CN105302898B - A kind of search ordering method and device based on click model - Google Patents
A kind of search ordering method and device based on click model Download PDFInfo
- Publication number
- CN105302898B CN105302898B CN201510697625.2A CN201510697625A CN105302898B CN 105302898 B CN105302898 B CN 105302898B CN 201510697625 A CN201510697625 A CN 201510697625A CN 105302898 B CN105302898 B CN 105302898B
- Authority
- CN
- China
- Prior art keywords
- result
- sequence
- score value
- result items
- items
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of search ordering methods based on click model, the method comprising the steps of: obtaining the first result sequence about inquiry, each result items have the first score value in first result sequence, and all result items sort according to the sequence of the first score value from high to low, the first score value is calculated according at least one predetermined characteristic;The second score value of each result items in the first result sequence is determined using click model;The result items that the second score value in the first result sequence is not 0 are chosen, based on the sequence of the selected result items of the second score value adjustment, obtain the second result sequence.The invention also discloses a kind of searching order device based on click model.
Description
Technical field
The present invention relates to search engine, especially a kind of search ordering method and device based on click model.
Background technique
Effect of the internet for economic society extends to driving big data and generates valence from abatement information asymmetry
Value.In this process, search is the important means that people obtain information and data from internet always, therefore becomes internet
Important entrance.The retrieval relevance for promoting search engine, is an important research direction of information retrieval field.In recent years,
Click model in information retrieval field achieves tremendous development.So-called click model is exactly drawn using user using search
Click data when holding up calculates the mathematical model of search result and user query correlation.It makes search system in ranking results
When, the ability using former user's click information is obtained, the result so as to make correlation high arranges to be located further forward.
Adjust power with good dependence query as a result, but clicking tune power and discomfort although carrying out click using click model
Close the weight order individually determined in search.Firstly, clicking rate height and correlation it is good be not fully the same thing;Secondly, theoretical
The essence of upper sequence is information computing its correlation that reasonable employment result includes, and clicking has the characteristics that sparsity, i.e.,
User clicked the sub-fraction that the inquiry of result is only all inquired, and the result that user clicked in an inquiry is
The sub-fraction of this query result, therefore it can be limited to calculate the information content of correlation contribution.To sum up, clicking adjusts power to answer
It should be used together, i.e., it must be dissolved into existing sequence frame with the existing other feature of search engine.But it clicks and adjusts
Power participate in sequence, often lead to " positive feedback " problem, that is, be clicked adjust power be discharged to front as a result, often obtaining higher point
It hits, to be easier to be discharged to front.
Therefore, how will click on tune power reasonably to incorporate in existing sequence frame, be this hair without causing adverse effect
It is bright to solve the problems, such as.
Summary of the invention
For this purpose, the present invention provides a kind of search ordering method and device based on click model, to try hard to solve or extremely
It is few to alleviate at least one existing problem above.
According to an aspect of the invention, there is provided a kind of search ordering method based on click model, this method include
Step: obtaining the first result sequence about inquiry, and each result items have the first score value, and all results in the first result sequence
Item sorts according to the sequence of the first score value from high to low, and the first score value is calculated according at least one predetermined characteristic;Using point
Hit the second score value that model determines each result items in the first result sequence;Choosing the second score value in the first result sequence is not 0
Result items obtain the second result sequence based on the sequence of the selected result items of the second score value adjustment.
Optionally, in the search ordering method according to the present invention based on click model, predetermined characteristic include viewing amount,
One or more of issuing time and money order receipt to be signed and returned to the sender quantity.
Optionally, in the search ordering method according to the present invention based on click model, click model is that series connection is clicked
Model.
Optionally, in the search ordering method according to the present invention based on click model, institute is adjusted based on the second score value
The step of sequence of the result items of selection include: for the second score value be greater than threshold value result items, according to the second score value from height to
The sequence at bottom sorts;It is not more than the result items of threshold value for the second score value, holding sequence is constant, and comes the second score value greater than threshold
After all result items of value.
Optionally, in the search ordering method according to the present invention based on click model, threshold value is series connection click model
Codomain mean value.
Optionally, it in the search ordering method according to the present invention based on click model, further comprises the steps of: the second knot
The first score value of the changed result items in position, is adjusted to the corresponding first result sequence in the existing position of the result items in infructescence column
First score value of result items in column.
According to another aspect of the present invention, a kind of searching order device based on click model is provided, which includes:
Acquiring unit, suitable for obtaining the first result sequence about inquiry, wherein each result items have first point in the first result sequence
Value;Computing unit, suitable for the first score value is calculated according at least one predetermined characteristic, is further adapted for determining institute using click model
State the second score value of each result items in the first result sequence;And sequencing unit, it is suitable for all results in the first result sequence
Item sorts according to the sequence of the first score value from high to low, is further adapted for choosing the result that the second score value in the first result sequence is not 0
, based on the sequence of the selected result items of the second score value adjustment, to obtain the second result sequence.
Optionally, in the searching order device according to the present invention based on click model, predetermined characteristic include viewing amount,
One or more of issuing time and money order receipt to be signed and returned to the sender quantity.
Optionally, in the searching order device according to the present invention based on click model, click model is that series connection is clicked
Model.
Optionally, in the searching order device according to the present invention based on click model, sequencing unit further includes judgement
Subelement, second score value selected suitable for judgement are not in 0 result items, and whether the second score value is greater than threshold value;Sequencing unit
It is further adapted for being greater than the second score value the result items of threshold value, sorts according to the sequence of the second score value from high in the end;And for second
Score value is not more than the result items of threshold value, and holding sequence is constant, and comes the second score value and be greater than after all result items of threshold value.
Optionally, in the searching order device according to the present invention based on click model, threshold value is series connection click model
Codomain mean value.
Optionally, in the searching order device according to the present invention based on click model, sequencing unit is further adapted for
First score value of the changed result items in position in two result sequences is adjusted to corresponding first knot in the existing position of the result items
First score value of result items in infructescence column.
According to another aspect of the present invention, a kind of information search engine system is provided, comprising: information bank, suitable for depositing
Store up information to be put;Searching order device based on click model as described above, suitable for the result sequence obtained to inquiry
It is ranked up;And information display device, suitable for showing query result in order.Searching based on click model according to the present invention
Rope sequencing schemes, in conjunction with the model feature of series connection click model, reasonably will click on tune power on the basis of old collating sequence
It is dissolved into the frame of searching order, changes position and the weight of result items, so that the final display order of query result, not only
Click model is reflected to the positive effect of correlation, it is thus also avoided that positive feedback disadvantage.
In addition, according to the solution of the present invention, maintaining the second result sequence also according to the property of old sequence, so that searching
All multioperations based on the property in holding up are indexed, such as this sequence is merged with the result sequence of other parallel search engines
Get up minor sort again together, after using this method still effectively.
Detailed description of the invention
To the accomplishment of the foregoing and related purposes, certain illustrative sides are described herein in conjunction with following description and drawings
Face, these aspects indicate the various modes that can practice principles disclosed herein, and all aspects and its equivalent aspect
It is intended to fall in the range of theme claimed.Read following detailed description in conjunction with the accompanying drawings, the disclosure it is above-mentioned
And other purposes, feature and advantage will be apparent.Throughout the disclosure, identical appended drawing reference generally refers to identical
Component or element.
Fig. 1 shows search engine according to an embodiment of the invention in the exemplary environments 100 wherein run;
The flow chart of Fig. 2 shows the according to an embodiment of the invention search ordering method 200 based on click model;
Fig. 3 shows click model schematic diagram according to an embodiment of the invention;
Fig. 4 shows the exemplary conceptual diagram of sort method according to an embodiment of the invention;And
Fig. 5 shows the schematic diagram of the searching order device 500 according to an embodiment of the invention based on click model.
Specific embodiment
Exemplary embodiments of the present disclosure are described in more detail below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
It is limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
It is fully disclosed to those skilled in the art.
Fig. 1 shows search engine can be in the exemplary environments 100 wherein run.Environment 100 includes by network 130, example
The one or more clients 110 being connected to each other such as internet, wide area network (WAN) or local area network (LAN) and one or more clothes
It is engaged in device 120 (usually " host ").Network 130 provides the access of the service to such as WWW (" web ") 131.
Web 131 allow client 110 access be included in for example safeguarded and serviced by server 120 webpage 121 (such as
Webpage or other documents) in the document based on text or multimedia content.In general, this is by executing in client 110
Web browser application program 114 is completed.The position of each page 121 can be by being such as input to web browser application journey
To access webpage 121 in sequence 114.Many webpages may include the hyperlink 123 to other webpages 121.Hyperlink 123 can also be with
It is the form of URL.Although there has been described the document realizations about the page it should be appreciated that environment 100 may include tool
There are the content that can be characterized and internuncial any link data object.
It will be appreciated by those skilled in the art that in general, search engine 140 corresponds to be calculated in one or more
The online service of trustship in machine and/or computing system, wherein said one or multiple computers or computing system are in entire net
In network 130 position and/or be distributed.The search inquiry that the search engine 140 receives and customer in response end 110 is submitted.Particularly,
In response to inquiry, which obtains related and/or related to received search inquiry (being defined by the item of search inquiry)
Search result information, i.e. result set 112.The result set 112 includes search result, i.e., to can from a variety of different network sites
The reference (typically, in the form of hyperlink) of obtained related and/or relevant content, wherein above-mentioned network site includes all
The content hosting website such as positioned in whole network 130.
As the skilled person will recognize, the trustship of content hosting website or storage pass through network 130 to client
Holding for 110 users is available and/or addressable content.By using the process for climbing the network sweep for content is grabbed, search
Index, which holds up 140, will be appreciated that at least one of the content of trustship on the multiple content hosting websites positioned in whole network 130
Point.Once located content, which will be equivalent to information bank 142 in content repository, store about trustship
The information of content.In response to inquiry, which extracts from information bank 142, returns to the item (example for meeting inquiry
Such as keyword) result set 112.
Since search engine 140 stores the page up to a million, especially when inquiry is loosely specified, result set 112
It may include the page of many qualifications.These pages can be related or unrelated with the actual information demand of user.Therefore, to client
The sequence for the result set 112 that end 110 is presented influences experience of the user about search engine 140.
In one implementation, a part that sequencer procedure can be used as the ranking engine 144 in search engine 140 is come real
It is existing, for example, the searching order device 500 in this programme.In some implementations, sequencer procedure can be based on click logs,
To improve the sequence of the page in result set 112, the page 113 relevant to specific topics can be identified more accurately in this way.Most
Afterwards, the page in result set 112 is presented to the user by improved sequence by information display device 146.
It has been found that user be more likely to click the higher page of ranking, but regardless of the page whether actually with inquiry phase
It closes.This is referred to as position deviation.Attempt to solve a kind of click model of position deviation to be position click model.The model hypothesis is only
When user is practical browse result and obtain a result and search for relevant conclusion when just click result.That is, when user browses
When as a result and thinking its correlation, user only perceive this result is that relevant, rather than knows really.Only when user's actual click knot
When fruit and browsing pages or document itself, whether user can understand result practical related.
The model that another kind is distinguished between the reality and perceived relevance of result is series connection click model.For one
Secondary inquiry, it is successively to browse to be attracted as a result, working as user by some result in sequence that model hypothesis user is clicked in series connection, user
It clicks on the result to be inquired, and can there is a probability to terminate inquiry and no longer browse.
Although above-mentioned click model solves the problems, such as position deviation, user clicks behavior cannot use click information completely
Amount is to explain.Therefore, this programme proposes a kind of search ordering method based on click model, and user is reasonably clicked behavior
Information content be dissolved into the frame of searching order.
The flow chart of Fig. 2 shows the according to an embodiment of the invention search ordering method 200 based on click model.
This method starts from step S210, when a user query is received, obtains the first result sequence about user query, and at this
Each result items all have the first score value in first result sequence, and all result items are arranged according to the sequence of the first score value from high to low
Sequence.Embodiment according to the present invention, the first score value are calculated according at least one predetermined characteristic, and predetermined characteristic includes
One or more of viewing amount, issuing time and money order receipt to be signed and returned to the sender quantity.That is, the first score value is result in the first result sequence
The measurement of item and the correlation of user search request.
Additionally, it should understand, the first score value is also possible to obtain based on other feature calculations for measuring search result relevance
Arrive, the present invention to predetermined characteristic with no restriction.
Then in step S220, of each result items in the first result sequence is determined in step S210 using click model
Two score values, the second score value here, that is, click feedback characteristic value.In this way, for each result in the first result sequence
, all there is first score value and second score value.Embodiment according to the present invention, click model selection series connection are clicked
Model.
What click model used clicks as a result, both generally being from former pages of recent search result, these results are previous
Front can be come in the sequence of section time, illustrate that they are the most correlated results that sort algorithm is thought at that time.Fig. 3 shows string
Join the schematic diagram of click model, briefly, for one query q and result sequence R (r1, r2 ..., rn), mould is clicked in series connection
Type assumes that user is resultful according to r1 → r2 →...→ rn sequence browsing institute, and can have a probability γ to terminate to inquire
No longer browse.During browsing, if user is attracted by result ri, user will click ri, and the probability of this behavior is ai.
If the ri result that user opens point is satisfied, user would not browse again, to terminate current inquiry, the probability of this behavior is
si.Then, the ai and si of each result just constitute sequence A (a1, a2 ..., an) and S (s1, s2 ..., sn), they and above
The user mentioned abandons the probability γ of search together, constitutes a probabilistic mathematical models, referred to as series connection click model.
It has been previously mentioned, series connection click model can make correlation high by the ability of acquisition user's click information
As a result it arranges to be located further forward, but is not suitable for the weight for individually determining to sort in search.Therefore, in searching order scheme of the invention
In, the feature of traditional measurement correlation is blended with result is clicked, the specific method for calculating weight order is shown in step
S230。
Then in step S230, the result items that the second score value in the first result sequence is not 0 are chosen, the second score value is based on
The sequence of the selected result items of adjustment, obtains the second result sequence, and query result is presented to user.According to described previously, by
There is sparsity in the behavior of click, have many results and do not have in the method by such there is no the influence by the behavior of click
There is the second score value of affected result items to be assigned a value of 0.
For example, if current, the sort algorithm of step S210 think to have a result r should come them it
Before, centainly illustrate that some features of r are improved in the recent period, for example viewing amount increases, article new especially, model reply number
Increase etc., so that r has higher correlation, need to be come front.At this point, if the result according to click model is arranged
Sequence has thus obliterated the promotion of this correlation after r is probably come the result that click model influences, that is,
It says, sequence generates positive feedback, i.e., the result clicked in the past can be more and more forward.It is therefore desirable to retain the sorting position of r, into
And the measurement for protecting other features (viewing amount, issuing time, money order receipt to be signed and returned to the sender quantity etc.) to promote correlation, mould will not be clicked
Type is covered.So in the method, only choosing the result items that click model influences whether, i.e. the second score value is not 0 result
, it resequences to them, and the result items that those second score values are 0, keep their positions in the first result sequence
It sets constant.
Later, for the sequence of selected result items, the codomain property of click model is used.For click model
Codomain is generally boundary with its midpoint, is upwards positive point, is downwards negative sense point.Forward direction point shows that result should be from present bit
It sets toward front row, negative sense point shows that result should be from current location toward heel row.In series connection click model, the forward direction of a result items
Point bigger, then it should be toward more front-seat;But the conclusion not opposite to negative sense point.Negative sense point only indicates that result is not suitable for
Current location is come, needs to be difficult to determine toward the degree for moving back, but moving back.
So in the method, choosing the codomain mean value of series connection click model first as threshold value, greater than codomain mean value
It is remaining to divide for negative sense branch for positive branch point, and to guarantee that the second score value is that positive point of result items come negative sense point
Before result items;Then, to having positive point of result items to sort according to the sequence of the second score value from high to low, at this time if gone out
The identical situation of existing second score value, just sorts according to the first score value;Finally, to all result items holding sequences with negative sense point
Constant, " holding sequence is constant " here is referred to keeps negative sense to divide result items by the sequence of the first score value in step S210
Relative ranks are constant, ensure that the result items of negative sense point will not cannot be clicked below very much because of being pressed into this way, thus
It is difficult again to be come up by row.
So far, the second result sequence after resequencing to the first result sequence, this when, by the second knot have just been obtained
The first score value of the changed result items in position, is adjusted to the corresponding first result sequence in the existing position of the result items in infructescence column
First score value of result items in column.That is, still to keep the first score value is by from high to low in the second result sequence
Sequence present.The property that the second result sequence is sorted from large to small also according to the first score value is thus maintained, so that
All multioperations in search engine based on the property, such as this sequence and the result sequence of other parallel search engines are closed
And the minor sort again together that gets up, after using this method still effectively.
For the explanation definitely for this method 200, Fig. 4 shows sequence side according to an embodiment of the invention
The exemplary conceptual diagram of method.
Wherein the sequence of the leftmost side represents the first result sequence to sort according to the first score value, in the first result sequence
Each is all one and includes<result items serial number, the first score value, the second score value>triple, and set the first score value and the second score value
Codomain be all [0,100], then codomain mean value be 50.The result that select the second score value not from the first result sequence be 0
, obtain one<result items serial number, the second score value>binary group, such as<2,11>,<3,99>... and then as described above, only
Change sequence of second score value greater than the result items of codomain mean value 50, and result items of second score value no more than 50 keep former sequence
It is constant, and guarantee that result items of second score value greater than 50 all come before the result items no more than 50.For example, item 2 and item 6
Relative ranks just there is no variation.Finally, according to sequencing, successively by the result items serial number of the above results item and second
Score value be put back into the second score value in the first result sequence be not 0 the corresponding position of result items in, without change the second score value be 0
Result items position, obtain right side the second result sequence.Note that the first result sequence and the second result sequence corresponding position
Result items the first score value it is identical, that is to say, that their the first score value all maintains sequence from high to low.
Using this method 200, on the basis of original collating sequence (i.e. the first collating sequence), in conjunction with series connection click model
Model feature, reasonably will click on tune power be dissolved into the frame of searching order, change position and the weight of result items so that
The final display order of query result not only reflects click model to the positive effect of correlation, it is thus also avoided that positive feedback lacks
Point.
For the effect for examining this method 200, this method is applied, in forum's search engine of the family of automobile with DBN
(dynamic bayesian network, dynamic bayesian network, a kind of tandem type click model) is click model, is calculated
Using several days after this method returning rates.Returning rate refers to the user for using product one at a time, whithin a period of time again
Carry out the ratio using it, is generally used to measure user to the viscosity or favorable rating of the product.In simple terms, search engine sorts
More preferably, higher returning rate can be obtained naturally;It adjusts power to cause positive feedback if clicked, user is made to be always what former pages were seen
It is that those were clicked as a result, returning rate can transfer to decline, therefore higher returning rate shows that positive feedback problem is smaller.The meter of returning rate
It is as follows to calculate formula:
If user used product at the 0th day, then
N-th day returning rate=
| ((the 1st day user) ∪ (the 2nd day user) ... ∪ (n-th day user)) ∩ (the 0th day user) | ÷ | the
0 day user |
Wherein, ∩ is to ask friendship, and ∪ is to ask simultaneously, | x | indicate the quantity of x.
It is compared using small amount data, the product returning rate of continuous preceding four days use/unuse this method is compared as follows:
Use this method | Without using this method | Promotion ratio |
8.4% | 6.6% | 27.3% |
12.1% | 9.6% | 26.0% |
19.1% | 14.7% | 29.9% |
23.0% | 17.7% | 29.9% |
As it can be seen that this method is obvious to the promotion of returning rate, illustrate the present invention can under the premise of not causing larger problem,
Effectively promote the correlation of search result.
Fig. 5 shows the schematic diagram of the searching order device 500 according to an embodiment of the invention based on click model.
The device includes: acquiring unit 510, computing unit 520 and sequencing unit 530.Wherein computing unit 520 respectively with acquisition
Unit 510 is mutually coupled with sequencing unit 530.
Acquiring unit 510 is suitable for obtaining the first result sequence about user query, wherein each in the first result sequence
Result items have the first score value.First score value is calculated by computing unit 520, and computing unit 520 is suitable for according at least one
A predetermined characteristic is calculated the first score value, embodiment according to the present invention, predetermined characteristic include viewing amount, issuing time and
One or more of money order receipt to be signed and returned to the sender quantity.Also, sequencing unit 530 is suitable for all result items in the first result sequence according to first
The sequence sequence of score value from high to low.
Then, computing unit 520 is further adapted for second point that each result items in the first result sequence are determined using click model
Value.According to one embodiment of present invention, click model is series connection click model.
Sequencing unit 530 is further adapted for choosing the result items that the second score value in the first result sequence is not 0, is based on second
The sequence of the selected result items of score value adjustment, to obtain the second result sequence.Specifically, sequencing unit 530 further includes being suitable for
The second selected score value of judgement is not in 0 result items, and whether the second score value is greater than the judgment sub-unit 532 of threshold value.According to
Embodiments of the present invention, the threshold value are the codomain mean value of series connection click model.Then sequencing unit 530 is further adapted for second point
Value is greater than the result items of threshold value, sorts according to the sequence of the second score value from high in the end;And threshold value is not more than for the second score value
Result items, holding sequence it is constant, and come the second score value greater than threshold value all result items after, here " holding sequence
It is constant " refer to that the relative ranks for the result items for keeping the second score value to be not more than threshold value by the sequence of the first score value are constant, in this way
Do is in order to which the result items for guaranteeing negative sense point will not cannot be clicked below very much because of being pressed into, thus difficult by row again
Come.Finally, sequencing unit 530 is further adapted for the first score value of the changed result items in position in the second result sequence, adjustment
For the first score value of result items in the corresponding first result sequence in the existing position of the result items, does so and ensure that the second result
The first score value is also by sequence sequence from big to small, so that many behaviour in search engine based on the ordering property in sequence
Make, such as the result sequence of this sequence and other parallel search engines is combined together minor sort again, is using this
After method still effectively.
About the specific steps and embodiment of sequence, it has been disclosed in detail in the description based on Fig. 4, it is no longer superfluous herein
It states.
It should be appreciated that in order to simplify the disclosure and help to understand one or more of the various inventive aspects, it is right above
In the description of exemplary embodiment of the present invention, each feature of the invention be grouped together into sometimes single embodiment, figure or
In person's descriptions thereof.However, the disclosed method should not be interpreted as reflecting the following intention: i.e. claimed hair
Bright requirement is than feature more features expressly recited in each claim.More precisely, as the following claims
As book reflects, inventive aspect is all features less than single embodiment disclosed above.Therefore, it then follows specific real
Thus the claims for applying mode are expressly incorporated in the specific embodiment, wherein each claim itself is used as this hair
Bright separate embodiments.
Those skilled in the art should understand that the module of the equipment in example disclosed herein or unit or groups
Part can be arranged in equipment as depicted in this embodiment, or alternatively can be positioned at and the equipment in the example
In different one or more equipment.Module in aforementioned exemplary can be combined into a module or furthermore be segmented into multiple
Submodule.
Those skilled in the art will understand that can be carried out adaptively to the module in the equipment in embodiment
Change and they are arranged in one or more devices different from this embodiment.It can be the module or list in embodiment
Member or component are combined into a module or unit or component, and furthermore they can be divided into multiple submodule or subelement or
Sub-component.Other than such feature and/or at least some of process or unit exclude each other, it can use any
Combination is to all features disclosed in this specification (including adjoint claim, abstract and attached drawing) and so disclosed
All process or units of what method or apparatus are combined.Unless expressly stated otherwise, this specification is (including adjoint power
Benefit require, abstract and attached drawing) disclosed in each feature can carry out generation with an alternative feature that provides the same, equivalent, or similar purpose
It replaces.
A6, method as described in a5, wherein further include: by of the changed result items in position in the second result sequence
One score value is adjusted to the first score value of result items in the corresponding first result sequence in the existing position of the result items.
B11, the device as described in B10, wherein threshold value is the codomain mean value of series connection click model.It is B12, as described in b11
Device, wherein sequencing unit is further adapted for for the first score value of the changed result items in position in the second result sequence being adjusted to
First score value of result items in the corresponding first result sequence in the existing position of the result items.
In addition, it will be appreciated by those of skill in the art that although some embodiments described herein include other embodiments
In included certain features rather than other feature, but the combination of the feature of different embodiments mean it is of the invention
Within the scope of and form different embodiments.For example, in the following claims, embodiment claimed is appointed
Meaning one of can in any combination mode come using.
In addition, be described as herein can be by the processor of computer system or by executing by some in the embodiment
The combination of method or method element that other devices of the function are implemented.Therefore, have for implementing the method or method
The processor of the necessary instruction of element forms the device for implementing this method or method element.In addition, Installation practice
Element described in this is the example of following device: the device be used for implement as in order to implement the purpose of the invention element performed by
Function.
As used in this, unless specifically stated, come using ordinal number " first ", " second ", " third " etc.
Description plain objects, which are merely representative of, is related to the different instances of similar object, and is not intended to imply that the object being described in this way must
Must have the time it is upper, spatially, sequence aspect or given sequence in any other manner.
Although the embodiment according to limited quantity describes the present invention, above description, the art are benefited from
It is interior it is clear for the skilled person that in the scope of the present invention thus described, it can be envisaged that other embodiments.Additionally, it should be noted that
Language used in this specification primarily to readable and introduction purpose and select, rather than in order to explain or limit
Determine subject of the present invention and selects.Therefore, without departing from the scope and spirit of the appended claims, for this
Many modifications and changes are obvious for the those of ordinary skill of technical field.For the scope of the present invention, to this
Invent done disclosure be it is illustrative and not restrictive, it is intended that the scope of the present invention be defined by the claims appended hereto.
Claims (11)
1. a kind of search ordering method based on click model, the method includes the steps:
The first result sequence about inquiry is obtained, each result items have the first score value in the first result sequence, and all
Result items sort according to the sequence of the first score value from high to low, and first score value is calculated according at least one predetermined characteristic
It arrives;
The second score value of each result items in the first result sequence is determined using click model;And
The result items that the second score value in the first result sequence is not 0 are chosen, the result items of threshold value are greater than for the second score value,
According to the sequence sequence of the second score value from high in the end, the result items of threshold value are not more than for the second score value, holding sequence is constant, and
After the second score value is come greater than all result items of threshold value, the second result sequence is obtained.
2. the method for claim 1, wherein the predetermined characteristic includes in viewing amount, issuing time and money order receipt to be signed and returned to the sender quantity
One or more.
3. method according to claim 1 or 2, wherein the click model is series connection click model.
4. method as claimed in claim 3, wherein the threshold value is the codomain mean value of series connection click model.
5. method as claimed in claim 4, wherein further include: by the changed result items in position in the second result sequence
The first score value, be adjusted to the first score value of result items in the corresponding first result sequence in the existing position of the result items.
6. a kind of searching order device based on click model, described device include:
Acquiring unit, suitable for obtaining the first result sequence about inquiry, wherein each result items tool in the first result sequence
There is the first score value;
Computing unit, suitable for the first score value is calculated according at least one predetermined characteristic, is further adapted for determining using click model
Second score value of each result items in the first result sequence;And
Sequencing unit, suitable for arranging all result items in the first result sequence according to the sequence of the first score value from high to low
Sequence is further adapted for choosing the result items that the second score value in the first result sequence is not 0, and the result of threshold value is greater than to the second score value
, it sorts according to the sequence of the second score value from high in the end, the result items of threshold value is not more than for the second score value, holding sequence is not
Become, and come the second score value and be greater than after all result items of threshold value, to obtain the second result sequence.
7. device as claimed in claim 6, wherein the predetermined characteristic includes in viewing amount, issuing time and money order receipt to be signed and returned to the sender quantity
One or more.
8. device as claimed in claims 6 or 7, wherein the click model is series connection click model.
9. device as claimed in claim 8, wherein the threshold value is the codomain mean value of series connection click model.
10. device as claimed in claim 9, wherein the sequencing unit is further adapted for occurring position in the second result sequence
First score value of the result items of variation is adjusted to first of result items in the corresponding first result sequence in the existing position of the result items
Score value.
11. a kind of information search engine system, comprising:
Information bank, suitable for storing information to be put;
The searching order device based on click model as described in any one of claim 6-10, suitable for the knot obtained to inquiry
Infructescence column are ranked up;And
Information display device, suitable for showing query result in order.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510697625.2A CN105302898B (en) | 2015-10-23 | 2015-10-23 | A kind of search ordering method and device based on click model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510697625.2A CN105302898B (en) | 2015-10-23 | 2015-10-23 | A kind of search ordering method and device based on click model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105302898A CN105302898A (en) | 2016-02-03 |
CN105302898B true CN105302898B (en) | 2019-02-19 |
Family
ID=55200168
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510697625.2A Active CN105302898B (en) | 2015-10-23 | 2015-10-23 | A kind of search ordering method and device based on click model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105302898B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107273112B (en) * | 2017-05-04 | 2021-02-02 | 武汉斗鱼网络科技有限公司 | Method and device for displaying gift list information |
CN113761368A (en) * | 2018-05-25 | 2021-12-07 | 重庆好德译信息技术有限公司 | Personalized service recommendation system and method based on environmental information |
CN110825939B (en) * | 2019-09-19 | 2023-10-13 | 五八有限公司 | Post score generation and ordering method and device, electronic equipment and storage medium |
CN111905378B (en) * | 2020-08-19 | 2024-04-02 | 上海莉莉丝网络科技有限公司 | Data updating system, data updating method and server |
CN113254810B (en) * | 2021-06-17 | 2021-10-29 | 浙江口碑网络技术有限公司 | Search result output method and device, computer equipment and readable storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101930438A (en) * | 2009-06-19 | 2010-12-29 | 阿里巴巴集团控股有限公司 | Search result generating method and information search system |
CN102004782A (en) * | 2010-11-25 | 2011-04-06 | 北京搜狗科技发展有限公司 | Search result sequencing method and search result sequencer |
CN103593353A (en) * | 2012-08-15 | 2014-02-19 | 阿里巴巴集团控股有限公司 | Information search method and display information sorting weight value determination method and device |
CN103970796A (en) * | 2013-02-04 | 2014-08-06 | 深圳市世纪光速信息技术有限公司 | Inquiry preference ordering method and device |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8713001B2 (en) * | 2007-07-10 | 2014-04-29 | Asim Roy | Systems and related methods of user-guided searching |
-
2015
- 2015-10-23 CN CN201510697625.2A patent/CN105302898B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101930438A (en) * | 2009-06-19 | 2010-12-29 | 阿里巴巴集团控股有限公司 | Search result generating method and information search system |
CN102004782A (en) * | 2010-11-25 | 2011-04-06 | 北京搜狗科技发展有限公司 | Search result sequencing method and search result sequencer |
CN103593353A (en) * | 2012-08-15 | 2014-02-19 | 阿里巴巴集团控股有限公司 | Information search method and display information sorting weight value determination method and device |
CN103970796A (en) * | 2013-02-04 | 2014-08-06 | 深圳市世纪光速信息技术有限公司 | Inquiry preference ordering method and device |
Also Published As
Publication number | Publication date |
---|---|
CN105302898A (en) | 2016-02-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105302898B (en) | A kind of search ordering method and device based on click model | |
Richardson et al. | Beyond PageRank: machine learning for static ranking | |
US7693901B2 (en) | Consumer-focused results ordering | |
US8930357B2 (en) | Domain expertise determination | |
AU2011202345B2 (en) | Methods and systems for improving a search ranking using related queries | |
US7797344B2 (en) | Method for assigning relative quality scores to a collection of linked documents | |
JP5341253B2 (en) | Generating ranked search results using linear and nonlinear ranking models | |
US7617208B2 (en) | User query data mining and related techniques | |
JP4746712B2 (en) | Calculate document importance by historical importance factoring | |
US9230024B2 (en) | Method and system for ranking web pages in a search engine based on direct evidence of interest to end users | |
US9922119B2 (en) | Navigational ranking for focused crawling | |
EP2248055B1 (en) | Determining quality of tier assignments | |
US20130297583A1 (en) | Operationalizing search engine optimization | |
CN104217031A (en) | Method and device for classifying users according to search log data of server | |
EP2573685A1 (en) | Ranking of heterogeneous information objects | |
Aktas et al. | Personalizing pagerank based on domain profiles | |
CN105678335A (en) | Click rate pre-estimation method, device and calculating equipment | |
Arzanian et al. | A multi-agent based personalized meta-search engine using automatic fuzzy concept networks | |
Wu et al. | A hybrid approach to personalized web search | |
Kirsch | Social information retrieval | |
CN104408156B (en) | Website page includes the detection method and device of quantity in a search engine | |
Rashidi et al. | Prediction of users’ future requests using neural network | |
Shen et al. | A content-based algorithm for blog ranking | |
AnigboguKenechukwu et al. | A Cohesive Page Ranking and Depth-First Crawling Scheme For Improved Search Results | |
Pawar et al. | Effective utilization of page ranking and HITS in significant information retrieval |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
TA01 | Transfer of patent application right | ||
TA01 | Transfer of patent application right |
Effective date of registration: 20180914 Address after: 100089 Beijing Haidian District Haidian District Dan Street 3 B block 11, 1110, 1111 rooms. Applicant after: Che Zhi interconnect (Beijing) Technology Co., Ltd. Address before: 300300 Tianjin Binhai New Area Airport International Logistics Area Second Street 1 Enterprise Service Center 311 room. Applicant before: TIANJIN CHESHIJIA TECHNOLOGY CO., LTD. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |