JP2000348066A - Similarity judgement system for trademark and similitude generation system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily decide whether or not the name of an article, etc., is similar by judging whether or not all trademarks are similar by one neural network irrelevantly to the number of different sounds. SOLUTION: There is, for example, a 3-layer perceptron as a neural network and constituted by coupling an input layer, an intermediate layer, and an output layer in series. In addition to a trademark A, a trademark B, and mutually different characters between the two trademarks, the network which judges whether or not a trademark is similar is structured by arranging in the input layer the word length of one of the two trademarks, the positions of the mutually different characters of one trademark in the whole trademarks, characters before and after the mutually different characters of one trademark, etc., and giving a similar/nonsimilar discrimination to the output layer. Data input, etc., to a neurolearnging program is performed by being normalized normally to 0 to 1. Namely, the Japanese syllabary, English letters, etc., are numerized, inputted and outputted by proper means, and discriminated. For example, they can be inputted and discriminated as character codes of binary data in a binary way.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a similarity judgment system for judging similarity in character names, keywords, naming, etc., especially in the names of trademarks, and a similar object for creating similar names (similar trademarks). It is about the system.

[0002]

BACKGROUND OF THE INVENTION In order to easily and objectively judge the similarity of trademarks and the like, the present inventor has previously filed Japanese Patent Application No. 11-21909.
As No. 0, a system using a neural network was proposed.

[0003] However, in the system of the earlier application, it is mainly determined whether or not trademarks having a difference of one sound or two sounds are similar to each other. In other words, basically, it is not possible to determine the similarity between trademarks having a difference of one sound, two sounds, or more than one difference in one neural network. Allowing one neural network to determine the similarity of all trademarks regardless of the number of sound differences has the advantage of requiring only one network construction. Therefore, it contributes to a reduction in learning time for network construction and simplification of the configuration of the entire system. In addition, since it is not necessary to select a neural network to be used for similarity judgment based on the number of differences, it is considered that usability is extremely improved.

[0004]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and is configured as described in the above-mentioned claims cited and introduced herein.

[0005]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, a system for judging similarity of a trademark or the like and a system for creating a similar product according to the present invention will be described in further detail.

[0006] The system of the present invention utilizes a neural network. Therefore, this will first be described slightly.

As a network model, for example, there is a three-layer perceptron. As shown in FIG. 1, the input layer, the intermediate layer, and the output layer are connected in series. Here, since the connection weight between the units is random, input signals are successively given to the input layer, and the connection weight is set so that the pattern can be correctly identified from the difference between the output signal guided by the network and the ideal output signal. If updated, any pattern can be identified.

Now, the output signal of the input layer unit i is represented by I _i
Assuming that the connection weight between the input layer unit i and the intermediate layer unit j is W _ji , the output H _j of the intermediate layer unit _j is derived as H _j = f (ΣW _ji · I _i −θ _j ). In the expression, i = 1 to n, n is the number of units in the input layer, and θ _j is a threshold value of the intermediate layer unit j.

Further, assuming that the coupling load between the intermediate layer unit j and the output layer unit k is V _kj , the output O _k of the output layer unit _k is expressed as follows: O _k = f (ΣV _ki · H _j −γ _k ) Become. In the equation, j = 1 to m, m is the number of units in the intermediate layer, and γ _k is a threshold value of the output layer unit k.

Here, f (x) is 1 when x is plus infinity and 1
And a sigmoid function having an S-shaped response characteristic in which x asymptotically approaches 0 at minus infinity, and f (x) = {1 + tanh (x / u)} / 2. By changing the slope (u> 0) of the sigmoid function, it is possible to contribute to creation of an optimal learning environment and construction of a model.

In order for the neural model to derive a correct output signal with respect to the input signal, the connection weight and the threshold must be set to appropriate values. Therefore, a method of forming data (pattern information) in which an input / output signal (input layer learning signal and output layer teacher signal) forms one pair and learning the data to form a network is generally performed. . As a supervised learning method, for example, a back propagation method has been proposed. For details, see, for example, "Neuro Information Processing Technology-Fundamentals and Applications-" (by Kazuo Yana and Yoshitake Suzuki, Kaibundo Publishing) and "Introductory and Practice Neurocomputer" (by Nakano, Iinuma, Neuron Net Group, Kiritani, Technical Review) The description is omitted here.

In a network for judging the similarity of a trademark name, a feature extracted from two contrasting trademarks (names) is given to an input layer, and a similarity degree (two similarities) of the two contrasting trademarks is given to an output layer. (Similar or dissimilar, or a numerical value of how similar) is given.

The features extracted from the two trademarks to be compared include different characters (sounds) between the two trademarks.
However, even if the different character combinations are the same,
It is thought that the difference between the characters before and after that, or where the difference character is in the trademark, such as the beginning or end of the word, affects the similarity judgment. Therefore, it is preferable to appropriately input not only different characters but also other items likely to affect similarity judgment.

FIG. 2 is a diagram showing an embodiment of a network for judging the similarity of trademarks. 3 to 5 show:
It is a figure which shows an example of the pattern information for this network construction, and these three figures comprise one table. In this pattern table, the word length and the position of the different character from the beginning of the word are indicated by numerical values before normalization.

[0015] As shown in these figures, the network of this embodiment includes (1) two different trademarks, trademark A and trademark B, each of which has a different character, and (2) two brands, The word length of one trademark (for example, trademark A having the longer word length); (3) the position of the different character in the entire trademark in the one trademark; and (4) the different character in the one trademark. Each character and the like before and after the character are arranged on the input layer, and the output layer is constructed by giving similar or dissimilar distinctions.

The construction of the neural network is performed by learning using a large amount of data of past trademark similarity judgment results of the JPO and the court. For example, learning data is collected from a collection of trademark decisions of the JPO.

Input / output to / from the neuro, such as data input to the neuro-learning program, is normally performed by normalizing to a value of 0 to 1. That is, kana characters, alphabets, and the like are usually converted into numerical values by appropriate means, input / output, and identified. For example, a character can be input and identified as a character code of binary data of 0 and 1 in a binary number.

Alternatively, in order to identify characters in each trademark, 5
A unit set consisting of a set of units corresponding to individual characters such as a zero tone and an alphabet can be provided for input and identification. In this case, among the units in the unit set, one of the numerical values such as 1 or 0 is applied to the corresponding unit, and the other numerical value not used for the one is used for the non-applicable unit. By inputting, characters can be identified.

Japanese characters are "A", "I",
Each stage of "U", "E", "O", "K", "Sa",
"Ta", "na", "ha", "ma", "ya", "la",
Each line of "wa" and sound-repellent "n", voiced point "゛", semi-voiced point "゜", prompting sound "tsu", characters for "m", "u",
"Yo", long sound "-", foreign words "a", "i",
Combining "e", "o", etc., one of the numerical values, such as 1 or 0, to the applicable one, and the numerical value of the other, such as 1 or 0, to the non-applicable, the other one not used in the above Can also be identified by inputting.

Note that, regardless of the turbid point "゛" and the semi-voiced point "゜", each stage (vowel) of "A", "I", "U", "E", "O", "Ka", "Sa", "ta", "na",
"C", "M", "Y", "L", "W", "G",
Each line (consonant) of “the”, “da”, “ba”, and “pa” may be used.

By the way, in the case of kanji and the like, the reading can be converted into kana and input. In the case of English characters, the reading can be similarly converted and input. Alternatively, a network dedicated to English trademarks can be created as a unit set consisting of 26 alphabets (or hyphens, commas, periods, etc.) for inputting English characters.

In any case, in order to cope with the case where the difference between the two trademarks is the presence or absence of a specific character, a unit may be provided for use when there is no space or no corresponding character.
However, as will be described later, in a trademark without a specific character, the unit set includes all 0s (the other number).
Can be dealt with by putting

In this embodiment, as shown in the pattern tables of FIGS. 3 to 5, kana characters (names) are identified by dividing them into rows (consonants) and columns (vowels). For example, in the case of “K”, since “K” row is “A” stage, “1” is input only to the units of “K” and “A”, and “0” is input to all subsequent units. For “A” to “O”, a “A” row may be separately prepared, but here, for “A”, only the vowel unit of “A” is set to 1 and the remaining units are set to 0. (All units in the row are 0). "N" is identified by preparing a separate unit of "N" and putting 1 in only that unit.

Further, characters having a cloud point such as "ga"
In order to input a character with a semi-voiced point such as "P", prepare a unit with a voiced and semi-voiced point, and if it is "Ga", enter 1 only in each of the units "Ka", "A" and Let them be identified. In the case of "P", "C" and "U" are distinguished by putting 1 only in each unit of the semi-voiced point.

Also, the prompting sound "tsu", the letter for the repetition sound "ya",
In order to enter characters with "u", "yo", "ヮ", long sound "-", foreign words "a", "i", "ゥ", "e", "o", Each of these units is also prepared. For example, in the case of "T", since "T" is in the "T" line "E" stage, there are three units of "T", "E", and the foreign language character "I". Put only 1 in. In addition, in the case of "D", it is dealt with by adding 1 to the unit of the cloud point.

By the way, in some cases, "ヰ", "ヱ",
"ヲ", "V", "K", "K", and hyphen "-"
Units such as various symbols may be separately prepared. In addition, voiced and semi-voiced voices, prompting sounds, letters for murmurs, letters for foreign words,
Characters with a long sound or the like may be one character with them attached, or may be one character with them alone. In particular, a character other than a cloud point or a semi-voice point (especially a prompt sound or a long sound) can be converted into a single character by itself.

Note that units of relatively infrequently used characters, such as the foreign language character "@", may be omitted. That is, the units to be provided depend on the characters in the (learning) pattern information. In addition, it is also possible to change the character to a certain extent by changing “va” to “ba”, “vi” to “bi”, and “ヂ” to “ji”.

In the present embodiment, each character of the 50 sounds of "A" to "N" is basically used as one character. It is determined that the character is one character in a state where the characters are attached. For example,
Of course, characters such as "B" and "Pi" with a cloud point or a semi-voice point,
Characters with a prompting sound, a character for repetitive sounds, a character for a foreign language, a long sound, and the like, such as “R”, “Kyo”, “Wo”, and “Ku”, are assumed to be one character with these characters attached. In addition, a character with two or more prompting sounds, long sounds, and the like, such as "ditch" and "tee", is, in principle, one character with all of them.

[0029] By forming a single character with a long sound or a booming sound, "Java" and "Jabbar" or "Shanti"
And "Shandy", "Sink rate" and "Sinlator", "Ustar" and "Aster", "Bimipu" and "Bidip" can be compared as one character difference. The nature becomes high. In addition, it is preferable to extract and determine the vicinity of the difference character as widely as possible.

However, in the present invention, since two trademarks to be compared are considered, there is an exception in the case where these different sounds are attached to different parts. In other words, according to the above rule, if one trademark can be made into one character with a long sound, etc., but if the other trademark has two characters corresponding to the character, the long sound, etc. I decided to use only one character.

For example, when comparing the two trademarks “Home Guard” and “Home Guide”, the differences are “−” of “Guard” and “A” of “Guide”. And
Observing only the "guard", it is possible to use "gar" as one character, but in such a case, if the corresponding character corresponding to "gar" is viewed with the other trademark, "guy" It becomes two characters of "Ga" and "I". Therefore, in this case, "gar" is divided into two characters "ga" and "-", and "-" and "i" are compared.

In the comparison between "Earthrat" and "Earthlight", the "Lat" part of one trademark can be one character, but the other corresponding character is "L". In such a case, the character "ra" is divided into two characters "ra" and "tsu", and "tsu" is also regarded as one character. Will be compared with the corresponding letter "i" in the other trademark.

Similarly, in the case of comparison between "Koyu" and "Koyu", "-" and "U" are compared as one character, and "Cartier" and "Carty" are compared. , "E" and "-" are compared as one character.

In the case of comparison between "corp" and "core", if "co" is one character, the corresponding character of the other trademark is "core". Therefore, in this case, "ko" is divided into "ko" and "-", and the "-" is compared with the corresponding character "a" in the other trademark.

On the other hand, when comparing "U-star" and "Aster", the difference is "U" and "A". If there is no possibility that the trademark will be two characters, it will not be disassembled.

Since the characters other than the parts different from each other are the same as the characters of the other trademark, there is no possibility that the corresponding portion of the other trademark becomes two characters even if a long sound is added. One character is added with a prompt or long sound.

That is, in principle, one character is added with a long sound and the like, and the exception rule is applied only to the vicinity of a different character. The judgment is as follows. Here,
For the sake of convenience, we are considering the comparison of trademarks that differ by one character. First, find the difference between the two trademarks.

(1) When the difference is a difference between specific characters:

(A) In the case of a difference between the Japanese syllabary characters (sometimes accompanied by a vocal sound, etc.): If the syllabary character has a vocal sound, etc., it is compared as one character including the vocal sounds.

For example, "thiochol" and "thymocol"
In the case of comparison with "o" and "mo" are different,
“O” and “MO” are extracted as comparison targets. Also, in the case of comparison between “Royen” and “Roy Fan”, “A” and “Fa” are extracted. Furthermore, in the case of comparing "chacco" and "sacco", "cha" and "sa" are different, but since "cha" is attached to each, "cha" and "sac" are compared. Extract as Similarly, when comparing "Microtech" and "Microdick",
“T” and “D” are extracted as comparison targets, and “G” and “T” are similarly extracted as comparison targets in the case of comparison between “Sync rate” and “Sinlator”.
Also, in the case of "FetchE" and "FetchB", "E" and "B" are extracted. Further, in the case of “Java” and “Java”, “Va” and “Bar” are extracted.

(B) In the case where there is a difference between the Japanese syllabary and the prompting sound, etc .: The syllabary character and the prompting sound are compared as one character.

For example, in the case of comparison between “Hitter” and “Hatter”, “I” and “tsu” are different, so “I”
And “tsu” are extracted as comparison targets. In the case of comparison between "Karoldo" and "Caroud", "Lu" and "-" are different, so "Lu" and "-" are extracted as comparison targets.
If “R” is extracted from the other trademark, “Lol” is extracted from the one trademark as a corresponding character, resulting in two characters.

(C) When there is a difference between prompting sounds and the like: The character including the 50-letter character preceding the prompting sound is compared as one character.

For example, in the case of comparing "Lelead" and "Lerid", "-" and "tsu" are different, but "Lee"
And “Li” are extracted as comparison targets. In the case of the comparison between “chakko” and “chaco”, “cha” and “cha” are different, but “cha” and “cha” are extracted as comparison targets. Note that in the case of “chaco” and “chaco”, it may be considered that there is a difference in the presence or absence of the prompting sound “tsu”, and even in such a case, the result is the same by a method described later.

(2) When the difference is in the presence or absence of a specific character:

(A) Presence or absence of Japanese syllabary characters (may be accompanied by a prompt, etc.): From one trademark having the specific character, the Japanese syllabary character is extracted, and the other trademark without the specific character is extracted. , Blanks are extracted as comparison targets in principle. At that time, if the syllabary character is accompanied by a prompt sound or the like in principle, it is one character including the syllabary sound.

For example, when comparing "Joana" and "Jona", there is a difference in the presence or absence of the character "A".
“A” and “_” (blank) are extracted as comparison targets.
Also, in the case of comparing "George Sunk" and "George Sunk", there is a difference in the presence or absence of the character "Ju",
“Ju” and “_” (blank) are extracted as comparison targets. Furthermore, in the case of "Ichiko" and "Ichiko", "I"
And "_" (blank) are extracted, and in the case of comparison between "es-eskay" and "es-esque", there is a difference in the presence or absence of the character of "ke", so "ke" and "_" (blank) Is extracted as a comparison target.

It should be noted that if there is no corresponding character ("_"), a special unit for identifying the character is provided in the unit set, and when there is no corresponding character, 1 is put in the corresponding unit. Good.

In the case of "Ichiko" and "Ichiko",
Of the two "I" in "Ichiko", either "A" can be a different character, but in this embodiment, it was considered that there was a leading "I". In this embodiment, as described above, one character is included as much as possible, such as a prompt sound or a long sound, and similarly, the purpose of extracting the contrast portion in a cohesive form as much as possible is maintained. . Therefore, in the case of "SSK" and "SSK", the common part of both is taken as much as possible,
"K" on the back side was extracted from "SSK".

On the other hand, in the comparison between "Sears" and "Seeds", there is a difference in the presence or absence of "A", but since "A" has a long sound "-", "A" is extracted. And "-" corresponding to "Ah" is extracted from the other trademark.

(B) In the case of presence or absence of only a sound or a long sound: From one trademark having such a sound or the like, one trademark including the 50 sounds before the sound or the like is extracted as one character, and the other trademark without the sound or the like is extracted. , The corresponding 50 tones are extracted as comparison targets.

For example, in the case of comparison between “Nike” and “Nike”, although there is a difference in the presence or absence of “T”, “Nik” is extracted including the character of “N” before “T”. From the other trademark, "d" is extracted as the corresponding character. Also,
Also in the case of comparison between “Age” and “Age”, although there is a difference in the presence or absence of “−”, “Gee” and “Ji” are extracted as comparison targets.

As described above, the characters of both trademarks are input. In this embodiment, as described above,
In the input layer, (1) each different character of both trademarks, (2) the word length of one trademark, (3) the position of the different part of the one trademark in the whole trademark, and (4) the one trademark Characters before and after the different part are arranged, and similar or dissimilar is given to the output layer.

The input of a different character and each character before and after the different character is dealt with by inputting 1 or 0 to each unit of the unit set for identification such as the Japanese syllabary. If there is a difference at the beginning of the word, the character before the difference character should be blank,
If there is a difference in the ending, the character following the difference character is blank, and if there is a difference in the presence or absence of the specific character, the difference character in the trademark without the specific character is blank. The blank is identified by putting all 0s in each unit of the unit set.

Further, the word length is normalized from 0 to 1 assuming that the number of characters of the trademark having the maximum number of characters is 1 and the number of characters of the trademark having the minimum number of characters is 0 in the pattern information. The normalization is performed by the following equation. Q = (P−P _min ) / (P _max −P _min )

Here, the normalized pattern information is indicated by Q, the numerical value of each item in the pattern information is indicated by P, the maximum value of each item is indicated by _Pmax , and the minimum value is indicated by _Pmin .

The position of the different character is represented by the arrangement ratio of the different part with respect to the entire trademark. That is, it is obtained as a ratio of the length from the beginning (or the end) to the different part with respect to the length (word length) of the entire trademark. Specifically, the position of the difference character is normalized to 0-1 when the difference between the two trademarks is at the beginning of the word and 0 when the difference is at the end of the word and is input. FIG. 6 shows an example of normalized data of the position of the difference character with respect to the trademark of the main character number. This normalization is defined by (the number of characters from the head -1) / (the word length of the trademark -1). You.
The position of the difference character can be dealt with by inputting the number of the character from the beginning of the word.

In this way, the word length of the trademark having the longer word length and the positions of the different characters in the trademark having the longer word length in the entire trademark are respectively set to the values normalized to 0-1. , Each of the different characters of the two trademarks, and each of the characters before and after the different characters in the trademark with the longer word length can be input with either 0 or 1 for each unit of a total of four unit sets. While placing
The output layer indicates 0 if the two trademarks are determined to be dissimilar,
If it is determined that they are similar, 1 is arranged to create pattern information.

By the way, in the network of this embodiment,
Not only a single character difference but also a difference of two or more characters can be determined. If there is a difference of two or more characters, separate pattern information may be created for each different part. Then, the degree of influence of each different part on the similarity of the entire trademark is also input to the input layer. Here, the degree of influence is simply calculated from the number of different points. That is, the degree of influence is calculated by 1 / (the number of different portions). Therefore, if the difference is one tone, the similarity of the trademark is determined only by the difference part, so that the degree of influence after normalization is 1, and if the difference is two sounds, it is assumed that the influence of each difference part is half. , The degree of influence becomes 0.5, and if there are three different sounds, the degree of influence of each different portion becomes 0.333.

When two or more different characters are handled, the front and rear parts of the different parts are not necessarily the same between the two trademarks because the different parts may be adjacent to each other. Therefore, in the present embodiment, a unit for identifying whether the front part and the rear part of the different part are common to both trademarks is also provided in the input layer. In the illustrated example, 1 is input if they are common, and 0 is input if they are not common.

In the case of common, the common character may be input to the unit set as the character before or after the different part. However, handling when the common character is not common becomes a problem. In this case, all 0s may be put in the unit set, but in the present embodiment, if there is a portion common to consonants, vowels, etc., only the common portion is input. For example, if the characters on the front side of the different part are "ka" and "da", they are not common. Since vowels are common to "ka" and "da",
The common vowel "A" is input to a unit set for character recognition in front of a different character. Note that the same applies to the character behind the different character.

On the premise of the above, first, the pattern information of the one-character difference trademark will be described. It should be noted that only one pattern information is created from two trademarks having one character difference. Therefore, the degree of influence of the pattern information on the similarity of the entire trademark is 1. In the case of a single character difference, the characters before and after the difference character should be common between the two trademarks, and therefore, a unit for identifying the character indicates that the character is common between the two trademarks.
Is inserted.

For example, when comparing “Avon” and “Avon”, the above (1) becomes “I” and “−”, (2) becomes 4, and (3) shows that the difference character is 2 from the beginning of the word.
It is 0.333 because it is in the letter (see FIG. 6),
(4) becomes "e" and "bo". If the two trademarks are determined to be similar in a past decision, 1 is put in the output layer in this embodiment.

Further, in the case of comparison between “Suiho” and “Suihou”, the above (1) becomes “S” and “Z”, and (2)
Is 4, and (3) is 0 because of the difference in the beginning of the word,
(4) is “_” (blank) and “i”. If the two trademarks are determined to be similar in the past decision, 1 is put in the output layer.

In the case of comparison between “Oaks” and “Oak”, the above (1) becomes “S” and “” (blank),
(2) becomes 3, (3) becomes 1 because of the difference in the ending, and (4) becomes "-" and "_" (blank). And
If the two trademarks are determined to be dissimilar in the past decision, 0 is entered in the output layer.

Further, in the case of comparison between "dim" and "gym", the above (1) becomes "di" and "ji", (2) becomes 2, and (3) becomes 0 because of the difference in the beginning of the word. (4)
Becomes "" (blank) and "mu". If the two trademarks have been determined to be dissimilar in the past decision, 0 is entered in the output layer.

Next, creation of pattern information from two trademarks having a difference of two or more characters will be described.

For example, “octanone” and “optaon”
In the case of comparison with "2", "ku" and "pu", "no" and "o"
It can be seen that there are differences. Therefore, two pieces of pattern information are created for each of these different parts. Therefore, the degree of influence of each pattern information is 0.5 at a time. In the first pattern information, the above (1) becomes "ku" and "pu", (2) becomes 5, and (3) becomes 0.25 because the different character is the second character from the beginning of the word ( FIG. 6) and (4) are "o" and "ta". If the two trademarks have been determined to be dissimilar in the past decision, 0 is entered in the output layer.

In the second pattern information, the above (1) becomes “no” and “o”, and (2) becomes 5,
(3) is 0.7 because the difference character is the fourth character from the beginning of the word
5, (4) becomes “ta” and “n”, and the output layer becomes dissimilar 0. In addition, in each pattern information, the character before and after the different character is common to both trademarks, and therefore, 1 is entered for each.

Next, in the comparison between "parasgin" and "paraxin", it can be seen that two differences, "su" and "ki" and "gi" and "shi" are adjacent to each other. Therefore, two pieces of pattern information are created for each of these different parts.
Therefore, the degree of influence of each pattern information is 0.5. In the first pattern information, the above (1) becomes "S" and "G", (2) becomes 5, and (3) becomes 0.5 because the different character is the third character from the beginning of the word. . Then, it can be seen that the character on the front side of (4) is common to “la”, but the character on the rear side is “gi” or “shi”, which is different between the two trademarks. In this case, non-common 0 is put in the identification unit for determining whether the character behind the different character is common or not.
Since the characters "Gi" and "Shi" on the rear side have the same vowel "I", "I" is inserted. If the two trademarks are determined to be dissimilar in the past trial, 0 is entered in the output layer.

Similarly, in the second pattern information, the above (1) becomes “gi” and “shi”, and (2) becomes 5,
(3) is 0.7 because the difference character is the fourth character from the beginning of the word
It becomes 5. Then, although the character on the rear side of (4) is known to be common with "n", the character on the front side is "su" or "ki",
There are differences between the two trademarks. In this case, non-common 0 is put in the identification unit for determining whether the character preceding the different character is common. Since the front characters "S" and "K" are not common to all vowels, consonants, and the like, nothing is put in the unit set of the front characters, and all become "0". If the two trademarks have been determined to be dissimilar in the past decision, 0 is entered in the output layer.

If there is a difference of three characters, three pieces of pattern information are created, and the degree of influence of each pattern information is 0.33.
Except for 3, it is created in the same manner as in the case of a two-character difference. In this way, a difference of four or more characters can be handled. The pattern tables of FIGS. 3 to 5 show examples of trademark pattern information up to a three-character difference.

In constructing the network, the results of a large number of past trademark decisions on the similarity of two trademarks are utilized, and the contents of the decision are converted into pattern information according to the above rules, and learned by the neural network learning program. Let it. As the specific learning method and learning program, various existing ones (for example, see the above-mentioned literature) can be used.
Here, a three-layer perceptron was considered, and learning was performed by a sequential correction method of back propagation. The number of intermediate layers, the number of units in the intermediate layer, and various parameters necessary for learning are appropriately set.

The learning for constructing the neural network is realized by, for example, a personal computer. That is, the personal computer is provided with a neural network learning means (program), a neural model construction means including a pattern information storage means storing a large number of pattern information necessary for learning, and the like, and a learning result is obtained. The coupling load and the like are stored in the coupling load accumulating means. In order to facilitate the creation of the pattern information, it is preferable to provide a normalization processing means (program) for automatically converting the pattern information only by entering the trademark (character) to be compared.

Then, pattern information is created and learned from a large number of past decision cases, and the connection weights and the like between the units at the time when the error is appropriately reduced are stored, thereby constructing a neural network. For example, a connection weight or the like when the sum of the square errors between the expected output and the net output is equal to or less than a predetermined value is used. Alternatively, there is a method of using an evaluation pattern (pattern information other than a teacher) in a learning process and using a connection weight or the like at the time when an error with respect to the evaluation pattern is minimized.

When it is desired to judge the similarity of the two trademarks, the input layer pattern information created from the two trademarks is input to the input layer of the network. Then, since the network outputs a value in the range of 0 to 1 to the output layer, the similarity between the two trademarks can be seen from the value. In the case of this embodiment, as described above, the dissimilarity is set to 0 and the similarity is set to 1 in the learning teacher data. Therefore, the closer the output layer value is to 1, the more similar the two trademarks are. I understand. In addition,
The program may be configured to automatically display any of “similar” or “dissimilar” from the value of the output layer. For example, if the value is 0 or more and less than 0.5, "dissimilar" may be displayed, and if the value is 0.5 or more and 1 or less, "similar" may be displayed.

If there is a difference of two or more characters, a plurality of pieces of pattern information are created, and a similarity determination is made using a neuro. Therefore, in that case, similarity may be determined based on the average value of each output value for each input layer pattern information. That is, after adding the output values for each different part,
Similarity can be determined from the value obtained by dividing by the number of differences (the number of patterns). In addition, in the case where the degree of influence is given with a light weight in consideration of a different character, a different place or the like regardless of the number of differences between the two trademarks, the degree of influence can be added. In that case, for example, a value obtained by multiplying the output value by the post-normalization degree of influence of the pattern may be added.

Next, an example of a method for producing a similar product such as a trademark of the present invention will be described. For creating a similar trademark, the neural network for similarity determination described in each of the above embodiments can be used.

For example, in the network shown in FIG. 2, that is, in the input layer, the different characters of the two trademarks to be compared and the characters before and after the two characters are put into the input layer, and the similarity of the similar character string is output to the output layer using the network. The creation method is described.

When a similar trademark is created, a plurality of character strings are developed from a specific first character string which is a basis for creating a similar character string, and each character string and the Japanese syllabary are included in an input layer. By inputting the second character (second character string / virtual name) in order, each similarity between them is output, and the second character whose similarity is equal to or greater than a predetermined value is used for the second character. Create a similar character string of one character string. Here, the second character is basically one character consisting of the Japanese syllabary or the like, but from the viewpoint of the nature of the network, the second character is substantially a character string (the second character string) with characters before and after the character. It can be recognized. This is because the network is configured to compare three characters of the different character and each of the common characters before and after the different character.

The creation of a similar character will be specifically described. For example, consider a method of creating a similar character string of a three-character trademark “guide”.

First, consider the creation of a similar trademark in which each character constituting the basic trademark "Guide" has been replaced with another character, and a similar trademark in which the basic trademark "Guide" has other characters at the beginning and end of the trademark. This is performed by arranging a part to be replaced or added to a different character in the input layer, and putting characters before and after the part in the input layer.

That is, “__ga” (“_” for the first character, “_” for one of the different characters, and “ga” for the second character) in the input layer and the other different characters are appropriately created. The second character is inserted in order. As the second character, 5 of "A" to "N"
Zero sound, voiced and semi-voiced characters, prompting sounds, long sounds, murmurs, foreign words, or characters with these characters are adopted in order.

Then, each similarity between “__ga” and each second character is output, and among them, the second character whose similarity is equal to or more than a predetermined value is adopted. The predetermined value of the degree of similarity can be appropriately set, but is, for example, a numerical value of 0.5 or more, and a value of 0.7 to 0.8 or more is suitably adopted. As the adopted second character, it is preferable to adopt not only the second character having the maximum similarity but also a plurality of characters whose similarity is equal to or more than a predetermined value. If there is no similarity equal to or greater than the predetermined value, the similarity may not be adopted at all.

In this way, "? Guide" with another word added to the beginning of "guide" is created as a similar character string of "guide". Note that “?” Is the second character adopted by the above method.

Also, by inserting “_ guy” and an appropriate second character in the input layer, a similar character string “? Id” in which “ga” in “guide” is replaced by another character is created.

Further, by inserting “guide” and an appropriate second character in the input layer, a similar character string “gadido” in which “a” of “guide” is replaced by another character is created.

Also, by putting "id_" and an appropriate second character in the input layer, a similar character string "Gai?" In which "do" of "guide" is replaced by another character is created.

Also, by putting “do__” and an appropriate second character in the input layer, “guide?” With another word at the end is created as a similar character string of “guide”.

Next, consider the creation of a similar character string in which another character is added between the constituent characters of the basic trademark "guide". In this case, an inter-character portion to be added is arranged in a different character in the input layer, and characters before and after the character are entered in the input layer.

To do this, first, “GA_A” and an appropriate second character are put in the input layer, so that “GA”
A similar character string "guide" in which another character is added between "" and "i" is created.

Also, by inserting “I_D” and an appropriate second character in the input layer, a similar character string “G” having another character added between “I” and “D” can be obtained. Guide "is created.

Further, each character in which one of the constituent characters of the "guide" is deleted is created, and the similarity between each of the created character strings and the "guide" is output. It is also possible to leave only the character string. In other words, "guide"
Similarity between the character strings of "id" with "ga" removed, "Gado" with "i" of "guide" removed, and "guy" with "do" removed of "guide" and "guide" I look at you.

In the case of the “guide” and the “id”, the difference is in the presence or absence of the “ga” (“ga” and “_”). What is necessary is just to look at the similarity.
In the case of “guide” and “gad”, the difference is in the presence or absence of “i” (“i” and “_”). You should look at.
Furthermore, in the case of "guide" and "guy", the difference is in the presence or absence of "do"("do" and "_"), so that "id_" and "_" are put in the input layer and similar Just look at the degree.
Then, only those having a predetermined similarity are left.

However, the deletion of the constituent characters is usually
Since it is sufficient to judge the replacement of the constituent characters, it can be omitted. That is, in this example, "id" should have been adopted if the degree of similarity is high when comparing "_ guy" with the second character "_". "Gado" and "Guy"
Similarly, if the similarity is high by comparing the “guide” or “id_” with the second character “_”, it should be adopted.

In the above description, the word length, the position of the different character, the commonality of the character before and after the different character, and the like have been omitted for the sake of convenience. However, it is needless to say that these can be considered.
For example, when creating a similar trademark in which the constituent characters of the basic trademark have been changed (including deletion) to other characters, the word length of the basic trademark may be entered, and the constituent characters of the basic trademark may be included. When other characters are added at the beginning, end, or between words, the word length of the basic trademark may be added by 1 as the word length. The position of the different character may be set according to the position to be changed or added. In addition, 1
When creating a similar trademark with a different letter, it is sufficient to put 1 in the degree of influence, and when creating a similar trademark with a difference of two or more characters, change the degree of influence accordingly and The characters before and after the difference character may be appropriately set.

As described above, the similar character string of "guide" is created. When searching these from the trademark database, not only the same as the "guide" itself but also the similar character string is used. , It may be configured to search for a trademark including them. When searching for a trademark that includes the word, it is possible to omit the above-described determination of the addition of the beginning or end of the trademark.

In the above description, creation of a similar trademark with a difference of one character has been described. However, the degree of influence is changed to 0.5 or the like, and a difference of two characters or more is obtained by using a plurality of second characters adopted as described above. You can also create a trademark. For example, "?
And so on.

Although the three-letter trademark "guide" has been described here, the similar technique can be applied to the creation of a similar trademark of a trademark having other characters. For example,
If you want to create a similar trademark of the four-character trademark "Segaste", "__se", "_sega", "segas", "gaste", "ste_", "te__", "se_ga", "Ga_
And the similarity between each character string of "su_te" and an appropriate second character, and the second character having the predetermined similarity,
"? Segaste", "? Gaste", "S? Ste", "Sega? Ste", "Segasu?", "Segaste?", "Sé? Gaste", "Sega? Ste", "Sega? Ste""Segaste"
Can be created as a similar character string. In addition,
The "?" Part may mean deletion in the same manner as in the example of the above "guide". For example, “Sesute” may include “Sesute”.

Next, a system for judging similarity of a trademark using the neural network of the present invention will be described. This system is mainly provided with an automatic similarity determination unit composed of a neural network constructed as described above.

FIGS. 7 and 8 are schematic diagrams showing one embodiment of the similarity / non-compliance determination system of the present invention.

The system of this embodiment judges whether a trademark registered or filed with the Patent Office is similar to a trademark which is to be used or filed in the future or in the future, and based on the result, the possibility of trademark registration is determined. It is a system that assists in determining such.

The present system is an automatic similarity determination unit for determining the similarity of a trademark using a neuro (similarity determination processing means).
And an input device (input means) including a trademark database 5 in which trademarks registered or applied to the JPO are stored, and a keyboard 2 for inputting a trademark whose similarity is to be determined.
And an output device (output means) including a display 3 and a printer 4 for outputting a database search result, a similarity determination result by the automatic similarity determination unit, and the like.

A system without the trademark database 5 can also be used. In this case, the two trademarks input from the input device 2 are converted into input layer pattern information by the feature extraction unit (normalization processing means). Then, based on the information, the automatic similarity determination unit determines similarity of the trademark and outputs the result to an output device. FIG. 9 shows the flow of processing in this case.

In the trademark database, for example, a trademark itself, one or more names thereof, an application number, a registration number, and the like are registered and stored together with classifications and similar groups.

The automatic similarity judgment section is provided with the neural network constructed for the similarity judgment method of the present invention as described above. The automatic similarity determination unit is realized by, for example, the personal computer 1. In this case, the personal computer is provided with a neuro execution means (program) for constructing a neuro using a connection weight or the like obtained by learning and for outputting an output to an output layer based on input layer pattern information. The personal computer 1 can be connected to a server or the like having the trademark database 5 via a communication line or the like. Alternatively, the trademark database 5 can be realized by a medium such as a CD-ROM. In this case, a CD-ROM or the like storing a registered trademark or the like is stored in a CD-R of a personal computer.
Start up and use it in an OM drive or the like.

Then, the predetermined program is started on the personal computer 1 and the trademark to be investigated (input trademark /
Specific name). Then, the personal computer 1 uses the data processing means to output the same trademark as the input trademark, a difference of one character from the input trademark (including addition or deletion, or a difference of two characters), a difference of two characters, and the like from the trademark database 5. Search for related trademarks such as those included in the section. If there is the same trademark, the fact is output to the output device.

The retrieved related trademark and the input trademark input from the input device 2 are extracted by a feature extracting unit to extract a characteristic portion suitable for a network to be used, such as a different portion between the two trademarks, and normalize the extracted trademark. It is converted into input layer pattern information using a processing means. Then, the automatic similarity determination unit performs similarity determination processing from the input layer pattern information (obtains an output of the output layer), and outputs the result to the output device 3. FIG. 10 shows the flow of a series of these processes.

An automatic similarity determination unit and a feature extraction unit for performing automatic similarity determination processing are provided on the trademark database side (a server or the like), and transmit a trademark input from an input device via a communication line to search for a related trademark. And the similarity judgment processing can be performed outside the personal computer. In this case, the personal computer controls input / output and communication. FIG. 11 shows an example of the system configuration in this case. In particular, by connecting to a server from a terminal equipped with a WWW browser via the Internet and inputting a name on a web page of the website, the server searches for a related name from a trademark database and automatically determines whether or not the name is similar. It is preferable that the result of the judgment is determined by the section and the result is displayed. Note that a system may be adopted in which a server having an automatic similarity determination unit determines the similarity based on the input of the names of the two trademarks on the webpage and displays the result on the webpage.

By the way, in case that there is a trademark which is the same as or similar to the input trademark among the related trademarks, a candidate trademark creating section for creating an unregistered and unfiled candidate trademark based on the related trademark is provided in the system. May be provided.

The candidate trademark creation section creates a trademark that is not the same as the related trademark based on the input trademark and / or the registered and pending related trademark retrieved from the trademark database based on the input trademark. Create unregistered and unfiled candidate trademarks. Specifically, a trademark that is different from the input trademark by one character or the like and is not in the trademark database (not a related trademark) is created as a candidate trademark.

For example, by designating one or more characters to be changed from among the constituent characters of the input trademark from the input device, the designated character is replaced with another character and is not in the trademark database. By extracting the trademark, an unregistered and unfiled candidate trademark is created.

Then, the similarity between the created candidate trademark and the related trademark or the like is determined by the automatic similarity determination unit. Then, a trademark having a relatively low similarity is finally left. Therefore, it can be said that the remaining candidate trademarks are unregistered, have not been filed, and are relatively likely to be registered. FIG. 12 shows an example of a series of these flows.

It should be noted that, among the candidate trademarks, for the trademark to be adopted, the related trademark is further searched, and the similarity between each related trademark and the adopted candidate trademark is determined by the automatic similarity determination unit. It can be said that it is more preferable to confirm the presence or absence of the trademark.

Next, a similar character string creation system will be described. This system is provided with a similar character string creation unit (similarity judgment processing means and similarity creation means) composed of a neural network or the like constructed as described above,
A similar character string such as a character string input from the input device is created and used appropriately.

For example, a system suitable for searching whether or not a trademark that is the same or similar to a trademark to be used or filed in the present or future and has already been registered or filed with the Patent Office can be obtained.

In this case, the system is provided with a similar character string creation unit having a neural network for similarity judgment, and this similar character string creation unit is preferably provided in the similarity judgment system. That is, in addition to the similar character string creating unit, the trademark database 5, an input device, an output device, and the like are provided. A system without the trademark database 5 may be used. In this case, based on the trademark input from the input device 2, the input layer pattern information is created by the normalization processing means of the feature extraction unit, and the information is created. A similar character string is created by the similar character string creating unit on the basis of the above, and the result is output to the output device.

As described in the similar character string creating method, the similar character string creating unit is constituted by a neural network constructed for a trademark similarity judging method. When a neural network constructed for a similarity determination method of a trademark is used, the automatic similarity determination unit in FIG. 8 functions as a similar character string creation unit.

Also in the case of this system, in the feature extraction unit, the first trademark (such as an input trademark), which is the basis for creating a similar trademark, is used.
A character string is developed based on the above, input layer pattern information is created by the normalization processing means by each character string and an appropriate second character, and is input to the input layer of the network of the similar character string creating unit. You. Then, the similar character string creating unit creates a similar character string of the first trademark based on each neuro output.

The input trademark and each similar trademark created by the similar character string creating unit are used by the trademark database 5 via the data processing means to obtain the same trademark as the trademark or a trademark partially including the trademark. Is searched. Then, a search result of the presence / absence of a trademark that is the same as or similar to the input trademark is output to the output device 3. FIG. 13 shows a flow of a series of these processes.

Note that, in this case as well, the similar character string creation unit
It is also possible to create a trademark on the trademark database side (server), send the trademark entered from the input device via a communication line, and create similar trademarks and search for them outside of a personal computer. This case can also be realized on an Internet website. That is, it is possible to receive an input of a trademark on a web page, create or search for a similar trademark based on the input, and display the result on a terminal.

By the way, the predetermined program, automatic similarity judgment processing, similar character string creation processing, candidate character string (trademark)
Each part of the creation process, or input / output control and communication control,
The combination of the feature extraction process (normalization process) is CD
-It can be distributed through various recording media such as a ROM.
Note that a configuration in which the connection load of the neuro execution means can be updated by an appropriate means is preferable.

Note that the similarity determination method / system of the present invention
The method and system for creating a similar object are not limited to the configurations of the above-described embodiments, and can be changed as appropriate. In particular, the configuration of the unit (unit set) for character identification (the method of inputting and identifying characters) and what constitutes one character (one sound) can be changed as appropriate. It is not bound by the description.

Further, in each of the above embodiments, the product category and the like are not taken into consideration. However, the search of the trademark database and the like can be performed in the similar group unit, the classification unit, or all classes. It is.

In the similarity judgment processing and the similar character string creation processing, the similarity, which is the output of the output layer of the neural network, is divided by a predetermined boundary value, so that less than the predetermined value is regarded as dissimilar. A value equal to or greater than a predetermined value was similar. Therefore, if the system is configured to be able to change the boundary value, the level of the similarity judgment criterion can be adjusted. By increasing the boundary value,
The number of trademarks judged to be similar can be reduced, and the number of extractions from the trademark database can be reduced. It is preferable that the boundary value is adjusted by an input device by an operator.

In the above embodiment, when a plurality of pieces of pattern information are created due to a plurality of different parts,
Although the similarity of the entire trademark is obtained by including the degree of influence for each pattern information, the unit of the degree of influence may be omitted and the output value of the output layer may be simply adjusted to take measures. For example, when two similar trademarks have two different parts, the output value of each pattern information is set to 0.5 or the like.

Further, the input layer may include other elements likely to affect the similarity judgment.
For example, the presence / absence of a prefix or suffix, the presence / absence of an idea possessed by the trademark, the relationship with the designated product, the remarkable difference in appearance, etc. may be considered.

[0128]

As described in detail above, according to the present invention,
It is easy to determine the similarity of a name such as a trademark.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a three-layer perceptron.

FIG. 2 is a diagram showing an example of a neural network for a similarity judgment system for a trademark or the like of the present invention.

FIG. 3 is a diagram showing a part of a table showing an example of pattern information for network construction of FIG. 2, and one table is constituted by FIG. 5;

FIG. 4 is a diagram showing a continuation of the table in FIG. 3;

FIG. 5 is a diagram showing a continuation of the table of FIG. 4;

FIG. 6 is a diagram illustrating an example of normalized data of a position of a different character in the network of FIG. 2;

FIG. 7 is a schematic diagram showing an example of a system for determining similarity of a trademark or the like according to the present invention.

FIG. 8 is a diagram showing an example of a system for judging similarity of a trademark or the like of the present invention.

FIG. 9 is a diagram showing an example of a processing flow of a similarity / unlikeness determination system for a trademark or the like of the present invention.

FIG. 10 is a diagram showing another example of the processing flow of the similarity / unlikeness determination system for a trademark or the like of the present invention.

FIG. 11 is a diagram showing another example of the similarity / unlikeness determination system for a trademark or the like of the present invention.

FIG. 12 is a diagram showing still another example of the processing flow of the trademark similarity determination system of the present invention.

FIG. 13 is a diagram showing an example of the processing flow of the trademark similar character string creation system of the present invention.

[Explanation of symbols]

 1 PC 2 Keyboard 3 Display 4 Printer 5 Trademark Database

Claims (11)

[Claims] 1. An input layer learning signal created based on two patent names to be compared and created based on a past trademark decision example of the JPO and an output layer teacher signal based on a result of similarity judgment between the two names. Equipped with a neural network constructed by learning a large number of learning pattern information. The neural network can input not only the difference between the two names to be compared but also the front and back of the difference in the input layer. In the case where there are a plurality of different parts, separate pattern information is provided for each of the different parts, and input layer pattern information created based on an alias for which similarity is to be determined is input to the input layer of the neural network. A similarity judgment system for a trademark or the like, wherein the similarity between the two names is obtained from the output of the output layer. 2. The apparatus according to claim 1, wherein the front part or the rear part of the different part is input based on one of the names and whether or not it is common with that of the other name. Similarity judgment system for trademarks and the like. 3. The different part and the respective names of the front part and the rear part of the different part are respectively input by being divided into elements such as consonants and vowels, and the front part and the rear part of the different part are between the two names. In the case of common, the common name is input, and in the case of non-common between the two names, when there is a common vowel or consonant, the common part is input. Item 2. Similarity determination system for trademarks and the like according to item 2. 4. When different pattern information is set for each different part based on the presence of a plurality of different parts, the different part represented by the pattern information for each of the pattern information is the entire trademark or the like. The similarity determination system for a trademark or the like according to claim 3, wherein the degree of influence on similarity is also input. 5. Among the names to be judged similar or not, one of the names is the name of the trademark to be searched for judging the possibility of registration of a trademark with the Patent Office, etc. Among the names, the other name is a related name searched from the trademark database in which the names of the trademarks registered and filed with the JPO are stored based on the name of the search object. A similarity determination system for a trademark or the like according to any one of claims 1 to 4. 6. An input layer learning signal which is created based on a past trademark decision example of the JPO and is based on two names to be compared, and an output layer teacher signal based on the similarity judgment result of the two names. Equipped with a neural network constructed by learning a large number of learning pattern information. The neural network can input not only the difference between the two names to be compared but also the front and back of the difference in the input layer. When there are a plurality of different parts, each of the different parts is set to separate pattern information, and a plurality of different parts are created based on a specific naming for which a similar naming is to be created and a virtual naming such as Japanese syllabary. The input layer pattern information is sequentially input to the input layer of the neural network, and the similarity between the specific name and each virtual name is determined from each output of the output layer. A similar naming system for a specific naming system based on the virtual naming system and the specific naming system whose similarity is equal to or greater than a predetermined value. 7. The specific name is a name of a trademark to be searched for judging the possibility of registration of a trademark with the JPO, and the name of the trademark registered and pending with the JPO is accumulated. 7. Searching the trademark database for related names such as the specific name and similar names to assist in judging the possibility of trademark registration with the JPO. A system for creating similar objects such as trademarks. 8. The system is realized by a web server on the Internet, by connecting to a website of the system from a terminal equipped with a WWW browser and inputting a trademark on the web page, 6. A system for judging the similarity of a trademark or the like according to any one of claims 1 to 5, wherein the similarity of the trademark or creation of a similar trademark is enabled. A system for creating similar items such as the described trademark. 9. A name for a similarity determination system for a trademark or the like according to any one of claims 1 to 5, or a similarity creation system for a trademark or the like according to claim 6 or 7. A computer-readable recording medium in which a program for causing a computer to execute an input receiving process, a similarity determination process based on a neural network based on the input receiving process, and a result output process is recorded. 10. A system for judging the similarity in the name of a trademark or the like using a computer, wherein the name of the trademark or the like is “A”, “I”, “U”, “E”,
Each row of "o" and "ka", "sa", "ta", "na",
Each line of "c", "ma", "ya", "la", and "wa", and sound-repelling "n", voiced point "゛", semi-voiced point "゜", prompting sound "tsu", and letter "ya" , "U", "yo", long sound "-", characters for foreign words "a", "i", "e", "o", etc. which are input or processed as binary data A similarity judgment system for trademarks and the like characterized by the following. 11. A system for judging the similarity in the name of a trademark or the like using a computer, wherein the positions of different parts of two trademarks or the like subject to similarity determination are determined by comparing the positions of the different parts with respect to the entire trademark or the like. A similarity judgment system for trademarks or the like characterized by being input or processed at the arrangement ratio.