CN103246740B - The Optimizing Search clicked on based on user of iteration and satisfaction method for improving and system - Google Patents

Abstract

The present invention proposes the Optimizing Search clicked on based on user and satisfaction method for improving and the system of a kind of iteration, belongs to Optimizing Search field.The method is；A, according to search key, the page of the click of this keyword is extracted from inquiry log, the keyword clicking on these pages is extracted from inquiry log, repeat above procedure until restraining, these keywords are become bigraph (bipartite graph) with corresponding page makeup of clicking on, using this figure as input, IA CLICK method is utilized to carry out similarity iterative computation；B, the similarity set of the page pair obtained based on step A, (q d) weighs formula and obtains key word of the inquiry and the degree of association of the page that user provides, scans for reordering of result to utilize degree of association R；C, utilize the click situation of user's reorder Search Results, utilize user satisfaction us to weigh formula and scan for evaluating；The present invention has the features such as burden for users is little, relatedness strong, evaluation framework is the most efficient；Improve verity and the motility of the effect of method.

Description

The Optimizing Search clicked on based on user of iteration and satisfaction method for improving and system

Technical field

The present invention relates to Optimizing Search field, particularly relate to the Optimizing Search clicked on based on user of a kind of iteration With satisfaction method for improving and system.

Background technology

At present in Optimizing Search field, the method that existing many realizes Optimizing Search, but there is no and carry weight The method of user satisfaction；Optimizing Search has the method utilizing user's inquiry log, this method to compare to connect Nearly user truly selects, but it is weak to the relatedness of Search Results to directly utilize inquiry log.Document [Optimizing Web Search Using Web Click-through Data] proposes for searching for Iterative Algorithm (referred to as IA) algorithm optimized.IA algorithm is kind of a classical chess game optimization Algorithm.IA algorithm had both considered to inquire about similarity, it is also considered that the similarity of the page, can be to a certain degree The upper association page that introduces, and filtering noise information；And iteration enhances the similarity of the page and inquiry, permissible Adapt to the new page or the inquiry produced at any time.Its shortcoming is not fully take into account the power between the page and inquiry The impact of value, causes the similarity that iterative process is too strengthened, introduces too much noise data, be unfavorable for Algorithmic statement and subsequent calculations.And the customer satisfaction evaluation method that presently, there are has the analysis block of complexity more Frame, is difficult to adapt to the search engine requirement to performance.

Summary of the invention

It is contemplated that at least solve technical problem present in prior art, the most innovatively propose one The Optimizing Search clicked on based on user of iteration and satisfaction method for improving and system.

In order to realize the above-mentioned purpose of the present invention, the invention provides the based on user's click excellent of a kind of iteration Change search and satisfaction method for improving, comprise the steps:

Step 1, the Search Results obtained according to user's search key collects the inquiry log of formation, utilizes Inquiry log builds key word of the inquiry bigraph (bipartite graph), using key word of the inquiry bigraph (bipartite graph) as input, utilizes IA-CLICK similarity based method scans for result iterative computation and obtains inquiring about similarity and Page resemblance；

Step 2, based on inquiring about similarity and Page resemblance in the key word of the inquiry bigraph (bipartite graph) that step 1 is obtained IA-CLICK similarity data, structure inquiry similarity and Page resemblance set, obtain system respectively Acquiescence degree of association and user define degree of association, then utilize degree of association R [q, d] to weigh formula and obtain comprehensive relevant Degree, and arrange the most successively Search Results according to the size of described synthesis pertinence；

Step 3, according to the user click condition of the Search Results that reorders, utilizes user satisfaction us to weigh Formula is calculated the performance evaluation of Optimizing Search result.

Technique scheme have the beneficial effect that with user's history click data for input, user zero bears, The renewal of implementing of history click data ensure that motility and the real-time of method；Iterative process considers inquiry Between similarity, there is relatedness, and can adapt to the dynamic of network；To between inquiry and the page The excessive enhancing of similarity has been cut down in the consideration of weight, it is possible to more efficient extraction match information；Dependency Calculate and consider all similar pages of the page and the dependency of inquiry, more rationally with comprehensive；User is satisfied The method quantifying to use sub-attribute assignment of degree, the most efficiently, can meet the search requirement to the time, carry High search efficiency.

The Optimizing Search clicked on based on user of described iteration and satisfaction method for improving, it is preferred that described Step 1 includes:

Step 1-1, the Search Results obtained according to user's search key, extracts this pass from inquiry log The page of the click of key word, extracts the keyword clicking on these pages from inquiry log, repeats above procedure Until convergence, by the keyword inquired about and corresponding click page makeup key word of the inquiry bigraph (bipartite graph)；

Step 1-2, using the key word of the inquiry bigraph (bipartite graph) of structure as input, carries out IA-CLICK similarity Calculate, utilize Page resemblance to calculate inquiry similarity, utilize inquiry Similarity Measure Page resemblance, directly To described inquiry similarity and the convergence of described Page resemblance.

Having the beneficial effect that the key word of the inquiry bigraph (bipartite graph) of structure as input of technique scheme, is carried out IA-CLICK Similarity Measure, utilizes Page resemblance to calculate inquiry similarity, utilizes inquiry similarity meter Calculate Page resemblance, by extraction match information that can be more efficient to the computing of similarity.

The Optimizing Search clicked on based on user of described iteration and satisfaction method for improving, it is preferred that described Step 1-2 includes:

Step 1-3, when utilizing Page resemblance to calculate inquiry similarity, the phase to key word of the inquiry qs and qt Like being calculated as follows of degree:

$S_Q[q_s,q_t]=\frac{C}{\left|\mathrm{AO}\left(q_s\right)\right|\left|\mathrm{AO}\left(q_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(q_s\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(q_t\right)\right|}{\mathrm{\Σ}}}W\×S_D[d_i,d_j]$

Wherein S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, it is similar that subscript Q represents key word of the inquiry Degree, q_sAnd q_tBeing any two the different keys word of the inquiry in key word of the inquiry bigraph (bipartite graph), s, t are the most whole Number, the value of s, t is different, and its minimum value is 1, and maximum occurrences is to inquire about in key word of the inquiry bigraph (bipartite graph) The number of keyword；| AO (q) | is the total degree of the page that key word of the inquiry q clicks on, and O (q) is inquiry pass The page set that key word q clicks on, C is harmonic factor, and W is the weighing factor of every pair of Page resemblance； S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, subscript D representing pages similarity, d_iAnd d_jIt is inquiry respectively Keyword q_sAnd q_tThe page clicked on；

Step 1-4, when utilizing inquiry Similarity Measure Page resemblance, (q d) is key word of the inquiry q point to t Hitting the number of times of page d, it is calculated as follows:

$W=t(q_s,d_i)t(q_t,d_j)-\left|{\log }_2\frac{t(q_s,d_i)}{t(q_t,q_j)}\right|$

Page d_sAnd d_tBeing calculated as follows of similarity:

$S_D[d_s,d_t]=\frac{C}{\left|\mathrm{AO}\left(d_s\right)\right|\left|\mathrm{AO}\left(d_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(d_s\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(d_t\right)\right|}{\mathrm{\Σ}}}W\×S_Q[q_i,q_j]$

Wherein S_D[d_s,d_t] it is page d_sAnd d_tSimilarity, d_sAnd d_tIt it is the key word of the inquiry two of described foundation Any two different pages in portion's figure, s, t are positive integer, and the value of s, t also differs, and it takes Value minimum is 1, and maximum is the number of the page in key word of the inquiry bigraph (bipartite graph)；| AO (d) | is that page d is by point The total degree hit, and O (d) is click on the set of key word of the inquiry of page d, C is harmonic factor, and W is The weighing factor of every pair of inquiry similarity, S_Q[q_i,q_j] it is key word of the inquiry q_iAnd q_jSimilarity, inquiry close Key word q_iAnd q_jIt is click on page d respectively_sAnd d_tKey word of the inquiry；

The algorithm initial value realizing IA-CLICK iterative process is as follows:

$S^0(d_s,d_t)=\left\{\begin{array}{c}0(d_s\≠d_t)\\ 1(d_s=d_t)\end{array}\right.;$

Wherein S⁰[d_s,d_t] it is page d_sAnd d_tSimilarity, d_sAnd d_tIt it is the key word of the inquiry two of described foundation Any two pages in portion's figure, s, t are positive integer, and its value minimum is 1, and maximum is key word of the inquiry The number of the page in bigraph (bipartite graph).

Having the beneficial effect that key word of the inquiry q of technique scheme_sAnd q_tThe result of calculation of similarity, Page d_sAnd d_tThe result of calculation of similarity the consideration of weight between inquiry and the page has been cut down similar The excessive enhancing of degree, it is possible to more efficient extraction match information.

The Optimizing Search clicked on based on user of described iteration and satisfaction method for improving, it is preferred that described Step 2 includes:

Step 2-1, system default degree of association is the basic of the Similarity Measure between key word of the inquiry and the page Computing, employing system default degree of association measurement formula:

The formula of the system default degree of association between key word of the inquiry and the page calculates:

$R(q,d)=C*\underset{k\∈p\left(q\right)}{\mathrm{\Σ}}S(d_k,d)*P\left(d_k\right|q)$

Wherein (q, d) is key word of the inquiry q and the degree of association of page d of system default to R, and key word of the inquiry q is The key word of the inquiry to be searched of user's input, d is the page in key word of the inquiry bigraph (bipartite graph), the value of j Scope is 1 to the number of the page in key word of the inquiry bigraph (bipartite graph), and C is harmonic factor, S (d_k, d) it is page d And d_kSimilarity, d_kIt is the page that in described key word of the inquiry bigraph (bipartite graph), key word of the inquiry q clicks on, P(d_k| it is q) that key word of the inquiry q clicks on page d respectively_kProbability, be that key word of the inquiry clicks on this page Number of times clicks on the total degree ratio of the page with key word of the inquiry, and subscript k is positive integer, and its value minimum is 1；

Step 2-2, if user submitted this key word of the inquiry to, then this inquiry just having this user is crucial The click record of the Search Results of word, it is necessary to consider that the user in user's eye defines degree of association；Otherwise step 2-2 and 2-3 need not calculate.It is as follows that user defines relatedness computation:

$R(q,d,u)=\frac{\left|\mathrm{Clicks}(q,d,u)\right|}{\left|\mathrm{Clicks}(q,\·,u)\right|+\mathrm{\β}};$

Wherein R (q, d, u) be key word of the inquiry q and the degree of association of page d in user's u eye, | Clicks (q, d, u) | It is that user u clicks on the number of times of page d for key word of the inquiry q, | and Clicks (q, u) | it is that user u is for inquiry The page sum that keyword q clicks on, β is smoothing factor；

Step 2-3, defines degree of association by system default degree of association and user and carries out unified, final degree of association Measurement formula is as follows:

Use weight α to balance system default degree of association and user defines degree of association weighing result

R [q, d]=(1-α) * R (q, d)+α * R (q, d, u),

Wherein R [q, d] is comprehensive key word of the inquiry q and the degree of association of page d, and (q d) is system default to R Key word of the inquiry q and the degree of association of page d, (q, d u) are key word of the inquiry q and page d in user's u eye to R Degree of association, α is weight factor；

Step 2-4, weighs height rearrangement user's Search Results of formula operation numerical value according to degree of association.

Technique scheme have the beneficial effect that respectively calculate inquiry with the page between system default be correlated with Degree, user define degree of association, system default degree of association and user are defined degree of association and unifies, more adduction Reason is with comprehensive.

The Optimizing Search clicked on based on user of described iteration and satisfaction method for improving, it is preferred that described Step 3 includes:

Step 3-1, user satisfaction us is analysis in terms of two, Optimizing Search and user's proficiency, and excellent Change search and comprise two sub-attributes, the page number of click and the average click location of the page of click,

The user satisfaction score of user's proficiency is as follows, if never using inquiry key before user Word, then the evaluation of Optimizing Search is likely to be affected by interface, layout etc. by user, and 5 is to judge a use Whether family is familiar with the threshold value of system,

Step 3-2, the user satisfaction score of the page number that inquiry is clicked on is as follows, num₁The page being click on Face number,

${\mathrm{us}}_2=\left\{\begin{array}{c}{1,\mathrm{num}}_1=1\\ 0.8,{\mathrm{num}}_1=2\\ 0.4,{\mathrm{num}}_1=3\\ 0.1,{\mathrm{num}}_1>3\end{array}\right.;$

Step 3-3, the score of the user satisfaction of the average click location of the page is as follows, avg₂The page being click on The average click location in face,

${\mathrm{us}}_3=\left\{\begin{array}{c}1,{\mathrm{avg}}_2=1\\ 0.8,{\mathrm{avg}}_2\∈(\mathrm{1,2}]\\ 0.4,{\mathrm{avg}}_2\∈\left(\mathrm{2,4}\right]\\ 0.1,{\mathrm{avg}}_2\∈\left(4,\∞\right]\end{array}\right.;$

Step 3-4, being calculated as follows of user satisfaction:

If the page in the order of the degree of association of the page obtained and user's eye Degree of association Ordinal Consistency increases, and user satisfaction increases.

Technique scheme have the beneficial effect that the analysis in terms of two of user satisfaction us, Optimizing Search and User's proficiency, and Optimizing Search comprises two sub-attributes, the page number of click and the page of click are average Click location, the method quantifying to use sub-attribute assignment of user satisfaction, the most efficiently, can meet and search The rope requirement to the time, improves search efficiency.

The Optimizing Search clicked on based on user and the satisfaction that invention additionally discloses a kind of iteration promote system, bag Include:

Bigraph (bipartite graph) structure loop module, collects formation for the Search Results obtained according to user's search key Inquiry log, utilizes inquiry log to build key word of the inquiry bigraph (bipartite graph), using key word of the inquiry bigraph (bipartite graph) as defeated Enter；

IA-CLICK iteration module, by scanning for based on result iteration according to IA-CLICK similarity computing Calculation obtains inquiring about similarity and Page resemblance；

Key word of the inquiry and page degree of association module, for by looking into of obtaining based on IA-CLICK iteration module Asking and inquire about similarity and the IA-CLICK similarity data of Page resemblance in keyword bigraph (bipartite graph), structure is looked into Ask similarity and Page resemblance set, obtain system default degree of association respectively and user defines degree of association, so After utilize degree of association R [q, d] to weigh formula to obtain comprehensive degree of association, and big according to described synthesis pertinence Little Search Results of arranging the most successively；

Customer satisfaction evaluation module, for the user click condition according to the Search Results that reorders, utilizes and uses Family satisfaction us is weighed formula and is calculated the performance evaluation of Optimizing Search result.

In sum, owing to have employed technique scheme, the invention has the beneficial effects as follows:

1. with user's history click data for input, user zero bears, and the enforcement of history click data updates to be protected Motility and the real-time of method are demonstrate,proved；

2. iterative process considers the similarity between inquiry, has relatedness, and can adapt to the dynamic of network State property；

3. pair the excessive enhancing of similarity has been cut down in the consideration of the weight between inquiry and the page, it is possible to higher The extraction match information of effect；

4. the calculating of dependency considers all similar pages of the page and the dependency of inquiry, more rationally and Comprehensively；

5. the method quantifying to use sub-attribute assignment of user satisfaction, the most efficiently, can meet search right The requirement of time, improves search efficiency.

The additional aspect of the present invention and advantage will part be given in the following description, and part will be retouched from following Become obvious in stating, or recognized by the practice of the present invention.

Accompanying drawing explanation

Above-mentioned and/or the additional aspect of the present invention and advantage are from combining the accompanying drawings below description to embodiment Will be apparent from easy to understand, wherein:

Fig. 1 is the Optimizing Search clicked on based on user and the flow process of satisfaction method for improving of iteration of the present invention Figure;

Fig. 2 be iteration of the present invention based on user click on Optimizing Search and satisfaction method for improving in two Portion's figure；Wherein q in Fig. 2₁…q_mIt is key word of the inquiry, and d₁…d_nIt it is the clicked page；Connect q_iWith d_jArrow prove user for key word of the inquiry q_iClick page d_j；M_mnIt is key word of the inquiry q_mClick on Page d_nNumber of times；

Fig. 3 is different in the Optimizing Search clicked on based on user of iteration of the present invention and satisfaction method for improving User satisfaction under data set size；

Fig. 4 is different in the Optimizing Search clicked on based on user of iteration of the present invention and satisfaction method for improving Click on the user satisfaction of the inquiry of entropy

Fig. 5 is the Optimizing Search clicked on based on user and the satisfaction lifting system schematic of iteration of the present invention.

Detailed description of the invention

Embodiments of the invention are described below in detail, and the example of described embodiment is shown in the drawings, wherein certainly Begin to same or similar label eventually represent same or similar element or there is the unit of same or like function Part.The embodiment described below with reference to accompanying drawing is exemplary, is only used for explaining the present invention, and can not It is interpreted as limitation of the present invention.

In describing the invention, it is to be understood that term " longitudinally ", " laterally ", " on ", D score, "front", "rear", "left", "right", " vertically ", " level ", " top ", " end " " interior ", " outward " etc. refer to The orientation shown or position relationship, for based on orientation shown in the drawings or position relationship, are for only for ease of description originally Invention and simplifying describes rather than instruction or the hint device of indication or element must have specific orientation, With specific azimuth configuration and operation, therefore it is not considered as limiting the invention.

In describing the invention, unless otherwise prescribed and limit, it should be noted that term " is installed ", " be connected ", " connection " should be interpreted broadly, for example, it may be mechanically connected or electrical connection, it is also possible to be The connection of two element internals, can be to be joined directly together, it is also possible to be indirectly connected to by intermediary, for For those of ordinary skill in the art, the concrete meaning of above-mentioned term can be understood as the case may be.

In order to solve the problem of IA in background technology and the problem that user satisfaction quantitatively evaluating is complicated, this The bright optimizing search method (referred to as IA-CLICK algorithm) clicked on based on user proposing iteration, at this On the basis of the measurement formula of degree of association, and the quantitative model of user satisfaction are proposed, form complete one repeatedly The Optimizing Search clicked on based on user in generation and satisfaction method for improving.

Fig. 1 is the Optimizing Search clicked on based on user and the flow chart of satisfaction method for improving of a kind of iteration, The degree of association comprising IA-CLICK (iterative algorithm click) algorithm, inquiry and the page is weighed and user satisfaction Quantify to weigh.Similarity Measure is with user's bigraph (bipartite graph) for input, and result similarity is used for the measurement of degree of association, Recommend related pages with the height of degree of association, complete Optimizing Search, and user satisfaction is used for realizing user and searches Rope is evaluated.

Below in conjunction with Fig. 1, Fig. 2, the Optimizing Search clicked on based on user and the satisfaction of the iteration of the present invention are described Degree method for improving.

For realizing goal of the invention, specifically comprising the following steps that of the inventive method

A, the Search Results obtained according to user's search key collect formation inquiry log, utilize IA-CLICK similarity based method scans for result iterative computation and obtains inquiring about similarity and Page resemblance； Utilize the bigraph (bipartite graph) building key word of the inquiry in inquiry log, using the bigraph (bipartite graph) of key word of the inquiry as input； Step later, inquiry log is not using, and the bigraph (bipartite graph) having been based on key word of the inquiry is carried out.

A1: according to search key, extract the page of the click of this keyword from inquiry log, from inquiry Daily record is extracted the keyword clicking on these pages, repeats above procedure until convergence (does not i.e. have new inquiry Keyword or the page occur), by the keyword inquired about and the two of corresponding click page makeup key word of the inquiry Portion's figure；

A2: using the key word of the inquiry bigraph (bipartite graph) of structure as input, carry out IA-CLICK Similarity Measure, Utilize Page resemblance to calculate inquiry similarity, utilize inquiry Similarity Measure Page resemblance, until described Inquiry similarity and the convergence of described Page resemblance.

Using the bigraph (bipartite graph) of key word of the inquiry as input, carry out IA-CLICK Similarity Measure,

To being calculated as follows of the similarity of key word of the inquiry qs and qt:

$S_Q[q_s,q_t]=\frac{C}{\left|\mathrm{AO}\left(q_s\right)\right|\left|\mathrm{AO}\left(q_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(q_s\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(q_t\right)\right|}{\mathrm{\Σ}}}W\×S_D[d_i,d_j]---\left(1\right)$

Wherein S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, it is similar that subscript Q represents key word of the inquiry Degree, q_sAnd q_tBeing any two the different keys word of the inquiry in key word of the inquiry bigraph (bipartite graph), s, t are the most whole Number, the value of s, t is different, and its minimum value is 1, and maximum occurrences is to inquire about in key word of the inquiry bigraph (bipartite graph) The number of keyword.| AO (q) | is the total degree of the page that key word of the inquiry q clicks on, and O (q) is inquiry pass The page set that key word q clicks on, C is harmonic factor.W is the weighing factor of every pair of Page resemblance, S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, subscript D representing pages similarity, d_iAnd d_jIt is inquiry respectively Keyword q_sAnd q_tThe page clicked on.T (q, d) be key word of the inquiry q click on page d number of times, its calculate such as Under:

$W=t(q_s,d_i)t(q_t,d_j)-\left|{\log }_2\frac{t(q_s,d_i)}{t(q_t,q_j)}\right|---\left(2\right)$

Page d_sAnd d_tBeing calculated as follows of similarity:

$S_D[d_s,d_t]=\frac{C}{\left|\mathrm{AO}\left(d_s\right)\right|\left|\mathrm{AO}\left(d_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(d_s\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(d_t\right)\right|}{\mathrm{\Σ}}}W\×S_Q[q_i,q_j]---\left(3\right)$

Wherein S_D[d_s,d_t] it is page d_sAnd d_tSimilarity, d_sAnd d_tIt it is the key word of the inquiry two of above-mentioned foundation Any two different pages in portion's figure, s, t are positive integer, and the value of s, t also differs, and it takes Value minimum is 1, and maximum is the number of the page in key word of the inquiry bigraph (bipartite graph).| AO (d) | is that page d is by point The total degree hit, and O (d) is click on the set of key word of the inquiry of page d.C is harmonic factor.W is The weighing factor of every pair of inquiry similarity, calculates ibid.S_Q[q_i,q_j] it is key word of the inquiry q_iAnd q_jSimilar Degree, computing formula such as (1), key word of the inquiry q_iAnd q_jIt is click on page d respectively_sAnd d_tInquiry crucial Word.The iterative process realizing IA-CKICK needs an algorithm initial value, as follows:

$S^0(d_s,d_t)=\left\{\begin{array}{c}0(d_s\≠d_t)\\ 1(d_s=d_t)\end{array}\right.---\left(4\right)$

Wherein S⁰[d_s,d_t] it is page d_sAnd d_tSimilarity, d_sAnd d_tIt it is the key word of the inquiry two of described foundation Any two pages in portion's figure, s, t are positive integer, and its value minimum is 1, and maximum is key word of the inquiry The number of the page in bigraph (bipartite graph).

In B, the key word of the inquiry bigraph (bipartite graph) obtained based on step A, the similarity set of every pair of page, utilizes Synthesis pertinence R [q, d] weighs formula and obtains key word of the inquiry and the degree of association of the page that user provides, and root Arrange the most successively Search Results according to the size of described synthesis pertinence；

B1, system default degree of association are the elementary operations of the Similarity Measure between key word of the inquiry and the page, System default degree of association is used to weigh formula

System default dependency equation below between inquiry and the page calculates:

$R(q,d)=C*\underset{k\∈p\left(q\right)}{\mathrm{\Σ}}S(d_k,d)*P\left(d_k\right|q)---\left(5\right)$

Here, (q d) is key word of the inquiry q and the degree of association of page d of system default to R.Inquiry key Word q is the key word of the inquiry to be searched that user inputs, and d is the page in key word of the inquiry bigraph (bipartite graph), j's Span is 1 to the number of the page in key word of the inquiry bigraph (bipartite graph).C is harmonic factor.S(d_k, d) it is page Face d and d_kSimilarity, computing formula such as (3).d_kIt is that in above-mentioned key word of the inquiry bigraph (bipartite graph), inquiry is closed The page that key word q clicks on, P (d_k| it is q) that key word of the inquiry q clicks on page d respectively_kProbability, be inquiry The number of times of this page and the total degree ratio of the key word of the inquiry click page clicked in keyword, and subscript k is the most whole Number, its value minimum is 1.

If B2 user submitted this key word of the inquiry to, then just there be the searching of this key word of the inquiry of this user The click record of hitch fruit, owing to optimizing field in personalization, the individual variation of user is fairly obvious, it is necessary to Consider that the user in user's eye defines degree of association；Otherwise step B2 and B3 need not calculate.User defines relevant Degree is calculated as follows:

$R(q,d,u)=\frac{\left|\mathrm{Clicks}(q,d,u)\right|}{\left|\mathrm{Clicks}(q,\·,u)\right|+\mathrm{\β}}---\left(6\right)$

Here, (q, d u) are key word of the inquiry q and the degree of association of page d in user's u eye to R. | Clicks (q, d, u) | it is that user u clicks on the number of times of page d for key word of the inquiry q.| Clicks (q, u) | it is to use The page sum that family u clicks on for key word of the inquiry q.β is smoothing factor (taking 0.5 here).

B3: in order to realize reordering of Search Results, it is necessary to realize system default degree of association and user defines phase The unification of Guan Du, it is as follows that final degree of association weighs formula, finally according to degree of association height rearrangement user Search Results.

The present invention uses weight α to come balancing user definition and the degree of association weighing result of system default, and it is:

R [q, d]=(1-α) * R (q, d)+α * R (q, d, u) (7)

Here, R [q, d] is comprehensive key word of the inquiry q and the degree of association of page d, and (q d) is system default to R Key word of the inquiry q and the degree of association of page d.(q, d u) are key word of the inquiry q and the page in user's u eye to R The degree of association of d.α is weight factor, and value is 0.7 here.

C, according to user click condition, utilize user satisfaction us weigh formula be calculated Optimizing Search knot The performance evaluation of fruit；Wherein user click condition record is among inquiry log, according to the inquired about key of user The historical record of word constantly updates inquiry log content.

In the present invention, user satisfaction as the evaluation index to Search Results, than search engine instantly Search Results evaluation index precision ratio from the point of view of, both considered user's efficiency, i.e. user's proficiency, also real Existing user zero bears.From many aspects overall merit Search Results, the result of user satisfaction can be the most anti- Mirror user and situation clicked in the experience of Search Results, it is possible to as pass judgment on a key word of the inquiry the need of Search again for the standard optimized.According to the level threshold value of chess game optimization, if key word of the inquiry is not up to its mark Quasi-threshold value, this key word of the inquiry again submitted to for user, it is necessary to carry out searching again for optimizing computing, as Fruit has reached level threshold value, then be not required to again optimize computing；Specifically it is calculated as follows:

User satisfaction is analysis in terms of two: Optimizing Search and user's proficiency.If a user is to system Being unfamiliar with, then when he is unsatisfied with the page searched out, sensation can be even worse.And Optimizing Search comprises two Sub-attribute, the page number of click and the mean place of the click page.

The user satisfaction score of user's proficiency is as follows.If using this to be before a user hardly System, then the evaluation of Optimizing Search is likely to be affected by interface, layout etc. by user.5 is to judge a use Whether family is familiar with the threshold value of system.

The user satisfaction score of the page number that inquiry is clicked on is as follows, num₁It it is the page click of inquiry Number.The page that user clicks on for an inquiry is the most, feels the most disappointed.Therefore this score superlinearity ground Decline.

${\mathrm{us}}_2=\left\{\begin{array}{c}{1,\mathrm{num}}_1=1\\ 0.8,{\mathrm{num}}_1=2\\ 0.4,{\mathrm{num}}_1=3\\ 0.1,{\mathrm{num}}_1>3\end{array}\right.---\left(9\right)$

The score of the user satisfaction of the average click location of the page is as follows, avg₂It is that the page inquired about averagely is clicked on Number.

${\mathrm{us}}_3=\left\{\begin{array}{c}1,{\mathrm{avg}}_2=1\\ 0.8,{\mathrm{avg}}_2\∈\left(\mathrm{1,2}\right]\\ 0.4,{\mathrm{avg}}_2\∈\left(\mathrm{2,4}\right]\\ 0.1,{\mathrm{avg}}_2\∈(4,\∞]\end{array}\right.---\left(10\right)$

Finally, being calculated as follows of user satisfaction:

$\mathrm{us}=\frac{1}{2}{\mathrm{us}}_1({\mathrm{us}}_2+{\mathrm{us}}_3)---\left(11\right)$

This is shown for the performance of user satisfaction and accuracy rate, and compared for the performance of several strategy: based on Reorder (group) that history is clicked on, IA, CM, the IA-CLICK algorithm of the present invention and degree of association are weighed Formula.

Fig. 3 is user satisfaction performance during different pieces of information collection size, and wherein data set is AOL search engine Inquiry log.Wherein in figure, CM represents existing search engine, GROUP represents based on history click Reorder, IA represents iterative algorithm, IA-Click represents the iterative algorithm similarity computing of the present invention, when When data set was less than 5 days, the performance of three kinds of algorithms all carries nice and high little, but when data set was more than 5 days, And the growth of IA-CLICK algorithm and the degree of association user satisfaction of weighing formula (8%) is the most fast IA(5%) Speed；When data set size rises to 10 days, curve is close to saturated.

Fig. 4 is the performance of the user satisfaction of the inquiry of different entropy.Wherein in figure, CM represents existing search Engine, GROUP represent based on history click on reorder, IA represents iterative algorithm, IA-Click represent The similarity computing of the present invention, the most that algorithm, along with the increase of inquiry entropy, user satisfaction is all Reducing, this is because user clicks multiple page；It will be seen that IA-CLICK on same entropy Algorithm and degree of association weigh the best of formula performance.Along with the increase of entropy, the satisfaction difference of three kinds of algorithms In increase tendency, especially IA-CLICK algorithm and degree of association weigh formula, and of the present invention changing is described The Optimizing Search clicked on based on user in generation and satisfaction method for improving are more suitable for there be looking into of big click entropy Ask.

As it is shown in figure 5, invention additionally discloses the Optimizing Search clicked on based on user and the satisfaction of a kind of iteration Lifting system, including:

Bigraph (bipartite graph) structure loop module, collects formation for the Search Results obtained according to user's search key Inquiry log, utilizes inquiry log to build key word of the inquiry bigraph (bipartite graph), using key word of the inquiry bigraph (bipartite graph) as defeated Enter；

The Search Results obtained according to user's search key, extracts the click of this keyword from inquiry log The page, from inquiry log extract click on these pages keyword, repeat above procedure until convergence, By the keyword inquired about and corresponding click page makeup key word of the inquiry bigraph (bipartite graph)；

Using the key word of the inquiry bigraph (bipartite graph) of structure as input, carry out IA-CLICK Similarity Measure, utilize Page resemblance calculates inquiry similarity, utilizes inquiry Similarity Measure Page resemblance, until described inquiry Similarity and the convergence of described Page resemblance.

IA-CLICK iteration module, by scanning for based on result iteration according to IA-CLICK similarity computing Calculation obtains inquiring about similarity and Page resemblance；Using the key word of the inquiry bigraph (bipartite graph) of structure as input, carry out IA-CLICK Similarity Measure, utilizes Page resemblance to calculate inquiry similarity, utilizes inquiry similarity meter Calculate Page resemblance, until described inquiry similarity and the convergence of described Page resemblance.

Key word of the inquiry similarity calculation module, when Page resemblance calculates inquiry similarity, to inquiry Keyword q_sAnd q_tBeing calculated as follows of similarity:

$S_Q[q_s,q_t]=\frac{C}{\left|\mathrm{AO}\left(q_s\right)\right|\left|\mathrm{AO}\left(q_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(q_s\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(q_t\right)\right|}{\mathrm{\Σ}}}W\×S_D[d_i,d_j]$

Wherein S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, it is similar that subscript Q represents key word of the inquiry Degree, q_sAnd q_tBeing any two the different keys word of the inquiry in key word of the inquiry bigraph (bipartite graph), s, t are the most whole Number, the value of s, t is different, and its minimum value is 1, and maximum occurrences is to inquire about in key word of the inquiry bigraph (bipartite graph) The number of keyword；| AO (q) | is the total degree of the page that key word of the inquiry q clicks on, and O (q) is inquiry pass The page set that key word q clicks on, C is harmonic factor, and W is the weighing factor of every pair of Page resemblance； S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, subscript D representing pages similarity, d_iAnd d_jIt is inquiry respectively Keyword q_sAnd q_tThe page clicked on；

Page resemblance computing module, when being used for inquiring about Similarity Measure Page resemblance, (q d) is inquiry to t Keyword q clicks on the number of times of page d, and it is calculated as follows:

$W=t(q_s,d_i)t(q_t,d_j)-\left|{\log }_2\frac{t(q_s,d_i)}{t(q_t,q_j)}\right|$

Page d_sAnd d_tBeing calculated as follows of similarity:

$S_D[d_s,d_t]=\frac{C}{\left|\mathrm{AO}\left(d_s\right)\right|\left|\mathrm{AO}\left(d_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(d_s\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(d_t\right)\right|}{\mathrm{\Σ}}}W\×S_Q[q_i,q_j]$

Wherein S_D[d_s,d_t] it is page d_sAnd d_tSimilarity, d_sAnd d_tIt it is the key word of the inquiry two of described foundation Any two different pages in portion's figure, s, t are positive integer, and the value of s, t also differs, and it takes Value minimum is 1, and maximum is the number of the page in key word of the inquiry bigraph (bipartite graph)；| AO (d) | is that page d is by point The total degree hit, and O (d) is click on the set of key word of the inquiry of page d, C is harmonic factor, and W is The weighing factor of every pair of inquiry similarity, S_Q[q_i,q_j] it is key word of the inquiry q_iAnd q_jSimilarity, inquiry close Key word q_iAnd q_jIt is click on page d respectively_sAnd d_tKey word of the inquiry；

The algorithm initial value realizing IA-CLICK iterative process is as follows:

$S^0(d_s,d_t)=\left\{\begin{array}{c}0(d_s\≠d_t)\\ 1(d_s=d_t)\end{array}\right.;$

Wherein S⁰[d_s,d_t] it is page d_sAnd d_tSimilarity, d_sAnd d_tIt it is the key word of the inquiry two of described foundation Any two pages in portion's figure, s, t are positive integer, and its value minimum is 1, and maximum is key word of the inquiry The number of the page in bigraph (bipartite graph).

Key word of the inquiry and page degree of association module, for by looking into of obtaining based on IA-CLICK iteration module Asking and inquire about similarity and the IA-CLICK similarity data of Page resemblance in keyword bigraph (bipartite graph), structure is looked into Ask similarity and Page resemblance set, obtain system default degree of association respectively and user defines degree of association, so After utilize degree of association R [q, d] to weigh formula to obtain synthesis pertinence, and according to the size of described synthesis pertinence Arrange the most successively Search Results；

System default relatedness computation module, is the elementary operation of key word of the inquiry and page degree of association module, Employing system default degree of association measurement formula:

The formula of the system default degree of association between key word of the inquiry and the page calculates:

$R(q,d)=C*\underset{k\∈p\left(q\right)}{\mathrm{\Σ}}S(d_k,d)*P\left(d_k\right|q)$

Wherein (q, d) is key word of the inquiry q and the degree of association of page d of system default to R, and key word of the inquiry q is The key word of the inquiry to be searched of user's input, d is the page in key word of the inquiry bigraph (bipartite graph), the value of j Scope is 1 to the number of the page in key word of the inquiry bigraph (bipartite graph), and C is harmonic factor, S (d_k, d) it is page d And d_kSimilarity, d_kIt is the page that in described key word of the inquiry bigraph (bipartite graph), key word of the inquiry q clicks on, P(d_k| it is q) that key word of the inquiry q clicks on page d respectively_kProbability, be that key word of the inquiry clicks on this page Number of times clicks on the total degree ratio of the page with key word of the inquiry, and subscript k is positive integer, and its value minimum is 1；

User defines relatedness computation module, the same queries keyword again submitted to for user, if User for the first time searches for this key word of the inquiry, and described user does not has the record of Search Results, this module and comprehensive Relatedness computation module need not be carried out.It is as follows that user defines relatedness computation:

$R(q,d,u)=\frac{\left|\mathrm{Clicks}(q,d,u)\right|}{\left|\mathrm{Clicks}(q,\·,u)\right|+\mathrm{\β}};$

Wherein R (q, d, u) be key word of the inquiry q and the degree of association of page d in user's u eye, | Clicks (q, d, u) | It is that user u clicks on the number of times of page d for key word of the inquiry q, | and Clicks (q, u) | it is that user u is for inquiry The page sum that keyword q clicks on, β is smoothing factor；

Final synthesis pertinence computing module, is carried out for system default degree of association and user are defined degree of association Unified, it is as follows that final synthesis pertinence weighs formula:

Use weight α to balance system default degree of association and user defines degree of association weighing result

R [q, d]=(1-α) * R (q, d)+α * R (q, d, u),

Wherein R [q, d] is comprehensive key word of the inquiry q and the degree of association of page d, and (q d) is system default to R Key word of the inquiry q and the degree of association of page d, (q, d u) are key word of the inquiry q and page d in user's u eye to R Degree of association, α is weight factor；The height rearrangement user of formula operation numerical value is weighed according to degree of association Search Results.

Customer satisfaction evaluation module, for the user click condition according to the Search Results that reorders, utilizes and uses Family satisfaction us is weighed formula and is calculated the performance evaluation of Optimizing Search result.

Satisfaction index generic module, for the analysis in terms of two of user satisfaction us, Optimizing Search and user are ripe White silk degree, and Optimizing Search comprises two sub-attributes, the page number of click and the page of click averagely click on position Put,

The user satisfaction score of user's proficiency is as follows, if never using inquiry key before user Word, then the evaluation of Optimizing Search is likely to be affected by interface, layout etc. by user, and 5 is to judge a use Whether family is familiar with the threshold value of system,

Clicking on page number module, the user satisfaction score of the page number for inquiring about click is as follows, num₁The page number being click on,

${\mathrm{us}}_2=\left\{\begin{array}{c}{1,\mathrm{num}}_1=1\\ 0.8,{\mathrm{num}}_1=2\\ 0.4,{\mathrm{num}}_1=3\\ 0.1,{\mathrm{num}}_1>3\end{array}\right.;$

The page averagely clicks on position module, for the average click location of the page user satisfaction score such as Under, avg₂The average click location of the page being click on,

${\mathrm{us}}_3=\left\{\begin{array}{c}1,{\mathrm{avg}}_2=1\\ 0.8,{\mathrm{avg}}_2\∈\left(\mathrm{1,2}\right]\\ 0.4,{\mathrm{avg}}_2\∈\left(\mathrm{2,4}\right]\\ 0.1,{\mathrm{avg}}_2\∈(4,\∞]\end{array}\right.;$

Satisfaction computing module, for the calculating of user satisfaction, its computing formula is:

If the page in the order of the degree of association of the page obtained and user's eye Face degree of association Ordinal Consistency increases, and user satisfaction increases.

Owing to have employed technique scheme, the optimization clicked on based on user of a kind of iteration of the present invention Search and satisfaction method for improving possess following beneficial effect:

1. with user's history click data for input, user zero bears, and the enforcement of history click data updates to be protected Motility and the real-time of method are demonstrate,proved；

2. iterative process considers the similarity between inquiry, has relatedness, and can adapt to the dynamic of network State property；

3. pair the excessive enhancing of similarity has been cut down in the consideration of the weight between inquiry and the page, it is possible to higher The extraction match information of effect；

4. the calculating of dependency considers all similar pages of the page and the dependency of inquiry, more rationally and Comprehensively；

5. the method quantifying to use sub-attribute assignment of user satisfaction, the most efficiently, can meet search right The requirement of time, improves search efficiency.

In the description of this specification, reference term " embodiment ", " some embodiments ", " example ", It is concrete that the description of " concrete example " or " some examples " etc. means to combine this embodiment or example describes Feature, structure, material or feature are contained at least one embodiment or the example of the present invention.In this theory In bright book, the schematic representation of above-mentioned term is not necessarily referring to identical embodiment or example.And, The specific features, structure, material or the feature that describe can be in any one or more embodiments or examples In combine in an appropriate manner.

Although an embodiment of the present invention has been shown and described, those of ordinary skill in the art can manage Solve: these embodiments can be carried out in the case of without departing from the principle of the present invention and objective multiple change, Amendment, replacement and modification, the scope of the present invention is limited by claim and equivalent thereof.

Claims (4)

1. an iteration based on user click on Optimizing Search and satisfaction method for improving, it is characterised in that, bag Include following steps:

Step 1, the Search Results obtained according to user's search key collects formation inquiry log, and utilization is looked into Ask daily record and build key word of the inquiry bigraph (bipartite graph), using key word of the inquiry bigraph (bipartite graph) as input, utilize IA-CLICK Similarity based method scans for result iterative computation and obtains inquiring about similarity and Page resemblance；

Step 1-1, the Search Results obtained according to user's search key, extracts this pass from inquiry log The page of the click of key word, extracts the keyword clicking on these pages from inquiry log, repeats above procedure Until convergence, by the keyword inquired about and corresponding click page makeup key word of the inquiry bigraph (bipartite graph)；

Step 1-2, using the key word of the inquiry bigraph (bipartite graph) of structure as input, carries out IA-CLICK similarity Calculate, utilize Page resemblance to calculate inquiry similarity, utilize inquiry Similarity Measure Page resemblance, directly To described inquiry similarity and the convergence of described Page resemblance；

Step 1-3, when utilizing Page resemblance to calculate inquiry similarity, to key word of the inquiry q_sAnd q_tPhase Like being calculated as follows of degree:

$S_Q\[q_s,q_t\]=\frac{C}{\left|AO\left(q_s\right)\right|\left|AO\left(q_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(q_s\right)\right|}{\Σ}}\underset{j=1}{\overset{\left|O\left(q_t\right)\right|}{\Σ}}W\×S_D\[d_i,d_j\]$

Wherein S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, it is similar that subscript Q represents key word of the inquiry Degree, q_sAnd q_tBeing any two the different keys word of the inquiry in key word of the inquiry bigraph (bipartite graph), s, t are the most whole Number, the value of s, t is different, and its minimum value is 1, and maximum occurrences is to inquire about in key word of the inquiry bigraph (bipartite graph) The number of keyword；| AO (q) | is the total degree of the page that key word of the inquiry q clicks on, and O (q) is inquiry pass The page set that key word q clicks on, C is harmonic factor, and W is the weighing factor of every pair of Page resemblance； S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, subscript D representing pages similarity, d_iAnd d_jIt is inquiry respectively Keyword q_sAnd q_tThe page clicked on, wherein the value of i, j is different, and i, j are positive integer, and its minimum value is 1, maximum occurrences is the number of key word of the inquiry in key word of the inquiry bigraph (bipartite graph)；

Step 1-4, when utilizing inquiry Similarity Measure Page resemblance, W is calculated as follows, and wherein (q d) is t The number of times of key word of the inquiry q click page d:

$W=t(q_s,d_i)t(q_t,d_j)-\left|{\log }_2\frac{t(q_s,d_i)}{t(q_t,d_j)}\right|$

Page d_iAnd d_jBeing calculated as follows of similarity:

$S_D\[d_i,d_j\]=\frac{C}{\left|AO\left(d_i\right)\right|\left|AO\left(d_j\right)\right|}\×\underset{i=1}{\overset{\left|O\left(d_i\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(d_j\right)\right|}{\mathrm{\Σ}}}W\×S_Q\[q_s,q_t\]$

Wherein S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, d_iAnd d_jIt it is the key word of the inquiry two of described foundation Any two different pages in portion's figure, i, j are positive integer, and the value of i, j also differs, its value Minimum is 1, and maximum is the number of the page in key word of the inquiry bigraph (bipartite graph)；| AO (d) | is that page d is clicked Total degree, and O (d) is click on the set of key word of the inquiry of page d, and C is harmonic factor, and W is every Weighing factor to inquiry similarity, S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, inquiry key Word q_sAnd q_tIt is click on page d respectively_iAnd d_jKey word of the inquiry, wherein the value of s, t is different, s, t For positive integer, its minimum value is 1, and maximum occurrences is the individual of key word of the inquiry in key word of the inquiry bigraph (bipartite graph) Number；

The algorithm initial value realizing IA-CLICK iterative process is as follows:

$S^0\[d_i,d_j\]=\left\{\begin{array}{c}0(d_i\≠d_j)\\ 1(d_i=d_j)\end{array}\right.;$

Wherein S⁰[d_i,d_j] it is page d_iAnd d_jSimilarity, d_iAnd d_jIt it is the key word of the inquiry two of described foundation Any two different pages in portion's figure, i, j are positive integer, and the value of i, j also differs, its value Minimum is 1, and maximum occurrences is the number of the page in key word of the inquiry bigraph (bipartite graph)；

Step 2, based on inquiring about similarity and Page resemblance in the key word of the inquiry bigraph (bipartite graph) that step 1 is obtained IA-CLICK similarity data, structure inquiry similarity and Page resemblance set, obtain system respectively Acquiescence degree of association and user define degree of association, then utilize degree of association R [q, d] to weigh formula and obtain comprehensive relevant Degree, and arrange the most successively Search Results according to the size of described synthesis pertinence；

Step 3, according to inquiry times and the user click condition of the Search Results that reorders, utilizes user satisfied Degree us weighs formula and is calculated the performance evaluation of Optimizing Search result.

The Optimizing Search clicked on based on user of iteration the most according to claim 1 and satisfaction promote Method, it is characterised in that described step 2 includes:

Step 2-1, system default degree of association is the basic of the Similarity Measure between key word of the inquiry and the page Computing, employing system default degree of association measurement formula:

The formula of the system default degree of association between key word of the inquiry and the page calculates:

$R(q,d)=C*\underset{k\∈p\left(q\right)}{\Σ}S(d_k,d)*P\left(d_k\right|q)$

Wherein (q, d) is key word of the inquiry q and the degree of association of page d of system default to R, and key word of the inquiry q is The key word of the inquiry to be searched of user's input, d is the page in key word of the inquiry bigraph (bipartite graph), the value of j Scope is 1 to the number of the page in key word of the inquiry bigraph (bipartite graph), and C is harmonic factor, S (d_k, d) it is page d And d_kSimilarity, d_kIt is the page that in described key word of the inquiry bigraph (bipartite graph), key word of the inquiry q clicks on, P(d_k| it is q) that key word of the inquiry q clicks on page d respectively_kProbability, be that key word of the inquiry clicks on this page Number of times clicks on the total degree ratio of the page with key word of the inquiry, and subscript k is positive integer, and its value minimum is 1；

Step 2-2, if user submitted this key word of the inquiry to, then this inquiry just having this user is crucial The click record of the Search Results of word, it is necessary to consider that the user in user's eye defines degree of association；Otherwise step 2-2 and 2-3 need not calculate, and it is as follows that user defines relatedness computation:

$R(q,d,u)=\frac{\left|Clicks(q,d,u)\right|}{\left|Clicks(q,\·,u)\right|+\mathrm{\β}};$

Wherein R (q, d, u) be key word of the inquiry q and the degree of association of page d in user's u eye, | Clicks (q, d, u) | It is that user u clicks on the number of times of page d for key word of the inquiry q, | and Clicks (q, u) | it is that user u is for inquiry The page sum that keyword q clicks on, β is smoothing factor；

Step 2-3, defines degree of association by system default degree of association and user and carries out unified, final comprehensive phase It is as follows that Guan Du weighs formula:

Use weight α to balance system default degree of association and user defines degree of association weighing result

R [q, d]=(1-α) * R (q, d)+α * R (q, d, u),

Wherein R [q, d] is comprehensive key word of the inquiry q and the degree of association of page d, and (q d) is system default to R Key word of the inquiry q and the degree of association of page d, (q, d u) are key word of the inquiry q and page d in user's u eye to R Degree of association, α is weight factor；

Step 2-4, weighs height rearrangement user's Search Results of formula operation numerical value according to degree of association.

The Optimizing Search clicked on based on user of iteration the most according to claim 1 and satisfaction promote Method, it is characterised in that described step 3 includes:

Step 3-1, user satisfaction us is analysis in terms of two, Optimizing Search and user's proficiency, and optimizes Search comprises two sub-attributes, the page number of click and the average click location of the page of click,

The user satisfaction score of user's proficiency is as follows, if never using inquiry key before user Word, then user is to the evaluation of Optimizing Search by interface, influence of arrangement, and 5 is to judge that a user is the ripest Know the threshold value of system,

Step 3-2, the user satisfaction score of the page number that inquiry is clicked on is as follows, num₁The page being click on Face number,

${\mathrm{us}}_2=\left\{\begin{array}{c}1,{\mathrm{num}}_1=1\\ 0.8,{\mathrm{num}}_1=2\\ 0.4,{\mathrm{num}}_1=3\\ 0.1,{\mathrm{num}}_1>3\end{array}\right.;$

Step 3-3, the score of the user satisfaction of the average click location of the page is as follows, avg₂The page being click on The average click location in face,

${\mathrm{us}}_3=\left\{\begin{array}{c}1,{\mathrm{avg}}_2=1\\ 0.8,{\mathrm{avg}}_2\∈(1,2\]\\ 0.4,{\mathrm{avg}}_2\∈(2,4\]\\ 0.1,{\mathrm{avg}}_2\∈(4,\mathrm{\∞}\]\end{array}\right.;$

Step 3-4, being calculated as follows of user satisfaction:

If the page in the order of the degree of association of the page obtained and user's eye Degree of association Ordinal Consistency increases, and user satisfaction increases.

4. the Optimizing Search clicked on based on user and the satisfaction of an iteration promotes system, it is characterised in that Including:

Bigraph (bipartite graph) structure loop module, collects formation for the Search Results obtained according to user's search key Inquiry log, utilizes inquiry log to build key word of the inquiry bigraph (bipartite graph), using key word of the inquiry bigraph (bipartite graph) as defeated Enter,

The Search Results obtained according to user's search key, extracts the click of this keyword from inquiry log The page, from inquiry log extract click on these pages keyword, repeat above procedure until convergence, By the keyword inquired about and corresponding click page makeup key word of the inquiry bigraph (bipartite graph)；

Using the key word of the inquiry bigraph (bipartite graph) of structure as input, carry out IA-CLICK Similarity Measure, utilize Page resemblance calculates inquiry similarity, utilizes inquiry Similarity Measure Page resemblance, until described inquiry Similarity and the convergence of described Page resemblance；

When utilizing Page resemblance to calculate inquiry similarity, to key word of the inquiry q_sAnd q_tThe calculating of similarity As follows:

$S_Q\[q_s,q_t\]=\frac{C}{\left|AO\left(q_s\right)\right|\left|AO\left(q_t\right)\right|}\×\underset{i=1}{\overset{\left|O\left(q_s\right)\right|}{\Σ}}\underset{j=1}{\overset{\left|O\left(q_t\right)\right|}{\Σ}}W\×S_D\[d_i,d_j\]$

Wherein S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, it is similar that subscript Q represents key word of the inquiry Degree, q_sAnd q_tBeing any two the different keys word of the inquiry in key word of the inquiry bigraph (bipartite graph), s, t are the most whole Number, the value of s, t is different, and its minimum value is 1, and maximum occurrences is to inquire about in key word of the inquiry bigraph (bipartite graph) The number of keyword；| AO (q) | is the total degree of the page that key word of the inquiry q clicks on, and O (q) is inquiry pass The page set that key word q clicks on, C is harmonic factor, and W is the weighing factor of every pair of Page resemblance； S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, subscript D representing pages similarity, d_iAnd d_jIt is inquiry respectively Keyword q_sAnd q_tThe page clicked on, wherein the value of i, j is different, and i, j are positive integer, its minimum value Being 1, maximum occurrences is the number of key word of the inquiry in key word of the inquiry bigraph (bipartite graph)；

When utilizing inquiry Similarity Measure Page resemblance, W is calculated as follows, and wherein (q d) is inquiry key to t The number of times of word q click page d:

$W=t(q_s,d_i)t(q_t,d_j)-\left|{\log }_2\frac{t(q_s,d_i)}{t(q_t,d_j)}\right|$

Page d_iAnd d_jBeing calculated as follows of similarity:

$S_D\[d_i,d_j\]=\frac{C}{\left|AO\left(d_i\right)\right|\left|AO\left(d_j\right)\right|}\×\underset{i=1}{\overset{\left|O\left(d_i\right)\right|}{\mathrm{\Σ}}}\underset{j=1}{\overset{\left|O\left(d_j\right)\right|}{\mathrm{\Σ}}}W\×S_Q\[q_s,q_t\]$

Wherein S_D[d_i,d_j] it is page d_iAnd d_jSimilarity, d_iAnd d_jIt it is the key word of the inquiry two of described foundation Any two different pages in portion's figure, i, j are positive integer, and the value of i, j also differs, its value Minimum is 1, and maximum is the number of the page in key word of the inquiry bigraph (bipartite graph)；| AO (d) | is that page d is clicked Total degree, and O (d) is click on the set of key word of the inquiry of page d, and C is harmonic factor, and W is every Weighing factor to inquiry similarity, S_Q[q_s,q_t] it is key word of the inquiry q_sAnd q_tSimilarity, inquiry key Word q_sAnd q_tIt is click on page d respectively_iAnd d_jKey word of the inquiry, wherein the value of s, t is different, s, t For positive integer, its minimum value is 1, and maximum occurrences is the individual of key word of the inquiry in key word of the inquiry bigraph (bipartite graph) Number；

The algorithm initial value realizing IA-CLICK iterative process is as follows:

$S^0\[d_i,d_j\]=\left\{\begin{array}{c}0(d_i\≠d_j)\\ 1(d_i=d_j)\end{array}\right.;$

Wherein S⁰[d_i,d_j] it is page d_iAnd d_jSimilarity, d_iAnd d_jIt it is the key word of the inquiry two of described foundation Any two different pages in portion's figure, i, j are positive integer, and the value of i, j also differs, its value Minimum is 1, and maximum occurrences is the number of the page in key word of the inquiry bigraph (bipartite graph)；

IA-CLICK iteration module, for scanning for result iteration according to IA-CLICK similarity computing It is calculated inquiry similarity and Page resemblance；

Key word of the inquiry and page degree of association module, for by looking into of obtaining based on IA-CLICK iteration module Asking and inquire about similarity and the IA-CLICK similarity data of Page resemblance in keyword bigraph (bipartite graph), structure is looked into Ask similarity and Page resemblance set, obtain system default degree of association respectively and user defines degree of association, so After utilize degree of association R [q, d] to weigh formula to obtain synthesis pertinence, and according to the size of described synthesis pertinence Arrange the most successively Search Results；

Customer satisfaction evaluation module, for the user click condition according to the Search Results that reorders, utilizes and uses Family satisfaction us is weighed formula and is calculated the performance evaluation of Optimizing Search result.