JP2005222244A - Word retrieval device, word retrieval method, and information providing system equipped with word retrieval device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve a problem that if an error exists in a keyword to be used for information search, a search result cannot be obtained in an information providing device using a database. <P>SOLUTION: A retrieval server 20 receives a search string from a user terminal device 16. A search section 22 searches an address information storage section 24 by using a search string as a key. When a search result is not obtained, a similar word request section 26 transmits a search string to a word retrieval device 100. The word retrieval device 100 extracts a similar word out of words contained in a dictionary based on the attribute of the search string and provides it to the similar word request section 26. A similar word providing section 28 transmits the similar word to the user terminal device 16. A user chooses the similar word to be used for search from the similar words. The search section 22 performs search again based on chosen similar word. Thereby, even when an error exists in the search string, search processing can be performed based on the similar word. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a technique for searching for a predetermined word, and more particularly to an apparatus for searching for a word similar to an inputted character string, a method thereof, and an information providing system including the word search apparatus.

With the progress of computer networking, various services are provided, for example, at search sites, shopping sites, dictionary sites, etc., using networks such as the Internet. Such service sites generally manage information using a database, and provide information associated with the search keyword by referring to the database using the search keyword received from the user as a key. However, if you enter "cold medicine" as "cold medicine" or "Internet" as "Internet" due to a user's typo, etc., those words are not registered in the database. Since there is no corresponding information, it is answered that there is no corresponding information. In order to cope with such a situation, there is a technique for searching for similar words in an input character string (for example, see Patent Document 1).
JP-A-10-105578

  The character string typed by the user is often different for a predetermined word as well as a typo. For example, katakana is generally very swaying, and “Web” may be typed as “Web” or “Web”. In this way, the database could not be flexibly searched despite the various notation methods for words including input errors.

  The present invention has been made in view of these points, and an object thereof is to provide a technique for searching for a word similar to a character string. Another object is to provide highly flexible information provision technology.

  One embodiment of the present invention is an apparatus for searching for a word. This device includes a receiving unit that accepts input of a character string, a storage unit that holds information about a plurality of words, and a similar word that is similar to a character string from words included in the storage unit based on the attributes of the character string And an analysis unit for extracting.

  According to this apparatus, words similar to the character string can be extracted based on the attribute of the character string. The “character string attribute” is, for example, the character type of the character included in the character string, the number of characters included in the character string, and the like, and may be information indicating the configuration of the character string.

  The analysis unit may include an extraction unit that extracts similar word candidates based on a character type of a character included in the character string. As a result, similar word candidates can be narrowed down based on the character type of characters included in the character string, so that similar words can be efficiently extracted.

  The extraction unit may extract similar words based on the number of characters included in the character string. Thus, similar word candidates can be narrowed down based on the number of characters included in the character string, so that similar words can be extracted efficiently.

  The storage unit holds a word table for each character type of characters included in the word, and the extraction unit specifies the character type of each character included in the character string, and refers to the word table corresponding to the specified character type to search for similar words May be extracted. Since a word table is prepared for each character type, the search range can be narrowed, and the amount of calculation for specifying similar words can be reduced.

  The storage unit holds a word table for each number of characters included in the word, and the extraction unit specifies the number of characters in the character string, and extracts similar word candidates by referring to the word table corresponding to the specified number of characters. Good. Since a word table is prepared for each number of characters, the search range can be narrowed, and the amount of calculation for specifying similar words can be reduced.

  The extraction unit may extract similar word candidates based on the degree of coincidence between the character block included in the character string and the character block included in the word. The “coincidence” may be simply a comparison between the number of characters included in a character string and the number of characters included in a word, or a comparison of the number of identical character blocks. . The “character block” is a character string created based on a predetermined condition, and may be, for example, two adjacent characters or two skipped characters.

  Another aspect of the present invention is a method for searching for a word. This method includes a step of receiving input of a character string from a predetermined search device, and a similar word similar to a character string from words included in a storage unit that holds information on a plurality of words based on the attribute of the character string And a step of outputting the extracted similar words to a search device.

  Yet another aspect of the present invention is a system for providing information. The system includes a receiving unit that receives input of a character string, a first storage unit that holds a predetermined word and predetermined information related to the word in association with each other, and a character that is received with reference to the first storage unit A search unit that searches for predetermined information corresponding to a column, and a second storage that holds information related to a plurality of words based on attributes of the character string when the predetermined information corresponding to the character string cannot be searched An extraction unit that extracts a similar word similar to the character string from the words included in the unit, and the search unit searches for predetermined information corresponding to the similar word. As a result, even if the search result cannot be obtained with the initially accepted character string, the search can be performed again based on the similar word automatically selected based on the character string.

  It should be noted that any combination of the above-described constituent elements and a conversion of the expression of the present invention between a method, an apparatus, a system, a recording medium, a computer program, etc. are also effective as an aspect of the present invention.

  ADVANTAGE OF THE INVENTION According to this invention, the apparatus which searches efficiently the word similar to a character string can be provided. In addition, a highly flexible information providing system can be provided.

  FIG. 1 is a configuration diagram of an information providing system 50 according to the embodiment. The user terminal device 16 is a terminal device such as a personal computer, a PDA (personal digital assistance), or a mobile phone, and is a computer that can communicate with the search server 20 via the network 10. The search server 20 is a device for realizing a search site, and includes a search unit 22, an address information storage unit 24, a similar word requesting unit 26, a similar word providing unit 28, and a communication unit 30. The search unit 22, the similar word request unit 26, and the similar word provision unit 28 communicate with each device connected to the network 10 via the communication unit 30. The search unit 22 searches the address information storage unit 24 for the address of the web page corresponding to the search character string received from the user terminal device 16, and provides the search result to the user terminal device 16. When there is no web page address corresponding to the search character string, the similar word requesting unit 26 requests the word search device 100 to search for a word similar to the search character string. The word search device 100 finds a word similar to the search character string and transmits it to the search server 20. And the similar word provision part 28 provides the similar word to the user terminal device 16, and makes a user select a desired word from the similar words. And the search part 22 searches again based on the similar word selected by the user. Thus, even when there is no information corresponding to the search character string in the database, an opportunity to search again based on a word similar to the search character string can be provided, so that a flexible information providing system can be provided.

  FIG. 2 is a diagram illustrating an example of the display screen 52 displayed on the user terminal device 16 when a similar word is found. When the similar word 54 is found, the search server 20 in FIG. 1 provides the user terminal device 16 with information for displaying the display screen 52. This display screen 52 instructs a similar word 54 similar to the search character string input by the user, a check box 56 for selecting a similar word used for the search, and a search again with the selected similar word. And a search start button 58. By displaying the similar word 54 in this way, even if the word stored in the database is not input due to, for example, the user's typing error or unclear storage, the search operation is not terminated and the user is not terminated. Searching can be continued by providing an opportunity to select similar words.

  FIG. 3 is an internal block diagram of the word search device 100 of FIG. In terms of hardware components, each component of the word search device 100 is a CPU, a memory of any computer, a program that implements the components shown in the figure loaded in the memory, and a storage unit such as a hard disk that stores the program. However, it will be understood by those skilled in the art that there are various modifications in the implementation method and apparatus. Each figure used for description of the embodiment shows a functional unit block, not a hardware unit configuration.

  The accepting unit 102 accepts a search character string to be searched for similar words from the search server 20. The authentication unit 108 authenticates the search server 20 that has transmitted the search character string. For example, the authentication unit 108 holds an IP address of a device that permits use of the word search device 100 in advance, and permits the search server 20 to which the IP address is assigned to use the word search device 100. The authentication method in the authentication unit 108 is not limited to this, and a known technique may be used. When the authentication is successful, the reception unit 102 supplies the search character string to the analysis unit 104. Although details will be described later, the analysis unit 104 searches for a word similar to the search character string and supplies it to the output unit 106. The output unit 106 transmits similar words to the search server 20. The output unit 106 may transmit similar words in an arbitrary format, for example, may be transmitted to the search server 20 in CSV format, or may be transmitted to the search server 20 in a format according to XML (eXtensible Markup Language). Also good. Further, when transmitting a plurality of similar words to the search server 20, the output unit 106 displays screen information for presenting similar words with priorities so as to specify the degree of similarity to the search character string, as shown in FIG. The similar words are transmitted to the search server 20 so that the search server 20 can generate them. For example, the output unit 106 associates similar words with information for specifying the degree of similarity of each similar word with respect to the search character string and outputs the information to the search server 20. In the case of the CSV format, the output unit 106 may arrange the similar words in descending order of similarity and transmit them to the search server 20.

  The log processing unit 110 stores the usage status of the word search device 100 in the log storage unit 112 in association with information specifying the search server 20 such as an IP address. For example, the log processing unit 110 receives the IP address and search character string of the search server 20 that has been successfully authenticated from the authentication unit 108 and stores the information in the log storage unit 112 in association with the reception date and time. Further, the log processing unit 110 may associate the similar word and the IP address of the transmission destination of the similar word from the output unit 106 and store them in the log storage unit 112 in association with the transmission date / time. The log processing unit 110 may store information necessary for the charging process by the charging unit 114 described later in the log storage unit 112. The accounting unit 114 performs accounting based on data held in the log storage unit 112. The billing form can be arbitrarily determined based on a business contract, and may be calculated based on, for example, the number of times of use and the period of use.

  FIG. 4 is an internal configuration diagram of the analysis unit 104 of FIG. The pre-processing unit 120 converts, for example, an old character form included in the search character string into a new character form, or unifies katakana and numbers into full-width or half-width characters to convert them into character strings suitable for subsequent analysis processing. The conversion data storage unit 160 holds the old font and the new font in association with each other. The pre-processing unit 120 determines whether or not there is an old font to be replaced by matching each character included in the search character string with the old font held in the conversion data storage unit 160. When the old font to be replaced is included, the preprocessing unit 120 replaces the old font in the search character string with the corresponding new font. After preprocessing the search character string, the preprocessing unit 120 supplies the search character string to the specific word search unit 122.

  The specific word search unit 122 searches the specific search character string for a similar word by matching the character string stored in the specific word list storage unit 130 with the search character string. By statistically analyzing the search character string and the word corresponding to the search character string, words that are frequently mistaken can be selected in advance. The specific word list storage unit 130 retains a word selected in advance and a character string that is easily mistaken for the word. As a result, similar words can be quickly identified for words that are likely to be mistaken. When the character string corresponding to the search character string is not registered in the specific word list storage unit 130, the specific word search unit 122 supplies the search character string to the extraction unit 124. Similar words in the search character string are selected by the processing in the extraction unit 124, the calculation unit 126, and the selection unit 128. Next, elemental techniques of processing realized by the extraction unit 124 and the calculation unit 126 will be described.

(Edit distance)
In general, similar words similar to a search character string can be extracted by using a method called edit distance. The edit distance method is also called edit distance, and the degree of similarity between character strings is calculated by calculating how many insertion, replacement, and deletion operations are required to convert an input character string into a predetermined character string. Is calculated. For example, when “Internet” is changed to “Internet”, it is necessary to insert “t” between “ne” and “t”. On the other hand, when “Internet” is changed to “Interchange”, “N” and “G” need to be replaced with “C” and “D” respectively, and “N” and “G” must be added. In the former and the latter, since the number of edits is less in the former, “Internet” is similar to “Internet”.

(Weighting according to editing content)
In the edit distance method, the similarity count value is calculated by changing the weighting value according to the character to be deleted, inserted, or replaced. In the present embodiment, the similarity count value is calculated by adding the weight value associated with the edited content every time editing is performed. For this reason, the smaller the similarity count value, the higher the similarity, that is, the similarity. In general, Katakana is greatly fluctuated, and for example, even a single word such as “user” or “user” has a different notation. When replacing, deleting, or inserting characters due to such shaking, a lighter value is set as a weighting value compared to replacing, deleting, or inserting other characters. For example, when “-” is deleted, a lighter value than that of “c” is set as a weighting value. Thereby, since a similarity count value can be appropriately calculated according to the character type, a similar word can be appropriately selected.

  By performing such work on all the words, similar words for the search character string can be specified, but it is not realistic to calculate a similarity count value for all the words, for example, several hundred thousand words. Therefore, before calculating the similarity count value by the edit distance method, candidates for similar words for which the similarity count value is to be calculated are narrowed down based on the character type and the number of characters. Then, the similarity count value is calculated for the narrowed-down words, and the similar words are selected in descending order of words having a high similarity, that is, words having a low similarity count value. Hereinafter, the process of extracting a word for which the similarity count value is to be calculated is referred to as “candidate selection process”, and the word selected by the candidate selection process is referred to as “similar word candidate”. Also, the process of calculating the similarity count value for similar word candidates is referred to as “similarity calculation process”.

(Bigram)
The candidate selection process is performed by checking the degree of matching between the character string included in the search character string and the character included in the word, and a different algorithm is used according to the character length which is one of the attributes of the search character string. . Specifically, an index in which a character and a word including the character are associated is created for all characters included in the dictionary. For example, an index is created by associating words such as “Mountain”, “Yamamoto”, “Arashiyama”, and “Mountain Range” with the word “mountain”. If the search character string is “Fujiyama”, check the index of “Wealth”, “Character”, “Mountain”, and extract words with a high degree of matching, that is, words that contain more characters in the search character string To do. Thereby, word candidates similar to the search character string can be selected. However, in the method of creating an index for each character, the number of matching words increases and the amount of calculation becomes enormous. Therefore, in order to perform this method more efficiently, an index is created with two adjacent characters, that is, a bigram. When an index is created with a bigram, for example, “Mt. Fuji” is recorded in two indexes “Fuji” and “Shiyama”.

(Stepping stone bigram)
If you search for “Fujiyama” using the bigram index, you cannot find similar word candidates. The indexes that can be created from “Fujiyama” are “Fuji” and “Jijiyama”, and they do not overlap at all with the index that is created from “Mt. Therefore, for example, in the case of a word composed of 3 or 4 characters, two skipped characters are used as an index instead of two adjacent characters (hereinafter simply referred to as “stepping stone bigram”). For example, in the case of “Mt. Fuji”, the index created by the stepping stone bigram is “Toyama”. In the case of “two birds with one stone”, the indexes created with the stepping stone bigram are “12” and “stone birds”. In the case of five characters or more, similar word candidates are selected using an index created with bigrams instead of stepping stone bigrams. In this embodiment, different algorithms are used according to the character length, such as using a stepping stone bigram when the character length is small and using a bigram when the character length is large.

  In this way, a character block is created with characters taken out according to a predetermined rule such as bigram or stepping stone bigram, and an index is created with the character block. For example, “Toyama” when the index of “Mt. Fuji” is created with a stepping stone bigram is a character block. By creating an index with character blocks, the degree of coincidence can be calculated in units of blocks. As a result, the processing speed can be increased as compared with the case where the matching degree is calculated for each character. “Calculating the degree of coincidence” means calculating how many character blocks the predetermined word includes the same character block as the character block included in the search character string. In calculating the degree of coincidence, it is not particularly necessary to consider whether or not the position of the character block in the word matches the position of the character block in the search character string.

  In general, it is rare that all characters are mistaken due to a typo or the like, and some of the characters included in the word are often wrong. From this point of view, similar word candidates can be efficiently extracted by calculating the degree of matching in character block units. Further, the number of characters included in the character block is not limited to two and may be arbitrarily set. For example, the number of characters may be determined according to the number of characters in the word. Also, a predetermined rule for creating a character block may be arbitrarily set. For example, a word having a large number of characters may be taken out every two characters to form a three-character block.

(Select algorithm according to character type and number of characters)
The candidate selection process is further performed based on the character type that is one of the attributes of the search character string. For example, Japanese is represented by a plurality of character types such as katakana, hiragana, kanji, numbers, and alphabets. From this point of view, similar words can be narrowed down according to the character type of characters included in the search character string. For example, when a 3-character word is used as a search character string, an 8-character character string is rarely input incorrectly. From this point of view, similar words can be narrowed down according to the number of characters in the search character string. For example, by preparing a dictionary for each attribute such as the character type of the word and the number of characters, the dictionary can be selected based on the attribute of the search character string. Then, by creating the above-mentioned index for each character block for each dictionary, the process of selecting similar word candidates can be performed quickly and efficiently.

  The extraction unit 124 performs candidate selection processing using the element technology described above, extracts similar word candidates, and outputs them to the calculation unit 126. The calculation unit 126 calculates a similarity count value with the search character string for each similar word candidate using the above-described edit distance method. Then, the calculation unit 126 associates the similar word candidate with the similarity count value and outputs the similar word candidate to the selection unit 128. The selection unit 128 selects words in descending order of the similarity count value and outputs them to the output unit 106 as similar words.

  The selection unit 128 supplies the selected similar word list to the update unit 132. The update unit 132 stores the similar word received from the selection unit 128 in the output word list storage unit 134. The updating unit 132 may calculate the frequency for each word based on the information held in the output word list storage unit 134 and add words with a high frequency to the specific word list storage unit 130. Thereby, the processing results in the extraction unit 124, the calculation unit 126, and the selection unit 128 can be reflected in the specific word list storage unit 130.

  The data storage unit 150 stores data necessary for processing in the extraction unit 124 and the calculation unit 126. The character type data storage unit 152 holds a character type dictionary and an index created based on each dictionary. The character length-specific data storage unit 154 holds a dictionary for each character length and an index created based on each dictionary. The dictionaries and indexes held in the character type data storage unit 152 and the character length data storage unit 154 have the same data structure. For example, the “dictionary” holds a word and a word ID that identifies the word in association with each other. The “index” is created for each dictionary and holds a character block determined based on a predetermined rule and a word ID of a word including the character block in association with each other.

  For example, the character type data storage unit 152 holds a katakana dictionary, a reading dictionary, an alphabet dictionary, and an index created based on each dictionary. The reading dictionary holds commonly used words in hiragana and includes words contained in the katakana dictionary. For example, the data storage unit 154 includes a dictionary that collects one-letter words, a dictionary that collects two-letter words, a dictionary that collects three-letter words, a dictionary that collects four-letter words, and a dictionary that collects five or more characters. It holds a dictionary that collects words and an index created based on each dictionary.

  The knowledge data storage unit 156 holds the edited content used when calculating the similarity and the weighting value corresponding to the edited content in association with each other. The reading data storage unit 158 holds a kanji, a sound reading of the kanji, and a kanji reading in association with each other.

(Matching Katakana character string)
When all the characters included in the search character string are Katakana characters, the extraction unit 124 uses a Katakana dictionary that collects words in which all characters are represented by Katakana characters, and a Katakana index created based on the Katakana dictionary. Candidate selection processing is performed. The katakana dictionary and the katakana index are stored in the character type data storage unit 152.

  In the first step, the extraction unit 124 determines whether the same word as the search character string is included in the Katakana dictionary. In the second step, the extraction unit 124 selects similar word candidates based on the matching degree of character blocks using the katakana index. The number of similar word candidates may be set arbitrarily. In the second step, the extraction unit 124 selects as similar word candidates in order from words with many matching character blocks. In the third step, the extraction unit 124 outputs the selected similar word candidate to the calculation unit 126.

  In the fourth step, the calculation unit 126 calculates a similarity count value with the search character string for each similar word candidate by the edit distance method. Based on the data held in the knowledge data storage unit 156, the calculation unit 126 calculates a similarity count value using a different weighting value for each editing content. In the fifth step, the calculation unit 126 outputs the calculated similarity count value and the similar word candidate in association with each other to the selection unit 128. In the sixth step, the selection unit 128 selects the similar words in order from the word with the lowest similarity count value.

  If the similar word candidate cannot be selected in the second step, the extraction unit 124 converts the katakana to hiragana and performs “reading matching”. When performing reading matching, the extraction unit 124 uses a hiragana dictionary and a hiragana index held in the character type data storage unit 152. This hiragana dictionary holds readings of all words. That is, when a similar word candidate cannot be selected in the Katakana dictionary, the extraction unit 124 uses a Hiragana dictionary and a Hiragana index that include more words. For example, if the personal name “Saito” is entered as “Kaito” in Katakana, the word “Saito” is not recorded in the Katakana dictionary, and therefore similar words cannot be searched. In order to deal with such a situation, when a similar word candidate cannot be selected in the Katakana dictionary, the Hiragana dictionary is used.

  Similarly to the case of Katakana when the characters included in the search character string are all hiragana, or when the characters included in the search character string are all alphabetic characters, the extraction unit 124 prepares a dictionary and index prepared for each character type. The same candidate selection process is performed using.

(Processing according to character length)
When a plurality of character types are included in the search character string, the extraction unit 124 performs candidate selection processing based on the number of characters included in the search character string. For example, if the character length of the search character string is one character, the extraction unit 124 performs candidate selection processing using a dictionary that collects one-character words held in the character-length data storage unit 154 and its index. . In consideration of the possibility that the length of the search character string is short, the extraction unit 124 performs candidate selection processing using a dictionary in which two-character words are collected and its index. That is, when performing the candidate selection process according to the character length, the extraction unit 124 performs the candidate selection process using a dictionary that collects the number of characters included in the search character string and the number of characters before and after the character string.

(Reading matching)
When the search character string includes kanji, the extraction unit 124 refers to the reading data storage unit 158 and generates all the kanji readings included in the search character string. When a plurality of kanji characters are included, the extraction unit 124 creates a combination of reading candidates for all kanji characters, performs candidate selection processing for all of these reading candidates, and selects similar word candidates. . When a combination of reading candidates is created, based on information on kanji readings and kanji readings, the readings of kanji readings or readings of kanji readings are often correct readings than readings combining readings of kanji and kanji readings. Therefore, the weighting value for the reading that combines the sound reading and the kinematic reading is set to a value larger than the weighting value for the reading that combines the sound readings or the kanji readings.

  For example, when the search character string is “Genji”, “Genji” and “Current” exist as the same reading. In this case, we want to make the similar count value of “Genji” larger than “current”. For this purpose, the calculation unit 126 compares the character included in the search character string with the character included in the word registered in the dictionary, and increases the weighting value for the word having a large number of different characters.

  FIG. 5 is a flowchart of processing in the extraction unit 124 of FIG. The extraction unit 124 identifies the attributes of the search character string, that is, the character type and the number of characters (S8). For example, the extraction unit 124 holds a character code range (hereinafter referred to as “character type range”) prepared based on the character code table, and the extraction unit 124 stores each character included in the search character string. The character type of each character may be specified by determining the character type range including the character code. When all the characters included in the search character string are Katakana, “−” and “−” variants (S10), “Matching Katakana character string” is performed (S12). When a similar word candidate is found (Y in S14), the extraction unit 124 outputs the similar word candidate to the calculation unit 126 (S44). If no similar word candidate is found in step 14 (N in S14), the extraction unit 124 performs “reading matching” (S16).

  If all the characters are not katakana in step 10 (N in S10), the extraction unit 124 determines whether all the characters are hiragana, “-” and “-” variants (S18). When all the characters are hiragana characters (Y in S18), “matching of reading” is performed. If all the characters are not hiragana in step 18 (N in S18), it is determined whether or not all the characters are alphabets and numbers (S22). If all characters are alphabets and numbers (Y in S22), “alphabet matching” is performed (S24), and the process proceeds to step 44.

  If all the characters are not alphabets or numbers in step 22 (N in S22), it is determined whether or not the character length of the search character string is one character (S26). If the character length is 1 character (Y in S26), candidate selection processing is performed when the search character string is 1 character (S28), and the process proceeds to step 44. If it is determined in step 26 that the character length is not one character (N in S26), it is determined whether the character length is two characters (S30). If the character length is 2 characters (Y in S30), candidate selection processing is performed when the search character string is 2 characters (S32), and the process proceeds to step 44. If it is determined in step 30 that the character length is not 2 characters (N in S30), it is determined whether or not the character length is 3 characters (S34). If the character length is 3 characters (Y in S34), candidate selection processing is performed when the search character string is 3 characters (S36), and the process proceeds to step 44. If it is determined in step 34 that the character length is not 3 characters (N in S34), it is determined whether or not the character length is 4 characters (S38). If the character length is 4 characters (Y in S38), candidate selection processing is performed when the search character string is 4 characters (S40), and the process proceeds to step 44. If the character length is not 4 characters (N in S38), candidate selection processing is performed when the search character string is 5 characters (S42), and the process proceeds to step 44.

  FIG. 6 is a diagram showing an example of a processing sequence between the devices in the information providing system 50 of FIG. The user terminal device 16 transmits a character string serving as a search keyword to the search server 20 (S50). The search server 20 performs a search based on the character string (S52). When the search result is not obtained (N in S54), the similar word requesting unit 26 in FIG. 1 transmits the character string to the word search device 100 (S56). The preprocessing unit 120 in FIG. 4 performs preprocessing on the character string (S58). Then, the extraction unit 124 of FIG. 4 selects similar word candidates (S60). The calculation unit 126 of FIG. 4 calculates a similarity count value for each selected similar word candidate (S62). The selection unit 128 of FIG. 4 selects a predetermined number of words in ascending order of similarity count values (S64) and transmits them to the search server 20 (S66).

  When a similar word is received from the word search device 100 (Y in S68), the search server 20 transmits screen information for displaying the display screen 52 of FIG. 2 to the user terminal device 16 (S70). The user terminal device 16 receives an instruction to select a similar word from the user (S72), and transmits the instructed similar word to the search server 20 (S74). The search server 20 searches again based on the word (S76). Then, the search server 20 transmits the search result to the user terminal device 16 (S78). In step 68, when a similar word is not received from the word search device 100 (N of S68), the user terminal device 16 is notified that there is no information corresponding to the character string (S78). If the search is successful in step 54 (Y in S54), the search result is output to the user terminal device 16 (S55).

  The present invention has been described based on the embodiments. This embodiment is an exemplification, and it will be understood by those skilled in the art that various modifications can be made to combinations of the respective constituent elements and processing processes, and such modifications are also within the scope of the present invention. is there. Examples of such modifications are as follows.

  As a first modification, the word search device 100 in FIG. 1 accepts a search character string from the search server 20 operated by an entity different from the operation entity of the word search device 100 and provides a similar word. It may be operated by the same operating entity.

  As a second modification, the word search device 100 is independent of the search server 20 and accepts a search character string via the network 10, but the word search device 100 is provided by being incorporated in the search server 20. Also good.

  As a third modification, the word search device 100 is not limited to the search server 20 but accepts a character string from a user such as a shopping site or a dictionary site, and provides predetermined information according to the character string. A character string may be received from a device that performs processing, and similar words of the character string may be output.

It is a block diagram of the information provision system which concerns on embodiment. It is a figure which shows an example of the display screen displayed on a user terminal device, when a similar word is found. It is an internal block diagram of the word search device of FIG. It is an internal block diagram of the analysis part of FIG. It is a flowchart of the process in the extraction part of FIG. It is a figure which shows an example of the sequence of the process between each apparatus in the information provision system of FIG.

Explanation of symbols

  10 network, 16 user terminal device, 20 search server, 22 search unit, 24 address information storage unit, 26 similar word request unit, 28 similar word providing unit, 30 communication unit, 50 information providing system, 100 word search device, 102 reception Unit, 104 analysis unit, 106 output unit, 108 authentication unit, 110 log processing unit, 112 log storage unit, 114 billing unit, 120 preprocessing unit, 122 specific word search unit, 124 extraction unit, 126 calculation unit, 128 selection unit , 130 Specific word list storage unit, 132 Update unit, 134 Output word list storage unit, 150 Data storage unit, 152 Character type data storage unit, 154 Character length data storage unit, 156 Knowledge data storage unit, 158 Reading data storage unit 160 Conversion data storage unit.

Claims (8)

A reception unit that accepts input of a character string;
A storage for holding information about multiple words;
Based on the attribute of the character string, an analysis unit that extracts similar words similar to the character string from words included in the storage unit;
A word search device comprising:   The word search apparatus according to claim 1, wherein the analysis unit includes an extraction unit that extracts candidates for the similar word based on a character type of a character included in the character string.   The word search apparatus according to claim 2, wherein the extraction unit extracts the similar words based on the number of characters included in the character string. The storage unit holds a word table for each character type of characters included in a word,
The said extraction part specifies the character type of each character contained in the said character string, The similar word candidate is extracted with reference to the word table corresponding to the specified character type. Word search device. The storage unit holds a word table for each number of characters included in a word,
4. The word search apparatus according to claim 3, wherein the extraction unit specifies the number of characters in the character string and extracts similar word candidates with reference to a word table corresponding to the specified number of characters.   The said extraction part extracts the candidate of a similar word based on the coincidence degree of the character block contained in the said character string, and the character block contained in the said word, The one of Claim 1 to 5 characterized by the above-mentioned. Word search device. Receiving an input of a character string from a predetermined search device;
Extracting a similar word similar to the character string from words included in a storage unit that holds information on a plurality of words based on the attribute of the character string;
Outputting the extracted similar words to the search device;
A word search method comprising: A reception unit that accepts input of a character string;
A first storage unit that holds a predetermined word and predetermined information related to the word in association with each other;
A search unit for searching for predetermined information corresponding to the character string with reference to the first storage unit;
If the predetermined information corresponding to the character string cannot be searched, it is similar to the character string from among the words included in the second storage unit that holds information related to a plurality of words based on the attribute of the character string. An extraction unit for extracting similar words;
The information providing system is characterized in that the search unit searches for predetermined information corresponding to the similar words.

Images