CN105335363B - A kind of Object Push method and system - Google Patents
A kind of Object Push method and system Download PDFInfo
- Publication number
- CN105335363B CN105335363B CN201410231465.8A CN201410231465A CN105335363B CN 105335363 B CN105335363 B CN 105335363B CN 201410231465 A CN201410231465 A CN 201410231465A CN 105335363 B CN105335363 B CN 105335363B
- Authority
- CN
- China
- Prior art keywords
- candidate target
- weighted value
- candidate
- target
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention provides a kind of Object Push method and system, this method comprises: obtain candidate target, the candidate target include client input object, according to the client input object retrieved after obtain matching object or according to the client click historical record obtain recommended；The weight absorptivity of the candidate target is obtained, the weight absorptivity of the candidate target is that the number for other candidate targets being directly linked with the candidate target is in the parameter value of inversely prroportional relationship；Wherein, other candidate targets being directly linked with the candidate target are other candidate targets for being greater than preset dependent thresholds with the degree of correlation of the candidate target；According to the weight absorptivity of the candidate target, the first weighted value of the candidate target is obtained；According to first weighted value and the candidate target, target object to be pushed is obtained.Technical solution provided in an embodiment of the present invention may be implemented to improve the accuracy and recall precision of information retrieval.
Description
[technical field]
The present invention relates to technical field of information retrieval more particularly to a kind of Object Push method and system.
[background technique]
PageRank algorithm is the proprietary algorithm of Google, for measuring particular webpage relative in search engine index
Then the significance level of other webpages is ranked up webpage according to the significance level of webpage.Personalized PageRank, can
When calculating the weighted value of webpage, it is contemplated that specific user, or the weighted value of other webpages is assessed for a webpage,
It thus can satisfy the demand of many practical applications.
However, PageRank algorithm is applied in search engine retrieving webpage, same category of webpage is had the disadvantage that
The weighted value of itself can be passed to the webpages of other classifications, be equivalent to search engine by a variety of different classes of webpages into
Row retrieval, therefore the accuracy of search result is lower and recall precision is lower.
[summary of the invention]
In view of this, may be implemented to improve information inspection the embodiment of the invention provides a kind of Object Push method and system
The accuracy and recall precision of rope.
In a first aspect, the embodiment of the invention provides a kind of Object Push methods, comprising:
Candidate target is obtained, the candidate target includes the input object of client, according to the input pair of the client
As the matching object obtained after being retrieved or according to the recommended of the client clicking historical record and obtaining；
The weight absorptivity of the candidate target is obtained, the weight absorptivity of the candidate target is and the candidate target
The number for other candidate targets being directly linked is in the parameter value of inversely prroportional relationship；Wherein, it is directly linked with the candidate target
Other candidate targets be other candidate targets for being greater than preset dependent thresholds with the degree of correlation of the candidate target；
According to the weight absorptivity of the candidate target, the first weighted value of the candidate target is obtained；
According to first weighted value and the candidate target, target object to be pushed is obtained.
In the first possible implementation of the first aspect, the weight absorptivity according to the candidate target,
Obtain the first weighted value of the candidate target, comprising:
Second weighted value of the candidate target is compared with preset screening threshold value；Second weighted value be with
Other candidate targets that the candidate target is directly linked pass to the weighted value of the candidate target；
If the second weighted value of the candidate target is greater than the screening threshold value, the weight according to the candidate target absorbs
Rate obtains the first weighted value of the candidate target；
If the second weighted value of the candidate target is less than or equal to the screening threshold value, the candidate target is not to pass
Other candidate targets of connection transmit third weighted value, and the third weighted value is equal to the second weighted value and the institute of the candidate target
The difference for stating the second weighted value of candidate target absorption, so that the third weighted value of the candidate target is right in associated candidate
Stop transmitting as between.
With reference to first aspect or the first possible implementation of first aspect, second in first aspect are possible
In implementation, the weight absorptivity according to the candidate target obtains the first weighted value of the candidate target, packet
It includes:
According to the number for other candidate targets being directly linked with the candidate target, and following formula is utilized, obtains institute
State the first weighted value of candidate target:
Wherein, the first weighted value of the S (i) ' expression candidate target i；The S (i) indicates the institute being obtained ahead of time
State the basic weighted value of candidate target i；The α indicates preset parameter value；It is describedIndicate the weight absorptivity, institute
Stating r (i) indicates the second weighted value of the candidate target i, and second weighted value is directly linked with the candidate target i
Other candidate targets pass to the weighted value of the candidate target i；It is describedIndicate what the candidate target i absorbed
Second weighted value；The d (i) indicates the number for other candidate targets that the candidate target i is directly linked.
With reference to first aspect or second of the first possible implementation of first aspect or first aspect possible
Implementation, it is in a third possible implementation of the first aspect, described according to first weighted value and the candidate
Object obtains target object to be pushed, comprising: is compared first weighted value with preset weight threshold, by institute
The first weighted value is stated greater than the candidate target of the weight threshold as the target object to be pushed.
With reference to first aspect or the first or second or the third possible implementation of first aspect, in first party
In the 4th kind of possible implementation in face, the method also includes:
According to the sequence that first weighted value is descending, the target object is ranked up, to obtain sequence knot
Fruit；
Push the ranking results.
Second aspect, the embodiment of the invention provides a kind of Object Push systems, comprising:
Object acquisition unit, for obtaining candidate target, the candidate target includes the input object of client, according to institute
State client input object retrieved after the matching object that obtains or the click historical record according to the client obtain
The recommended obtained；
First processing units, for obtaining the weight absorptivity of the candidate target, the weight of the candidate target absorbs
Rate is that the number for other candidate targets being directly linked with the candidate target is in the parameter value of inversely prroportional relationship；Wherein, with institute
Other candidate targets for stating candidate target direct correlation are to be greater than preset dependent thresholds with the degree of correlation of the candidate target
Other candidate targets；
The second processing unit obtains the first of the candidate target for the weight absorptivity according to the candidate target
Weighted value；
Object screening unit, for obtaining target pair to be pushed according to first weighted value and the candidate target
As.
In the first possible implementation of the second aspect, described the second processing unit is specifically used for:
Second weighted value of the candidate target is compared with preset screening threshold value；Second weighted value be with
Other candidate targets that the candidate target is directly linked pass to the weighted value of the candidate target；
If the second weighted value of the candidate target is greater than the screening threshold value, the weight according to the candidate target absorbs
Rate obtains the first weighted value of the candidate target；
If the second weighted value of the candidate target is less than or equal to the screening threshold value, the candidate target is not to pass
Other candidate targets of connection transmit third weighted value, and the third weighted value is equal to the second weighted value and the institute of the candidate target
The difference for stating the second weighted value of candidate target absorption, so that the third weighted value of the candidate target is right in associated candidate
Stop transmitting as between.
In conjunction with the possible implementation of the first of second aspect or second aspect, second in second aspect is possible
In implementation, described the second processing unit is specifically used for:
According to the number for other candidate targets that the candidate target is directly linked, and following formula is utilized, described in acquisition
First weighted value of candidate target:
Wherein, the first weighted value of the S (i) ' expression candidate target i；The S (i) indicates the institute being obtained ahead of time
State the basic weighted value of candidate target i；The α indicates preset parameter value；It is describedIndicate the weight absorptivity, institute
Stating r (i) indicates the second weighted value of the candidate target i, and second weighted value is directly linked with the candidate target i
Other candidate targets pass to the weighted value of the candidate target i；It is describedIndicate what the candidate target i absorbed
Second weighted value；The d (i) indicates the number for other candidate targets that the candidate target i is directly linked.
Second in conjunction with the possible implementation of the first of second aspect or second aspect or second aspect is possible
Implementation, in the third possible implementation of the second aspect, the object screening unit are specifically used for: by described
One weighted value is compared with preset weight threshold, and the candidate that first weighted value is greater than the weight threshold is right
As the target object.
In conjunction with the first or second or the third possible implementation of second aspect or second aspect, in second party
In the 4th kind of possible implementation in face, the system also includes: object output unit, for according to first weighted value
Descending sequence is ranked up the target object, to obtain ranking results；Push the ranking results.
As can be seen from the above technical solutions, the embodiment of the present invention has the advantages that
In the embodiment of the present invention, other candidates that weight absorptivity and the candidate target of candidate target are directly linked are right
The number of elephant is in inversely prroportional relationship, therefore, other candidate targets that the candidate target according to edge in incidence relation is directly linked
The lesser principle of number, the weight absorptivity of the candidate target at edge is larger, therefore can guarantee weight to a certain extent
It is worth when reaching the edge candidate target of certain classification, less weighted value is only passed to the candidate target of other classifications, it can
Guarantee that weighted value transmits between same category of different candidate targets, in this way, even if the candidate target of certain classification is to other classes
Weighted value is not delivered, the candidate target of other classifications is as the weighted value obtained is smaller, without by as wait push
Target object, so as to guarantee to be retrieved inside the generic of input object to a certain extent, with obtain to
The target object of push, and then the cluster of information retrieval may be implemented, it can be improved the accuracy and retrieval effect of search result
Rate.
[Detailed description of the invention]
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for this field
For those of ordinary skill, without any creative labor, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is the flow diagram of Object Push method provided by the embodiment of the present invention；
Fig. 2 (a)~Fig. 2 (c) is to belong to the distribution of the weighted value of same category of candidate target in the embodiment of the present invention to show
It is intended to；
Fig. 3 is the distribution schematic diagram of candidate target provided by the embodiment of the present invention；
Fig. 4 is the transfer probability schematic diagram of weight between candidate target provided by the embodiment of the present invention；
Fig. 5 is the distribution schematic diagram of the search result of the prior art；
Fig. 6 is the distribution schematic diagram of search result provided by the embodiment of the present invention；
Fig. 7 (a)~Fig. 7 (d) is the first schematic diagram of image retrieval provided by the embodiment of the present invention；
Fig. 8 (a)~Fig. 8 (c) is the second schematic diagram of image retrieval provided by the embodiment of the present invention；
Fig. 9 (a)~Fig. 9 (c) is the second schematic diagram of image retrieval provided by the embodiment of the present invention；
Figure 10 (a)~Figure 10 (d) is that search result is shown under difference weighted average density provided by the embodiment of the present invention
It is intended to；
Figure 11 is the functional block diagram of Object Push system provided by the embodiment of the present invention；
Figure 12 is the structural schematic diagram of Object Push system provided by the embodiment of the present invention.
[specific embodiment]
For a better understanding of the technical solution of the present invention, being retouched in detail to the embodiment of the present invention with reference to the accompanying drawing
It states.
It will be appreciated that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Base
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts it is all its
Its embodiment, shall fall within the protection scope of the present invention.
It is retrieved in mass data (such as webpage and picture), during obtaining and pushing search result, database
In data volume it is often bigger, but only sub-fraction data in mass data relevant to retrieval, for example, with one
Other relevant webpages of webpage are the sub-fraction of entire internet data, so if can only consider in the process of retrieval
Sub-fraction data in mass data will greatly improve the efficiency of retrieval and pushed information.
The embodiment of the present invention provides a kind of Object Push method, referring to FIG. 1, it is right provided by the embodiment of the present invention
As the flow diagram of method for pushing, as shown, method includes the following steps:
Step 101, candidate target is obtained, the candidate target includes the input object of client, according to the client
Input object retrieved after obtain matching object or according to the client click historical record obtain recommendation
Object.
Specifically, client obtains candidate target, the candidate target includes the input object of client, according to the visitor
The matching object or click what historical record obtained according to the client that the input object at family end obtains after being retrieved
Recommended.
Preferably, the client may include browser, application program or image retrieval program etc..
Preferably, the input object of the client includes the keyword inputted in the client or the client
The pixel of upper input.
Preferably, it can be retrieved in the database according to the input object, to obtain of the input object
With object.The matching object may include webpage, application message, Item Information or pixel.Wherein, the matching object
Other objects including other objects being directly linked with the input object or with the input object indirect association；With institute
Other objects for stating input object direct correlation are other for being greater than preset dependent thresholds with the degree of correlation of the input object
Object, other objects with the input object indirect association are other objects by being directly linked with the input object,
With the associated object of the input object.For example, input object A, matching object B and matching object C, wherein input object A with
It matches object B to be directly linked, matching object B is directly linked with object C is matched, then matches object C and input object A and close indirectly
Connection.
Preferably, the recommended includes Item Information or application message.
For example, firstly, from the historical record of browser, obtaining in a period of time recently and using in nearest a period of time
The webpage that family is clicked in a browser；Then the number of clicks of each webpage is counted, and according to descending suitable of number of clicks
Sequence, the webpage clicked to user are ranked up；Finally, by least one webpage in the top in ranking results, it is clear as this
The corresponding recommended of device of looking at.The incidence relation between webpage and webpage is preserved in database, such as webpage A is directed toward webpage B,
It then according to the corresponding webpage of browser, is retrieved in the database, the webpage being directly linked with the webpage can be obtained, in turn
Webpage associated by the webpage with the direct correlation is obtained, the direct correlation net of webpage corresponding with browser may finally be obtained
Page and indirect association webpage.
For example, user has been when mobile terminal has just opened the clients such as Taobao, Amazon, can be pushed to user pair
Image information such as recommends certain article, then can determine that nearest a period of time of user is bought according to the corresponding historical record of client
Article；Then the number of clicks of each article is counted, and is sorted according to the descending sequence of number of clicks, user is clicked
Article be ranked up；Finally, by least one article in the top in ranking results, as the corresponding recommendation of the client
Object.Incidence relation between article and article is preserved in database, if article A and article B are bought simultaneously, article A and object
Product B attribute is similar/mutually same, then according to the corresponding Item Information of client, is retrieved in the database, can obtain and be somebody's turn to do
The article that article is directly linked, so obtain with article associated by the article of the direct correlation, may finally obtain and client
Hold the direct correlation article and indirect association article of corresponding article.
It should be noted that the mode of above-mentioned acquisition candidate target, can obtain several input objects, input object is improved
Diversified degree, realize input distribution, guarantee the diversity of search result.
Step 102, obtain the weight absorptivity of the candidate target, the weight absorptivity of the candidate target be with it is described
The number for other candidate targets that candidate target is directly linked is in the parameter value of inversely prroportional relationship；Wherein, with the candidate target
Other candidate targets being directly linked are right greater than other candidates of preset dependent thresholds with the degree of correlation of the candidate target
As.
Specifically, in the embodiment of the present invention, it is also necessary to be calculated for each candidate target at least one candidate target
One corresponding first weighted value, and before calculating the first weighted value, need first to determine the weight absorptivity of each candidate target.
Wherein, the weight absorptivity of the candidate target refers to that other candidates being directly linked with the candidate target are right
The number of elephant is in the parameter value of inversely prroportional relationship.
Preferably, it can use following formula and obtain weight absorptivity:
In the formula, the α indicates preset parameter value；The d (i) indicates and candidate target i direct correlation
The number of other candidate targets；It can be seen that other candidate targets being directly linked with the candidate target according to the formula
The value of number d (i) is bigger, and the value of weight absorptivity is smaller, the number for other candidate targets being directly linked with the candidate target
The value of mesh d (i) is smaller, and the value of weight absorptivity is bigger, therefore, the weight absorptivity of the candidate target and the candidate target
The number for other candidate targets being directly linked is in inversely prroportional relationship.
Fig. 2 (a) is please referred to, the distribution of the weighted value to belong to same category of candidate target in the embodiment of the present invention is shown
It is intended to, as shown in Fig. 2 (a), a candidate target can be associated with other candidate targets, will form Fig. 2 (a) between candidate target
Shown in association structure, in the embodiment of the present invention, the weight absorptivity of the candidate target and the candidate target are directly linked
Other candidate targets number be in inversely prroportional relationship, as shown in Fig. 2 (a), candidate target 1 direct correlation other candidate targets
Number be 3, i.e., candidate target 2, candidate target 3 and candidate target 4 respectively with candidate target 1 be directly linked；Similarly, candidate right
The number for other candidate targets being directly linked as 2 is 2, i.e., candidate target 5 and candidate target 6 are direct with candidate target 2 respectively
Association.Wherein, the association structure may include the incidence relation between the candidate target i and other candidate targets；It is preferred that
, which can show as reticular structure or tree etc., and details are not described herein.
Referring to FIG. 3, its distribution schematic diagram for candidate target provided by the embodiment of the present invention, as shown, for
For candidate target, the number for other candidate targets that the candidate target closer to association structure center is directly linked is bigger, more
The number for other candidate targets that candidate target close to association structure edge is directly linked is smaller.
Step 103, according to the weight absorptivity of the candidate target, the first weighted value of the candidate target is obtained.
Specifically, each candidate target can have the weighted value of an input, referred to as the second weighted value, second weight
Value is the weighted value that other candidate targets being directly linked with the candidate target pass to the candidate target；Such as Fig. 2 (a) institute
Show, input object, if the second weighted value of both candidate nodes 1 can be pre-assigned, and affiliated partner, such as candidate target 2 is waited
The second weighted value of object 3 etc. is selected to can be the weighted value for passing to oneself with itself other associated candidate target, Mei Gehou
The second weighted value of part can be absorbed by selecting object all, and then candidate target passes to remaining second weighted value with itself directly
Other associated candidate targets, other candidate targets reabsorb the second weighted value that part passes to itself, then by remaining
Second weighted value passes to other candidate targets being directly linked with itself again, and so on；That is, for each candidate
Object can receive second weighted value, then obtain the first weighted value of candidate target, will finally remove in the second weighted value
Part other than first weighted value passes to other candidate targets of direct correlation.
As shown in Fig. 2 (a)~Fig. 2 (c), if the second weighted value to be regarded as to the input flow rate of candidate target, this is defeated
Inbound traffics will be flowed along the incidence relation between candidate target, and as shown in Fig. 2 (a), candidate target 1 has input flow rate, such as be filled out
The circle for filling shade, as shown in Fig. 2 (b), candidate target 1 absorbs the input flow rate of part, then by the input of rest part
Flow passes to candidate target 2, candidate target 3 and the candidate target 4 being directly linked with itself, candidate right as shown in Fig. 2 (c)
Then it is right to be passed to the candidate being directly linked with itself by the input flow rate for absorbing part acquisition as 2 again for remaining input flow rate
As 5 and candidate target 6, the operation of candidate target 3 and candidate target 4 is similar with candidate target 2, and which is not described herein again.
In the embodiment of the present invention, whenever there is candidate target to obtain the second weighted value, just by the second weighted value of candidate target
It is compared with preset screening threshold value；If the second weighted value of the candidate target is greater than the screening threshold value, according to described in
The weight absorptivity of candidate target obtains the first weighted value of the candidate target；If the second weighted value of the candidate target
Less than or equal to the screening threshold value, the candidate target does not transmit third weighted value, institute to other associated candidate targets
Difference of the third weighted value equal to the second weighted value that the second weighted value of the candidate target and the candidate target absorb is stated,
So that the third weighted value of the candidate target stops transmitting between associated candidate target, that is to say, that input flow rate
It will stop transmitting between associated candidate target.Here, threshold value is screened by setting, weighted value can be prevented in association structure
In unlimitedly transmit, so as to reduce the calculation amount of weighted value, for example, the screening threshold value can be 0.001.In addition,
If including that at least two candidate targets can calculate time when calculating the first weighted value of candidate target simultaneously in object set
The first weighted value of object is selected, the first weighted value of candidate target can also be successively calculated.
In the embodiment of the present invention, the second weighted value of the affiliated partner in candidate target is directly linked with affiliated partner
Other candidate targets pass to the affiliated partner, and the second weighted value of the input object in candidate target is to allocate in advance
's.It since the second weighted value of input object is pre-assigned is then each input if there is at least two input objects
The method that object distributes the second weighted value can be mean allocation, i.e., the second weighted value of each input object is identical, can also be with
Ranking according to input object distributes the second weighted value, and the second weighted value of input object in the top is larger, ranks behind
Input object the second weighted value it is smaller, without specifically limiting in the embodiment of the present invention.
Preferably, according to the weight absorptivity of the candidate target, the side of the first weighted value of the candidate target is obtained
Method may include:
According to the number for other candidate targets being directly linked with the candidate target, and following formula (1) is utilized, obtained
First weighted value of the candidate target:
Wherein, the first weighted value of the S (i) ' expression candidate target i；The S (i) indicates the institute being obtained ahead of time
State the basic weighted value of candidate target i；The α indicates preset parameter value；It is describedIndicate the weight absorptivity, institute
Stating r (i) indicates the second weighted value of the candidate target i, and second weighted value is directly linked with the candidate target i
Other candidate targets pass to the weighted value of the candidate target i,Indicate the candidate target i absorbs second
Weighted value；The d (i) indicates the number with the candidate target i other candidate targets being directly linked.Wherein, it is obtained ahead of time
The basic weighted value S (i) of the candidate target i can be understood as candidate target i before obtaining the second weighted value r (i)
The weighted value of itself, in this way, needing to carry out more itself current basic weighted value S (i) after obtaining the second weighted value
Newly, to obtain S (i) '.
It is understood that other candidates that the weight absorptivity of the candidate target and the candidate target are directly linked
The number of object is in inversely prroportional relationship, it may be assumed that the second weighted value is considered as an entirety, the candidate target i be directly linked its
The numerical value of the number d (i) of his candidate target is bigger, and the weight absorptivity of candidate target i is smaller, thus what candidate target i absorbed
Second weighted value is just smaller, and remaining the second weighted value not absorbed by candidate target i is bigger, is left not by candidate target
The second weighted value that i absorbs will be passed to other candidate targets of direct correlation by candidate target i, therefore candidate target i is transmitted
It is bigger to the second weighted value of other candidate targets of direct correlation；Similarly, other times that the candidate target i is directly linked
Select the numerical value of the number d (i) of object smaller, the weight absorptivity of candidate target i is bigger, thus the second of candidate target i absorption
Weighted value is bigger, and remaining the second weighted value not absorbed by candidate target i is just smaller, remaining not inhaled by candidate target i
The second weighted value received will be passed to other candidate targets of direct correlation by candidate target i, therefore candidate target i is passed to directly
The second weighted value of other associated candidate targets is connect with regard to smaller.As shown in figure 3, for candidate target, closer to association
The weight absorptivity of the candidate target of structure centreSmaller, the weight closer to the candidate target at association structure edge is inhaled
YieldIt is bigger, thus it is guaranteed that waiting during the extrorse candidate target of candidate target at association structure center
Select the weight absorptivity of object i increasing, the second weighted value of residue of the candidate target transmitting at edge can be fewer, and belongs to
Same category of candidate target forms an association structure, thereby may be ensured that weighted value is retained in each of current class as far as possible
In candidate target, once weighted value reaches the candidate target at association structure edge, as the weight of the candidate target at edge
Absorptivity is larger, keeps the weighted value for the candidate target for being transmitted to other classifications very small, therefore, ensure that power to a certain extent
Weight values can be retained on the candidate target of current class, can greatly reduce the classification of retrieval, improve the accurate of search result
Degree and recall precision.
Referring to FIG. 4, the transfer probability schematic diagram of its weight between candidate target provided by the embodiment of the present invention, such as schemes
Shown, by taking candidate target k, candidate target i and candidate target j as an example, weighted value is transferred to candidate target i again by candidate target k
The probability of happening for being transferred to candidate target j meets following second order Markov chain models:
T moment, the second weighted value are located at candidate target k, and candidate target k retains the second weighted value of part, and in t+1
It carves candidate target k and remaining second weighted value is passed into candidate target i, then, candidate target i retains remaining second power again
Part in weight values, and the remainder in remaining second weighted value is passed into candidate target in t+2 moment candidate target i
J, the probability of happening of the transmitting event of above-mentioned second weighted value can as shown in above-mentioned second order Markov chain model, as i=j and
When i=k, indicate there was only candidate target k, i.e. t moment, t+1 moment and t+2 moment, the second weighted value all rests on candidate always
Object k, therefore, the probability of happening of the transmitting event of above-mentioned second weighted value is 1；As i ≠ j and i=k, t moment and when t+1
It carves, the second weighted value rests on candidate target k, and therefore, the second weighted value will not be transferred to candidate target j by candidate target k, because
, at the t+2 moment, the probability of happening of candidate target k to the event of the second weighted value of candidate target j transmitting portions is 0 for this；When i ≠
When k, no matter as i=j or i ≠ j, at the t+1 moment, the second weighted value can be transferred to candidate target i by candidate target k, in t+2
At the moment, the probability of happening that the second weighted value can be transferred to the event of candidate target j by candidate target i is Pij。
Preferably, it using following formula, obtains candidate target i and passes to other times being directly linked with candidate target i
Select the weighted value r (j) ' of object j:
R (j) '=r (j)+r (i) w (i, j)/(α+d (i))
Wherein, the r (j) indicates the basic weighted value of candidate target j；The r (i) indicates the of the candidate target i
Two weighted values；The α indicates preset parameter value；The d (i) indicates other candidates being directly linked with the candidate target i
The number of object.
It should be noted that in the candidate target of a classification, if association structure G=(V, W), V indicate candidate target
Set, W=w (i, j) indicate the degree of association matrix of candidate target i and candidate target j, and the degree of association matrix w (i, j) is for indicating
Candidate target i is associated with tightness with candidate target j's, if candidate target i and candidate target j is directly linked, w (i, j) etc.
In 1, if candidate target i and candidate target j are not to be directly linked, w (i, j) is equal to 0.The d (i) of candidate target i be equal to w (i,
J) cumulative and.
Alternatively, the weight absorptivity according to the candidate target obtains the candidate target using following formula (2)
The method of first weighted value may include:
S (i) '=[α (α × I+L)-1r0]i
Wherein, the α indicates preset parameter value, the r0Indicate the candidate target i obtain the second weighted value it
Itself preceding basic weighted value, second weighted value are that other candidate targets being directly linked with the candidate target i transmit
To the weighted value of the candidate target i；I indicates unit matrix；L=D-W, L indicate the Laplacian Matrix of association structure,
In, D=diag (d (1) ..., d (n)), wherein D=diag () indicates that diagonal matrix, d (i) indicate right with the candidate
As i be directly linked other candidate targets number, i=1,2 ..., n；The total number of n expression candidate target.
It should be noted that in formula (2), [α (α × I+L)-1r0] indicate that the first weighted value of all candidate targets forms
Matrix, [α (α × I+L)-1r0]iMean that the first weighted value of i-th of candidate target in the matrix.
Formula (2) is suitable for association structure the case where candidate target negligible amounts, if candidate target in association structure
Quantity is more, is also calculated using formula (2), it is bigger to will lead to computational throughput；So if candidate in association structure
Number of objects is more, can choose the first weighted value that candidate target is calculated using formula (1).
Step 104, according to first weighted value and the candidate target, target object to be pushed is obtained.
Specifically, needing after the first weighted value for obtaining candidate target by first weighted value and preset weight
Threshold value is compared, and first weighted value is greater than the candidate target of the weight threshold as the mesh to be pushed
Object is marked, the candidate target that the first weighted value is less than or equal to weight threshold is screened out.
Optionally, sequence that can be descending according to first weighted value, is ranked up the target object, with
Obtain ranking results；Push the ranking results.
Referring to FIG. 5, its distribution schematic diagram for the search result of the prior art, as shown, point indicated by circle 1
Indicate that input object, all points indicate the search result of output, i.e., above-mentioned target object, however, point indicated by circle 2 is
The most close search result with input object incidence relation is the most desirable search result of user；Circle 2, circle 3 and circle 4
In, the weighted value closer to intermediate point is higher, and the weighted value closer to the point at edge is lower, moreover, in all search results,
Incidence relation wherein more between the point and input object of right-hand component is smaller, if examined as search result by reducing
The accuracy of hitch fruit.
Referring to FIG. 6, its distribution schematic diagram for search result provided by the embodiment of the present invention, as shown, in Fig. 6
All points indicate the search result of output, i.e., above-mentioned target object, and the horizontal axis in Fig. 6 indicates the number of candidate target, such as wait
1~candidate target of object 300,300~candidate target of candidate target 600, candidate target 600 to candidate target 900 are selected, Fig. 6's
Vertical pivot indicates the size for the second weighted value that candidate target absorbs.As shown in fig. 6, what 1~candidate target of candidate target 300 absorbed
Second weighted value is bigger, indicate candidate target~candidate target 300 be with the most close candidate target of input object relationship, because
This, it can be seen that 100~candidate target of candidate target 300 absorbs most of second weighted value, so that output is to other
Second weighted value of classification is fewer, therefore search method provided by the embodiment of the present invention can embody the cluster of candidate target
Retrieval.
It is the first signal of image retrieval provided by the embodiment of the present invention in addition, please referring to Fig. 7 (a)~Fig. 7 (d)
Figure, Fig. 7 (a) indicate that original image, Fig. 7 (b) indicate input object, can be in all pixels in Fig. 7 (a) according to input object
Matching retrieval is carried out in point, input object is several scattered pixels in bird pattern, and Fig. 7 (c) indicates that weighted value is being associated with
The schematic diagram of candidate target when balanced in structure can probably find out that the profile of search result, Fig. 7 (d) indicate final and need
The target object of push.
Fig. 8 (a)~Fig. 8 (c) is please referred to, is the second schematic diagram of image retrieval provided by the embodiment of the present invention, Fig. 8
(a) original image is indicated, wherein cross indicates input object, retrieval can match in Fig. 8 (a) according to the input object, defeated
Entering object is a pixel, and Fig. 8 (b) indicates the schematic diagram of candidate target when weighted value is balanced in association structure, greatly
It can generally find out that the profile of search result, Fig. 8 (c) indicate target object.
Fig. 9 (a)~Fig. 9 (c) is please referred to, is the second schematic diagram of image retrieval provided by the embodiment of the present invention, Fig. 9
(a) it indicates that original image, Fig. 9 (b) indicate input object, retrieval can be matched in Fig. 9 (a) according to the input object, input
The quantity of object is more, causes image relatively fuzzyyer, Fig. 9 (c) indicates target object.Figure 10 (a)~Figure 10 (d) is please referred to,
For the schematic diagram of search result under difference weighted average density provided by the embodiment of the present invention, Figure 10 (a)~Figure 10 (d) can
The robustness for embodying the technical solution of the embodiment of the present invention well is used as threshold value for 1 times of weighted mean, to candidate right
As being screened, obtain target object shown in Figure 10 (a), be used as threshold value for 1.5 times of weighted mean, to candidate target into
Row screening, obtains target object shown in Figure 10 (b), is used as threshold value for 2 times of weighted mean, sieves to candidate target
Choosing obtains target object shown in Figure 10 (c), is used as threshold value for 2.5 times of weighted mean, screens to candidate target,
Target object shown in Figure 10 (d) is obtained, here, although the input of retrieval is blurred picture, it is clear still to be able to obtain
Ground target image, and constantly becoming larger with threshold value, the target image of acquisition is more and more clear, the number of the target object of acquisition
Measure fewer and fewer, the weighted value of the pixel (pixel in Figure 10 (d)) inside the target image of acquisition is noticeably greater than weight
Average value.
The embodiment of the present invention, which further provides, realizes the Installation practice of each step and method in above method embodiment.
Figure 11 is please referred to, is the functional block diagram of Object Push system provided by the embodiment of the present invention.As shown,
The system includes:
Object retrieval unit 11, for obtaining candidate target, the candidate target includes the input object of client, foundation
The matching object or the click historical record according to the client that the input object of the client obtains after being retrieved
The recommended of acquisition；
First processing units 12, for obtaining the weight absorptivity of the candidate target, the weight of the candidate target is inhaled
Yield is that the number for other candidate targets being directly linked with the candidate target is in the parameter value of inversely prroportional relationship；Wherein, with
Other candidate targets that the candidate target is directly linked are to be greater than preset dependent thresholds with the degree of correlation of the candidate target
Other candidate targets；
The second processing unit 13 obtains the of the candidate target for the weight absorptivity according to the candidate target
One weighted value；
Object screening unit 14, for obtaining target to be pushed according to first weighted value and the candidate target
Object.
Preferably, described the second processing unit 13 is specifically used for:
Second weighted value of the candidate target is compared with preset screening threshold value；Second weighted value be with
Other candidate targets that the candidate target is directly linked pass to the weighted value of the candidate target；
If the second weighted value of the candidate target is greater than the screening threshold value, the weight according to the candidate target absorbs
Rate obtains the first weighted value of the candidate target；
If the second weighted value of the candidate target is less than or equal to the screening threshold value, the candidate target is not to pass
Other candidate targets of connection transmit third weighted value, and the third weighted value is equal to the second weighted value and the institute of the candidate target
The difference for stating the second weighted value of candidate target absorption, so that the third weighted value of the candidate target is right in associated candidate
Stop transmitting as between.
Preferably, described the second processing unit 13 is specifically used for:
According to the number for other candidate targets that the candidate target is directly linked, and following formula is utilized, described in acquisition
First weighted value of candidate target:
Wherein, the first weighted value of the S (i) ' expression candidate target i；The S (i) indicates the institute being obtained ahead of time
State the basic weighted value of candidate target i；The α indicates preset parameter value；It is describedIndicate the weight absorptivity, institute
Stating r (i) indicates the second weighted value of the candidate target i, and second weighted value is directly linked with the candidate target i
Other candidate targets pass to the weighted value of the candidate target i；It is describedIndicate what the candidate target i absorbed
Second weighted value；The d (i) indicates the number for other candidate targets that the candidate target i is directly linked.
Preferably, the object screening unit 14, is specifically used for: by first weighted value and preset weight threshold into
Row compares, and first weighted value is greater than the candidate target of the weight threshold as the target object.
Preferably, the system also includes:
Object output unit 15, for the sequence descending according to first weighted value, to the target object into
Row sequence, to obtain ranking results；Push the ranking results.
Figure 12 is please referred to, is the structural schematic diagram of Object Push system provided by the embodiment of the present invention.As shown,
The equipment includes:
Memory 20, for storing one or more groups of program codes；
Processor 21 is coupled with memory 20, for calling the program code stored in memory 20, to execute with the following figure
Method shown in 1, specifically includes: obtaining candidate target, the candidate target includes the input object of client, according to the visitor
The matching object or click what historical record obtained according to the client that the input object at family end obtains after being retrieved
Recommended；The weight absorptivity of the candidate target is obtained, the weight absorptivity of the candidate target is right with the candidate
As the number of other candidate targets of direct correlation is in the parameter value of inversely prroportional relationship；Wherein, it is directly closed with the candidate target
Other candidate targets of connection are other candidate targets for being greater than preset dependent thresholds with the degree of correlation of the candidate target；Foundation
The weight absorptivity of the candidate target obtains the first weighted value of the candidate target；According to first weighted value and institute
Candidate target is stated, target object to be pushed is obtained.
Method shown in FIG. 1 is able to carry out by processor in this present embodiment, the part that the present embodiment is not described in detail,
It can refer to the related description to Fig. 1.
The technical solution of the embodiment of the present invention has the advantages that
The number for other candidate targets that the weight absorptivity of candidate target and the candidate target are directly linked is in inverse ratio
Example relationship, therefore, the lesser original of number for other candidate targets that the candidate target according to edge in incidence relation is directly linked
Reason, the weight absorptivity of the candidate target at edge is larger, therefore can guarantee that weighted value reaches certain classification to a certain extent
Edge candidate target when, less weighted value is only passed to the candidate target of other classifications, weighted value can be guaranteed same
It is transmitted between a kind of other different candidate targets, in this way, even if the candidate target of certain classification delivers weighted value to other classifications,
The candidate targets of other analogies is as the weighted value obtained is smaller, without by as target object to be pushed, thus
It can guarantee to be retrieved inside the generic of input object to a certain extent, to obtain target object to be pushed,
And then the cluster of information retrieval may be implemented, it can be improved the accuracy and recall precision of search result.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the present invention.
Claims (8)
1. a kind of Object Push method, which is characterized in that the described method includes:
Obtain candidate target, the candidate target include client input object, according to the client input object into
The matching object obtained after row retrieval or the recommended clicking historical record and obtaining according to the client；
The weight absorptivity of the candidate target is obtained, the weight absorptivity of the candidate target is direct with the candidate target
The number of other associated candidate targets is in the parameter value of inversely prroportional relationship；Wherein, with the candidate target be directly linked its
His candidate target is other candidate targets for being greater than preset dependent thresholds with the degree of correlation of the candidate target；
According to the weight absorptivity of the candidate target, the first weighted value of the candidate target is obtained；
According to first weighted value and the candidate target, target object to be pushed is obtained.
2. the method according to claim 1, wherein the weight absorptivity according to the candidate target, is obtained
Obtain the first weighted value of the candidate target, comprising:
Second weighted value of the candidate target is compared with preset screening threshold value；Second weighted value be with it is described
Other candidate targets that candidate target is directly linked pass to the weighted value of the candidate target；
If the second weighted value of the candidate target is greater than the screening threshold value, according to the weight absorptivity of the candidate target,
Obtain the first weighted value of the candidate target；
If the second weighted value of the candidate target is less than or equal to the screening threshold value, the candidate target is not to associated
Other candidate targets transmit third weighted value, second weighted value of the third weighted value equal to the candidate target and the time
Select object absorb the second weighted value difference so that the third weighted value of the candidate target associated candidate target it
Between stop transmitting；
Wherein, the weight absorptivity according to the candidate target, obtains the first weighted value of the candidate target, comprising:
According to the number for other candidate targets being directly linked with the candidate target, and following formula is utilized, obtains the time
Select the first weighted value of object:
Wherein, the first weighted value of the S (i) ' expression candidate target i；The S (i) indicates the time being obtained ahead of time
Select the basic weighted value of object i；The α indicates preset parameter value；It is describedIndicate the weight absorptivity, the r
(i) indicate that the second weighted value of the candidate target i, second weighted value are its being directly linked with the candidate target i
His candidate target passes to the weighted value of the candidate target i；It is describedIndicate that the candidate target i absorbs the
Two weighted values；The d (i) indicates the number for other candidate targets that the candidate target i is directly linked.
3. the method according to claim 1, wherein described right according to first weighted value and the candidate
As obtaining target object to be pushed, comprising:
First weighted value is compared with preset weight threshold, first weighted value is greater than the weight threshold
The candidate target as the target object to be pushed.
4. the method according to claim 1, wherein the method also includes:
According to the sequence that first weighted value is descending, the target object is ranked up, to obtain ranking results；
Push the ranking results.
5. a kind of Object Push system, which is characterized in that the system comprises:
Object acquisition unit, for obtaining candidate target, the candidate target includes the input object of client, according to the visitor
The matching object or click what historical record obtained according to the client that the input object at family end obtains after being retrieved
Recommended；
First processing units, for obtaining the weight absorptivity of the candidate target, the weight absorptivity of the candidate target is
The number for other candidate targets being directly linked with the candidate target is in the parameter value of inversely prroportional relationship；Wherein, with the time
Other candidate targets for selecting object to be directly linked are other for being greater than preset dependent thresholds with the degree of correlation of the candidate target
Candidate target；
The second processing unit obtains the first weight of the candidate target for the weight absorptivity according to the candidate target
Value；
Object screening unit, for obtaining target object to be pushed according to first weighted value and the candidate target.
6. system according to claim 5, which is characterized in that described the second processing unit is specifically used for:
Second weighted value of the candidate target is compared with preset screening threshold value；Second weighted value be with it is described
Other candidate targets that candidate target is directly linked pass to the weighted value of the candidate target；
If the second weighted value of the candidate target is greater than the screening threshold value, according to the weight absorptivity of the candidate target,
Obtain the first weighted value of the candidate target；
If the second weighted value of the candidate target is less than or equal to the screening threshold value, the candidate target is not to associated
Other candidate targets transmit third weighted value, second weighted value of the third weighted value equal to the candidate target and the time
Select object absorb the second weighted value difference so that the third weighted value of the candidate target associated candidate target it
Between stop transmitting；
Wherein, described the second processing unit is specifically used for:
According to the number for other candidate targets that the candidate target is directly linked, and following formula is utilized, obtains the candidate
First weighted value of object:
Wherein, the first weighted value of the S (i) ' expression candidate target i；The S (i) indicates the time being obtained ahead of time
Select the basic weighted value of object i；The α indicates preset parameter value；It is describedIndicate the weight absorptivity, the r
(i) indicate that the second weighted value of the candidate target i, second weighted value are its being directly linked with the candidate target i
His candidate target passes to the weighted value of the candidate target i；It is describedIndicate that the candidate target i absorbs the
Two weighted values；The d (i) indicates the number for other candidate targets that the candidate target i is directly linked.
7. system according to claim 5, which is characterized in that the object screening unit is specifically used for:
First weighted value is compared with preset weight threshold, first weighted value is greater than the weight threshold
The candidate target as the target object.
8. system according to claim 5, which is characterized in that the system also includes:
Object output unit is ranked up the target object for the sequence descending according to first weighted value,
To obtain ranking results；Push the ranking results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410231465.8A CN105335363B (en) | 2014-05-28 | 2014-05-28 | A kind of Object Push method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410231465.8A CN105335363B (en) | 2014-05-28 | 2014-05-28 | A kind of Object Push method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105335363A CN105335363A (en) | 2016-02-17 |
| CN105335363B true CN105335363B (en) | 2018-12-07 |
Family
ID=55285907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410231465.8A Active CN105335363B (en) | 2014-05-28 | 2014-05-28 | A kind of Object Push method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105335363B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107545450A (en) * | 2016-06-27 | 2018-01-05 | 百度在线网络技术（北京）有限公司 | A kind of Transaction Information method for pushing and device |
| CN107945079A (en) * | 2016-10-12 | 2018-04-20 | 普天信息技术有限公司 | A kind of poverty alleviation object selection method and device |
| CN108108419B (en) * | 2017-12-15 | 2021-01-26 | 百度在线网络技术（北京）有限公司 | Information recommendation method, device, equipment and medium |
| US10460359B1 (en) | 2019-03-28 | 2019-10-29 | Coupang, Corp. | Computer-implemented method for arranging hyperlinks on a graphical user-interface |
| CN110413751B (en) * | 2019-07-18 | 2023-11-07 | 腾讯科技（深圳）有限公司 | Medicine searching method, device, terminal equipment and storage medium |
| CN111738754A (en) * | 2020-06-01 | 2020-10-02 | 上海晶赞融宣科技有限公司 | Object recommendation method and device, storage medium and computer equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1996299A (en) * | 2006-12-12 | 2007-07-11 | 孙斌 | Ranking method for web page and web site |
| CN101105799A (en) * | 2007-08-02 | 2008-01-16 | 华中科技大学 | Method for evaluating file importance degree |
| CN101770521A (en) * | 2010-03-11 | 2010-07-07 | 东华大学 | Focusing relevancy ordering method for vertical search engine |
| CN102270246A (en) * | 2011-09-08 | 2011-12-07 | 胡辉 | Method for calculating importance of web page |
| CN103559258A (en) * | 2013-11-04 | 2014-02-05 | 同济大学 | Webpage ranking method based on cloud computation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110270847A1 (en) * | 2010-05-01 | 2011-11-03 | Adam Etkin | Method and system for appraising the extent to which a publication has been reviewed by means of a peer-review process |
-
2014
- 2014-05-28 CN CN201410231465.8A patent/CN105335363B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1996299A (en) * | 2006-12-12 | 2007-07-11 | 孙斌 | Ranking method for web page and web site |
| CN101105799A (en) * | 2007-08-02 | 2008-01-16 | 华中科技大学 | Method for evaluating file importance degree |
| CN101770521A (en) * | 2010-03-11 | 2010-07-07 | 东华大学 | Focusing relevancy ordering method for vertical search engine |
| CN102270246A (en) * | 2011-09-08 | 2011-12-07 | 胡辉 | Method for calculating importance of web page |
| CN103559258A (en) * | 2013-11-04 | 2014-02-05 | 同济大学 | Webpage ranking method based on cloud computation |
Non-Patent Citations (3)
| Title |
|---|
| "Personalized Web Page Ranking Using Trust and Similarity";Lara Srour ET AL;《IEEE》;20071231;全文 * |
| "基于PageRank算法的中文垂直搜索引擎的研究与应用";杨晨;《中国优秀硕士学位论文全文数据库 信息科技辑》;20140515;全文 * |
| "基于链接结构的中文网页排序算法研究";药成刚;《中国优秀硕士学位论文全文数据库 信息科技辑》;20071215;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105335363A (en) | 2016-02-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105335363B (en) | A kind of Object Push method and system | |
| US9886517B2 (en) | Ranking product information | |
| US8332775B2 (en) | Adaptive user feedback window | |
| CN104965905B (en) | A kind of method and apparatus of Web page classifying | |
| CN106815263B (en) | The searching method and device of legal provision | |
| US20150213368A1 (en) | Information recommendation method, apparatus, and server | |
| CN103902597B (en) | The method and apparatus for determining relevance of searches classification corresponding to target keyword | |
| WO2015192667A1 (en) | Advertisement recommending method and advertisement recommending server | |
| CN106570144A (en) | Method and apparatus for recommending information | |
| CN105653701B (en) | Model generating method and device, word assign power method and device | |
| CN102163230B (en) | Method for implementing personalized information retrieval system for customizing privacy protection | |
| CN105468598A (en) | Friend recommendation method and device | |
| CN108319672A (en) | Mobile terminal malicious information filtering method and system based on cloud computing | |
| CN108733756A (en) | Data preload method, apparatus, electronic equipment and readable storage medium storing program for executing | |
| CN106777143A (en) | A kind of news Aggreagation method and news Aggreagation server | |
| CN105574030A (en) | Information search method and device | |
| CN106095939A (en) | The acquisition methods of account authority and device | |
| CN106130756A (en) | A kind of method and device predicting access content clicking rate | |
| CN106936778A (en) | The abnormal detection method of website traffic and device | |
| CN103955480A (en) | Method and equipment for determining target object information corresponding to user | |
| CN109740621B (en) | Video classification method, device and equipment | |
| CN106021423B (en) | META Search Engine personalization results recommended method based on group division | |
| CN104123321B (en) | A kind of determining method and device for recommending picture | |
| CN107026755A (en) | A kind of service recommendation method based on sequence study | |
| CN105786810A (en) | Method and device for establishment of category mapping relation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |