspect (pattern example) of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of an evaluation system according to the first aspect of the present invention. The evaluation system according to the first aspect is particularly effective when evaluating the similarity between an input character string having the same number of characters and a stored character string. As shown in FIG. 1, the system 1 according to the first embodiment includes a character string input means 2 for inputting a character string, and the number of characters constituting the input character string input by the character string input means. The number-of-characters calculating unit 3 calculates the number of characters calculated by the number-of-characters calculating unit, and for each character of the input character string input by the character string input unit, the first character is set to 0 and the last character is set to 1. Numerical value assigning means 4 for assigning a numerical value of n / N to the nth character of the character string, so that the number of characters is N, so that the characters between the beginning and the end are equally spaced; For each character in the stored character string and the stored character string, the first character is 0, the last character is 1, and the characters between the first and last characters are equally spaced. When the number of characters in the stored character string is M , Storage means 5 for allocating and storing a numerical value of m / M for the m-th character of the stored character string, each character constituting the input character string, and a stored character stored in the dictionary When the character matching judgment means 6 for judging whether or not each character constituting the column matches, and when the character matching judgment means judges that there is a matching character, the numerical value assigned to each character of the input character string The difference means 7 for obtaining the difference, the square means 8 for squaring the difference value obtained by the difference means, and the evaluation value by calculating the sum of the values for each character obtained by the square means An evaluation system for evaluating the similarity between an input character string and a stored character string, comprising an evaluation value calculation means 9 for obtaining. In the figure, reference numeral 10 denotes a memory such as a ROM, a RAM, an external memory, and a hard disk, and 11 denotes an output device such as a printer, a monitor, and an external terminal.

  Note that the evaluation system of the present invention may be a computer such as a personal computer, or a computer server connected to the Internet or an intranet. FIG. 2 is a diagram showing an example of the system of the present invention which is a computer connected to the Internet. As shown in FIG. 2, the computer 21 of the present invention is connected to another computer 23 via the Internet network 22. The connection mode is not particularly limited, and may be connected by a wireless LAN or the like, and may be configured to exchange information with a mobile phone or the like. Such a computer includes, for example, a main memory storing a program that causes the computer to function as each of the above means, an input device for inputting information to the system, and an output device for outputting information to the outside of the system. And a storage device such as a database for temporarily or semi-permanently storing information, and a central processing unit (CPU) for performing various arithmetic processes.

  Below, the operation example of the system of this invention which concerns on a 1st aspect is demonstrated. FIG. 3 is a flowchart for explaining this operation. In FIG. 3, S means a step. In the system of the present invention, first, a character string input means such as an input device inputs a character string into the system (S101). This character string is called "input character string". For example, a character string may be input to a server connected to the Internet, or a character string may be input to a computer or the like. Here, it is assumed that the character string “ABC” is input to the computer.

  Next, the character number calculating means calculates the number of characters in the input character string (S102). The character string may be calculated by hardware such as a character number calculation circuit for calculating the character string. Further, a character number calculating means such as a CPU may calculate the number of characters in the input character string in response to a command from a control program in the main memory. In this specification, the number of characters in the input character string is N. If the input character string is “ABC”, the number of characters is calculated as 3. In this case, for example, the input character string “ABC” may be separated character by character, and the number of characters that can be separated may be used as the number of characters. The n-th character is assigned the order of n. For example, in the previous example, each character is assigned the order from the first to the third.

  For each character of the input character string input by the character string input means, the numerical value assigning means sets the first character to 0, the last character to 1, and the character between the first and last characters. Assuming that the number of characters calculated by the number-of-characters calculation means is N so as to be equally spaced, a numerical value (coordinate value) of (n−1) / (N−1) is obtained for the nth character in the character string. Is assigned (S103). Character number calculation means such as a CPU may be assigned the above numerical values upon receiving a command from a control program in the main memory. For example, the values of n-1 and N-1 may be obtained by a difference circuit or a difference program, and the value of (n-1) / (N-1) may be obtained by a division circuit or a division program. A table storing the number of characters and the coordinate value of each character in the number of characters may be prepared, and the coordinate value may be read from the table. N is a number of 20 or less, preferably 15 or less. For example, for the input character string “ABC”, 0 is assigned to the character A, 1/2 to 0.5 are assigned to the character B, and 1 is assigned to the character C. This is expressed as (A, B, C) = (0, 0.5, 1). If the input character string is “ABCDE”, a numerical value may be assigned such as (A, B, C, D, E) = (0, 0.25, 0.5, 0.75, 1). .

  On the other hand, various storage character strings are stored in storage means such as a storage dictionary or a database. For each character in the storage character string, the first character is set to 0 and the last character is set to 1. , (M−1) for the mth character of the stored character string, where M is the number of characters in the stored character string so that the characters between the beginning and the end are equally spaced. A numerical value of / (M−1) is assigned and stored. Such information may be stored in a hard disk such as a computer, a ROM, or the like. The stored character string is preferably stored in association with the stored character string, each character constituting the stored character string, the number of characters in each character string, and the coordinate value of each character constituting the character string. Is preferably stored as. When information is stored as a database, a known data model such as a hierarchical data model, a network data model, or a relational data model can be used as a data model for classifying and organizing data. Among these, the relational data model is particularly preferable in that data can be easily made independent from an application, a new table corresponding to input information can be easily created, and data manipulation is easy.

  Here, the relational data model is a data model that uses the concept of a two-dimensional table (table) and manages data so that a vertical column corresponds to an item and a horizontal column corresponds to a record. A new table (view table or the like) can be created by extracting arbitrary data from the table using arithmetic processing such as projection, selection, join, and quotient. It is also possible to create a new table by extracting data from a plurality of tables. The table consists of a plurality of sets. Each set is composed of a plurality of pieces of information and constitutes a row. In the vertical column of the table, data having the same attribute is organized to form a column. A numerical value (index) for management is assigned to each group, and it is preferable that the numerical values of a plurality of sets constitute a domain (domain). Setting an index is preferable because it quickly reaches the storage location on the data storage device. When the index is not assigned, the database is preferably systematically configured so that information can be searched using keywords or the like. Each command in the relational database may be created in a language such as SQL.

  For example, it is assumed that “ABC” and “ACB” are stored in the storage means. In this case, (A, B, C) = (0, 0.5, 1) is stored, and (A, C, B) = (0, 0.5, 1) is stored. When the above information is stored in a relational database, for example, the stored character string ABC, the number of characters 3, the coordinate value (0) of A, the coordinate value (0.5) of B, and the coordinate value (1) of C are associated with each other. Each of which is stored.

  The character matching judgment unit judges whether each character constituting the input character string matches each character constituting the stored character string stored in the dictionary (S104). This character matching determination means may be configured by hardware such as a circuit for determining character matching. In addition, the CPU or the like may receive a command from a control program in the main memory and determine character matching. Such a program can be easily created using a known program such as Excel (registered trademark) or Access (registered trademark).

  When the character matching judgment means judges that any of the characters in the input character string matches the characters constituting the stored character string, the difference means stores the numerical value assigned to the character in the input character string, The difference between the character constituting the character string and the numerical value of the character that matches the character of the input character string is obtained (S105). The difference means may be a difference circuit for obtaining a numerical difference value, or the difference value may be obtained by a table storing the difference value. In any case, two coordinate values are input to the circuit or table, and a difference value corresponding to the two coordinate values is obtained. The difference means may be achieved by a program that causes a computer to function as a means for obtaining a difference value. Such a program can be easily created using a known program.

  For example, when both the input character string and the stored character string are “ABC”, the difference value Δ of each character is Δ = (0, 0, 0). On the other hand, when the input character string is “ABC” and the stored character string is “ACB”, Δ = (0, −0.5, 0.5).

  The square means squares the difference value obtained by the difference means to obtain the square value of the difference value (S106). The square means may be a square circuit for obtaining a square value, or may be configured by a table storing the square value of the input value, and the square value may be obtained. Further, the square means may be one in which a CPU or the like obtains a square value in response to an instruction of a control program in the main memory. Such a program can be easily created using a known program.

  For example, when the input character string and the stored character string are both “ABC”, the difference value Δ of each character is Δ = (0, 0, 0). Therefore, the value obtained by squaring the difference value of each character is also 0. On the other hand, when the input character string is “ABC” and the stored character string is “ACB”, the difference value Δ = (0, −0.5, 0.5). Therefore, the square value of each character of the input character string “ABC” is (0, 0.25, 0.25), respectively.

  The evaluation value calculation means calculates an evaluation value by calculating the sum of the values for each character obtained by the square means (S107). Such an evaluation value calculation means may be a circuit for obtaining the sum, or may be hardware such as a table storing the sum value. Alternatively, the CPU or the like may receive a command from a control program in the main memory and obtain the sum of the difference values. Such a program can be easily created using a known program. The obtained evaluation value is stored in a memory or the like in association with the stored character string. Then, the obtained evaluation value or the stored character string is output by the output device. Specifically, it is displayed on a monitor or printed on a printer. Information may also be uploaded to a website.

  For example, when both the input character string and the stored character string are “ABC”, the values obtained by squaring the difference value of each character are all zero. Therefore, the evaluation value which is the sum of them is also zero. On the other hand, when the input character string is “ABC” and the stored character string is “ACB”, the square value of the difference value for each character of the input character string “ABC” is (0, 0.25, 0.25). Therefore, the evaluation value is 0.5 because it is the sum of them. That is, when the input character string and the stored character string completely match, the evaluation value becomes 0, and when the input character string deviates from this, the evaluation value increases. Thus, according to this system, the similarity between the input character string and the stored character string can be evaluated. In this system, the difference between the coordinate values is obtained, the value is squared, and the squared value is added. This looks as if the distance (the square) is compared with the distance (the square). This is why the system of the present invention evaluates character similarity based on the concept of distance.

  In the above example, the difference between the coordinate values is obtained, the value is squared, and the squared value is added to obtain an evaluation value. However, the evaluation value may be a temporary evaluation value, and the evaluation value may be obtained by dividing the temporary evaluation value by a predetermined number (for example, N, M, or N + M). Alternatively, the square root of the temporary evaluation value may be used as the evaluation value. Furthermore, it is good also considering the value which divided 1 or a certain number by the temporary evaluation value as an evaluation value. In this case, the greater the evaluation value, the higher the similarity.

  According to such a system, the similarity between the input character string and the stored character string can be evaluated from the stored character strings stored in the database. It can be preferably used for a search engine.

  In a preferred aspect of the evaluation system according to the first aspect, an evaluation value setting means for setting an evaluation value, a calculated evaluation value that is an evaluation value calculated by the table value calculation means, and the evaluation value setting means are set. An evaluation value comparing means for comparing with the set evaluation value which is the table value, and as a result of comparison by the evaluation value comparing means, the calculated table value is smaller than the set table value In this case, the evaluation system selects the stored character string. In the evaluation system according to this aspect, the stored character string associated with the input character string can be selected by setting the evaluation value.

  In the system of this aspect, the evaluation value is set in advance by the evaluation value setting means. This set value can be appropriately changed by the input means inputting a predetermined value to the server or computer. That is, an evaluation value is set by inputting an evaluation value from a client or the like to a server or a computer. This set evaluation value may be temporarily stored in a RAM or the like, for example, or may be stored in a ROM or a hard disk for a long time. Further, it may be designed in hardware as a circuit relating to the set value.

  As a setting example of the evaluation value setting means, the set evaluation value is 0.7 or 0.3. For example, a setting evaluation value of 0.7 is input by a pointing device such as a keyboard and is input to the system by an input device of a computer.

  The evaluation value comparison means compares the calculated evaluation value, which is the evaluation value calculated by the table value calculation means, with the set evaluation value, which is the table value set by the evaluation value setting means. This evaluation value comparison means may be configured as a comparison circuit. Alternatively, the CPU or the like may receive a command from a control program in the main memory, read the stored set evaluation value and the calculated evaluation value, and compare the calculated evaluation value and the set evaluation value. Such a program can be easily created using a known program. This comparison result is stored in a memory or the like. Then, the CPU or the like as the selection means receives a control program command in the main memory, reads the comparison result stored in the memory or the like, and compares the evaluation value comparison means with the result that the calculated table value is the setting value. If the value is smaller than the table value, the stored character string is selected. In this way, the set value is cleared, and the stored character string determined to be similar to the input character string is selected.

  For example, if the input character string and the stored character string are both “ABC”, the evaluation value is 0. On the other hand, when the input character string is “ABC” and the stored character string is “ACB”, the evaluation value is 0.5. If so, when the set evaluation value is 0.7, since the calculated evaluation values of both the stored character strings “ABC” and “ACB” are 0.7 or less, both character strings are selected. On the other hand, when the set evaluation value is 0.3, only the stored character string “ABC” is selected as a similar stored character string because the calculated table value is smaller than the set table value.

  Another preferred aspect of the system is a system having sorting means for rearranging the calculated evaluation values of each stored character string in ascending order when there are a plurality of stored character strings selected by the selecting means. In such a system, a plurality of stored character strings selected by the selection means can be arranged in a similar order (that is, in the order of small evaluation values). For example, the calculated evaluation value is stored in a memory or the like in association with the stored character string, and the CPU or the like receives a command from the control program in the main memory, compares the calculated evaluation value, and the value is They are arranged in order from the smallest, and according to the value, they are arranged in order from the stored character string with the smallest calculated evaluation value.

  For example, if the input character string and the stored character string are both “ABC”, the evaluation value is 0. On the other hand, when the input character string is “ABC” and the stored character string is “ACB”, the evaluation value is 0.5. If so, when the set evaluation value is 0.7, since the calculated evaluation values of both the stored character strings “ABC” and “ACB” are 0.7 or less, both character strings are selected. Then, the calculation evaluation values are rearranged in ascending order. That is, the order is “ABC” and “ACB”.

  Another preferable aspect of the present system is an evaluation system in which when the stored character string does not include any character of the input character string, the difference value for the character not included is set to 1. Since the evaluation value is calculated in this way, for example, the evaluation value can be calculated even when the input character string is “ABC” and the stored character string is “ABD” or “ADB”. For example, each character of the input character string is compared with each character of the stored character string, and if there is an input character string that is not included in the stored character string, the information is stored in a memory or the like. Then, the CPU or the like receives an instruction from the control program in the main memory, reads the information stored in the memory, and gives 1 as the difference value of the character. Information having a difference value of 1 is stored in a memory or the like.

  In the previous example, the characters A and B in the input character string also exist in the stored character string. On the other hand, since the character C is not included in the stored character string among the characters constituting the input character string, the difference value for the character C is set to 1. That is, since the difference value Δ of “ABD” is (0, 0, 1), the evaluation value is 1. On the other hand, since the difference value of “ADB” is (0, −0.5, 1), the evaluation value is 1.25. For example, when the set evaluation value is 0, 7 or the like, none of these stored character strings is selected.

  Then, the evaluation value and the character string given the evaluation value are temporarily or semi-permanently stored in evaluation value information storage means such as an evaluation value table. Then, the evaluation value and the stored character string (or only the stored character string) to which the evaluation value is given are read in response to a command from the control program in the main memory, and are displayed on a display means such as a display. Become.

  In another aspect of the system according to the first aspect, when the character matching judgment means judges that there is a matching character, the difference means corresponds to the numerical value of each character in the stored character string and the character string corresponding thereto. The difference of the numerical value of each character of the input character string is obtained. For example, when the input character string is “ABC” and the stored character string is “ABD”, for example, the character D is not included in the input character string. Therefore, the difference value for the character string “ABD” is (0, 0, 1), and the evaluation value of the input character string based on the recorded character string is 1.

  In the evaluation system of the present invention according to the second aspect, the difference means assigns a numerical value assigned to each character of the input character string and a numerical value assigned to the same character as each character of the input character string in the stored character string. A difference between a numerical value assigned to each character of the stored character string and a numerical value assigned to the same character as each character of the stored character string in the input character string, the difference Square the difference between the numerical value assigned to each character of the input character string obtained by the means and the numerical value assigned to each character of the input character string in the stored character string, and obtain the sum of them The input-derived evaluation value calculating means for calculating the input-derived evaluation value by calculating the temporary evaluation value and dividing the temporary evaluation value by N, the numerical value assigned to each character of the stored character string, and the input Each sentence of the memory string in the string The value of the difference between the numerical values assigned to the same character is squared and the sum of them is obtained to obtain a temporary evaluation value, and the evaluation value is M. The evaluation system according to the above, wherein the calculation means, the input-derived evaluation value calculated by the input-derived evaluation value calculating means, and the memory-derived evaluation value calculated by the memory-derived evaluation value calculating means are combined into an evaluation value. .

  The step of obtaining a temporary evaluation value is the same as the step of obtaining the evaluation value in the first aspect. For each temporary evaluation value, a value obtained by dividing by N or M is obtained by a division means such as a division circuit or a division program, and an evaluation value is obtained by a summation means such as an addition circuit or an addition program. In such a system, the evaluation value is normalized to 2 regardless of the number of characters in the input character string or the number of stored characters, so that the similarity can be appropriately evaluated. In this system, division or multiplication means may be appropriately combined so that the evaluation value is corrected to 1 or 100 at the maximum.

For example, it is assumed that the input character string is “ABC” and the similarity with the stored character string “ABDE” is evaluated. The coordinate value of “ABDE” is (A, B, D, E) = (0, 0.33, 0.66, 1). The difference value Δ of “ABDE” based on the input character string “ABC” is (A, B, C) = (0, 0.17, 1). Therefore, a temporary evaluation value is 0 ² +0.17 ² +1 ² = 1.0289. As a value obtained by dividing this by 3, 0.343 is obtained. On the other hand, the difference value Δ of “ABC” based on the character string “ABDE” is (A, B, D, E) = (0, −0.17, 1, 1). Therefore, the provisional evaluation value is 0 ² +0.17 ² +1 ² +1 ² = 2.0289. Dividing this value by 4 gives 0.572. Therefore, this evaluation value is 0.343 + 0.572 = 0.915.

  In the evaluation system according to the third aspect of the present invention, the difference means includes a numerical value assigned to each character of the input character string and a numerical value assigned to the same character as each character of the input character string in the stored character string. A difference between a numerical value assigned to each character of the stored character string and a numerical value assigned to the same character as each character of the stored character string in the input character string, the difference Square the difference between the numerical value assigned to each character of the input character string obtained by the means and the numerical value assigned to each character of the input character string in the stored character string, and obtain the sum of them Then, a temporary evaluation value is obtained, and the difference between the numerical value assigned to each character of the stored character string and the numerical value assigned to the same character as each character of the stored character string in the input character string is squared. , By calculating the sum of them, Calculated values, obtains a sum of the evaluation values of the temporary, the evaluation system described in that the evaluation value calculated values obtained by dividing the sum by (N + M). The process described in the first embodiment can be used as the process for obtaining a temporary evaluation value.

For example, it is assumed that the input character string is “ABC” and the similarity with the stored character string “ABDE” is evaluated. The coordinate value of “ABDE” is (A, B, D, E) = (0, 0.33, 0.66, 1). The difference value Δ from the input character string “ABC” is (A, B, C) = (0, 0.17, 1). Therefore, a temporary evaluation value is 0 ² +0.17 ² +1 ² = 1.0289. On the other hand, the difference value Δ from the character string “ABDE” is (A, B, D, E) = (0, −0.17, 1, 1). Therefore, the provisional evaluation value is 0 ² +0.17 ² +1 ² +1 ² = 2.0289. The sum of the temporary evaluation values is 3.0578. Dividing this by 7 gives 0.4368, which is the evaluation value.

  In another preferred embodiment of the system of the present invention, in the above system, when the input character string or the recorded character string includes two or more same characters, a difference between numerical values assigned to the respective characters is obtained. This is an evaluation system that uses the character with the smaller absolute value of the difference to find the numerical difference. Since it has such a structure, even if it is a case where two or more same characters are contained in the said input character string or the said recording character string, an evaluation value can be calculated | required appropriately.

  Examples of the operation of the system according to this aspect include the following. A determination means for determining whether or not each character string includes two or more of the same characters determines whether or not each character string includes two or more of the same characters. More specifically, the process of determining whether the first character is the same as the next character, and determining whether the first character is the same as the next character is repeated. If it is determined that the first character is not the same as all characters, it is determined whether the second character is the same as the third character, and further, it is determined whether the second character is the same as the fourth character. Such processing is sequentially repeated. When any two or more characters constituting the character string are determined to be the same, a coordinate value is assigned to each character as in the normal case. Then, when obtaining the difference value, the absolute value of the difference value of each character is obtained, and the difference value is compared by comparison means such as a comparison circuit or a comparison program for comparing the absolute value of the difference value. Then, the smallest difference value is adopted as the difference value.

For example, when the input character string is “ABB”, the first character A is different from the second character and the third character. However, the second character B matches the third character B. Therefore, since there is the same character constituting the character string, processing by a circuit or a program is performed to process it. In this case as well, coordinate values are given as in the normal case. That is, (A, B1, B2) = (0, 0.5, 1). On the other hand, when the recorded character string is “ABC”, since (A, B, C) = (0, 0.5, 1), the difference value based on the input character string (A, B, B) is It is calculated as follows. The difference value Δ of (A, B1, B2) = (0, 0, 0.5). Therefore, the evaluation value is 0.25. On the other hand, it is considered to obtain the difference value of the recorded character string (A, B, C). The coordinate of A in the input character string is 0, and there is no C character in the input character string. Therefore, the difference values between the characters A and C in the recorded character string are 0 and 1, respectively. On the other hand, there are two characters B1 and B2 in the input character string corresponding to the character B in the recorded character string. Therefore, in this embodiment, a difference value for these two characters is obtained. That is, the difference value from the coordinate value of the character B1 is 0.5−0.5 = 0. On the other hand, the difference value from the coordinate value of the character B2 is 0.5-1 = −0.5. Since the absolute values thereof are 0 and 0.5, respectively, the absolute values are compared, and 0 having the smaller absolute value is adopted as the difference value of the character B. That is, since the difference value Δ based on the recorded character string (A, B, C) is (0, 0, 1), the evaluation value of the recorded character string (A, B, C) is 0 ² +0 ² +1. ² = 1. In the processing as described above, for example, a CPU or the like may read information necessary for calculation from the memory, receive an instruction of a control program in the main memory, and obtain an evaluation value based on the read information.

  In the system that performs the above processing, for example, when comparing a character string (ABCDEF) with a character string (ABCDEFA), when calculating a difference value based on (ABCDEF), the coordinates of A at the end of (ABCDEA) Since the situation where the values are compared can be prevented, an appropriate evaluation value can be calculated.

  In another preferred embodiment of the present invention, there is provided replacement means for replacing an input character string with a character string having the smallest evaluation degree among recorded character strings. According to such replacement means, it is possible to replace the input character string with the character string having the lowest evaluation degree among the recorded character strings. In the processing as described above, the CPU may read predetermined information from the memory, receive an instruction from the control program in the main memory, and convert the input character string into a recorded character string.

  The system of this aspect can be used as, for example, a word processor autocorrect system. In a questionnaire using a website, for example, character strings such as “Louis Vuitton (registered trademark)”, “Louis Vuitton”, and “Vuitton” are input as words indicating the same concept. Since such an input character string can be corrected to "Louis Vuitton" (registered trademark) which is a recorded character string registered in advance, questionnaire data in which various words are input for words having the same idea This is effective for systems that easily count

  In another preferred embodiment of the present invention, the storage means stores various stored character strings, and for each character of the stored character string, the leading character is set to 0 and the trailing character is Is 1 and the number of characters in the stored character string is M, the mth character of the stored character string is (m The evaluation system further includes a numerical value assigning means for assigning a numerical value of -1) / (M-1).

  The evaluation system according to this aspect does not assign a numerical value to a stored character string in advance, but performs a calculation to assign a numerical value not only to an input character string but also to a stored character string when obtaining an evaluation value. Therefore, it is assumed that the processing speed is slower than the evaluation system according to the aspect described above. However, in the system according to this aspect, since it is not necessary to set the coordinate value of the stored character string in advance, the similarity between the character string stored in any database or the like and the input character string can be evaluated.

  As a system of the present invention, a system for obtaining an evaluation value in consideration of the strength of the co-occurrence relationship is another preferable aspect of the present invention. The co-occurrence relationship expresses the frequency of occurrence of a statistical combination of a plurality of attributes when a language is operated. The more frequently a combination of attributes appears, the stronger the co-occurrence of those attributes. The attribute importance level expresses how much the searcher wants the attribute included in the search expression specified by the searcher to be included in a similar search expression generated by the system. The attribute that the searcher strongly desires that the attribute included in the search expression specified by the searcher is included in the similar search expression generated by the system is more important.

  That is, when a certain search expression includes a plurality of attributes A1,..., Ar, a certain attribute As and other attributes A1,..., As-1 As + 1,. The importance is calculated so that the importance of the attribute As increases as the co-occurrence relationship between Ar increases, and the calculation of the similarity between different search formulas uses the evaluation scale of similarity as the attribute appearance. If the retrieval formula includes the plurality of attributes A1,..., Ar, among the plurality of attributes A1,..., Ar, one attribute As and other attributes A1,. .., Ar, the co-occurrence of the co-occurrence relationship with Ar + 1, the arbitrary attribute At different from the certain attribute As, and the other attributes A1,..., As-1 As + 1 As + 1,. The smaller the difference in the strength of the relationship, the higher the similarity between the attribute As and the attribute At. Either replace the less important attribute with another attribute or remove the less important attribute from the search expression to obtain a similar search expression with a high similarity to the search expression A similar search method (for example, the invention described in claim 1 of Japanese Patent Laid-Open No. 9-153068, “similar search method and apparatus”) may be used as appropriate.

  When obtaining a similar search expression, either the lower importance attribute is replaced with another attribute, or the lower importance attribute is removed from the search expression, and the search entered by the searcher A similarity search formula having a high similarity to the formula is obtained. Here, the similarity evaluation scale is the similarity of how the attribute appears. That is, if the search expression includes a plurality of attributes A1,..., Ar, among these attributes A1,..., Ar, a certain attribute As and other attributes A1, ..., As-1 As + 1,. The difference between the strength of the co-occurrence relationship with Ar and the strength of the co-occurrence relationship between an arbitrary attribute At different from a certain attribute As and the other attributes A 1,..., As-1 As + 1,. The smaller the value, the higher the similarity between the attribute As and the attribute At.

  In this way, it is not necessary for the system operator to set the importance and similarity between attributes in advance, and the operation of the device does not become difficult even if the number of attributes increases, and is similar to the importance. The effect is obtained that the feelings of many searchers agree on the degree.

  By using the evaluation system of the present invention as a similarity calculation unit, a similar character string search system can be provided.

Such a similar character string search system includes, for example, a keyword extraction unit that extracts a keyword from an input character string, and from a focused item in a database having a hierarchical structure to a hierarchy lower than a plurality of predetermined levels. Based on the reading unit that sequentially reads all items of the item, the item read by the reading unit, the similarity calculation unit that calculates the similarity to the keyword, and the similarity calculated by the similarity calculation unit, This is a similar character string search system including an item specifying unit for specifying.

  The keyword extraction unit is composed of a key decomposition unit and a command determination unit. The key decomposition unit extracts keywords from the input sentence. The command determination unit 2 determines commands (search, registration, update, deletion, etc.) from the keywords extracted from the input sentence by the key decomposition unit. The item retrieval unit retrieves an item corresponding to a keyword from data, and includes a reading unit, a similarity calculation unit, and an item specifying unit. The reading unit sequentially reads all the items registered in the hierarchy from the item of interest to the lower predetermined level in the data. The similarity calculation unit calculates an evaluation value that is the similarity of the item (stored character string) read out from the data by the reading unit with respect to the keyword (dormitory character string), and the minimum evaluation value and the character at this time A column (item character string) is stored in a similar table. The item specifying unit specifies an item corresponding to the keyword based on the maximum similarity calculated by the similarity calculating unit 5 and the maximum similar character string at this time. The data is items stored in the database. The similarity table stores the maximum similarity calculated by the similarity calculation unit and the maximum similarity character string at this time.

  Next, the operation of this similar character string search system will be described. First, the key decomposition unit, which is an optional element, decomposes the input character string into keywords. Then, the keyword is notified to the command determination unit 2. Upon receiving these keyword notifications, the command determination unit determines the command, for example, determines the processing to be executed for the input sentence based on the input character string input as the keyword (search command), Notify The reading unit constituting the item search unit sequentially reads out candidate items corresponding to the keyword, for example, all items from the item of interest to the lower level of the second level from the data 7 and notifies the similarity calculation unit. The similarity calculation unit calculates the similarity between the notified item and the keyword, and stores the value of the minimum evaluation value and the character string of the item at this time in the similarity table. The item specifying unit specifies an item to be noted based on the minimum evaluation value stored in the similar table and the maximum similar character string at this time. Similarly, for the next keyword, all items from the focused item to the lower two levels are read out, the minimum evaluation value and the maximum similar character string at this time are stored in the similarity table, and the corresponding item is Identify.

  As described above, with regard to the keywords extracted from the input sentence, all items from the sequentially focused item to the hierarchy lower than the predetermined level are read, the most similar item is identified, and then the identified item is focused. Similarly, by repeatedly specifying the most similar item, it is possible to easily search for the item most similar to the keyword extracted from the input sentence using a simple system.

  In Example 1, the validity of the similarity evaluation value was verified. Using the distribution song list of a certain karaoke model, the operability of the system of the present invention was confirmed. Specifically, a program for functioning the system of the present invention was stored in a personal computer, and the computer was functioned as each means.

  The algorithm described above was used.

  A dictionary consisting of a relational database storing about 8000 titles was created on the computer's hard disk. Then, the word “Dancing Summer Festival” was entered into the system of the present invention to search for a song with a high degree of similarity.

The results were as follows.
Rank Song name Similarity 1 Dancing! Summer festival 0.113746
2 Danjiri 0.757937
3 Darling 0.795599
4 Twilight Dancing 0.819262
5 Summer Festival 0.917304
6 Summer Festival 0.917304
7 Cynthia 0.949830
8 Dancing All Night 0.974581
9 Linda Linda 1.020226
10 Gambling 1.037901
(Display only 10 songs from the nearest)

  Example 2 is an example in which the evaluation system of the present invention is used as a questionnaire system for evaluating which input character string is closest to which is registered in the dictionary. In this example, the system of the present invention is configured as a server. And on the website, I asked the question “Please answer your favorite brand” and asked for an answer.

  According to the system of the present invention, for example, input words such as “Louis Vuitton (registered trademark)”, “Louis Vuitton”, and “Vuitton” can be corrected to the previously registered word “Louis Vuitton” (registered trademark). , It can be used as a system for easily counting questionnaire data in which various words are input for words having the same concept.

  In addition, since the system of the present invention can be corrected to pre-registered words, it can be used as word processing software having an automatic word correction function. In the system of the present invention, the similarity between each word stored in a database including various words and the input word can be obtained, and the similarity of the selected word can be set. It can be used as a search system in a search engine.

FIG. 1 is a block diagram of an evaluation system according to the first aspect of the present invention. FIG. 2 is a diagram showing an example of the system of the present invention which is a computer connected to the Internet. FIG. 3 is a flowchart for explaining an operation example of the system of the present invention according to the first aspect.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 System based on 1st Embodiment 2 Character string input means 3 Character number calculation means 4 Numerical value assignment means 5 Storage means 6 Character matching judgment means 7 Difference means 8 Square means 9 Evaluation value calculation means 10 Memory 11 Output device 21 Computer 22 Internet network 23 Computer

Claims (11)

A character string input means for inputting a character string;
A character number calculating means for calculating the number of characters constituting the input character string input by the character string input means;
For each character of the input character string input by the character string input means, if the leading character is 0, the trailing character is 1, and the number of characters calculated by the character count calculating means is N, the leading character is Numerical value assigning means for assigning a numerical value of (n-1) / (N-1) to the nth character of the input character string so that the characters between and the end are equally spaced; ,
For various stored character strings and each character of the stored character string, when the leading character is 0, the trailing character is 1, and the number of characters in the stored character string is M, Storage means for assigning and storing a numerical value of (m−1) / (M−1) for the mth character of the stored character string so that the characters between the end and the end are equally spaced; ,
Character matching judgment means for judging whether or not each character constituting the input character string matches each character constituting the stored character string stored in the dictionary,
A difference means for calculating a difference between a numerical value assigned to each character of the input character string and a numerical value assigned to the same character as each character of the input character string in the stored character string;
A squaring means for squaring the value of each character difference in the input character string obtained by the difference means;
An evaluation value calculating means for calculating an evaluation value by calculating a sum of square values of characters of the input character string obtained by the square means;
With
An evaluation system for evaluating the similarity between an input character string and a stored character string. Evaluation value setting means for setting an evaluation value in advance;
An evaluation value comparing means for comparing a calculated evaluation value that is an evaluation value calculated by the table value calculating means with a set evaluation value that is a table value set by the evaluation value setting means;
2. The evaluation system according to claim 1, further comprising selection means for selecting the stored character string when the calculated table value is smaller than the set table value as a result of the comparison by the evaluation value comparison unit.   3. The evaluation system according to claim 2, further comprising: a sorting unit that rearranges the calculated evaluation values of each stored character string in ascending order when there are a plurality of stored character strings selected by the selecting unit.   The difference value about the character which is not contained is set to 1 in any one of the said input character strings, when the said character string is not contained in the said memory character string. The described evaluation system.   The difference means is a means for obtaining a difference between a numerical value assigned to each character of the stored character string and a numerical value assigned to the same character as each character of the stored character string in the input character string. The evaluation system according to any one of claims 1 to 4. The difference means is:
The difference between the numerical value assigned to each character of the input string and the numerical value assigned to the same character as each character of the input character string in the stored character string,
Means for determining a difference between a numerical value assigned to each character of the stored character string and a numerical value assigned to the same character as each character of the stored character string in the input character string;
The difference between the numerical value assigned to each character of the input character string obtained by the difference means and the numerical value assigned to the same character as each character of the input character string in the stored character string is squared, and the sum of them is calculated. An input-derived evaluation value calculating means for obtaining a temporary evaluation value and determining an input-derived evaluation value by dividing the temporary evaluation value by N;
By squaring the value assigned to each character of the stored character string and the value assigned to the same character as each character of the stored character string in the input character string, and obtaining the sum thereof, A memory-derived evaluation value calculating means for obtaining a temporary evaluation value and obtaining a memory-derived evaluation value by M being the evaluation value;
The input-derived evaluation value calculated by the input-derived evaluation value calculating means and the memory-derived evaluation value calculated by the memory-derived evaluation value calculating means are combined into an evaluation value. Evaluation system. The difference means is:
The difference between the numerical value assigned to each character of the input string and the numerical value assigned to the same character as each character of the input character string in the stored character string,
Means for determining a difference between a numerical value assigned to each character of the stored character string and a numerical value assigned to the same character as each character of the stored character string in the input character string;
The difference between the numerical value assigned to each character of the input character string obtained by the difference means and the numerical value assigned to the same character as each character of the input character string in the stored character string is squared, and the sum of them is calculated. To obtain a temporary evaluation value, and square the difference between the numerical value assigned to each character of the stored character string and the numerical value assigned to the same character as the stored character string of the input character string. Then, by calculating the sum of them, a temporary evaluation value is obtained,
6. The evaluation system according to claim 1, wherein a sum of these provisional evaluation values is obtained, and a value obtained by dividing the sum by (N + M) is obtained as an evaluation value.   When the input character string or the recorded character string includes two or more of the same characters, when calculating the difference between the numerical values assigned to each character, use the character with the smaller absolute value of the difference. The evaluation system according to claim 1, wherein a difference is obtained. The storage means is means for storing various stored character strings,
For each character in the stored character string, the number of characters in the stored character string is such that the leading character is 0, the trailing character is 1, and the characters between the leading and trailing are equally spaced. A numerical value assigning means for assigning a numerical value of (m-1) / (M-1) to the mth character of the stored character string when M is M. Item 9. The evaluation system according to any one of Items 8.   The program for functioning a computer or a server as an evaluation system in any one of Claims 1-9. A recording medium storing the program according to claim 10.

Images