JP5590610B2 - Synonym determining device, synonym determining method and program - Google Patents

Description

  The present invention relates to a technique for determining a synonym of a keyword used for a search.

  In a search service that acquires a keyword and provides an address of a Web page that includes the acquired keyword, various keywords are searched. The keywords may be input in various notation even if they have the same meaning, such as katakana, hiragana, kanji, and romaji. In addition, the notation such as “server” and “server” may fluctuate, or an input error may occur when a keyword is input, for example, only one character of the input keyword may be input incorrectly. For users of search services, it is desirable to provide information that the user wants to obtain even if there is a difference in notation, a notation or an incorrect input. For this purpose, it is desirable to provide the user with search results of synonyms that are considered synonymous with the input keyword even if the keyword is input with a difference in notation, a notation fluctuation, or an incorrect spelling.

Various techniques have been proposed for obtaining synonyms.
For example, in Patent Document 1, a first related text related to one first character string to be a synonymous determination target and a second related text related to the other second character string are acquired, and a second character string in the first related text is acquired. A technique for determining whether the first character string and the second character string are synonymous based on the appearance frequency of the first character string and the appearance frequency of the first character string in the second related text is disclosed.
Further, in Patent Document 2, synonymous expressions are analyzed by analyzing dependency expressions, common expressions that are commonly included in a pair of documents, and differential expressions that are included only in one document. Techniques for extraction are disclosed.
There is also a technique that uses reading for synonymous determination. For example, in Patent Document 3, synonym candidate candidates are extracted from input text, and a pair of synonym candidates is generated after normalizing notation and reading. However, a technique for determining synonyms is disclosed.

JP 2008-293231 A JP 2008-234175 A JP 2009-223463 A

  According to the techniques disclosed in Patent Documents 1 to 3, it is possible to determine synonyms, but it is necessary to collect documents in order to determine synonyms, and many synonyms are determined. In order to do so, a large amount of documents are required, and the amount of processing and time spent for the determination are enormous.

  The present invention has been made under the above-described background, and an object thereof is to provide a technique for determining synonyms without collecting documents.

The present invention provides an acquisition means for acquiring a search keyword and a log of a set of addresses of information accessed from the search result of the keyword, and an access probability of accessing the address paired with the keyword in the set from the search result of the keyword And an access probability calculation means for calculating the number of appearances of the keyword in the log and the number of appearances of the set in the log, and a guidance probability derived from the search result of the keyword to an address paired with the keyword in the set. , The number of appearances of the address in the log, and the guidance probability calculating means for calculating from the number of appearances of the set in the log, the first set from the log, and the first set from the log and the same keyword and different keywords Obtain a second set, and the similarity between the first set of keywords and the second set of keywords Using the similarity calculated from the number of appearances in the log of the first set, the number of appearances in the log of the second set, and the similarity calculated by the similarity calculation unit wherein the first set of keywords a second set of keywords have a determination means whether synonyms, the similarity calculation means, said access probability according to the first set of keywords, Provided is a synonym determination device for calculating a similarity from the guidance probability related to the second set of keywords .
The present invention also provides an acquisition means for acquiring a log of a search keyword and a set of addresses of information accessed from the search result of the keyword, and accessing the address of a pair with the keyword in the set from the search result of the keyword The access probability calculating means for calculating the access probability from the number of appearances of the keyword in the log and the number of appearances of the set in the log, the first set from the log, and the address of the first set are the same. A second set having different keywords is acquired, the similarity between the first set of keywords and the second set of keywords is calculated, the number of appearances in the log of the first set, and the second set. Similarity calculation means for calculating based on the number of appearances in the log and the access probability calculated by the access probability calculation means, and the similarity calculation Wherein said first set of keywords using the degree of similarity calculated by means a second set of keywords provides synonyms determination apparatus having a determination unit whether synonyms.

  In the present invention, when the determination means determines that the synonym is a synonym, a keyword having a large number stored in the log among the keywords of the first group and the keyword of the second group is defined as a normal notation word. A configuration may be adopted in which keywords with a small number stored in the log are determined to be synonyms of the normal notation word.

  Further, in the present invention, there is provided a first distance calculating means for calculating a distance of reading the first set of keywords and the second set of keywords, and the determining means is calculated by the similarity calculating means. It is good also as a structure which determines whether the said 1st group of keywords and the said 2nd group of keywords are synonyms using the calculated similarity and the distance calculated by the said 1st distance calculation means.

  Also, in the present invention, the determination means uses the distance calculated by the first distance calculation means and the length of reading of the first set of keywords and the second set of keywords and the second It is good also as a structure which determines whether the keyword of this group is a synonym.

  In the present invention, there is provided a second distance calculating means for calculating a distance between the first set of keywords and the second set of keywords, and the determining means is calculated by the similarity calculating means. It is good also as a structure which determines whether the said 1st set of keywords and the said 2nd set of keywords are synonyms using the calculated similarity and the distance calculated by the said 2nd distance calculation means.

  In the present invention, the determination means uses the distance calculated by the second distance calculation means and the number of characters in the first set of keywords to describe the first set of keywords and the second set. It is good also as a structure which determines whether these keywords are synonyms.

In addition, the present invention provides an acquisition step in which a control unit of a computer apparatus acquires a log of a search keyword and an address set of information accessed from the keyword search result, and the control unit includes a keyword in the set. An access probability calculating step of calculating an access probability of accessing a pair of addresses from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the set in the log; and the control means includes the group A guidance probability calculation step of calculating a guidance probability induced by the keyword search result to an address paired with the keyword from the number of appearances of the address in the log and the number of appearances of the set in the log; and the control means However, the first set from the log has the same address as the first set. A second set having different keywords is acquired, the similarity between the first set of keywords and the second set of keywords is calculated, the number of appearances in the log of the first set, and the second set. A similarity calculation step that is calculated from the number of appearances in the log, and the control means uses the similarity calculated in the similarity calculation step to determine the first set of keywords and the second set of keywords. have a synonymous whether determination step, the similarity calculation step, said access probability according to the first set of keywords, the similarity from the inductive probability according to the second set of keywords A synonym determination method for calculating
In addition, the present invention provides an acquisition step in which a control unit of a computer apparatus acquires a log of a search keyword and an address set of information accessed from the keyword search result, and the control unit includes a keyword in the set. An access probability calculating step of calculating an access probability of accessing a pair of addresses from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the pair in the log; and the control means includes the log To obtain a first set and a second set having the same address and different keywords as the first set, and calculating the similarity between the first set of keywords and the second set of keywords, The number of appearances in the log of the first set, the number of appearances in the log of the second set, and the access probability calculating means A similarity calculation step that is calculated based on the issued access probability, and the control means uses the similarity calculated in the similarity calculation step to use the first set of keywords and the second set of keywords. And a determination step for determining whether or not is a synonym.

In addition, the present invention provides a computer that acquires a log of a set of search keywords and an address of information accessed from the keyword search results, and each time the keyword and address are acquired by the acquisition unit. Storage unit control means for storing the acquired keyword and address pair in the storage unit as a log, and the access probability of accessing the address paired with the keyword in the set from the keyword search result in the log of the keyword An access probability calculating means for calculating the number of appearances and the number of appearances in the log of the set, and the induction probability derived from the search result of the keyword to an address paired with the keyword in the set, the appearance probability in the log of the address Guidance calculated from the number of times and the number of appearances in the log of the set A rate calculating unit, a first set from the log, the first set and the address to get the different second set the same keyword, the first set of keywords and the second set of keywords The similarity calculated by the number of appearances in the log of the first set and the number of appearances in the log of the second set, and the similarity calculated by the similarity calculation unit a program for the second set of keywords to function as a synonym whether determination hand stage and the first set of keywords using the similarity calculation means, said first Provided is a program for calculating a similarity degree from the access probability related to one set of keywords and the guidance probability related to the second set of keywords .
In addition, the present invention provides a computer that acquires a log of a set of search keywords and an address of information accessed from the keyword search results, and each time the keyword and address are acquired by the acquisition unit. Storage unit control means for storing the acquired keyword and address pair in the storage unit as a log, and the access probability of accessing the address paired with the keyword in the set from the keyword search result in the log of the keyword An access probability calculating means for calculating the number of appearances and the number of appearances of the set in the log, and obtaining a first set from the log and a second set having the same address and different keyword as the first set. , The degree of similarity between the first set of keywords and the second set of keywords is expressed as the log of the first set. Similarity calculation means calculated based on the number of appearances in the log, the number of appearances in the log of the second set, and the access probability calculated by the access probability calculation means, and the similarity calculated by the similarity calculation means A program for functioning as a determination means for determining whether or not the first set of keywords and the second set of keywords are synonyms using the degree is provided.

  According to the present invention, synonyms can be determined without collecting documents.

1 is a diagram showing an overall configuration of a communication system 1 according to an embodiment of the present invention. The block diagram which showed the hardware constitutions of the synonym determination apparatus 40. FIG. The block diagram which showed the structure of the function implement | achieved in the synonym determination apparatus 40. FIG. The sequence diagram for demonstrating operation | movement of embodiment. The flowchart which showed the flow of the process which CPU402 performs. The figure which showed the contents of the log. The figure which showed the format of table TB10. The figure which showed the format of table TB11. The flowchart which showed the flow of the process of CPU402 which concerns on the modification of 1st Embodiment. The figure which showed the format of table TB12. The flowchart which showed the flow of the process of CPU402 which concerns on 2nd Embodiment. The figure which showed the format of table TB20. The flowchart which showed the flow of the process of CPU402 which concerns on 3rd Embodiment.

[First Embodiment]
(overall structure)
FIG. 1 is a diagram illustrating an overall configuration of a communication system 1 according to an embodiment of the present invention. The communication line 20 is the Internet in this embodiment. In FIG. 1, one terminal device 10 is connected to the communication line 20, but a plurality of terminal devices 10 can be connected to the communication line 20. The terminal device 10 is a personal computer in which a Web browser is installed, and performs data communication with other computer devices via the communication line 20. The terminal device 10 is not limited to a personal computer, and may be a computer device in which a Web browser is installed, such as a smartphone, a mobile phone, or a PDA (Personal Digital Assistant).

  The search server device 30 acquires a keyword, and provides a search service that provides an address (location) of information related to the acquired keyword (for example, a Web page including the keyword) to the user of the terminal device 10. A server device having a type search engine. The search server device 30 is connected to the communication line 20 and performs data communication with the terminal device 10 via the communication line 20. The search server device 30 has a function of searching for a Web page related to the keyword sent from the terminal device 10 and providing the terminal device 10 with the address of the Web page found by the search. Note that the search server device 30 may provide not only a Web page but also an address of a document file including information related to a keyword.

  The server device 35 is a computer device that is connected to the communication line 20 and performs data communication with the terminal device 10 via the communication line 20. The server device 35 acquires the keyword used when the user of the terminal device 10 uses the search service and the information indicating the address of the Web page the user is going to browse, and stores the acquired keyword / address pair as a log. To do. The URL (Uniform Resource Locator) of the server device 35 is “http: // redirect_server” in this embodiment.

  The synonym determination device 40 is connected to the server device 35 and acquires the log stored by the server device 35 from the server device 35 by data communication. Note that the synonym determination device 40 may be connected to the communication line 20 and the log may be acquired from the server device 35 via the communication line 20. The synonym determination device 40 extracts, from the log acquired from the server device 35, the keyword used when the user of the terminal device 10 uses the search service and information indicating the address of the Web page that the user is going to browse. It has a function to extract synonyms of keywords.

(Configuration of synonym determination device 40)
FIG. 2 is a block diagram illustrating a hardware configuration of the synonym determination device 40. As shown in FIG. 2, each unit of the synonym determination device 40 is connected to a bus 401, and data is exchanged between the units via the bus 401.

  The communication unit 408 functions as an interface for performing data communication with the server device 35. The communication unit 408 is connected to the server device 35 by a communication cable (not shown). The operation unit 406 includes a keyboard (not shown) and a mouse (not shown) operated by the user. When the mouse or the keyboard is operated, the operation unit 406 outputs a signal indicating the operated button or key to a CPU (Central Processing Unit) 402. The display unit 407 has a liquid crystal display, and displays a character and graphic screen, a menu screen for operating the synonym determination device 40, and the like under the control of the CPU 402.

  The storage unit 405 includes a device (for example, a hard disk device) that permanently stores data, and stores a program that realizes the function of the operating system in the synonym determination device 40. In addition, the storage unit 405 stores an application program that implements a function of acquiring a keyword used when the user of the terminal device 10 uses the search service and extracting a synonym of the acquired keyword.

  The ROM 403 stores an IPL (Initial Program Loader). The RAM 404 is used as a work area when the CPU 402 executes a program. The CPU 402 is a control unit that controls each unit of the synonym determination device 40. When the synonym determination device 40 is turned on, the CPU 402 reads the IPL from the ROM 403 and executes it. After executing the IPL, the CPU 402 reads out a program that realizes the function of the operating system from the storage unit 405 and executes it. When the CPU 402 executes a program that realizes the function of the operating system, the synonym determination device 40 functions as a server of the client server system and can perform data communication with other computer devices. Further, the CPU 402 executes an application program stored in the storage unit 405. When the CPU 402 executes the application program, the various functions shown in FIG. 3 are realized, and the keyword used when the user of the terminal device 10 uses the search service and the address of the Web page that the user is going to browse are acquired. A function to extract synonyms of acquired keywords is realized.

(Functional configuration of the synonym determination device 40)
FIG. 3 is a block diagram showing a configuration of functions realized in the synonym determination device 40 by executing the transmission program.
The acquisition unit 450 acquires information received from the server device 35 by the communication unit 408. Specifically, the log generated by the server device 35, that is, the keyword used when the user of the terminal device 10 uses the search service and the information indicating the address of the Web page that the user intends to browse is acquired. The storage unit control unit 451 stores the log acquired by the acquisition unit 450 in the storage unit 405. The access probability calculating means 452 calculates the access probability of accessing the address paired with the keyword in the log from the search result of the keyword from the number of appearances in the log of the keyword and the number of appearances in the log of the pair of the keyword and the address. To do. The similarity calculation means 453 acquires two keywords from the keywords stored as logs, and calculates the similarity between the acquired keywords from the number of appearances of each acquired keyword in the log. The determination unit 454 determines whether the two keywords are synonyms using the similarity calculated by the similarity calculation unit.

(Operation of the embodiment)
First, the exchange of information between the devices related to the communication system 1 will be described. FIG. 4 is a sequence diagram for explaining the operation of the embodiment. When the URL of the search server device 30 is input by the user in the terminal device 10 in which the Web browser is executed and an operation for instructing access to the search server device 30 is performed, the HTTP is sent from the terminal device 10 to the search server device 30. A (HyperText Transfer Protocol) request is sent (step S101). When receiving the HTTP request, the search server device 30 sends a Web page for inputting a search keyword to the terminal device 10 (step S102). When the terminal device 10 receives this Web page, the terminal device 10 displays a Web page for inputting a search keyword.

  When the user of the terminal device 10 inputs a search keyword on this page and performs an operation of sending the keyword to the search server device 30, the input keyword is sent to the search server device 30 (step S103). Upon receiving this keyword, the search server device 30 searches for a Web page related to the received keyword. Next, the search server device 30 generates a plurality of URLs including the host name of the server device 35, the received keyword and the address of the searched Web page as arguments, and displays a Web page that displays the generated URL as a link as a terminal device. 10 (step S104). For example, when the keyword is “server”, an example of the generated URL is “http: // redirect_server /? Q = server & url = xxxxx”. Here, “server” between “q =” and “&” is the keyword received, and “xxxxx” after “url =” is the address of the Web page related to the keyword. “Xxxxx” is the URL of the Web page.

  When receiving the Web page sent from the search server device 30, the terminal device 10 displays the received Web page. Thereby, the URL generated by the search server device 30 is displayed as a link. When the user of the terminal device 10 performs an operation of clicking the displayed link, an HTTP request is sent to the server device 35 based on the URL related to the clicked link (step S105). For example, when the URL related to the clicked link is “http: // redirect_server /? Q = server & url = xxxxx”, the argument “q = server & url = xxxxx” is sent to the server device 35 whose host name is “redirect_server”. An HTTP request containing "is sent.

  Upon receiving this HTTP request, the server device 35 extracts the keyword “server” and the address “xxxxx”. For example, as shown in FIG. 6, the reception date and time of the HTTP request, the keyword q of the argument, and the address s Are associated and stored as a log. The server device 35 includes an address “xxxxx”, generates an HTTP redirect message to be redirected to this address, and sends the generated message to the terminal device 10 (step S106). Upon receiving this message, the terminal device 10 transmits an HTTP request for requesting a Web page to an address included in the message. When a web page is sent from the destination to which the HTTP request is sent, the sent web page is displayed on the terminal device 10.

  Thus, in the present embodiment, a message for redirecting to a Web page related to a keyword is sent from the server device 35 to the terminal device 10, and the terminal device 10 accesses the server device that is the redirect destination in response to this message. It has a configuration. In the above-described embodiment, the search server device 30 is configured to send not only the URL combining the host name and the path name but also the host name when sending the address of the Web page related to the keyword to the terminal device 10. There may be.

  Next, the operation of the synonym determination device 40 will be described. FIG. 5 is a flowchart showing the flow of processing performed by the CPU 402 of the synonym determination device 40. The synonym determination device 40 acquires a log generated by the server device 35 from the server device 35 before performing the processing of FIG. The log acquisition timing may be acquired from the server device 35 every time a predetermined time elapses, or may be acquired from the server device 35 every time a new keyword / address pair is added to the log. Good.

First, the CPU 402 counts the number of appearances N (q, s) of each pair in the log for each pair of the keyword q and the address s from the log acquired from the server device 35 (step SA1). Since the terminal device 10 requests a Web page from the address paired with the keyword in the log as described above, the number of appearances N (q, s) is determined by the user of the terminal device 10 as the keyword q. It can be said that the number of accesses to the address s is based on this. Next, the CPU 402 obtains the number of appearances N (q) of each keyword q in the log, and calculates the access probability P (s | q) of accessing the keyword s and the paired address s by the following equation (1) (step) SA2).

The CPU 402 obtains the access probability P (s | q) for each pair of the keyword q and the address s, and as shown in FIG. 7, the keyword q, the address s, the number of appearances N (q, s), and the access probability P ( A table TB10 associated with s | q) is generated (step SA3). Next, the CPU 402 acquires two keywords from the stored keywords, and calculates the similarity Sim (q _a , q _b ) for one keyword q _a and the other keyword q _b by the following equation (2). (Step SA4).

Sim (q _a , q _b ) calculated by the formula 2 takes a value between 0 and 1, and the closer to 1, the higher the similarity between the keyword q _a and the keyword q _b. To do. Note that the similarity Sim (q _a , q _b ) need not be obtained for all keyword combinations stored in the log. For example, one record is specified in the table TB10, a record having a different address from the specified record but having the same address is specified, the number of appearances N (q _a , s) of one specified record, When the record appearance count N (q _b , s) is equal to or greater than a predetermined number, the similarity Sim (q _a , q _b ) between the keyword q _a and the keyword q _b may be obtained. Further, one record is identified in the table TB10, a record having a different address from the identified record but having the same address is identified, the access probability P (s | q _a ) of the identified one record, and the other record When the record access probability P (s | q _b ) is equal to or higher than a predetermined probability, the similarity Sim (q _a , q _b ) between the keyword q _a and the keyword q _b may be obtained.

For example, in the case of the table TB10 of FIG. 7, if the addresses of the records in the first line and the record in the sixth line are different for the keyword “server” and “server”, the appearance of the record in the first line When the number of times N (q _a , s) is equal to or greater than the predetermined number of times, and the number of appearances N (q _b , s) of the record in the sixth row is also equal to or greater than the predetermined number of times, the keyword “server” and the keyword The similarity “Sim” (q _a , q _b ) of “server” is obtained. When the access probability P (s | q) is used, the access probability P (s | q _a ) of the record in the first row is equal to or higher than a predetermined probability, and the access probability P of the record in the sixth row. When (s | q _b ) is also equal to or higher than a predetermined probability, the similarity Sim (q _a , q _b ) between the keyword q _a “server” and the keyword q _b “server” is obtained.

Next, the CPU 402 determines that the combination of the keyword q _a and the keyword q _b for which the obtained similarity Sim (q _a , q _b ) is equal to or greater than a predetermined threshold (for example, 0.5) is _a synonym, and is synonymous. Extracted as a set of words (step SA5).
Then, the number of appearances N (q _a ) of the keyword q _{a in} the log is compared with the number of appearances N (q _b ) of the keyword q _{b in} the log. Here, CPU402, the keyword q _a and out of the keyword q _b, definitive appearance number of times there are more keywords a positive notation word Q1 of the log, the keyword of more frequent appearances were limited as a synonym Q2, FIG. 8 (a ), A table TB11 is generated in which the correct notation word Q1 is associated with the synonym Q2 with the smaller number of appearances (step SA6). In the table TB11, as shown in FIG. 8B, the number of appearances N (Q2) of the synonym Q2 in the log, the number of appearances N (Q1) of the regular notation word Q1 in the log, and the synonym Q2 The similarity Sim (Q2, Q1) of the positive notation word Q1 may be associated.

  According to this table TB11, although the notation is different from the normal notation word Q1, it is possible to obtain a keyword that can access the same Web page, that is, the same address as the search result, and is synonymous with the correct notation word Q1 from the similarity. The synonym Q2 which can be considered can be specified. If the search server device 30 acquires the synonym Q2 from the terminal device 10 and sends the search result of the correct notation word Q1 to the terminal device 10 as well, the correct notation word that is considered synonymous with the input synonym Q2 The search result of Q1 can also be provided to the user.

(Modification of the first embodiment)
FIG. 9 is a flowchart illustrating a process flow of the CPU 402 according to the modification of the first embodiment. The synonym determination device 40 may obtain the similarity between two keywords according to the flowchart of FIG. Specifically, the CPU 402 first counts the number of appearances N (q, s) of each pair in the log for each pair of the keyword q and the address s (step SB1). Further, the CPU 402 obtains the number of appearances N (q) of the keyword q in the log, and calculates the access probability P (s | q) for accessing the address s of the keyword and the pair according to the equation 1 (step SB2). The processing in steps SB1 and SB2 so far is the same as the processing in steps SA1 and SA2.

Next, the CPU 402 calculates, based on the address s stored as a log, a guidance probability P (q | s) that the user accessing the address s stored in the log is guided by the search result of the keyword q. (Step SB3). The guidance probability P (q | s) is obtained by the following equation (3), and N (s) in the equation (3) is the number of appearances of the address s in the log. When the CPU 402 obtains the guidance probability (q | s), as shown in FIG. 10, the number of appearances N (q, s) and the guidance probability P in the log of the keyword q, the address s, and the combination of the keyword q and the address s. A table TB12 associated with (q | s) is generated (step SB4).

Next, the CPU 402 calculates the similarity Sim (q _a → q _b ) from the keyword q _a to the keyword q _{b according} to the equation 4 from the obtained access probability P (s | q) and guidance probability P (q | s). ) Is calculated (step SB5).

The similarity Sim (q _a → q _b ) takes a value between 0 and 1, and the higher this value, the more the address s accessed by the keyword q _a is accessed by the keyword q _b . Show. Further, CPU 402 calculates the similarity Sim _{(q a} → _{q a)} of the keyword _{q a} by the following formula having 5 to keyword _{q a} (step SB6).

The CPU 402 calculates the similarity Sim (q _a → q _b ) and the similarity Sim (q _a → q _a ) for various combinations of the keyword q _a and the keyword q _b . If Sim (q _a → q _b )> α * Sim (q _a → q _a ), the set of the keyword q _a and the keyword q _a is determined as a synonym and extracted as a set of synonyms (step SB7). ). Here, α is a real number larger than 1, and is 2 in this modification. The similarity Sim (q _a → q _b ) takes a larger value than the similarity Sim (q _a → q _a ), which means that the keyword q _b is accessible to the same address as the keyword q _a and many Indicates that the keyword is used by someone who is. Accordingly, CPU 402 is positive notation word Q1 of the keyword _{q b} keyword _{q a,} the keyword _{q a} is synonymous Q2, to create a table TB11 that associates a positive notation word Q1 synonym Q2 (step SB8).

Even when the condition of Sim (q _a → q _b )> α * Sim (q _a → q _a ) is satisfied, the similarity Sim (q _a → q _b ) is a predetermined threshold value β (for example, If it is smaller than 0.01), there is a high possibility that the keyword q _a and the keyword q _b are completely unrelated words, so they may not be registered in the table TB12.
The similarity Sim (q _a → q _b ) need not be calculated for all keyword pairs. For example, one record is identified in the table TB10, and the address of the identified record and the keyword is different. identify the record is the same, number of occurrences N (q _{a, s)} of the specified one of the record and, number of occurrences N (q _{b, s)} of the other records in the case of more than the number of times the predetermined these You may make it obtain | require the similarity of the keyword contained in this record. Further, the similarity may be obtained when the access probability or the guidance probability is equal to or higher than a predetermined probability.

[Second Embodiment]
Next, a second embodiment of the present invention will be described. In the second embodiment of the present invention, the communication system 1 according to the second embodiment has the same hardware configuration as that of the first embodiment, and the processing performed in the synonym determination device 40 is different from that of the first embodiment. . For this reason, description about the same structure as 1st Embodiment is abbreviate | omitted, and the difference with 1st Embodiment is demonstrated below.

  FIG. 11 is a flowchart showing a process flow of the CPU 402 according to the second embodiment. First, the CPU 402 counts the number of appearances N (q, s) of each pair in the log for each pair of the keyword q and the address s from the log (step SC1). Next, the CPU 402 obtains the number of appearances N (q) of the keyword q in the log, and calculates the access probability P (s | q) for accessing the address s of the keyword and the pair according to the equation 1 (step SC2). The processing in steps SC1 and SC2 up to this point is the same as the processing in steps SA1 and SA2.

  Next, as shown in FIG. 12, the CPU 402 generates a table TB20 in which the keyword q, the reading of the keyword q, the address s, the number of appearances N (q, s), and the access probability P (s | q) are associated. (Step SC3). In the present embodiment, the storage unit 405 stores a dictionary in which words and word readings are associated with each other, and the CPU 402 specifies keyword readings using the dictionary and associates the specified readings with keywords. For example, using this dictionary, the keyword “server” is read as “server”.

  Next, based on the address s stored as a log, the CPU 402 calculates a guidance probability P (q | s) that the user who accesses the address s stored in the log is guided by the keyword q (step SC4). ). The guidance probability P (q | s) is obtained by the equation (3). When determining the guidance probability (q | s), the CPU 402 associates the number of appearances N (q, s) and the guidance probability P (q | s) in the log of the keyword q, the address s, and the set of the keyword q and the address s. The table TB12 is generated (step SC5).

Next, the CPU 402 calculates the similarity Sim (q _a → q _b ) from the keyword q _a to the keyword q _{b according} to the equation 4 from the obtained access probability P (s | q) and guidance probability P (q | s). ) as well as calculating a degree of similarity is calculated from the keyword _{q a} by the numerical formula 5 to the keyword _{q a} Sim _{(q a} → _{q a)} (step SC6).

Further, the CPU 402 calculates an editing distance d (q _a , q _b ) for reading the keyword q _a and the keyword q _b (step SC7). For example, the keyword “note PC” whose reading is “note PC” and the keyword “note PE” whose reading is “note PC” will be the same reading if “i” is replaced with “shi”. Therefore, the edit distance d (q _a , q _b ) = 1.

When the CPU 402 finishes calculating the access probability P (s | q), the guidance probability P (q | s), and the edit distance d (q _a , q _b ), Sim (q _a → q _b )> α * Sim ( q _a → q _a ), and d (q _a , q _b ) ≦ β (β is a number greater than or equal to 0, for example, 1), and a pair of keyword q _a and keyword q _b is determined as a synonym and extracted (Step SC8). When extracting a set of conditions are satisfied keyword _{q a} and keyword _{q b,} positive notation word Q1 keyword _{q a} keyword _{q b,} the keyword _{q a} as synonyms Q2, the positive notation word Q1 synonymous Q2 The associated table TB11 is generated (step SC9).
According to the present embodiment, the edit distance of reading, that is, the closeness of reading between keywords is added to the reference for extracting synonyms, so that synonyms can be extracted with high accuracy.

(Modification of the second embodiment)
In the second embodiment described above, synonyms are extracted using the editing distance of reading, but for example, the difference between uppercase and lowercase letters such as “a” and “a”, and the presence or absence of the long sound code “-” Is not a big difference in reading, so the edit distance may be set to “0”. In this case, keywords with similar pronunciation are more likely to be extracted as synonyms.

In the second embodiment described above, the editing distance for reading between keywords is used, but synonyms may be extracted using the reading length. For example, the edit distance d (q _a → q _b ) of the keyword q _a with respect to the keyword q _b is defined as d (q _a → q _b ) = d (q _a , q _b ) / Length (q _a ). Note that Length (q _a ) represents the length of reading of the keyword q _a . For example, when the keyword q _a is “note PE”, the reading is “note pe-e”, so that Length (q _a ) = 7. If the keyword q _a is “note PE” and the keyword q _b is “note PC”, d (q _a → q _b ) is 1 / 7≈0.14. When the edit distance d (q _a → q _b ) is a predetermined threshold γ (a value greater than or equal to 0, for example 0.5), the keywords q _a and q _b are extracted as synonyms. When Sim (q _a → q _b ) −d (q _a → q _b )> δ (δ is a predetermined value), the keywords q _a and q _b are determined as synonyms and determined to be synonyms. And may be extracted.

[Third Embodiment]
Next, a third embodiment of the present invention will be described. Also in the third embodiment of the present invention, the hardware configuration of the communication system 1 according to the embodiment is the same as that of the first and second embodiments, and the processing performed in the synonym determination device 40 is different from that of the second embodiment. . For this reason, description about a hardware configuration is abbreviate | omitted and the difference with 2nd Embodiment is demonstrated below.

FIG. 13 is a flowchart showing a processing flow of the CPU 402 according to the present embodiment. Note that the processing from step SD1 to step SD6 is the same as the processing from step SC1 to step SC6.
Next, in the second embodiment described above, synonyms are extracted by using the editing distance d (q _a , q _b ) of how to read between keywords. Synonyms are extracted using the edit distance. For example, if the keyword q _a is “note PE” and the keyword q _b is “note PC”, only the “E” and “C” are different, and therefore the edit distance e (q _a , q _b on the notation. ) Is 1. In the present embodiment, the CPU 402 calculates the similarity Sim (q _a → q _b ), the similarity Sim (q _a → q _a ), and the edit distance e (q _a , q _b ) (step SD7). Then, the CPU 402 satisfies Sim (q _a → q _b )> α * Sim (q _a → q _a ), and e (q _a , q _b ) ≦ β (β is a number of 0 or more, for example, 1) consisting keyword _{q a} and extracting it is determined that the set of synonyms for keywords _{q b} (step SD8). CPU402 is, extracting a set of conditions are satisfied keyword _{q a} and keyword _{q b,} keyword _{q b} keyword _{q a} positive notation word Q1, the keyword _{q a} as synonymous Q2, positive notation word Q1 synonymous Q2 Is created (step SD9).

(Modification of the third embodiment)
In the third embodiment, the edit distance e (q _a , q _b ) on the notation between the keywords is used, but synonyms may be extracted using the number of characters in the notation. For example, the edit distance e (q _a → q _b ) of the keyword q _a with respect to the keyword q _b is defined as e (q _a → q _b ) = e (q _a , q _b ) / Length (q _a ). Note that Length (q _a ) represents the number of characters represented by the keyword q _a . For example, when the keyword q _a is “note PE”, since the number of characters is 6, Length (q _a ) = 5. If the keyword q _a is “note PE” and the keyword q _b is “note PC”, e (q _a → q _b ) = 1/5 = 0.2. When the edit distance e (q _a → q _b ) is a predetermined threshold γ (a value greater than or equal to 0, for example 0.5), the keywords q _a and q _b are extracted as synonyms. When Sim (q _a → q _b ) −e (q _a → q _b )> δ (δ is a predetermined value), the keywords q _a and q _b are determined as synonyms and determined to be synonyms. And may be extracted.

[Modification]
As mentioned above, although embodiment of this invention was described, this invention is not limited to embodiment mentioned above, It can implement with another various form. For example, the above-described embodiment may be modified as follows to implement the present invention, or may be implemented in combination with each modification.

  In the embodiment described above, the synonym determination device 40 acquires logs from the server device 35 and extracts synonyms, but the server device 35 includes the functions of the synonym determination device 40, and the server device 35 Synonyms may be extracted. Further, the search server device 35 may include the function of the synonym determination device 40 and the function of the server device 35, and the search server device 30 may extract the synonym.

  Programs that realize the function of extracting synonyms include magnetic recording media (magnetic tape, magnetic disk (HDD (Hard Disk Drive), FD (Flexible Disk)), etc.), optical recording media (optical disc (CD (Compact Disc), DVD (Digital Versatile Disk), etc.), a magneto-optical recording medium, a semiconductor readable recording medium such as a semiconductor memory, and the like, and may be installed in the synonym determination device 40 or the search server device 30. Alternatively, it may be downloaded and installed via a communication line.

DESCRIPTION OF SYMBOLS 1 ... Communication system, 10 ... Terminal device, 20 ... Communication line, 30 ... Search server device, 35 ... Server device, 40 ... Synonym determination device, 401 ... Bus, 402 ... CPU, 403 ... ROM, 404 ... RAM, 405 ... storage unit, 406 ... operation unit, 407 ... display unit, 408 ... communication unit, 450 ... acquisition unit, 451 ... storage unit control unit, 452 ... access probability calculation unit, 453 ... similarity calculation unit, 454 ... determination unit

Claims (11)

An acquisition means for acquiring a log of a set of search keywords and an address of information accessed from a search result of the keywords;
An access probability calculating means for calculating an access probability of accessing an address paired with a keyword in the set from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the set in the log;
A guidance probability calculating means for calculating a guidance probability induced by a search result of the keyword to an address paired with the keyword in the set from the number of appearances of the address in the log and the number of appearances of the set in the log;
The first set and the second set having the same address and different keywords as the first set are acquired from the log, and the similarity between the first set keyword and the second set keyword is obtained. Similarity calculation means for calculating the number of appearances in the log of the first set and the number of appearances in the log of the second set;
Wherein said first set of keywords using the degree of similarity calculated by the similarity calculation means a second set of keywords have a determination means whether synonyms,
The similarity calculation means is a synonym determination device that calculates a similarity from the access probability related to the first set of keywords and the guidance probability related to the second set of keywords . An acquisition means for acquiring a log of a set of search keywords and an address of information accessed from a search result of the keywords;
An access probability calculating means for calculating an access probability of accessing an address paired with a keyword in the set from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the set in the log;
The first set and the second set having the same address and different keywords as the first set are acquired from the log, and the similarity between the first set keyword and the second set keyword is obtained. Similarity calculation means for calculating based on the number of appearances in the log of the first set, the number of appearances in the log of the second set, and the access probability calculated by the access probability calculation means;
Determination means for determining whether or not the first set of keywords and the second set of keywords are synonyms using the similarity calculated by the similarity calculation means;
A synonym determining device. When the determination unit determines that the word is a synonym, the keyword having a large number stored in the log among the keywords of the first group and the keyword of the second group is set as a normal notation word and stored in the log. synonyms determination apparatus according to claim 1 or claim 2, wherein determining that synonyms number keyword of the positive notation word less that is. First distance calculating means for calculating a distance of reading the first set of keywords and the second set of keywords;
The determination means uses the similarity calculated by the similarity calculation means and the distance calculated by the first distance calculation means to determine whether the first set of keywords and the second set of keywords are synonymous. The synonym determination device according to any one of claims 1 to 3, wherein a determination is made as to whether or not. The determination means uses the distance calculated by the first distance calculation means and the length of reading of the first set of keywords, and the first set of keywords and the second set of keywords are synonymous. The synonym determining device according to claim 4, wherein the synonym determining device according to claim 4 is determined. A second distance calculating means for calculating a distance between the first set of keywords and the second set of keywords;
The determination means uses the similarity calculated by the similarity calculation means and the distance calculated by the second distance calculation means to determine whether the first set of keywords and the second set of keywords are synonymous. The synonym determination device according to any one of claims 1 to 3, wherein a determination is made as to whether or not. The determination means uses the distance calculated by the second distance calculation means and the number of characters in the first set of keyword descriptions to determine whether the first set of keywords and the second set of keywords are synonyms. The synonym determination device according to claim 6, wherein: An acquisition step in which the control means of the computer device acquires a log of a set of addresses of information accessed from a search keyword and a search result of the keyword;
Access probability calculation in which the control means calculates the access probability of accessing the address paired with the keyword in the group from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the keyword in the log Steps,
The control means calculates the induction probability induced by the search result of the keyword to an address paired with the keyword in the group from the number of appearances of the address in the log and the number of appearances of the group in the log A calculation step;
The control means obtains a first set from the log and a second set having the same address and different keyword as the first set, and the first set of keywords and the second set of keywords. The similarity calculation step of calculating the similarity with the number of appearances in the log of the first set and the number of appearances in the log of the second set;
Said control means, said and said using the degree of similarity calculated by the similarity calculating step a first set of keywords a second set of keywords have a a determination step of determining whether synonyms,
The similarity calculation step is a synonym determination method of calculating similarity from the access probability related to the first set of keywords and the guidance probability related to the second set of keywords . Computer
An acquisition means for acquiring a log of a set of search keywords and an address of information accessed from a search result of the keywords;
A storage unit control unit that stores a set of the acquired keyword and address in a storage unit as a log each time the keyword and address are acquired by the acquisition unit;
An access probability calculating means for calculating an access probability of accessing an address paired with a keyword in the set from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the set in the log;
A guidance probability calculating means for calculating a guidance probability induced by a search result of the keyword to an address paired with the keyword in the set from the number of appearances of the address in the log and the number of appearances of the set in the log;
The first set and the second set having the same address and different keywords as the first set are acquired from the log, and the similarity between the first set keyword and the second set keyword is obtained. Similarity calculation means for calculating the number of appearances in the log of the first set and the number of appearances in the log of the second set;
Wherein using the similarity calculated by the similarity calculation means a first set of keywords and the second set of keywords synonym whether determination hand stage
A program to function as a,
The similarity calculation means calculates a similarity from the access probability related to the first set of keywords and the guidance probability related to the second set of keywords. An acquisition step in which the control means of the computer device acquires a log of a set of addresses of information accessed from a search keyword and a search result of the keyword;
Access probability calculation in which the control means calculates the access probability of accessing the address paired with the keyword in the group from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the keyword in the log Steps,
The control means obtains a first set from the log and a second set having the same address and different keyword as the first set, and the first set of keywords and the second set of keywords. The similarity is calculated based on the number of appearances in the log of the first set, the number of appearances in the log of the second set, and the access probability calculated by the access probability calculation means A calculation step;
A determining step in which the control means determines whether the first set of keywords and the second set of keywords are synonyms using the similarity calculated in the similarity calculation step;
A synonym determination method. Computer
An acquisition means for acquiring a log of a set of search keywords and an address of information accessed from a search result of the keywords;
A storage unit control unit that stores a set of the acquired keyword and address in a storage unit as a log each time the keyword and address are acquired by the acquisition unit;
An access probability calculating means for calculating an access probability of accessing an address paired with a keyword in the set from the search result of the keyword from the number of appearances of the keyword in the log and the number of appearances of the set in the log;
The first set and the second set having the same address and different keywords as the first set are acquired from the log, and the similarity between the first set keyword and the second set keyword is obtained. Similarity calculation means for calculating based on the number of appearances in the log of the first set, the number of appearances in the log of the second set, and the access probability calculated by the access probability calculation means;
Determination means for determining whether or not the first set of keywords and the second set of keywords are synonyms using the similarity calculated by the similarity calculation means;
Program to function as.

Images