KR100525616B1 - Method and system for identifying related search terms in the internet search system - Google Patents

Abstract

The present invention relates to a method and system for extracting correlated search queries, and more particularly, to a number of search sessions in which each search query is input and a pair of search queries including the search query. The present invention relates to a related search query extraction method and system that can measure the number of search sessions and determine whether the search sessions are related to each other.

In accordance with another aspect of the present invention, there is provided a method for extracting an associated search query, the method including: maintaining a database including a search session and a record relating to a search query received from a user terminal during the search session (the record is generated at predetermined time intervals; Recorded in the database), generating a total number of search sessions by counting the total number of search sessions set per time interval with reference to the database, and referring to the database, a first search query per time interval Counting the number of search sessions received to generate first search session number information, and counting the number of search sessions for which a second search query was received per time interval with reference to the database for a second search session. Generating number information, referring to the database for each time interval; Generating third search session number information by counting the number of search sessions in which the first search query and the second search query have been received, using the first search session number information and the third search session number information Generating conditional probability information, and performing correlation using the total search session number information, the first search session number information, the second search session number information, and the third search session number information. ) Generating information, and determining whether the first search query and the second search query are related to each other based on the conditional probability information and the correlation information.

According to the present invention, by efficiently collecting and analyzing the data related to the search query received from the user to automatically determine whether or not the association between the search query to quickly extract the correct associated search query to provide a higher quality service to the user An associated search query extraction method and system are provided.

Description

{METHOD AND SYSTEM FOR IDENTIFYING RELATED SEARCH TERMS IN THE INTERNET SEARCH SYSTEM}

The present invention relates to a method and system for extracting correlated search queries, and more particularly, to a number of search sessions in which each search query is input and a pair of search queries including the search query. The present invention relates to a related search query extraction method and system that can measure the number of search sessions and determine whether the search sessions are related to each other.

In general, a search service system that provides a search service, when a search query is input from a user, a search result corresponding to the search query (for example, a web site including the search query, an article including the search query, and the search). An image with a file name containing the query).

Meanwhile, the recent search service system provides a related search query providing service that extracts a search query related to a search query input from the user and provides the search query to the user so that the user can find the desired information more quickly and accurately. . That is, the search service system generally provides users with very different search results according to the input search query. For example, a search result obtained by a searcher by entering "car" and a search result obtained by entering "car" are different from each other. Therefore, the searcher tries to input a search query that is more relevant to the information he / she wants in order to obtain the information he / she wants, but it is difficult for the searcher to think of such a search query by himself. Therefore, a recent search service system provides a user with search queries related to the input search query by using the search query and statistical information input by the user so that the user can search by using another search query. Providing a service.

Here, the related search query is, for example, a search query corresponding to a higher concept or a lower concept of a search query input from a user ("foreign language" when the Japanese language is input, or "foreign language"). "Japanese" if you enter "), a search query that is synonymous with the search query you entered (" Bookstore "if you typed" bookstore "), or a search query that is synonymous with the search query you entered ( `` Tails '' if you type "tail"), or a search query that is related to a search query entered by a user ("saw, seen, seeing" if you typed "see") Can mean a query. However, the relevant search query does not simply mean a search query that has a semantic relation, for example, if the user inputs "Park Chan-ho", his occupation "baseball" and his league "major league" , His college "Hanyang University", his team "Texas Rangers", and Korean baseball player "Kim Byung-hyun" in the same major league.

However, according to the related art search query providing service, a service operator had to sort and store another search query that is related to one search query, and thus there was a problem in that it had to take time and economic loss. .

In addition, in order to extract the associated search query with less time and cost, the concurrent classification method defined as the probability of occurrence of relevance among terms simultaneously, and the document classification method defined by related terms as terms that appear mainly in each group after classifying documents In addition, a grammar classification method for identifying relationships between terms by using linguistic knowledge and co-occurrence characteristics in documents has emerged. However, these methods mainly deal with statistical relationships without considering semantic relationships between terms. There was a problem that a user could not understand the association between the extracted association search queries.

In order to solve this problem, Korean Patent Registration No. 10-0372078 automatically generates a related glossary using a query log input from a user, and searches for a semantically relevant related word using the generated related glossary. A related word search method is disclosed.

The related word search method registers some of the query words made in a specific session among the query words input by the user as related words, and generates a related glossary. When a query is input into the related word search apparatus, the related words of the query words are identified and sorted, and the user is input. To provide.

However, although the related word search method can determine whether or not a query is related to each other by using a query log made within a specific session, the related art problem is as follows.

In other words, even when using the related word search method, 1) if a term appears more than once in the same session, the user may not be able to understand the relationship between the extracted related words because all of them are registered as related words. Because it is not a systematic analysis of accumulated data over a period of time, the related words are coincidentally registered in the same session. 3) In general, although the input from a user is a search query, not a query, In addition, the method of extracting terms from query terms requires unnecessary time to identify correlations between search queries. 4) If a term appears more than once in the same session, all of them are registered as related words. There are still problems such as that may exist.

As a result, even with the related term search method, it is still highly likely that a search query that is not related to a search query input from a user is extracted, and thus, a user who wants to search for desired information more quickly and accurately by providing a high quality service. There is a disadvantage that cannot satisfy the desire.

Therefore, a new technology that can provide a higher quality service to the user by effectively collecting data about the search query received from the user, by analyzing the collected data systematically and using the same to accurately determine the association between the search query Has been called for.

The present invention has been made to solve the problems of the prior art as described above, by constructing a system that can automatically determine the association between the search query by effectively collecting and analyzing data about the search query received from the user It is an object of the present invention to provide an associated search query extraction method and system that can reduce the time and economic loss caused to a service operator by classifying and storing another search query that is related to one search query.

In addition, the related search query extraction method and system according to the present invention maintains a database that systematically records the association index information between search queries, so that when the search query is input from the user, the degree of association is higher by using the association index information. Its purpose is to provide a search query to the user first.

In addition, the related search query extraction method and system according to the present invention extracts only useful data that has undergone a systematic preprocessing process among the search data inputted from users and maintains an appropriate number of related search queries, thereby creating a truly meaningful related search query. The purpose of this is to provide users with higher quality related search query service.

In addition, the related search query extraction method and system according to the present invention aggregates the data accumulated for a certain period of time to determine whether or not the association between the search query, providing a user with a search query that maintains a constant association for a long time Its purpose is to meet the needs of users who want to provide accurate associated search query services.

In order to achieve the above object and to solve the above-mentioned problems of the prior art, the related search query extraction method according to the present invention includes a search session and a record relating to a search query received from a user terminal during the search session. Maintaining a database (the record is generated at predetermined time intervals and recorded in the database), and counting the total number of search sessions set per time interval with reference to the database to obtain the total number of search session information. Generating a first search session number information by counting the number of search sessions in which a first search query is received per time interval with reference to the database; and generating the first search session number information with reference to the database. Modifying a second search session by counting the number of search sessions for which a second search query was received Generating a third search session number information by counting the number of search sessions in which the first search query and the second search query have been received per time interval, with reference to the database; Generating conditional probability information using session number information and third search session number information, the total search session number information, the first search session number information, the second search session number information, and Generating correlation information by using the third search session number information, and determining whether to correlate between the first search query and the second search query based on the conditional probability information and the correlation information; Characterized in that it comprises a step.

According to one aspect of the invention, the step of determining whether or not the association between the first search query and the second search query based on the conditional probability information is determined whether the association only when the conditional probability information is a predetermined value or more. And the numerical value is varied based on a predetermined function using the first search session number information as a variable.

According to another aspect of the present invention, the step of determining whether or not the association between the first search query and the second search query based on the conditional probability information and the correlation information is the conditional probability information and the correlation Generating association index information between the first search query and the second search query using the information; and determining the association based on the association index information, and as a result of the association determination, If determined, recording the association index information in a second database in association with the first search query and the second search query, receiving a third search query from a user terminal, and referring to the second database. Extracting at least one fourth search query associated with a third search query; And sorting the extracted fourth search query to generate a related search query list, and providing the generated related search query list to the user terminal. Extraction methods are provided.

In addition, the association search query extraction system according to the present invention is a database comprising a search session and a record relating to a search query received from a user terminal during the search session, generating the record at predetermined time intervals to the database Database management means for recording, counting the total number of search sessions set per said time interval, with reference to said database to generate total search session number information, and a search session for which a first search query has been received per said time interval; Counting the number of times to generate first search session number information, counting the number of search sessions from which a second search query was received per time interval, to generate second search session number information, and per second time interval. A third search count by counting the number of search sessions in which the first search query and the second search query have been received; Counter means for generating number information, conditional probability information generating means for generating conditional probability information using the first search session number information and the third search session number information, the total search session number information, the Correlation information generating means for generating correlation information using first search session number information, second search session number information, and third search session number information, and the conditional probability information and the correlation And association means for determining association between the first search query and the second search query based on the information.

Hereinafter, an association search query extraction method and system according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating a network connection of an association search query extraction system according to an embodiment of the present invention. Users access the associated search query extraction system 100 through a wired or wireless communication network using the user terminal 110a or 110b and input a search query. The association search query extraction system 100 transmits an association search query corresponding to the search query to the user terminal 110a or 110b. In addition, the association search query extraction system 100 according to another embodiment of the present invention may further transmit the ranking information according to the association index information of the association search query to the user terminal 110a or 110b. The related search query extracting system of the present invention can be integrated and operated in the Internet search service system. Therefore, when the user accesses the Internet search service system and inputs a search query, the search result for the search query is provided to the user. The associated search query of the search query may also be provided to the user.

2 is a flowchart illustrating a method of extracting a related search query according to an embodiment of the present invention. The related search query extraction method according to the present embodiment is provided in a predetermined related search query extraction system.

In step 201 the associative search query extraction system according to the invention maintains a database containing a search session and a record relating to a search query received from a user terminal during the search session.

According to an embodiment of the present invention, the search session is established when a search box is first provided to the user terminal, terminates when there is no data transmission from the user terminal for a predetermined time, and after the search session is terminated. If a new search query is received from the terminal, a related search query extraction method is provided, which starts a new search session.

In the present embodiment, the search session is a new type of search session set up in a different manner from the conventional search session. The conventional search session simply means a search activity performed by a user for a predetermined time, and the time until the search session is established and terminated is predetermined. For example, assuming that the time is 10 minutes, the conventional search session is terminated when the user performs a search activity for 10 minutes, and the search activity for 10 minutes after the termination is performed in a new search session. It is a conventional search session establishment method.

In contrast, the search session in this embodiment is set when the search window is first provided to the user terminal, and is terminated when there is no data transmission from the user terminal for a predetermined time. For example, if the time is 5 minutes, the search session is started when a user accesses a search service web page through a user terminal and the search window is first opened, and the user inputs a search query or selects a search result. The search session is terminated when waiting for 5 minutes without performing any search activity from the "last time" of performing the search activity. That is, if the user waits without taking any action for 5 minutes from the time of the user's last search activity, the search session is terminated. In addition, when a new search query is received from the user terminal after the termination of the search session, a new search session is started.

When a search result using a search query does not satisfy the user, the user generally inputs another search query related to the search query to perform a search activity. Therefore, the related search query extraction system according to the present embodiment performs a final search. It is determined that the user waits without taking any action for a predetermined time from the activity time and finishes the search activity related to the specific subject. Therefore, if the new search query is input after the elapsed time, the associated search query extraction system regards the search query related to the new specific subject as input and starts a new session.

On the other hand, even when a long time elapses, when a continuous search activity is performed by a user, the related search query extraction system according to the present embodiment determines that the user inputs various search queries related to a specific subject according to one intention. do. Thus, the associated search query extraction system does not start a new search session in this case, and maintains one search session.

Compared to the conventional method of fixedly setting the time from start to end of a search session and registering all search queries appearing in the search session as associative search queries, according to the present embodiment, the search session is defined in the above-described new manner. As a result, the correlation between search queries received in the same search session can be reliably obtained to a very high degree.

In addition, according to the present invention, instead of registering all the search queries received in the same search session as the related search query, as described below, by counting the number of search sessions in which a pair of search queries appeared, it is determined whether the search queries are related to each other. By utilizing as one of the possible factors, an association search query extraction method and system for more accurately determining whether an association between search queries is provided is provided.

Hereinafter, an embodiment of the step 201 of maintaining a database including records relating to the search session and the search query will be described in detail using the new definition of the search session as described above. 3 is a flowchart illustrating a process of maintaining a database according to the present embodiment.

In step 301, the associated search query extraction system according to the present embodiment generates a first search session identifier associated with a first search session and records it in the database. The associated search query extraction system transmits the first search session identifier and visual information about each search time to a user terminal whenever a search activity is made from a user, and the user terminal receives the received first search session identifier and the The visual information may be stored in a predetermined position in the user terminal in the form of a cookie.

When a final search activity is made from the user, in step 302, the association search query extraction system transmits the first search session identifier and first time information about the last search time to the user terminal. The first search session identifier and the first visual information may also be stored in a predetermined position in the user terminal in the form of a cookie.

In step 303, the association search query extraction system receives a search query from the user terminal, and in step 304 compares the second time information and the first time information from which the search query was received.

If it is determined in step 305 that the gap between the two pieces of time information exceeds a predetermined time, the associated search query extraction system generates a second search session identifier associated with the second search session in step 306. And in step 307 record a record relating to the second search session identifier and the received search query to the database.

On the other hand, if it is determined in step 305 that the gap between the two pieces of time information is less than or equal to a predetermined time, the associated search query extraction system is configured in step 308 for the received search query in association with the first search session identifier. Write a record to the database.

According to the related search query extraction method according to the present embodiment, it is possible to efficiently manage a search session by maintaining a database through the systematic process as described above, and when the search session is efficiently managed, it is received in the same search session. You can achieve a highly reliable effect on the relevance between one search query.

4 is a diagram illustrating an example of a record included in a database according to one embodiment of the present invention. As shown in FIG. 4, the record may include a search session identifier 401 and information about a search query received from a user terminal during a search session associated with the search session identifier 401. Referring to reference numeral 402 illustrated in FIG. 4, "Park Chan-Ho", "Major League", which is a search query received from a user terminal during a search session to which a search session identifier "sessionId1" and a search session identifier "sessionId1" are assigned. It can be seen that "," baseball "and the like are recorded in the record.

According to an embodiment of the present invention, the associated search query extraction system generates the records at predetermined time intervals and records them in the database. The time interval may be predetermined by the service operator such as "day", "two days", "weekly", etc., and the service operator may change the existing time interval to another time interval. According to the present exemplary embodiment, the data may be utilized in determining whether the search queries are related to each other by collecting data according to a predetermined time interval, and continuously checking whether the search queries may change over time. You can do it. For example, if the record is generated at an interval of one day, two days ago, a plurality of search sessions in which the search queries "Park Chan Ho" and "major league" are simultaneously received are recorded, and based on this, the related search query extraction system of yesterday. May have judged "Park Chan-ho" and "major league" as an associative search query, but yesterday's records rarely recorded search sessions that received both "Park Chan-ho" and "major league" at the same time. On the basis of this, it can be determined that "Park Chan-ho" and "major league" are not related search queries. Therefore, there is an advantage that can provide a user with an associated search query extracted based on the most recent data.

The association search query extracting system according to an embodiment of the present invention may map the search session or the received search query to a number in step 201 and generate the record using the mapped number. have.

FIG. 5 is a diagram showing an example of a record in which a search session and a search query are mapped to numbers in this embodiment. Referring to reference numeral 501 illustrated in FIG. 5, a number “56” is mapped to a search session identifier “sessionId1” compared to reference numeral 402 illustrated in FIG. 4, and “Park Chan Ho”, “Major”. It can be seen that the numbers "18759", "18760", and "18761" are mapped to the search query for "league" and "baseball."

When performing the steps according to the present invention using the data mapped to the number according to the present embodiment, compared to extracting the correlated search query using the data consisting of a string, when writing the data to the database Using less memory not only saves memory, but also makes processing easier compared to strings, resulting in faster processing speed.

According to an embodiment of the present invention, the associated search query extracting system may receive the received information during the specific search session and the specific search session when more than a predetermined number of search queries are received during the specific search session in step 201. A related search query extraction method is provided which does not include records relating to a search query in the database. In the present embodiment, the associated search query extraction system counts the number of search queries generated in one search session, and when the number of the search queries exceeds a predetermined number, the data received in the search session is in a general manner. Rather, it can be judged as data coming in unexpected ways. If the number of search queries received in one search session is too large, it is very unlikely that all of those search queries are related search queries, so you can extract more accurate related search queries by not writing this data to the database. To get

In step 202, the associated search query extraction system counts the total number of search sessions set per time interval with reference to the database to generate total search session number information. For example, if the time interval is one day, the associated search query extraction system may count the total number of search sessions established during the day.

In step 203, the associated search query extraction system generates a first search session number information by counting the number of search sessions in which the first search query has been received per time interval with reference to the database, In operation 204, the second search session number information is generated by counting the number of search sessions in which the second search query is received per time interval. For example, if the time interval is one day, the associated search query extraction system counts the number of search sessions in which a search query called "Park Chan-ho" was received during the day, and the search session in which a search query called "major league" was received. You can count the number of.

In step 205, the associated search query extraction system counts the number of search sessions in which both the first search query and the second search query have been received per time interval, referring to the database, to obtain third search session number information. Create For example, when the time interval is one day, the associated search query extraction system may count the number of search sessions in which both a search query called "Park Chan Ho" and a search query called "major league" are received during the day.

According to an embodiment of the present invention, the associated search query extraction system determines the number of the third search session only when the first search session number information and the second search session number information are greater than or equal to a predetermined number in step 205. An associated search query extraction method is provided which generates information. That is, in the present embodiment, the associated search query extraction system may not generate the third search session number information when the first search session number information or the second search session number information does not reach a predetermined number. have. If each search query receives too few search sessions, these search queries are very unlikely to be related search queries, so search queries that do not appear above a predetermined number are used to determine the associated search query. By not generating, it is possible to obtain an effect that greatly contributes to improving the performance of the associated search query extraction system.

According to one embodiment of the present invention, an associated search query extraction method using a hash-tree data structure in counting the number of search sessions is provided in step 205.

Hash-tree data structure is a kind of data structure used to store and find data. It is a method of finding the location of data by using a value obtained by processing a string to be searched by a specific function. Known as The hash-tree data structure has a characteristic that the number of data has little influence on the speed of finding data, which can not only locate data efficiently and quickly, but also greatly save memory in the system. can do.

6 is a diagram showing an example of a hash tree data structure used for counting the number of search sessions in this embodiment. FIG. 6 illustrates an example of counting the number of search sessions using a hash tree data structure when a search session including a search query composed of a search query "Dae Jang Geum" and a search query "Lee Young Ae" exists in the database. Is shown.

In step 206, the associated search query extraction system generates conditional probability information using the first search session number information and the third search session number information.

The conditional probability information may be used as one factor for evaluating correlation between search queries. For example, the association between a search query called "Dae Jang Geum" and a search query called "Lee Young Ae" can provide a single piece of probability information about the number of search sessions in which "Dae Jang Geum" appeared. Can be determined by using as a factor. In other words, if a large number of "Lee Young-ae" appeared in the search session that "Daejanggeum" appeared, it can be used as a factor that can have a strong influence in determining the relationship between the two search queries.

The following shows an example of a formula that can be used to generate the conditional probability information.

<Equation 1. Conditional Probability Information>

As shown in Equation 1, the "B" search query is divided by the probability that both "A" and "B" search queries appear in the same search session by the probability that the "B" search query appears in the search session. Probability information may also be generated for an "A" search query to appear in the appeared session.

In step 207, the association search query extraction system correlates using the total search session number information, the first search session number information, the second search session number information, and the third search session number information. ) Generate information.

The correlation information may be used as another factor for evaluating whether the search query is related to each other. In the case of a search query with a high frequency input from a user, the generated conditional probability information may be high even though there is no actual correlation. Therefore, the correlation information is used as another factor for evaluating the correlation between search queries. Doing so can greatly contribute to determining whether or not the association is more accurate.

In probability theory, there is an independence judgment, and the correlation information is used for the independence judgment. That is, when the correlation information has a value close to 1, it can be used as a strong factor that can be determined that both search queries are not related and independent, and when the correlation information has a value significantly larger than 1, it is used. It can be used as a strong factor to determine that both search queries are correlated.

The following shows an example of a formula that can be used to generate the correlation information.

Equation 2. Correlation Information

As shown in Equation 2, the probability that both "A" and "B" search queries appear in the same search session is determined by the probability that "A" search query appears in the search session and "B" search query in the search session. The correlation information may be generated by dividing the probability of appearance by the product of multiplication. In addition, when the formula is developed, the correlation information may be calculated by multiplying the total number of search sessions by the number of search sessions in which both the "A" search query and the "B" search query are present and the number of search sessions in which the "A" search query appears. It is calculated by dividing the number by the number of search sessions in which the "B" search query appeared. Accordingly, the associated search query extraction system may generate correlation information by using the total search session number information, the first search session number information, the second search session number information, and the third search session number information. It can be created.

In step 208, the association search query extraction system determines whether the first search query and the second search query are related based on the conditional probability information and the correlation information. For example, a predetermined index may be calculated by multiplying both pieces of information by a predetermined value, and determining whether the index is related by using the index. In addition, it is also possible to calculate a certain index by multiplying both pieces of information by a predetermined value and multiplying it again, and determining whether or not it is related using the index. In addition to the above method, it is apparent to those skilled in the art that various embodiments may exist in the method for determining the association using the conditional probability information and the correlation information.

According to an embodiment of the present invention, the related search query extraction system provides a related search query extraction method, characterized in that in step (208) determines the association only when the conditional probability information is a predetermined value or more. do.

In the present embodiment, if the conditional probability information is very low so that the conditional probability information falls below a predetermined value, the association search query does not determine whether or not the association by determining that the association is very low In addition, it is possible to obtain an effect of reducing unnecessary memory requirements and improving performance of the system.

According to an embodiment of the present invention, an association search query extraction method is provided, wherein the predetermined value in the embodiment is changed based on a predetermined function using the first search session number information as a variable. .

This is because if the first search session number information corresponds to a very low value, the conditional probability information may not be obtained normally, which may far exceed a fixed predetermined value. For example, if the number of search sessions in which both the "A" search query and the "B" search query appeared is "1" and the number of search sessions in which the "A" search query appeared is "5", the conditional probability information is "1". It is calculated as / 5 "so that a very high number can be recorded. In this case, although the degree of association between the actual "A" search query and the "B" search query is low, there is a possibility that the association between the two search queries is recognized, thereby providing the user with an incorrect related search query. Therefore, as in the present embodiment, it is necessary to vary the predetermined numerical value according to the first search session number information, thereby providing a more accurate related search query to the user.

According to an embodiment of the present invention, the predetermined function in the embodiment is a function having a percentage value obtained by dividing 100 percent ("1" as a value) by the square root of the first search session number information as a function value. An associated search query extraction method is provided. According to the present embodiment, the function may be expressed as "100 / √ (root) first search session number information".

According to another embodiment, the predetermined function may be expressed by the following general formula.

y (%) = a * x ^-b

<Equation 3. Function>

In Equation 3, y is the predetermined value, and x is the first search session number information. In addition, in Equation 3, the constant a and the degree b may be obtained through an experimental procedure for deriving the best equation for obtaining the numerical value. For example, a may be 80 and b may be 1.

In the above-described embodiments, the association search query extracting method according to an embodiment of the present invention, wherein the association is determined only when the conditional probability information is equal to or greater than a predetermined value. Since the predetermined number becomes higher as the number of search session information is lower, the related search query extraction system increases the number of cases in which the related search query extraction system does not determine the association, thereby reducing unnecessary memory requirements and speeding up the performance of the system. The effect of improving can be obtained.

According to an embodiment of the present invention, the related search query extraction system provides a related search query extraction method, characterized in that in step 208, the correlation information is determined only when the correlation information is more than a predetermined value. do.

In the present embodiment, the association search query extraction system is already between search queries when the correlation information corresponds to a value that is too low to fall below a predetermined value (for example, when it is almost close to "1"). By determining that the association is very low and not determining whether the association is very low, it is possible to obtain an effect of reducing unnecessary memory requirements and improving performance of the system.

According to the related search query extraction method according to an embodiment of the present invention, the related search query extraction system records when both search queries are determined to be related, and extracts a more accurate related search query by performing a toggle error check. This embodiment will be described below.

In general, a toggle key in a keyboard refers to a key capable of performing two or more functions with one key. Representative examples of the toggle key include "Insert", "Korean / English", "Caps Lock", "Num Lock", and "Scroll Lock".

As used herein, "toggle error check" relates to a check associated with a "Korean / English" conversion key of the toggle keys. For example, a user who wants to input "next" through the Korean keyboard may enter "ekdma" through the English keyboard according to the setting of the "Korean / English" conversion key. As shown in the above example, a user may frequently enter a Korean search query in English keyboard or an English search query in Korean keyboard according to the setting of the "Korean / English" conversion key. have. In this case, the user may re-enter the correct search query, which may cause the original search query and the search query due to the toggle error to appear in the same search session, so that the association between the two search queries may be recognized. In the case of the above example, there is a possibility that "ekdma" which is a toggle error of "next" is designated as an associated search query of "next".

If a search error is specified as an associative search query due to a toggle error, unnecessary memory is required to record it, which may cause a problem that the performance of the system is slowed down and is also inaccurate. A related search query may be provided to the user, thereby lowering the reliability of the service. However, the related search query extraction system according to the present embodiment can solve the above-mentioned problem by performing the above-described toggle error check.

If the association search query extraction system according to the present embodiment determines that the association result is determined in step 208, the association search query extraction system designates the first search query and the second search query as an association search query and records the association in the second database. Can be.

The association search query extraction system may perform a toggle error check on the first search query and the second search query by referring to the second database. According to an embodiment of the present invention, the association search query extraction system may perform the toggle error check using morphological analysis.

If the check result indicates that the first search query and the second search query are in a toggle error relationship, the associated search query extraction system may delete a record associated with the associated search query specification from the second database.

According to the related search query extraction method according to an embodiment of the present invention, the related search query extraction system may generate association index information between search queries and determine the association based on this, and provide the related search query list to the user. have.

FIG. 7 is a flowchart illustrating a process for providing a related search query list by using related index information. According to the related search query extraction method according to the present embodiment, step 208 may include step 701 and step 702.

In step 701, the association search query extraction system generates association index information between the first search query and the second search query using the conditional probability information and the correlation information. For example, the association index information may be generated by multiplying and adding both information to a predetermined value, or multiplying both information by a predetermined value and multiplying the information. In addition to the above method, it will be apparent to those skilled in the art that various embodiments may exist in the method for generating the correlation index information using the conditional probability information and the correlation information.

According to an embodiment of the present invention, the association search query extraction system generates first weight information associated with the conditional probability information and generates second weight information associated with the correlation information, thereby generating the first weight information and An association search query extraction method capable of generating the association index information using the second weight information is provided. In this case, the first weight information or the second weight information may vary according to a predetermined criterion.

According to the present exemplary embodiment, the influence of the conditional probability information on the association index information or the correlation information on the association index information can be appropriately adjusted according to a situation, thereby generating more accurate association index information. You can get the effect. In addition, the associated search query extraction system may vary the weight information. For example, when the first search session number information or the second search session number information is too low, the conditional probability information or the correlation information may become very large, in which case the associated search query extraction system may be By adjusting the influence of the conditional probability information or the correlation information, more accurate association index information can be generated.

In step 702, the association search query extraction system determines whether the association is based on the association index information. For example, when the association index information is equal to or greater than a predetermined value, the association between the two search queries may be recognized.

According to the related search query extraction method according to an embodiment of the present invention, the related search query extraction system may further perform steps 703 to 707.

In step 703, the association search query extraction system records the association index information in a second database in association with the first search query and the second search query when it is determined that the association is related.

8 is a diagram showing an example of a second database in which association index information is recorded in this embodiment. As shown in FIG. 8, association index information is recorded in the second database in association with a pair of search queries.

In step 704 the association search query extraction system receives a third search query from a user terminal and in step 705 extracts one or more fourth search queries associated with the third search query with reference to the second database. do.

In step 706, the related search query extraction system sorts the extracted fourth search query according to the related index information to generate a related search query list. For example, a method of generating the related search query list by sorting only the extracted search queries whose related index information is equal to or greater than a predetermined value may be used, or by sorting the search queries extracted in ascending or descending order of the related index information. You can use this method to generate a list of related search queries. In addition to the above generation method, it is apparent to those skilled in the art that various embodiments may exist in a method of generating a related search query list by sorting the search query extracted according to the related index information.

In step 707, the related search query extraction system provides the generated related search query list to the user terminal.

According to the present embodiment, the user may be provided with only a high relevance search query, and may also be provided with a search query in the order of high or low relevance. There is an advantage that can be found more quickly.

According to an embodiment of the present invention, the related search query extraction system generates relevant index information between search queries at predetermined time intervals and accumulates them in a predetermined database, and uses the accumulated recorded related index information to search the query. A method of extracting an association search query for determining whether an association between objects is provided.

FIG. 9 is a flowchart illustrating a process of determining whether an association is made using accumulated association index information. According to the present embodiment, step 208 may include steps 901 to 905.

In step 901, the association search query extraction system generates association index information between the first search query and the second search query every second time interval using the conditional probability information and the correlation information. In this case, the second time interval may be the same as or different from the predetermined time interval for generating records relating to the search session and the search query.

In step 902, the association search query extraction system accumulates the generated association index information in association with the first search query and the second search query in a second database at every second time interval.

FIG. 10 is a diagram showing an example of a second database in which cumulative correlation index information and its recording time point are recorded in this embodiment. As shown in FIG. 10, each recording time point at which the pair of search queries, cumulative correlation index information corresponding to the pair of search queries, and the cumulative correlation index information are recorded is recorded in the second database. Although the second time interval is “1 day” in FIG. 10, this is only an example and the association index information may be generated and recorded at various time intervals.

In step 903, the associated search query extraction system extracts first cumulative correlation index information and second cumulative correlation index information with reference to the second database, and in step 904, the first cumulative correlation index information and the The second association index information is generated using the second cumulative association index information. For example, the second correlation index information may be generated by multiplying and adding both information to a predetermined value, or multiplying both information by a predetermined value and multiplying the information. In addition to the above method, there may be various embodiments in the method of generating the second association index information using the first cumulative association index information and the second cumulative association index information. Self-explanatory

According to an embodiment of the present invention, the associated search query extraction system generates first weight information associated with the first cumulative correlation index information, and generates second weight information associated with the second cumulative correlation index information. An association search query extraction method capable of generating the second association index information using the first weight information and the second weight information is provided. For example, when the second time interval is "1 day", since the first cumulative correlation index information and the second cumulative correlation index information represent a state of one day as a result that occurs every day, the first cumulative correlation index information occurs. Chooses how to synthesize the result. For example, you can synthesize the results from today's seven days ago to today's results to generate the relevant index information as the final result.

According to the present embodiment, it is possible to provide a user with a search query that maintains a constant association over a long period of time, thereby obtaining an effect of satisfying a user's desire to receive a more accurate related search query service.

In the present exemplary embodiment, the first weight information is generated based on a time point at which the first cumulative correlation index information is recorded in the second database, and the second weight information is the second cumulative correlation index information. 2 can be generated based on the time points recorded in the database. When generating the weight information, a linear method may be used, or a non-linear method may be used.

According to the present embodiment, by generating weighted association index information by weighting more recently and recorded cumulative association index information, it is possible to provide a user with an association search query that reflects a more recent association relationship. The effect of satisfying the desire of the user who wants to receive the service can be obtained.

In step 905, the association search query extraction system determines whether the association is based on the second association index information. For example, when the second correlation index information is equal to or greater than a predetermined value, the correlation between the two search queries may be recognized.

According to an embodiment of the present invention, the related search query extraction system may determine and record the association between the search queries, and provide the related search query to the user by using the related search query extraction system. The associated search query extraction method according to the present embodiment may further include steps 209 to 212.

In step 209, the related search query extraction system determines that the first search query and the second search query are related search queries to the second database if it is determined that they are related. Record it.

In step 210 the association search query extraction system receives a third search query from a user terminal and in step 211 extracts one or more fourth search queries associated with the third search query with reference to the second database. do.

In step 212, the association search query extraction system provides the extracted fourth search query to the user terminal.

According to the related search query extraction method according to the present invention, through a series of processes as described above, it is possible to extract a truly meaningful related search query to provide a higher quality related search query service to the user. .

In addition, embodiments of the present invention include computer-readable media containing program instructions for performing various computer-implemented operations. The program recorded on the computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts.

Hereinafter, a related search query extraction system according to another embodiment of the present invention will be described. 11 is a block diagram illustrating an association search query extraction system according to the present embodiment.

The association search query extraction system 1100 according to the present embodiment includes a database 1101, a database management means 1102, a counter means 1103, conditional probability information generating means 1104, correlation information generating means 1105, and the like. And association means 1106.

The database 1101 includes a record relating to a search session and a search query received from a user terminal during the search session.

The search session may be established when a search window is first provided to the user terminal, and may end when there is no data transmission from the user terminal for a predetermined time. In addition, when a new search query is received from the user terminal after the termination of the search session, a new search session is started.

3 is a flowchart illustrating an embodiment of a process of maintaining the database 1101, and FIG. 4 is a diagram illustrating an example of a record included in the database 1101. Since FIG. 3 and FIG. 4 have already been described herein, a detailed description thereof will be omitted.

According to an embodiment of the present invention, the associated search query extraction system 1100 maps the search session or the received search query to numbers, and includes a database including records generated using the mapped numbers. 1110 may be maintained. 5 illustrates an example of a record in which a search session and a search query are mapped to numbers. Since FIG. 5 has already been described herein, a detailed description thereof will be omitted.

The database management means 1102 generates the record at predetermined time intervals and records the records in the database 1101.

The counter means 1103 generates a total search session number information by counting the total number of search sessions set per time interval, referring to the database 1101, and receives a first search query per time interval. Counting the number of search sessions to generate first search session number information, counting the number of search sessions for which a second search query was received per time interval, to generate second search session number information, and the time The third search session number information is generated by counting the number of search sessions in which the first search query and the second search query have been received per interval.

According to one embodiment of the invention, the counter means 1103 may use a hash-tree data structure in counting the number of search sessions.

Hash-tree data structure is a kind of data structure used to store and find data. It is a method of finding the location of data by using a value obtained by processing a string to be searched by a specific function. Known as 6 is a diagram showing an example of a hash tree data structure used for counting the number of search sessions in this embodiment. Since FIG. 6 has already been described herein, a detailed description thereof will be omitted.

The conditional probability information generating means 1104 generates conditional probability information using the first search session number information and the third search session number information.

Correlation information generating means 1105 is configured to correlate correlation information using the total search session number information, the first search session number information, the second search session number information, and the third search session number information. Create

The association determining unit 1106 determines whether the first search query and the second search query are related based on the conditional probability information and the correlation information.

Related search query extraction system 1120 according to another embodiment of the present invention may further include a predetermined recording means 1107. The recording means 1107 can record a predetermined function using the first search session number information as a variable.

In the present embodiment, the correlation determining unit 1106 determines the association only when the conditional probability information is equal to or greater than a predetermined value, and the numerical value is changed based on the function recorded in the recording unit 1107. Can be.

This embodiment is for the case where the conditional probability information is not normally obtained when the first search session number information corresponds to a very low value, and far exceeds a fixed predetermined value. For example, if the number of search sessions in which both the "A" search query and the "B" search query appeared is "1" and the number of search sessions in which the "A" search query appeared is "5", the conditional probability information is "1". It is calculated as / 5 "so that a very high number can be recorded. In this case, although the degree of association between the actual "A" search query and the "B" search query is low, there is a possibility that the association between the two search queries is recognized, thereby providing the user with an incorrect related search query. Therefore, as in the present embodiment, it is necessary to vary the predetermined numerical value according to the first search session number information, thereby providing a more accurate related search query to the user.

According to still another embodiment of the present invention, there is provided an association search query extraction system that can determine and record an association between search queries, and provide an association search query to a user using the same. The association search query extraction system according to the present embodiment may further include predetermined devices corresponding to the reference numeral 1130 in addition to the association search query extraction system 1100 according to the embodiment.

Reference numeral 1130 denotes a second database 1108, a second database management means 1109, a search query receiving means 1110, a search query extracting means 1111, and a search query providing means 1112.

The second database 1108 includes records related to associative search queries. 8 is a diagram illustrating an example of the second database 1108. Since FIG. 8 has already been described herein, a detailed description thereof will be omitted.

The second database management means 1109 designates the first search query and the second search query as an association search query when the association determination means 1106 determines that the association result is related, and thus the second database 1108. ).

The search query receiving means 1110 receives a third search query from the user terminal, and the search query extracting means 1111 refers to the second database 1108 to generate one or more fourth search queries associated with the third search query. Extract.

The search query providing unit 1112 provides the extracted fourth search query to the user terminal.

According to the related search query extraction system according to the present embodiment, through a series of processes as described above, it is possible to extract a truly meaningful related search query to provide a higher quality related search query service to the user. have.

12 is an internal block diagram of a general purpose computer device that may be employed to perform the associated search query extraction method in accordance with the present invention.

Computer device 1200 includes one or more processors 1210 coupled with a main memory device including random access memory (RAM) 1020 and read only memory (ROM) 1230. The processor 1210 is also called a central processing unit (CPU). As is well known in the art, the ROM 1230 serves to transfer data and instructions to the CPU unidirectionally, and the RAM 1220 typically transfers data and instructions bidirectionally. Used to. RAM 1220 and ROM 1230 may include any suitable form of computer readable media. Mass storage 1240 is bi-directionally coupled to processor 1210 to provide additional data storage capabilities, and may be any of the computer readable recording media described above. The mass storage device 1240 is used to store programs, data, and the like, and is a secondary memory device such as a hard disk which is generally slower than the main memory device. Certain mass storage devices such as CD ROM 1260 may also be used. The processor 1210 may include one or more input / output interfaces, such as a video monitor, trackball, mouse, keyboard, microphone, touchscreen display, card reader, magnetic or paper tape reader, voice or handwriting reader, joystick, or other known computer input / output device. 1250 is connected. Finally, the processor 1210 may be connected to a wired or wireless communication network through the network interface 1270. Through this network connection, the procedure of the method described above can be performed. The apparatus and tools described above are well known to those skilled in the computer hardware and software arts.

The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention.

As described above, although the present invention has been described with reference to limited embodiments and drawings, the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains various modifications and variations from such descriptions. This is possible.

Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the claims below, but also by the equivalents of the claims.

According to the related search query extraction method and system according to the present invention, by building a system that can automatically determine whether or not the association between the search query by effectively collecting and analyzing data about the search query received from the user, one search query By sorting and storing other search queries that are related to, you can reduce the time and economic loss caused to service operators.

In addition, according to the related search query extraction method and system according to the present invention, by maintaining a database that systematically records the association index information between the search query, the degree of association by using the association index information when the search query is input from the user The effect of providing a higher search query to the user first can be obtained.

In addition, according to the method and system for extracting the relevant search query according to the present invention, by extracting only the useful data that has undergone a systematic preprocessing process among the search data input from the users to maintain an appropriate number of related search queries, a truly meaningful related search By extracting the query, it is possible to provide the user with a higher quality related search query service.

In addition, according to the method and system for extracting the relevant search query according to the present invention, by combining the accumulated data for a certain period of time to determine whether the search query is related, providing a search query that maintains a constant association for a long time to the user In this way, the user's desire to receive a more accurate associated search query service can be obtained.

1 is a diagram illustrating a network connection of an association search query extraction system according to an embodiment of the present invention.

2 is a flowchart illustrating a method of extracting a related search query according to an embodiment of the present invention.

3 is a flowchart illustrating a process of maintaining a database according to an embodiment of the present invention.

4 is a diagram illustrating an example of a record included in a database according to one embodiment of the present invention.

5 illustrates an example of a record in which a search session and a search query are mapped to numbers according to an embodiment of the present invention.

FIG. 6 illustrates an example of a hash tree data structure used to count the number of search sessions according to an embodiment of the present invention.

FIG. 7 is a flowchart illustrating a process for providing a related search query list using related index information according to an embodiment of the present invention.

8 is a diagram illustrating an example of a second database in which association index information is recorded according to an embodiment of the present invention.

9 is a flowchart illustrating a process of determining whether an association is made using accumulated association index information according to an embodiment of the present invention.

FIG. 10 illustrates an example of a second database in which cumulative correlation index information and a recording time point are recorded according to an embodiment of the present invention.

11 is a block diagram illustrating an association search query extraction system according to another embodiment of the present invention.

12 is an internal block diagram of a general purpose computer device that may be employed to perform the associated search query extraction method in accordance with the present invention.

<Explanation of symbols for the main parts of the drawings>

1100: Association Search Query Extraction System

1101: database 1102: database management means

1103: counter means 1104: conditional probability information generating means

1105: correlation information generating means 1106: association determination means

Claims (27)

delete delete In a method of extracting a correlated search query, Maintaining a database comprising a search session and a record relating to a search query received from a user terminal during the search session, wherein the record is generated and recorded in the database at predetermined time intervals; Generating total search session number information by counting the total number of search sessions set per time interval with reference to the database; Generating first search session number information by counting the number of search sessions in which the first search query has been received per time interval with reference to the database; Generating second search session number information by counting the number of search sessions for which a second search query was received per time interval, with reference to the database; Generating third search session number information by counting the number of search sessions in which the first search query and the second search query have been received per time interval with reference to the database; Generating conditional probability information using the first search session number information and the third search session number information; Generating correlation information using the total search session number information, the first search session number information, the second search session number information, and the third search session number information; And Determining an association between the first search query and the second search query based on the conditional probability information and the correlation information. Including, The determining of the association between the first search query and the second search query based on the conditional probability information and the correlation information may include determining whether the association occurs when the conditional probability information is greater than or equal to a predetermined value. The numerical value is changed based on a predetermined function that decreases as the first search session number information increases. The search session is established when a search window is provided to the user terminal, terminates when there is no data transmission from the user terminal for a predetermined time, and when a new search query is received from the user terminal after the termination of the search session. Starting a search session Associated search query extraction method characterized in that. The method of claim 3, The step of maintaining a database comprising a search session and a record relating to a search query received from a user terminal during the search session, includes: Generating and recording a first search session identifier associated with a first search session in the database; Transmitting first time information about the first search session identifier and the last search time to a user terminal; Receiving a search query from the user terminal; Comparing the first visual information with the second visual information from which the search query was received; And As a result of the comparison, when the gap between the first time information and the second time information exceeds a predetermined time, a second search session identifier associated with a second search session is generated to generate the second search session identifier and the received search. Recording a record relating to the query in the database and recording in the database a record relating to the received search query in association with the first search session identifier if the gap is less than a predetermined time Association search query extraction method characterized in that it comprises a. The method of claim 3, The step of maintaining a database comprising a record relating to the search session and a search query received from a user terminal during the search session, Mapping the search session or the received search query to a number; And Generating the record using the mapped number Association search query extraction method characterized in that it comprises a. The method of claim 3, The step of maintaining a database comprising a record relating to the search session and a search query received from a user terminal during the search session, If more than a predetermined number of search queries are received during a particular search session, records relating to the specific search session and search queries received during the specific search session are not included in the database. . The method of claim 3, The step of generating third search session number information by counting the number of search sessions in which the first search query and the second search query have been received per time interval with reference to the database, And generating the third search session number information only when the first search session number information and the second search session number information are more than a predetermined number. delete delete The method of claim 3, The determining of the association between the first search query and the second search query based on the conditional probability information and the correlation information may include: An association search query extraction method, wherein the association is determined only when the correlation information is equal to or greater than a predetermined value. The method of claim 3, If it is determined that the association result is related, designating the first search query and the second search query as an association search query and recording the same in a second database; Performing a toggle error check on the first search query and the second search query by referring to the second database; And Deleting a record associated with the associated search query designation from the second database if the first search query and the second search query are in a toggle error relationship as a result of the checking. Association search query extraction method characterized in that it further comprises. The method of claim 3, The determining of the association between the first search query and the second search query based on the conditional probability information and the correlation information may include: Generating correlation index information between the first search query and the second search query using the conditional probability information and the correlation information; And Determining whether the association is based on the association index information; Association search query extraction method characterized in that it comprises a. The method of claim 12, If it is determined that the association result is related, recording the association index information in a second database in association with the first search query and the second search query; Receiving a third search query from a user terminal; Extracting one or more fourth search queries associated with the third search query with reference to the second database; Generating a related search query list by sorting the extracted fourth search query according to the related index information; And Providing the generated related search query list to the user terminal; Association search query extraction method characterized in that it further comprises. The method of claim 12, The generating of the correlation index information between the first search query and the second search query based on the conditional probability information and the correlation information may include: Generating first weight information associated with the conditional probability information; Generating second weight information associated with the correlation information; And Generating the association index information using the first weight information and the second weight information. Association search query extraction method characterized in that it comprises a. The method of claim 14, And the first weight information or the second weight information is changed according to a predetermined criterion. The method of claim 3, The determining of the association between the first search query and the second search query based on the conditional probability information and the correlation information may include: Generating correlation index information between the first search query and the second search query every second time interval using the conditional probability information and the correlation information; Accumulating and recording the generated correlation index information in a second database at each second time interval in association with the first search query and the second search query; Extracting first cumulative correlation index information and second cumulative correlation index information with reference to the second database; Generating second association index information by using the first cumulative association index information and the second cumulative association index information; And Determining whether the association is based on the second association index information; Association search query extraction method characterized in that it comprises a. The method of claim 16, The generating of the second association index information by using the first cumulative association index information and the second cumulative association index information may include: Generating first weight information associated with the first cumulative correlation index information; Generating second weight information associated with the second cumulative correlation index information; And Generating the second association index information using the first weight information and the second weight information. Including, The first weight information is generated based on a time point when the first cumulative correlation index information is recorded in the second database, and the second weight information is a time point when the second cumulative correlation index information is recorded in the second database. The associated search query extraction method, characterized in that it is generated based on. In a method of extracting a correlated search query, Maintaining a database comprising a search session and a record relating to a search query received from a user terminal during the search session, wherein the record is generated and recorded in the database at predetermined time intervals; Generating first search session number information by counting the number of search sessions in which the first search query has been received per time interval with reference to the database; Generating second search session number information by counting the number of search sessions in which the first search query and the second search query have been received per time interval with reference to the database; Generating conditional probability information using the first search session number information and the second search session number information; And Determining whether the first search query is related to the second search query based on the conditional probability information; Including, The determining of the association between the first search query and the second search query based on the conditional probability information may include determining whether the association is performed when the conditional probability information is equal to or greater than a predetermined value. The numerical value is changed based on a predetermined function that decreases as the first search session number information increases. The search session is established when a search window is provided to the user terminal, terminates when there is no data transmission from the user terminal for a predetermined time, and when a new search query is received from the user terminal after the termination of the search session. Starting a search session Associated search query extraction method characterized in that. delete delete The method according to claim 3 or 18, wherein If it is determined that the association result is related, designating the first search query and the second search query as an association search query and recording the same in a second database; Receiving a third search query from a user terminal; Extracting one or more fourth search queries associated with the third search query with reference to the second database; And Providing the extracted fourth search query to the user terminal; Association search query extraction method characterized in that it further comprises. 19. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 3-7 and 10-18. In a system for extracting correlated search queries, A database comprising a search session and a record relating to a search query received from a user terminal during the search session; Database management means for generating the record at predetermined time intervals and recording the record in the database; Referring to the database, the total number of search sessions set per the time interval is counted to generate total search session number information, and the number of search sessions for which the first search query is received per time interval is counted. Generate first search session number information, generate second search session number information by counting the number of search sessions in which the second search query has been received per time interval, and generate the first search query and Counter means for counting the number of search sessions for which the second search query was received to generate third search session number information; Conditional probability information generating means for generating conditional probability information using the first search session number information and the third search session number information; Correlation information generating means for generating correlation information using the total search session number information, the first search session number information, the second search session number information, and the third search session number information; And Associating means for determining association between the first search query and the second search query based on the conditional probability information and the correlation information Including, The association determining unit determines whether the association is performed when the conditional probability information is equal to or greater than a predetermined value; The numerical value is changed based on a predetermined function that decreases as the first search session number information increases. The search session is established when a search window is provided to the user terminal, terminates when there is no data transmission from the user terminal for a predetermined time, and when a new search query is received from the user terminal after the termination of the search session. Starting a search session Association search query extraction system, characterized in that. In a system for extracting correlated search queries, A database comprising a search session and a record relating to a search query received from a user terminal during the search session; Database management means for generating the record at predetermined time intervals and recording the record in the database; Referring to the database, the number of search sessions in which the first search query is received per time interval is counted to generate first search session number information, and the first search query and the second search per time interval. Counter means for counting the number of search sessions for which a query was received to generate second search session number information; Conditional probability information generating means for generating conditional probability information using the first search session number information and the second search session number information; And Association determination means for determining association between the first search query and the second search query based on the conditional probability information; Including, The association determining unit determines whether the association is performed when the conditional probability information is equal to or greater than a predetermined value; The numerical value is changed based on a predetermined function that decreases as the first search session number information increases. The search session is established when a search window is provided to the user terminal, terminates when there is no data transmission from the user terminal for a predetermined time, and when a new search query is received from the user terminal after the termination of the search session. Starting a search session Association search query extraction system, characterized in that. delete delete The method of claim 23 or 24, A second database containing records relating to the associated search query; Second database management means for designating the first search query and the second search query as an associated search query and recording the result in the second database when it is determined that the association is related; Search query receiving means for receiving a third search query from a user terminal; Search query extracting means for extracting one or more fourth search queries associated with the third search query with reference to the second database; And Search query providing means for providing the extracted fourth search query to the user terminal Association search query extraction system further comprises.