JP2010176198A - Retrieval server, information retrieval method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a retrieval result in order of priority in consideration of trend with fine accuracy. <P>SOLUTION: A retrieval server 100 calculates effective periods of contents, based on content types such as MIME types and the number of content update times retained in content preservation servers 300-1 through 300-n. By the calculation of a weight coefficient based on each effective period and the lapse of time which is time elapsing after the last retrieval of a content, the ritrieval server 100 calculates a content retrieval sequence, based on the weight coefficient and the number of click times by which the content is retrieved. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a search server, an information search method, and a program for searching for information.

  In recent years, the amount of information handled in a company has increased, and there is a tendency that structured data and unstructured data of the information are scattered in a plurality of servers. In addition, there is a growing need to search for such information across the board for information utilization and operational efficiency.

  One of the important elements when searching for information across the board is that the desired information is searched for in the higher rank of the search results. As for the search for documents in the company, an effective ranking algorithm unique to the search in the company (Enterprise Search) different from the search on the Internet has been demanded.

  In-house search, unlike Internet search, has the characteristic of searching across various forms of structured data and unstructured data such as databases, file servers, and business applications. Therefore, it is necessary to apply a ranking algorithm specialized for the characteristics of in-house search, not the link algorithm well-known in Internet search (an algorithm that determines the search order for the search keyword based on the number of links placed on the site). There is.

  A click feedback method is considered as one of ranking algorithms in in-house search (see, for example, Patent Documents 1 and 2). This is based on the idea that content that is viewed many times by many people in company information (clicked on search results) is more important. This algorithm is reflected in the score of subsequent search results.

  For example, if there is a personnel event or the like that is targeted for everyone in the company, a search request for contents such as forms necessary for the event will increase. At this time, the click feedback method is expected to have an effect that a lot of employees refer to the form so that the form is ranked higher in the search result. As a supplement, this click feedback method plays a role like a weighting factor in consideration of the above consideration for the search rank calculated mainly by the existing scoring engine, rather than calculating the score alone.

  In addition to the click feedback method, various similar methods have been conceived that try to use (feedback) past user search operations for the next search.

  For example, a method is considered in which a large amount of information such as a searcher's physical location information and network information is stored in a history, and how to display the search results from various viewpoints (views) in one search at the next search. (For example, see Patent Document 3).

  In addition, a technique is considered in which a final ranking is determined in consideration of the concept of “user preference” after determining a temporary ranking using a search keyword (see, for example, Patent Documents 4 and 5).

  In addition to the ranking based on the search keyword as in other methods, a technique is considered in which a profile is prepared for each user and the compatibility with the profile is reflected in the ranking (see, for example, Patent Document 6). .)

  Further, a technique is considered in which the search keyword group closest to the present from the search keyword history information is fed back to the score calculation for the next and subsequent times based on the concept of topic (for example, see Patent Document 7).

  In addition, a technique for calculating a higher priority of search keywords during a preset effective period has been considered (see, for example, Patent Document 8).

JP-A-10-307845 JP 2006-331295 A Japanese Patent Laid-Open No. 08-044759 JP 2006-092557 A JP 2006-244212 A JP 2002-032401 A JP 2007-140605 A JP 2007-188508 A

  However, the above-described click feedback method does not consider the trend toward content. Therefore, unless the information on the number of clicks held is cleared, content that has been popular in the past will continue to be ranked higher even if there is no reference thereafter.

  In addition, the technique described in Patent Document 3 has a problem in that the trend of content cannot be reflected in the search result as in the above-described problem.

  In Patent Documents 4 and 5, “user preference” is a general-purpose and conceptual description, and the user must define it in advance, which takes time. is there.

  Moreover, in the technique described in Patent Document 6, it is necessary to determine a profile for each user in advance, which is troublesome.

  Further, in the technique described in Patent Document 7, it is not known from the history information of the search keyword which content is actually referred to. It's not technology.

  Further, in the technique described in Patent Document 8, only whether or not it is in the effective period is used as a priority criterion, and therefore, there is a problem in that the priority cannot be calculated with fine accuracy. is there.

  An object of this invention is to provide the search server, the information search method, and program which solve the subject mentioned above.

The search server of the present invention
A search server that searches for content based on a search keyword transmitted from a terminal,
A content validity period calculation processing unit that calculates the validity period of the content based on the type of the content such as MIME type and the number of updates of the content;
A click score update processing unit that calculates a weighting factor based on the effective period and an elapsed time that is an elapsed time from when the content was last searched;
A click score reflection processing unit for calculating a search order of the content based on the weighting factor and the number of clicks for searching the content;

Moreover, the information retrieval method of the present invention includes:
An information search method for searching content based on a search keyword transmitted from a terminal,
A process for calculating the validity period of the content based on the type of the content such as MIME type and the number of updates of the content;
A process of calculating a weighting factor based on the effective period and an elapsed time that is an elapsed time from when the content was last searched;
Processing for calculating the search order of the content based on the weighting factor and the number of clicks on which the content was searched.

The program of the present invention is
A program for causing a search server to search for content based on a search keyword transmitted from a terminal,
A procedure for calculating the validity period of the content based on the type of the content such as MIME type and the number of updates of the content;
A procedure for calculating a weighting factor based on the validity period and an elapsed time that is an elapsed time since the content was last searched;
Based on the weighting factor and the number of clicks on which the content has been searched, a procedure for calculating the search order of the content is executed.

  As described above, in the present invention, the effective period of the content is calculated based on the content type such as the MIME type and the number of times the content is updated, and the effective period and the time elapsed since the last time the content was searched. The weighting factor is calculated based on a certain elapsed time, and the content search order is calculated based on the weighting factor and the number of clicks on which the content has been searched. Can provide search results.

It is a figure which shows one Embodiment of the search server of this invention. It is a figure which shows an example of the update frequency of each content and MIME type which the content effective period calculation process part shown in FIG. 1 confirmed. It is a figure which shows an example of the content effective period information calculated about the content shown in FIG. 2, and stored in the content effective period information database shown in FIG. It is a figure which shows an example of the click history information regarding the 1st content stored in the click history information database shown in FIG. It is a figure which shows an example of the click history information regarding the 2nd content stored in the click history information database shown in FIG. It is a figure which shows an example of the click history information regarding the 3rd content stored in the click history information database shown in FIG. It is a figure which shows an example of the click score information calculated based on the click history information shown in FIGS. 4-6, and was stored in the click score information database shown in FIG. It is a figure which shows an example of the score of the temporary ranking determined by the temporary ranking determination part shown in FIG. It is a figure which shows an example of the final score of the content calculated in the click score reflection process part shown in FIG. It is a figure which shows other embodiment of the search server of this invention.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a diagram showing an embodiment of a search server according to the present invention.

  As shown in FIG. 1, the present embodiment includes a search server 100, a terminal 200, and content storage servers 300-1 to 300-n (n is an integer), which are connected to each other via a network 400. Has been.

  The terminal 200 transmits the search keyword input by the client as a search request to the search server 100 via the network 400.

  The search server 100 searches for content based on the search keyword transmitted from the terminal 200.

  The content storage servers 300-1 to 300-n are servers that store content to be searched.

  Further, as shown in FIG. 1, the search server 100 includes a client request reception / return processing unit 104, a temporary ranking determination unit 105, a click score reflection processing unit 106, a click feedback reception processing unit 107, a content validity period. A calculation processing unit 108, a click score update processing unit 109, a click score information database 110, a click history information database 111, and a content valid period information database 112 are provided.

  Here, the temporary ranking determining unit 105 can use a general-purpose scoring engine, and information necessary for determining the temporary ranking is already stored in the search server 100.

  In addition, the content validity period calculation processing unit 108 confirms the number of times each content is updated and the MIME type that is the file type to the content storage servers 300-1 to 300-n. Therefore, it is necessary to grasp the number of times content is updated. This is because means for introducing a document management system capable of managing the number of updates to the content storage servers 300-1 to 300-n, and the content storage servers 300-1 to 300-n are used as content valid periods each time content is updated This can be realized by using a means for notifying the calculation processing unit 108 and keeping the number of updates in the search server 100.

  Further, the content valid period calculation processing unit 108 calculates the content valid period from the number of times each content is updated and the MIME type so that a longer period is set for a useful content that does not change much in the long term. Calculate This specific calculation method will be described later. Then, content validity period information indicating the calculated validity period of the content is stored in the content validity period information database 112. At this time, if the content valid period information of the same content is already set in the content valid period information database 112, it is overwritten. This processing needs to be performed before searching for content, and is executed periodically.

  In this embodiment, a method for setting the content validity period from the number of updates and the MIME type is described. However, the present invention is not limited to this method, and it is likely to affect the validity period of the content such as the content access protocol and server type. If there exists, an equivalent calculation method may be used.

  Next, search processing using search keywords will be described.

  At the time of search, the terminal 200 transmits the search keyword received from the client to the client request reception / return processing unit 104.

  Then, the client request reception / return processing unit 104 passes the received search keyword to the temporary ranking determination unit 105, and the temporary ranking determination unit 105 calculates a temporary ranking score.

  Subsequently, the temporary ranking score calculated by the temporary ranking determination unit 105 is transferred to the click score reflection processing unit 106, and the click score reflection processing unit 106 recalculates the score considering the trend of the content. This specific calculation method will be described later.

  Then, the click score reflection processing unit 106 passes the recalculated score to the client request reception / return processing unit 104, and the client request reception / return processing unit 104 returns the search result to the terminal 200.

  When the client operating the terminal 200 that has received the search result then selects (clicks) the content to be browsed from the search results displayed on the terminal 200, the content is stored in the content storage servers 300-1 to 300-n. You can browse the content that you have. At that time, the terminal 200 transmits the clicked content / date information to the click feedback reception processing unit 107.

  The click feedback reception processing unit 107 stores information transmitted from the terminal 200 in the click history information database 111 as click history information.

  Then, the click score update processing unit 109 acquires the content effective period information stored in the content effective period information database 112 and discards information that exceeds the effective period in the click history information database 111.

  Then, the click score update processing unit 109 calculates a weight coefficient for the click history information stored in the click history information database 111 so that the click history information closer to the search date is reflected in the score. The calculated weighting coefficient is stored in the click score information database 110. This specific calculation method will be described later. In addition, it is preferable that this processing be performed automatically at least once at the beginning of the day, aiming at a time period when the client does not search using the terminal 200 (executed at night when the search is not executed). The setting depends on the operation.)

  Hereinafter, the processing in the form shown in FIG. 1 will be described in detail. In the following description, for convenience, the date and time are described in units of days. However, the present invention can be applied without depending on units of time such as minutes or seconds. Similarly, the features used for calculating the effective period of the content have been described using the number of updates and the MIME type, but the present invention can be applied to other features.

  First, as search pre-processing, the content validity period calculation processing unit 108 confirms the number of updates of each content and the MIME type to the content storage servers 300-1 to 300-n. Then, the effective period of each content is calculated, and the calculated content effective period is stored in the content effective period information database 112 as content effective period information.

  FIG. 2 is a diagram illustrating an example of the number of updates of each content and the MIME type confirmed by the content validity period calculation processing unit 108 illustrated in FIG.

  As shown in FIG. 2, for example, confirm that the first content (http: //aaa/bbb/abab.doc) has been updated 7 times and the MIME type is “application / msword”. Yes. Further, it is confirmed that the second content (smb: //xyz/abc/zzz/yyyy.xls) has been updated 50 times and the MIME type is “application / msexcel”. Confirm that the third content (file: /// C: /aaa/bbb/ccc/abc.pdf) has been updated 950 times and the MIME type is “application / pdf”. Yes.

  FIG. 3 is a diagram showing an example of content valid period information calculated for the content shown in FIG. 2 and stored in the content valid period information database 112 shown in FIG.

  As shown in FIG. 3, the content validity period information database 112 stores the validity periods calculated for the three contents. These are calculated using (Equation 1).

ここで、「ｍａｘ（更新回数）」は、各コンテンツの更新回数の中の最大値を意味する。また、「ＭＩＭＥタイプ別更新平均回数」は、ＭＩＭＥタイプ単位で更新回数の統計をとって平均した値である。また、「ｍａｘ（ＭＩＭＥタイプ別更新平均回数）」は、ＭＩＭＥタイプ別更新平均回数の中の最大値を意味する。また、ｘ，ｙは、更新回数、ＭＩＭＥタイプ別更新平均回数に対する重みで、ｘ＋ｙ＝１となる数値である。また、「基準期間」は、有効期間を計算する上であらかじめ設定された基準を示す期間であり、システム上で適切な値を１つ設けるものである（コンテンツ毎に設定しなければならないものではない）。

Here, “max (update count)” means the maximum value among the update counts of each content. The “average number of updates by MIME type” is a value obtained by averaging statistics of the number of updates for each MIME type. Further, “max (MIME type update average number of times)” means the maximum value among MIME type update average times. Also, x and y are weights for the number of updates and the average number of updates for each MIME type, and are numerical values such that x + y = 1. In addition, the “reference period” is a period indicating a standard set in advance for calculating the effective period, and one appropriate value is provided on the system (in the case where it must be set for each content) Absent).

  As shown in FIG. 3, for example, the validity period of the first content (http: //aaa/bbb/abab.doc) is 950 days, whereas the third content (file: /// C: /aaa/bbb/ccc/abc.pdf) is valid for 7 days. That is, the third content is highly trending content that considers only the click history information for seven days compared to the first content. On the other hand, the first content is considered to be valid content because the number of updates is small and the content is unchanged, and the click history information referred to in the past is also effective.

  Further, as another pre-processing, at least once at the beginning of the day, “elapsed days”, “weighting coefficient”, “click count” indicating the elapsed time of the click history information stored in the click history information database 111 The click score update processing unit 109 calculates the “correction value” using (Expression 2), (Expression 3), and (Expression 4). Further, the click score update processing unit 109 calculates “click score” of the click score information stored in the click score information database 110 using (Equation 5).

  If the frequency of performing this calculation is increased, more recent click history information can be taken into account, and the accuracy of the search result is improved. On the other hand, since the database table is locked during these calculations, the search cannot be performed. Therefore, it is necessary to set an appropriate calculation timing according to the operation. Note that data whose “weighting factor” is “0” during the following processing is information whose valid period has expired, and is discarded from the click history information database 111.

Elapsed days = Current date (search date)-Click date (Formula 2)
Weighting factor = (Content valid period−Elapsed days) / Content valid period (Formula 3)
Click count correction value = Click count x Weighting factor (Formula 4)
Click score = Σ click count correction value (Formula 5)
Here, when the calculation result of (Expression 3) becomes a negative value, the weighting coefficient is set to “0”.

  FIG. 4 is a diagram showing an example of click history information related to the first content stored in the click history information database 111 shown in FIG.

  FIG. 5 is a diagram illustrating an example of click history information relating to the second content stored in the click history information database 111 illustrated in FIG. 1.

  FIG. 6 is a diagram showing an example of click history information regarding the third content stored in the click history information database 111 shown in FIG.

  FIG. 7 is a diagram showing an example of click score information calculated based on the click history information shown in FIGS. 4 to 6 and stored in the click score information database 110 shown in FIG.

  In the example shown in FIGS. 2 to 7, the first content is often referenced several years ago, but has not been frequently referenced recently, and the third content has been frequently referenced recently. The content is the content, and the second content is the content that is continuously referred to. Looking at the calculation results, it can be seen that the click score has a larger value as the content is referenced more recently.

  After that, when the client request reception / return processing unit 104 receives a search request from the terminal 200, the temporary ranking determination unit 105 calculates a temporary ranking score. The calculation method in the provisional ranking determination unit 105 is arbitrary, and the details are omitted because it is not included in the appeal range in the present invention.

  FIG. 8 is a diagram illustrating an example of a temporary ranking score determined by the temporary ranking determination unit 105 illustrated in FIG. 1.

  As shown in FIG. 8, the provisional ranking score (provisional score) increases in the order of the first, second, and third contents. In the case of a normal search to which the present invention is not applied, this order is returned to the terminal 200 as a search result as it is.

  Subsequently, the click score reflection processing unit 106 calculates a final score reflecting the click score (content trend) using (Equation 6), and returns the search result to the terminal 200 in the order of the final score.

Final score = provisional score × (click score / max (click score of all contents)) (Formula 6)
Here, “max (click score of all contents)” means the maximum value among the click scores of each content.

  FIG. 9 is a diagram illustrating an example of the final score of content calculated in the click score reflection processing unit 106 illustrated in FIG.

  As shown in FIG. 9, the first content having the largest value in the provisional score (provisional search order) is the content that has not been referred to recently (low trend), and the final score (search order). Has the smallest value. In addition, it can be seen that the third content having the smallest value in the provisional score is the content that is frequently referred to recently (having a high trend), and thus has the highest final score.

  The terminal 200 that has received the search result displays the search result. When the content to be browsed is selected (clicked) by the client, the contents of the content storage servers 300-1 to 300-n are referred to. At this time, the selected click history information (content / date) is transmitted to the click feedback reception processing unit 107 of the search server 100. Then, the click feedback reception processing unit 107 updates the click history information database 111 based on the transmitted click history information. At this time, “1” is added to the number of clicks of the content stored in the click history information database 111.

  In the form shown in FIG. 1, the score is calculated based on the information selected (clicked) on the search result. However, even if the information to be used has more versatility, the score of the present invention is similarly applied. Techniques can be applied.

  FIG. 10 is a diagram showing another embodiment of the search server of the present invention.

  In this embodiment, as shown in FIG. 10, an access log management reception processing unit 113 is used instead of the click feedback reception processing unit 107 in the configuration shown in FIG. 1, and an access log information database 114 is used instead of the click history information database 111. Is a form provided in the search server 100.

  As described above, the access log information is used instead of the click history information for the feature to be used. In the case of an access log, an access history other than reference information with respect to a search result (reference not through a search such as normal Web access or FTP (File Transfer Protocol) access) can be reflected in the search order, and is more general. .

  The access log information stored in the access log information database 114 is information managed by a normal Web server or the like. Therefore, the premise of the present invention is that each content storage server 300-1 to 300-n already holds an access log, and the access log management reception processing unit 113 performs the access log as a pre-processing of the search operation. The access log information stored in the information database 114 is updated. Here, the present invention can be applied in a similar manner by replacing information such that the click date of the click history information shown in FIGS. 4 to 6 is the access date and the click count is the access count. .

  Note that the processing of the search server 100 described above may be performed by a logic circuit manufactured according to the purpose. Further, a program in which processing contents are described as a procedure may be recorded on a recording medium readable by the search server 100, and the program recorded on the recording medium may be read by the search server 100 and executed. The recording medium readable by the search server 100 is a removable recording medium such as a floppy disk (registered trademark), a magneto-optical disk, a DVD, and a CD, and a memory such as a ROM and a RAM built in the search server 100. And HDD. The program recorded on the recording medium is read by a CPU (not shown) in the search server 100, and the same processing as described above is performed under the control of the CPU. Here, the CPU operates as a computer that executes a program read from a recording medium on which the program is recorded.

  Further, differences and advantages between the technique described in the above-described patent document and the present invention will be described below.

  The technique described in Patent Document 3 and the present invention are similar in that the history information of past search operations is fed back, but the feedback target is not a search ranking score. The present invention, which is set and considering the concept of time from the search date and time, effectively feeds back history information.

  In addition, the technique described in Patent Documents 4 and 5 and the present invention have a process flow in which a final ranking is determined in consideration of the concept of “user preference” after determining a temporary ranking based on a search keyword. It is similar. However, in the techniques described in Patent Documents 4 and 5, the “user preference” is defined in advance by the user, whereas in the present invention, it is automatically and dynamically changed. The point is greatly different.

  In the technique described in Patent Document 6, it is necessary to determine a profile for each user in advance. On the other hand, in the present invention, dynamically fed back information does not change. The point is very different.

  In general, the above-described similar method is an approach of personalizing search results for each user in consideration of the user's preference / preference, whereas the present invention utilizes click history information for all users. The approach is to reflect the trend of content.

  Further, the technique described in Patent Document 7 pays attention to whether content is referred to within a certain period as in the present invention. In addition, past search keywords are recorded and categorized in units of content, and the keyword group closest to the present is fed back to the score calculation from the next time on the concept of topic. Are similar. However, the history information / score calculation method is greatly different from the present invention. Since it is not known from the search keyword history information which content is actually referred to, the present invention is more useful in the case of a search for a need to search for a content that has been frequently referenced recently.

  The present invention described above has the following effects.

  The first effect is that a search result is obtained such that content frequently referred to in the past is ranked higher.

  The reason is that the date and number of times referred to in the past in content units are stored in the database, and this information is reflected in the score at the time of search.

  The second effect is that a search result is obtained in which not only the number of times of reference is high but also the trend of content is taken into consideration.

  The reason is that the effective period is automatically calculated for each content unit, and the more recent reference times during that period are weighted and reflected in the score.

DESCRIPTION OF SYMBOLS 100 Search server 104 Client request reception / return processing part 105 Temporary ranking determination part 106 Click score reflection processing part 107 Click feedback reception processing part 108 Content effective period calculation processing part 109 Click score update processing part 110 Click score information database 111 Click history information Database 112 Content validity period information database 113 Access log management reception processing unit 114 Access log information database 200 Terminal 300-1 to 300-n Content storage server 400 Network

Claims (12)

A search server that searches for content based on a search keyword transmitted from a terminal,
A content validity period calculation processing unit that calculates the validity period of the content based on the type of the content such as MIME type and the number of updates of the content;
A click score update processing unit that calculates a weighting factor based on the effective period and an elapsed time that is an elapsed time from when the content was last searched;
A search server comprising: a click score reflection processing unit that calculates a search order of the content based on the weighting factor and the number of clicks by which the content has been searched. The search server according to claim 1,
The click score reflection processing unit calculates a click score based on the weighting factor and the number of clicks, and calculates the search order based on the click score and a temporary search order calculated by an arbitrary method. Feature search server. The search server according to claim 1,
A search server, comprising: a client request reception return processing unit that transmits the search result to the terminal in the search order. The search server according to claim 1,
The content validity period calculation processing unit acquires the number of updates of the content from a content storage server in which the content is stored. An information search method for searching for content based on a search keyword transmitted from a terminal,
A process of calculating a validity period of the content based on the type of the content such as MIME type and the number of updates of the content;
A process of calculating a weighting factor based on the effective period and an elapsed time that is an elapsed time from when the content was last searched;
An information search method comprising: processing for calculating a search order of the content based on the weighting factor and the number of clicks for searching the content. The information search method according to claim 5,
Processing for calculating a click score based on the weighting factor and the number of clicks;
And a process for calculating the search order based on the click score and a temporary search order calculated by an arbitrary method. The information search method according to claim 5,
An information search method comprising a process of transmitting the search result to the terminal in the search order. The information search method according to claim 5,
An information search method comprising: a process of acquiring the number of updates of the content from a content storage server storing the content. To search server that searches contents based on search keyword sent from terminal,
A procedure for calculating the validity period of the content based on the type of the content such as MIME type and the number of updates of the content;
A procedure for calculating a weighting factor based on the validity period and an elapsed time that is an elapsed time since the content was last searched;
A program for executing a procedure for calculating a search order of content based on the weighting factor and the number of clicks by which the content has been searched. The program according to claim 9,
Calculating a click score based on the weighting factor and the number of clicks;
A program for executing a procedure for calculating the search order based on the click score and a temporary search order calculated by an arbitrary method. The program according to claim 9,
A program for executing a procedure for transmitting the search result to the terminal in the search order. The program according to claim 9,
The program for performing the procedure which acquires the update frequency of the said content from the content storage server in which the said content is preserve | saved.

Images