JPH08339376A - Foreign language search device and information search system - Google Patents

Abstract

(57) [Summary] [Purpose] Efficiently search foreign words registered in the database using katakana words. [Structure] A phonetic symbol / katakana correspondence table 7 that stores the correspondence between each phonetic symbol and katakana characters, and the phonetic symbols of registered data consisting of foreign words and phonetic symbols input from the registered data input unit 1 Phonetic symbol katakana conversion means 3 for converting to katakana words using the katakana correspondence table, a database 3 for storing and holding foreign words of registered data and converted katakana words, and a search keyword input from the search keyword input unit 4. Word similarity calculation means 5 for calculating the word similarity Ri for each katakana word stored in the katakana word database 8 of
And an output unit 6 for outputting, as a search result for a search keyword, a foreign word corresponding to a katakana word having a calculated word similarity Ri of a specified value or more.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a foreign language search device capable of searching a foreign language word registered in a database with a search keyword consisting of katakana words, and an information search system incorporating the foreign language search device.

[0002]

2. Description of the Related Art In a general information retrieval system, when a new document or the like is registered in a database, a plurality of keywords related to the contents of the document included in the document are registered in this database. This keyword is generally composed of one word.

When a searcher searches a document in which necessary contents (information) are described using this information search system, he searches a database for a keyword related to the necessary information, and this keyword is registered. Information that identifies the relevant document, such as the document name of the document, publisher, author, issue date, and holding place, is output as the search result.

In such an information retrieval system,
When a Japanese document is registered, the word that is a keyword included in this document is Japanese. Therefore, it is sufficient to register this Japanese word in the database as it is.

When a document in a foreign language is registered in the database, the words as keywords included in this document are naturally foreign languages. If the foreign language word of this keyword is a general noun, then this foreign language word may be translated into Japanese and registered.

Generally, in the case of proper nouns such as a person's name or a place name, or when a noun for which no proper Japanese translation exists is written in Japanese, katakana notation is used, but reading of proper nouns is very complicated. There may be multiple types of katakana notation. There are also foreign words that are difficult to write in katakana accurately. In such cases, foreign words may be registered in the database.

[0007]

However, the information retrieval system in which foreign language words are registered in the database as described above still has the following problems. That is, when a user of this information retrieval system needs to retrieve a document in which the foreign language word is described, he / she needs to input the foreign language word written in a foreign language as a search keyword.

However, general users of the information retrieval system do not always know the correct foreign language spelling of foreign words. Therefore, the user needs to perform a search operation after checking the correct spelling using a dictionary or the like, which increases the burden on the user.

Further, in many cases, proper nouns and personal names are not listed in a dictionary or the like. In such a case, it cannot be dealt with at all. In order to eliminate such an inconvenience, it is sufficient to register the katakana word in which the foreign language word is converted to katakana notation in the database. However, as described above, there are multiple types of katakana notation, and if the user does not enter the katakana word of the same katakana notation registered as the search keyboard, the target katakana word and this katakana word There arises a problem that information specifying a document containing a word cannot be retrieved.

The present invention has been made in view of the above circumstances, and by inputting a foreign language word and a phonetic symbol or a voice when the foreign language word is registered, the foreign language word and the katakana word of the corresponding word are stored in the database. A foreign language search device and a foreign language search device that can automatically search for a correct foreign language word even if you search with a katakana word that is automatically registered and has a different notation from the registered katakana word. The purpose is to provide an information retrieval system.

[0011]

In order to solve the above-mentioned problems, in the foreign language search device according to claim 1, registered data consisting of a foreign word written in a foreign language and a phonetic symbol of this foreign word is provided. The registered data input section to be input, the phonetic symbol / katakana correspondence table in which each phonetic symbol and the corresponding Katakana character corresponding to the phonetic symbol are stored, and the phonetic symbol of the registered data input via the registered data input section. Phonetic symbol katakana conversion means for converting into katakana words using the phonetic symbol / katakana correspondence table, and stores and retains foreign words of each registration data input via the registration data input section and katakana words converted to katakana The database, the search keyword input section where the search keywords consisting of katakana words are input, and the search keyword input section where the search keywords are input A word similarity calculation means for calculating the word similarity for each katakana word stored in the database of katakana words of the search keyword, and a foreign word corresponding to the katakana word for which the calculated word similarity is a specified value or more. And a search result determination output unit that outputs the result as a search result.

Further, in the foreign language search device according to claim 2, the foreign language word written in the foreign language and the registered data input unit for inputting the registered data consisting of the voice of this foreign language word, and each voice corresponding. The phonetic / phonetic transcription correspondence table in which each phonetic symbol for the voice is stored, and the voice of the registration data input via the registration data input unit
A phonetic phonetic symbol conversion means for converting phonetic symbols using a phonetic transcription table, a phonetic symbol / katakana table storing phonetic symbols corresponding to each phonetic symbol and the corresponding phonetic symbol, and phonetic phonetic symbol conversion. Phonetic symbol katakana conversion means for converting phonetic symbols obtained by means into katakana words using phonetic symbol / katakana correspondence table and foreign words and katakana of each registered data input via the registered data input section A database that stores and stores the specified Katakana words, a search keyword input section in which a search keyword consisting of Katakana words is input, and each Katakana word stored in the Katakana word database of the search keywords input from this search keyword input section The word similarity calculation means for calculating the word similarity to And a search result determination output unit that outputs the search results for the search keyword a foreign language word corresponding to the above katakana word value.

Further, according to claim 3, the word similarity calculating means in each of the foreign language search devices includes each katakana character forming a katakana word of a search keyword and each katakana character forming a katakana word stored in a database. The character similarity indicating the similarity for each character is calculated by weighting it according to the character type, and the word similarity is calculated using a predetermined formula from the calculated character similarity for each character. I am trying to do it.

In the information retrieval system of claim 4,
The registration data input part where the registration data consisting of the foreign language word written in a foreign language, the pronunciation symbol of this foreign language word and the information about the foreign language word is input, and each pronunciation symbol and the katakana character corresponding to the corresponding pronunciation symbol are Registered phonetic symbol / katakana conversion table and phonetic symbol / katakana conversion means for converting phonetic symbols of registration data input via the registered data input section to katakana words using the phonetic symbol / katakana correspondence table, and registration. A database for storing and holding foreign words, information and katakana words converted into katakana of each registration data input through the data input unit;
A search keyword input section for inputting a search keyword composed of katakana words, and a word similarity calculation means for calculating a word similarity for each katakana word stored in a database of katakana words of the search keyword input from the search keyword input section. And a search result determination output unit that outputs, as a search result for a search keyword, a foreign language word and information corresponding to a katakana word whose calculated word similarity is equal to or higher than a specified value.

[0015]

In the foreign language search device according to claim 1 configured as described above, when registration data including a foreign language word and a phonetic symbol is input to the registration data input unit, the katakana word in which the foreign language word is written in katakana from the phonetic symbol is written. Has been created,
Stored in a database together with foreign words.

When searching for a foreign language word, the katakana word written in katakana is input to the search keyword input section as a search keyword. As described above, Katakana words that exactly match the Katakana words entered as search keywords do not always exist in the database, so the word similarity for each Katakana word stored in the database of search keyword Katakana words is calculate. Then, the foreign language word corresponding to the katakana word whose word similarity is equal to or higher than the specified value is set as the search result for the search keyword.

Therefore, even if the Katakana word of the search keyword and the Katakana word indicating the same foreign language word in the database do not completely match, it is possible to search for the correct target foreign language word with a very high probability. .

In the foreign language search device according to the second aspect, when the registration data consisting of the foreign language word and the voice is input to the registration data input unit, the voice is once converted into a phonetic symbol,
Converted to katakana words. The katakana word is stored and held in the database together with the foreign language word.

The method of searching for a foreign language word in a database in which katakana words corresponding to foreign words are stored and held with a search keyword composed of katakana words is the same as the foreign language search device of claim 1.

In the foreign language search apparatus having such a configuration, when a foreign language word is registered in the database, it is only necessary to input a voice in addition to the corresponding foreign language word, so that the registration work efficiency is greatly improved. .

Further, in the foreign language search device according to the third aspect, as a word similarity calculation method, a character similarity indicating a similarity is calculated for each katakana character forming two katakana words to be compared and compared. . Then, it is calculated from the respective character similarities of the respective katakana characters forming the word similarity between the katakana words composed of a plurality of katakana characters.

In this case, each character similarity is calculated by weighting it according to the type of each character. For example, comparing the case where the consonant rows match and the case where the vowel strings match, the similarity of the entire katakana word is higher when the consonant strings match, so the consonant similarity The weight is set higher than the weight of the similarity of vowels.

In the information retrieval system of claim 4,
In the database, in addition to foreign words and katakana words, various kinds of information about the corresponding foreign words are stored. Therefore, when a katakana word is input as a search keyword, necessary information about this foreign language word is output from the database together with the corresponding foreign language word.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of an information search system incorporating a foreign language search device of the embodiment. This information retrieval system is composed of a kind of information processing device such as a computer. Specifically, the registration data input unit 1,
It comprises an input data buffer 2, a phonetic symbol / katakana conversion unit 3, a search keyword input unit 4, a search processing unit 5, a search result output unit 6, a phonetic symbol / Katakana correspondence table 7 and a database 8.

As shown in FIG. 2, the phonetic symbol / katakana correspondence table 7 stores katakana characters corresponding to each phonetic symbol. Specifically, when katakana characters are arranged in a Japanese syllabary matrix, a consonant correspondence table 7a indicating which line each consonant of a phonetic symbol corresponds to, and which column each vowel of a phonetic symbol corresponds to. A vowel correspondence table 7b indicating whether they correspond is stored.

Then, in principle, one katakana character is specified by a pair of a consonant and a vowel. For example, since the phonetic symbol of [ba] is designated as [A row] of [C row] from each of the tables 7a and 7b, the Katakana character [BA] is specified from the Japanese syllabary matrix array.

However, consonants and vowels are not always paired in a series of phonetic symbols indicating one foreign language word. Such an exception is also registered in the consonant correspondence table 7a. For example, when the phonetic symbol of the consonant [d] appears alone, it becomes the katakana character of [do] instead of [to].

As described above, in the phonetic symbol / katakana correspondence table 7, corresponding katakana characters are determined for all combinations or single occurrences of phonetic symbols when a foreign word is written by phonetic symbols, and Information is stored that can convert a word into a katakana word that is uniquely determined by its phonetic symbol.

The database 8 is formed in, for example, an HDD, and as shown in FIG. 1, a plurality of areas with area numbers i are formed therein, and each area has a foreign word. The katakana word corresponding to the foreign language word and the information for identifying the relevant document such as the document name, the publisher, the author, the date of publication, and the holding place of the document in which the foreign language word is described are stored.

The registration data input unit 1 is composed of, for example, a keyboard and a CRT display device, etc., and can input ordinary alphanumeric characters, hiragana and katakana characters as well as phonetic symbols displayed on the CRT display device using, for example, a mouse. You can specify and enter. The search keyword input unit 4 is composed of, for example, a keyboard and inputs a search keyword composed of katakana words. Further, the search result output unit 6 is composed of, for example, a CRT display device or the like, and displays and outputs the searched foreign language word and information of the foreign language word.

In the actual system, the registration data input unit 1, the search keyword input unit 4, and the search result output unit 6 are composed of one man-machine device including a keyboard and a CRT display device.

Next, the operation of each unit will be described in order. When registering a new foreign language document in the database 8 of this information retrieval system, the registrant extracts a foreign language word that is a keyword that appropriately represents the contents of this document, and uses this foreign language word by using, for example, a dictionary. Examine the phonetic symbols of. Then, the registration data including the foreign word, the phonetic symbol, and the information for specifying the document is input to the registration data input unit 1.

Registration data consisting of foreign language words, phonetic symbols, and information input to the registration data input unit 1 is temporarily stored in the input data buffer 2. Then, the phonetic symbols in the one registration data stored in the input data buffer 2 are transmitted to the phonetic symbol katakana conversion unit 3.

The phonetic symbol katakana conversion unit 3 converts a series of phonetic symbols indicating the pronunciation of one input foreign language word into each table 7a, 7b of the phonetic symbol / katakana correspondence table 7.
Is converted into one katakana word by using the method described above. Then, the foreign word stored in the input data buffer 2, the katakana word converted this time, and the information stored in the input data buffer 2 are set in the empty area of the database 8.

This completes the registration process for registering one foreign language word and information in the database 8. For example, as shown in FIG. 4, when the foreign language word is [library] and the phonetic symbol is [laiberri], the converted katakana word is [liverelli]. The foreign language word is [Einstein] and the phonetic symbol is [ainstein].
n], the converted katakana word becomes [Einstein].

Next, the operation of each part when searching a foreign language word and information using this information search system will be described. The searcher inputs the katakana word of the foreign language word to be searched as a search keyword into the search keyword input unit 4. The search keyword input unit 4 transfers the input search keyword consisting of one katakana word to the search processing unit 5.

The search processing unit 5 searches the database 8 with the search keyword input according to the flow chart shown in FIG. The flow chart starts, and in S (step) 1, the area number i of the database 8 is initialized to 1. next,
In S2, the katakana word in the i-th area is read. Then, the katakana word of the search keyword and the character number j indicating the character position from the first character of the katakana word read this time are initialized to 1.

Then, the character similarity Rj indicating the similarity between the characters of the j-th katakana character of the search keyword and the j-th katakana character of the read katakana word is calculated. Specifically, the character similarity Rj is determined according to the following rules.

(A) When the two comparison characters are exactly the same character and the character is a consonant ... 100/100 (b) When the two comparison characters are exactly the same character and the character is a vowel ... 40/40 (c) The two comparison characters are Different, consonant strings are the same and vowel stages are different ... 60/100 (d) Two comparison characters are different, consonant strings are different and vowel stages are the same ... 40/100 (e) Two comparison characters are different and consonants When the columns and vowel stages are also different ... 0/100 However, the consonant sequences and vowel stages correspond to columns and rows (columns) when the above-mentioned Katakana characters are arranged in the Japanese syllabary matrix.

In this way, the consonant string is weighted by 60%, and the vowel stage is weighted by 40%.
That is, the character similarity Rj is set to be higher when the consonant strings match each other than when the vowel steps match each other.

Note that lowercase letters such as [a], [i], [a], etc. are weighted by 20%. Also, no consonants [A]
Vowels such as [a] ... are given a weight of 40%. further,
The difference between [-] and [-] is considered the same.

For example, assuming that the fifth characters [Lee] and [Li] of [Live Reli] and [Library] shown in FIG. 4 are the same, the character similarity Rj (= R ₅ ) is 100/100. . When the character similarity Rj for the jth character is obtained in S4, the process proceeds to S5, and accumulation is performed to obtain the word similarity Ri of all katakana words. Specifically, each numerator R _1j of the character similarity Rj of each character from the first character to the j-th character
To the calculated ratio between the cumulative value .SIGMA.R _0j of the cumulative value .SIGMA.R _1j and the denominator R _0j.

Ri = ΣR _1j / ΣR _0j Next, when there is still a character to be compared, 1 is added to the comparison character number j (S7), the process returns to S4, and the character similarity of the next character is obtained. The degree Rj is calculated.

When the characters to be compared are completed in S7, S
Proceed to 8. If the number of characters in the search keyword does not match the number of characters in the read katakana word, the number j is updated until the number of characters in the longer katakana word ends, and the character similarity Rj Is 0/100
And

For example, as shown in FIG. 4, when the katakana word stored in the database 8 is [livery] and the search keyword is [library], the word similarity Ri is Ri = ΣR _1j / ΣR _0j = 400 / 460 = 87%.

When the katakana word stored in the database 8 is [Einstein] and the search keyword is [Einstein], the word similarity Ri is Ri = ΣR _1j.
/ ΣR _0j = _360/400 = 90%.

In step S8, if the cumulative word similarity Ri of all katakana words at the time when the characters to be compared are finished is less than a predetermined value, the search keyword is the i-th area read this time. When it is determined that the word does not correspond to the registered foreign language word, 1 is added to the area number i of the area of the database 8 in S9. If the area number i after addition has not reached the maximum area number (S1
0), the process returns to S2 to read the katakana word in the area of the area number i after addition.

Then, in S8, when the word similarity Ri is equal to or greater than the specified value, it is determined that the search keyword corresponds to the foreign language word registered in the i-th area read this time,
Proceed to S11. In step S11, the foreign word and information in the area of the area number i of the relevant katakana word are read from the database 8 and displayed as a search result on the CRT display device of the search result output unit 6 for output.

When the reading of the katakana words in all areas in the database 8 is completed in S10, it is determined that the foreign language word corresponding to the search keyword is not registered, and the fact is output by the search result output unit. Display output to the CRT display device of No. 6.

In the information retrieval system configured as described above, when registering a foreign word and information relating to this foreign word in the database 8, the registrant automatically inputs the phonetic symbol together with the foreign word into the database 8. Foreign words and corresponding katakana words are registered.

When searching for a foreign language word, if the searcher inputs the katakana word written in katakana as a search keyword into the search keyword input section 4, the searcher inputs the foreign word corresponding to the relevant search keyword and this foreign language. Information specifying the relevant document, such as the document name of the document in which the word is registered, the publisher, the author, the date of publication, and the holding place, is output as the search result.

In this case, as described above, the katakana word input as the search keyword does not always completely match the katakana word registered in the database 8. However, in the embodiment system, each word similarity Ri for each katakana word stored in the database 8 of the katakana word of the search keyword is calculated. Then, a foreign word corresponding to a katakana word having a word similarity Ri of a specified value or more is set as a search result for the search keyword.

Therefore, even if the Katakana word of the search keyword and the Katakana word indicating the same foreign language word in the database 8 do not completely match, it is highly likely that the foreign word and the corresponding foreign word are correct. You can search for information about.

Further, in the embodiment system, when calculating the word similarity Ri, the character similarity Rj between the respective Takatana characters forming the Katakana word is calculated, and these katakana characters are accumulated to form one katakana. The word similarity Ri for a word is calculated. Then, in the process of calculating each character similarity Rj, as described above, there is a difference in the degree of influence on the word similarity Ri between the case where the consonant strings match and the case where the vowel stages match. It is provided.

Specifically, since the similarity is set higher when the consonant strings match than when the vowel stages match, a more accurate word similarity Ri can be obtained. , Further improving the accuracy of searching foreign words.

FIG. 5 is a block diagram showing a schematic structure of an information search system incorporating a foreign language search device according to another embodiment of the present invention. The same parts as those of the embodiment system shown in FIG. 1 are designated by the same reference numerals. Therefore, detailed description of the overlapping portions is omitted.

The registration data input unit 1 in the system of this embodiment
The character a is composed of a character input unit 9a such as a keyboard and a microphone 9b for inputting voice. And
The foreign language word input from the character input unit 9a and information about this foreign language word are once written in the input data buffer 2a. On the other hand, the voice of the corresponding foreign language word input from the microphone 9b is converted into, for example, a digital voice signal and written in the voice buffer 2b.

Further, in the system of this embodiment, a voice / phonetic symbol correspondence table 11 is provided. In the voice / phonetic symbol correspondence table 11, for example, for each unit voice such as [A] and [Sha], a single phonetic symbol or a combination phonetic symbol corresponding to the corresponding unit voice is stored.

The phonetic pronunciation symbol conversion unit 10 divides the voice of one foreign language word stored in the voice buffer 2b into a plurality of unit voices, and the voice / phonetic symbol correspondence table 11 is obtained.
To search for a single phonetic symbol or a combination phonetic symbol for the corresponding unit voice. Then, it has a function of editing the voice of the input foreign language word into a series of phonetic symbols and sending them to the next phonetic symbol katakana conversion unit 3.

The signal waveform obtained when voice is converted into an electric signal is the same as when the same foreign language word is pronounced.
Although there are some individual differences, the phonetic pronunciation symbol conversion unit 10 absorbs the individual differences by using a pattern recognition method and specifies the corresponding pronunciation symbols.

Further, in order to improve the conversion accuracy, the person to be registered is limited to a specific person, and the relationship between the voice signal of the voice of this specific person and the phonetic symbol is registered in the voice / phonetic symbol correspondence table 11. You may.

In the information retrieval system configured as described above, when newly registering a foreign word and information in the database 8, the character input unit 9 of the registration data input unit 1a is used.
While inputting foreign words and information to a,
The voice of a foreign language word is input to the microphone 9b.

Then, the voice of the corresponding foreign language word input from the microphone 9b is input to the voice pronunciation symbol conversion unit 10 via the voice buffer 2b. Then, the voice of this foreign language word is converted into a series of phonetic symbols using the phonetic / phonetic symbol conversion table 11. The converted phonetic symbols are transmitted to the phonetic symbol katakana conversion unit 3. Then, the phonetic symbols of the corresponding foreign language word are converted into katakana words using the phonetic symbol / katakana correspondence table 7.

Then, the foreign language word stored in the input data buffer 2a, the information, and the Katakana word corresponding to the converted foreign language word are written in the empty area of the database 8. Therefore, in each area of the database 8, foreign language words, katakana words, and information having the same contents as the database 8 of the embodiment system shown in FIG. 1 are stored.

The search processing method for the database 8 using the input search keyword in the search processing section 5 is the same as the search processing method of the search processing section 5 in the embodiment system shown in FIG.

That is, when a searcher inputs a katakana word of a foreign language word to be searched into the search keyword input section 4, the search processing section 5 searches the Katakana word in each area of the database 8 with the search keyword to obtain the word. Foreign words and information of katakana words having a similarity Ri of a specified value or more are displayed and output as search results.

In the information retrieval system configured as described above, when registering a foreign language word and information in the database 8, the registrant inputs a voice of the foreign language word in addition to the relevant foreign language word and information. Since only that is required, the work efficiency of registration is significantly improved as compared with the case where the phonetic symbols shown in FIG.

In each of the above-described embodiments, the database 8 stores and holds the katakana word corresponding to the foreign language word and the information, but the database 8 corresponds to the foreign language word without storing the information. It is also possible to store only katakana words. In this case, this information search system becomes a foreign language search device, and can be used as a dictionary for checking the correct spelling of a foreign language word from katakana words, for example.

As an application example of the present invention, a search keyword is input by voice, the input voice is converted into a katakana word, and the converted katakana word is used as a new search keyword to search the database. Is also possible.

[0070]

As described above, in the foreign language search device and the information search system incorporating the foreign language search device of the present invention, the foreign language word and the phonetic symbol or voice are input when the foreign language word is registered. As a result, the foreign language word and the corresponding katakana word are automatically registered in the database.

Further, when searching the database with the search keyword, the word similarity between the search keyword and the katakana word stored in the database is calculated,
A foreign word corresponding to a katakana word whose calculated word similarity is equal to or higher than a specified value is used as a search result.

Therefore, even if a katakana word that is slightly different from the katakana word registered in the database is searched, the correct foreign language word that is the target can be searched. Also,
When calculating the word similarity between the search keyword and the katakana word, it is obtained from the character similarity weighted for each katakana character according to the character type. Therefore, the search accuracy can be further improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an information search system incorporating a foreign language search device according to an embodiment of the present invention.

FIG. 2 is a diagram showing stored contents of a phonetic symbol / katakana correspondence table of the system of the embodiment.

FIG. 3 is a flowchart showing a search operation in a search processing unit of the system of the embodiment.

FIG. 4 is a diagram showing an example of a calculation procedure of a search keyword and a word similarity of katakana words stored in a database in the system of the embodiment.

FIG. 5 is a block diagram showing a schematic configuration of an information search system incorporating a foreign language search device according to another embodiment of the present invention.

[Explanation of symbols]

1, 1a ... Registered data input unit, 2, 2a ... Input data buffer, 3 ... Phonetic symbol Katakana conversion unit, 4 ... Search keyword input unit, 5 ... Search processing unit, 6 ... Search result output unit, 7
... phonetic symbol / katakana correspondence table, 7a ... consonant correspondence table, 7b ... vowel correspondence table, 8 ... database,
9a ... Character input section, 9b ... Microphone, 10 ... Voice phonetic symbol conversion unit, 11 ... Voice / phonetic symbol correspondence table

Claims (4)

[Claims] 1. A registration data input section for inputting registration data composed of a foreign language word written in a foreign language and a phonetic symbol of the foreign language word, and a phonetic character corresponding to each phonetic symbol and the corresponding phonetic symbol is stored. A phonetic symbol / katakana conversion table, and a phonetic symbol katakana conversion means for converting the phonetic symbols of the registration data input through the registration data input unit into katakana words using the phonetic symbol / katakana correspondence table; A database for storing and holding foreign words of each registration data input via the registration data input unit and katakana words converted to the katakana; and a search keyword input unit for inputting a search keyword composed of katakana words, The katakana word of the search keyword input from this search keyword input section is stored in the database. Word similarity calculation means for calculating a word similarity for each katakana word, and a search result determination for outputting a foreign language word corresponding to the katakana word having the calculated word similarity of a specified value or more as a search result for the search keyword A foreign language search device with an output unit. 2. A registration data input unit for inputting registration data consisting of a foreign language word written in a foreign language and a voice of the foreign language word, and a voice in which each voice and each phonetic symbol for the corresponding voice are stored. A phonetic-phonetic symbol correspondence table; and a phonetic-phonetic symbol conversion means for converting the voice of the registration data input via the registered-data input unit into a phonetic symbol by using the voice-phonetic-phonograph corresponding table. A phonetic symbol / katakana correspondence table in which phonetic symbols and katakana characters corresponding to the corresponding phonetic symbols are stored, and phonetic symbols obtained by the phonetic phonetic symbol conversion means are converted into katakana words using the phonetic symbol / katakana table. Phonetic symbol katakana conversion means for converting, foreign language words of each registration data input via the registration data input section, and katakana converted to katakana A database for storing and holding words, a search keyword input section for inputting a search keyword consisting of katakana words, and a word for each katakana word stored in the database for the katakana words of the search keyword input from this search keyword input section A word similarity calculation unit for calculating a similarity; and a search result determination output unit for outputting, as a search result for the search keyword, a foreign word corresponding to a Katakana word having the calculated word similarity of a specified value or more. Foreign language search device. 3. The word similarity calculation means indicates the similarity of each character between each katakana character forming the katakana word of the search keyword and each katakana character forming the katakana word stored in the database. The character similarity is calculated by weighting it according to the character type, and the word similarity is calculated from the calculated character similarity for each character using a predetermined formula. 1 or 2
Foreign language search device described. 4. A registration data input section for inputting registration data comprising a foreign word written in a foreign language, a phonetic symbol of the foreign word and information on the foreign word, each phonetic symbol and a corresponding phonetic symbol. A phonetic symbol / katakana correspondence table in which katakana characters corresponding to are stored, and a phonetic symbol of registration data input via the registration data input unit is converted into a katakana word using the phonetic symbol / katakana correspondence table. Phonetic symbol katakana conversion means, a database that stores and stores foreign language words and information of each registration data input through the registration data input unit, and katakana words converted into the katakana, and a search keyword composed of katakana words. The search keyword input section to be input and the katakana of the search keyword input from this search keyword input section A word similarity calculating means for calculating a word similarity for each katakana word stored in the database of words; and a foreign language word and information corresponding to the katakana word for which the calculated word similarity is a specified value or more An information search system including a search result determination output unit that outputs a search result for a keyword.