CN104951460B - Method and device is determined based on the ranking parameter value of keyword clustering - Google Patents
Method and device is determined based on the ranking parameter value of keyword clustering Download PDFInfo
- Publication number
- CN104951460B CN104951460B CN201410119451.7A CN201410119451A CN104951460B CN 104951460 B CN104951460 B CN 104951460B CN 201410119451 A CN201410119451 A CN 201410119451A CN 104951460 B CN104951460 B CN 104951460B
- Authority
- CN
- China
- Prior art keywords
- keyword
- user
- information
- categories
- word
- 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
This application involves a kind of ranking parameter values based on keyword clustering to determine method and device.The scheme of the application is clustered by the multiple keywords selected the first user to obtain one or more keyword categories, the determining keyword to match with second user query information currently entered, its corresponding information to be put, the keyword categories belonging to it, and operational feedback information of the second user in the determining corresponding information to be put of the keyword categories, adjustment institute matched keyword target ranking parameter value, and then adjustment the corresponding information to be put of matched keyword ranking.The keyword for belonging to the same keyword categories is considered as a keyword by the application, and determine unified target ranking parameter value, it can solve the problem of same type of keyword under long-tail keyword changing effect is unstable and broad match and phrase matching influences each other.
Description
Technical field
This application involves data processing fields, and it is determining to relate more specifically to a kind of ranking parameter value based on keyword clustering
Method and device.
Background technique
For the network interaction platform of current mainstream, such as e-commerce platform, search engine bring flow (i.e. user
Amount of access or user's volumes of searches) account for the 20% to 30% of platform total flow.Introducing flow from search engine is very to optimization data
It is necessary.And in order to obtain more search engine flows, generally use the method for increasing magnanimity long-tail keyword, this method
Although effectively, the characteristics of long-tail keyword, is that flow is few, effect is random and unstable, in addition search engine data volume is huge,
It is more difficult to manage the data of corresponding search engine.Thus, it is desirable to which search engine data are analyzed and are optimized, with optimization
Data management improves treatment effeciency.Illustrate defect existing for existing account data management by taking advertiser is SEM as an example below.
For example, search engine keywords bid advertisement system AdWords(is also known as " sponsor link ", Chinese is commonly called as " right side
Advertisement "), be a kind of quick and easy mode of purchase advertising service, this advertising service it is with strong points, it is no matter pre- more at last
Few, it, which is all pressed, clicks charging (CPC) every time.It is often necessary to the advertiser for delivering advertisement can open Adwords account,
Keyword can be bought in Adwords account, each keyword can bind an advertisement.When a user query word passes through certain
When kind matching way is matched to the keyword of advertiser's purchase, the advertisement accordingly bound will be in query word search result displayed page
Display.There are three types of Keywords matching modes, accurate to match, phrase matching, broad match.Wherein, accurate matching is exactly when inquiry
Word is identical with a certain keyword in Adwords account, then query word and word successful match of bidding, search engine triggering
Advertisement corresponding to the keyword of Adwords account successful match.Phrase matching is exactly when query word completely includes keyword
When constituent, no matter its each constituent occurs in what order, then query word and Keywords matching success, search engine
Trigger advertisement corresponding to the keyword of successful match in Adwords.For example, being directed to query word " mp3 price ", pass through phrase
The corresponding advertisement of keyword " mp3 " can be triggered with mode.Broad match, matching range are matched with phrase matching more than accurate
Extensively, query word and keyword need to only have correlation semantically, do not require identical or have inclusion relation, very
It is entirely different on to letter can also be with.Such as " mp3 " and " player " may make up broad match relationship.In search account
In, advertiser mainly considers the Roi of the keyword when deciding whether to buy a certain keyword.Roi(return over
Investment, input-output ratio) that is, income (such as the flow introduced by the corresponding advertisement of the keyword) divided by cost,
When Roi is greater than given threshold, it is believed that profit；It may be considered loss when Roi is less than given threshold.
In SEM(Search Engine Marketing, search engine marketing) promote when, in order to increase account search stream
Amount, can usually be added a large amount of long-tail keywords.Long-tail keyword flow is low, and changing effect is random, and fluctuation is big, and advertiser is difficult root
A reasonable ranking is determined according to its Roi for its corresponding advertisement.Therefore, the unstable long-tail keyword pair of a large amount of effect
Advertisement main management Adwords account brings challenge.In addition, search engine generally supports broad match and phrase matching at present.
Under the mode of broad match or phrase matching, a query word (query) occurs, and can match more in the account of advertiser
A keyword, this, which causes to generate between same type of keyword, influences each other.Advertiser is by promoting keyword in account
Investment (for example increasing the expense of the keyword, to promote the ranking of the corresponding advertisement of the keyword) more multithread can be introduced
Amount, but will lead to the reduction of account entirety Roi.Can achieve whole promotion account by way of reducing the investment of keyword
The purpose of family Roi, but this needs the adjustment taken turns just to can reach purpose more, excessively slowly.
Currently, generalling use rule-based method management Adwords account, that is, the keyword good to Roi promotes investment
Expense, the poor keyword of Roi reduces the expense of investment.This rule-based approach only from single keyword angle,
It can not effectively manage and possess magnanimity long-tail keyword Adwrods account, due to the unstability and conversion effect of long-tail keyword
The random fluctuation of fruit needs to do keyword frequent adjustment, not only will increase Adwords account management cost, but will cause
The cost fluctuations of advertiser's investment are excessive, unstable, and this rule-based management method can not also solve broad match
It is influenced with phrase matching bring.
Summary of the invention
In view of the above-mentioned defects in the prior art, the application provides a kind of improved ranking parameter based on keyword clustering
Be worth and determine method and device, with solve long-tail keyword changing effect existing for search engine in the prior art it is unstable and by
It is interactional between the same type of keyword caused by two kinds of matching ways of broad match and phrase matching of keyword
Problem.
According to the one aspect of the application, a kind of ranking parameter value based on keyword clustering is provided and determines method, comprising:
Multiple key word informations of the first user selection and the ranking parameter value of each keyword are received, the ranking parameter value is for adjusting
Ranking of the corresponding information to be put of the whole keyword in the page；Multiple keywords of first user selection are clustered,
To obtain one or more keyword categories；According to keyword and its ranking parameter value that the first user selects, show it is described to
Impression information, to obtain operational feedback information of the second user in the corresponding information to be put of each keyword categories；It receives
Second user query information currently entered, determination are corresponding with the matched keyword of the query information, the matched keyword
Information to be put and the matched keyword belonging to keyword categories；Using second user in the determining key
Operational feedback information in the corresponding information to be put of word class, adjustment matched keyword target ranking parameter value；Benefit
With the target ranking parameter value, adjustment the corresponding information to be put of matched keyword ranking.
Further, cluster includes: one by keyword each in the multiple keyword and the corresponding triggering keyword
A or multiple queries word carries out word segmentation processing, to obtain multiple word information set corresponding with the multiple keyword respectively；Root
According to obtained multiple word information set corresponding with the multiple keyword respectively, determine each described in the multiple keyword
The degree of correlation between keyword；The multiple keyword is clustered according to the degree of correlation between each keyword, with
Obtain one or more keyword categories.
Wherein, include in the corresponding word information set of each keyword in the multiple keyword: with the key
The weight of word corresponding one or more lexical items and one or more of lexical items in the keyword.
Wherein, by keyword each in the multiple keyword and the corresponding one or more inquiries for triggering the keyword
Word carry out word segmentation processing, with obtain multiple word information set corresponding with the multiple keyword respectively include: will be the multiple
Each keyword and the corresponding one or more query words for triggering the keyword carry out word segmentation processing in keyword, with obtain with
The corresponding one or more lexical items of the keyword；Determine one or more lexical items corresponding with the keyword in the key
Weight in word；According to one or more lexical items corresponding with the keyword and one or more of lexical items in the pass
Weight in keyword constructs the corresponding word information set of the keyword.
Wherein it is determined that weight of the one or more lexical item corresponding with the keyword in the keyword, further
It include: to be occurred in the keyword and the corresponding one or more query words for triggering the keyword according to each lexical item
Number and the number for the keyword of the lexical item occur, determine weight of the lexical item in the keyword.
Wherein, it according to obtained multiple word information set corresponding with the multiple keyword respectively, determines each described
The degree of correlation between keyword further comprises: obtaining the potential class of one or more relevant to the multiple keyword；It is based on
The corresponding word information set of each keyword in the multiple keyword, it is every in the multiple keyword to calculate each potential class
Weight in a keyword；It is determined according to weight of each potential class in each keyword each in the multiple keyword
The degree of correlation between a keyword.
Wherein, it based on the corresponding word information set of keyword each in the multiple keyword, calculates each potential class and exists
Weight in the multiple keyword in each keyword, comprising: include according in the corresponding word information set of each keyword
Weight of the one or more lexical items corresponding with the keyword in the keyword, calculate each potential class described more
The probability that occurs in a keyword and the probability that each keyword occurs under each potential class；According to each potential class every
The probability that occurs in a keyword and the probability that each keyword occurs under each potential class calculate each potential class in institute
State the weight in multiple keywords in each keyword.
Wherein, the multiple keyword is clustered according to the degree of correlation between each keyword, to obtain one
A or multiple keyword categories further comprise: the keyword that the degree of correlation in the multiple keyword is more than predetermined threshold is drawn
Enter same keyword categories, to obtain one or more keyword categories.
According to further aspect of the application, a kind of ranking parameter value determining device based on keyword clustering is provided, is wrapped
Include: the first receiving module, the multiple key word informations and the first user for receiving the first user selection are to each keyword
Ranking parameter value, the ranking parameter value is for adjusting row of the corresponding first user information to be put of the keyword in the page
Name；Cluster module, multiple keywords for selecting the first user cluster, to obtain one or more crucial parts of speech
Not；Display module, the ranking parameter of multiple keywords and the first user for being selected according to the first user to each keyword
Value, shows the information to be put of corresponding first user of the multiple keyword, to obtain second user in each crucial part of speech
Operational feedback information in not corresponding information to be put；Second receiving module, for receiving, second user is currently entered to be looked into
Information is ask, determine the matched keyword of the query information, the corresponding information to be put of the matched keyword of institute and is matched
Keyword belonging to keyword categories；Determining module, for corresponding in the determining keyword categories using second user
Information to be put on operational feedback information, adjustment matched keyword target ranking parameter value；Module is adjusted, is utilized
The target ranking parameter value, adjustment the corresponding information to be put of matched keyword ranking.
Wherein, cluster module further comprises: word segmentation processing module, is used for keyword each in the multiple keyword
And the corresponding one or more query words for triggering the keyword carry out word segmentation processing, with obtain respectively with the multiple keyword
Corresponding multiple word information set；Degree of correlation determining module, for corresponding with the multiple keyword respectively according to obtaining
Multiple word information set determine the degree of correlation in the multiple keyword between each keyword；Keyword clustering module,
For being clustered according to the degree of correlation between each keyword to the multiple keyword, to obtain one or more passes
Keyword classification.
Wherein, include in the corresponding word information set of each keyword in the multiple keyword: with the key
The weight of word corresponding one or more lexical items and one or more of lexical items in the keyword.
Wherein, the word segmentation processing module further comprises: word segmentation module, is used for pass each in the multiple keyword
Keyword and the corresponding one or more query words for triggering the keyword carry out word segmentation processing, corresponding with the keyword to obtain
One or more lexical items；Weight determination module, for determining one or more lexical items corresponding with the keyword described
Weight in keyword；Module is constructed, for according to one or more lexical items corresponding with the keyword and one
Or weight of multiple lexical items in the keyword, construct the corresponding word information set of the keyword.
Wherein, the weight determination module is further used for: according to each lexical item in the keyword and corresponding triggering
The number of the keyword of the number and the appearance lexical item that occur in one or more query words of the keyword, determines institute
Weight of the predicate item in the keyword.
Wherein, the degree of correlation determining module further comprises: potential class obtains module, for obtaining and the multiple pass
The relevant potential class of one or more of keyword；Weight calculation module, for based on each keyword pair in the multiple keyword
The word information set answered calculates weight of each potential class in the multiple keyword in each keyword；First determines mould
Block, for according to weight of each potential class in each keyword determine in the multiple keyword each keyword it
Between the degree of correlation.
Wherein, the weight calculation module further comprises: the first computational submodule, for corresponding according to each keyword
Word information set in include weight of the one or more lexical items corresponding with the keyword in the keyword, calculate
The probability that each potential class occurs in the multiple keyword and the probability that each keyword occurs under each potential class；
Second computational submodule, probability for being occurred in each keyword according to each potential class and under each potential class it is every
The probability that a keyword occurs calculates weight of each potential class in the multiple keyword in each keyword.
Wherein, the keyword clustering module is further used for: being more than predetermined threshold by the degree of correlation in the multiple keyword
The keyword of value is divided into same keyword categories, to obtain one or more keyword categories.
Compared with prior art, using the technical solution of the application, first user's keyword is clustered respectively by clustering
At a series of classifications, using operational feedback information of the second user on of all categories, the first user is adjusted on the whole to same
The variation of the ranking parameter value of all keywords in classification is believed with reasonably determining the first user to the inquiry that second user inputs
The target ranking parameter value of the matched keyword of breath institute, so as to so that the first user can be effective rapidly to the keyword of selection
Ground reaches target, so can the matching of Optimizing Search engine, avoid long-tail keyword changing effect unstable and wide
Same type of keyword influences each other under general matching and phrase matching, the problem of effectively to reach target rapidly.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present application, constitutes part of this application, this Shen
Illustrative embodiments and their description please are not constituted an undue limitation on the present application for explaining the application.In the accompanying drawings:
Fig. 1 be the invention relates to the flow chart that method is determined based on the ranking parameter value of keyword clustering；
Fig. 2 is being clustered based on the degree of correlation to multiple keywords of the first user selection according to the embodiment of the present application,
With the specific flow chart for the step of obtaining one or more keyword categories；
Fig. 3 is that keyword each in the multiple keyword and correspondence are triggered the key according to the embodiment of the present application
One or more query words of word carry out word segmentation processing, to obtain multiple word information collection corresponding with the multiple keyword respectively
The flow chart of the step of conjunction；
Fig. 4 is multiple word information collection corresponding with the multiple keyword respectively that the basis of the embodiment of the present application obtains
The specific flow chart for the step of closing, determining the degree of correlation between each keyword；
Fig. 5 is according to an embodiment of the present application based on the corresponding word information collection of keyword each in the multiple keyword
The flow chart for the step of closing, calculating weight of each potential class in the multiple keyword in each keyword；
Fig. 6 be the invention relates to the ranking parameter value determining device based on keyword clustering structural block diagram,
And
Fig. 7 is operational feedback information of the application according to second user in the corresponding information to be put of the keyword categories
Determine the flow chart of the first user couple with an embodiment of the target ranking parameter value of the matched keyword of query information.
Specific embodiment
It clusters, obtains the main idea of the present application lies in that being based on the degree of correlation to multiple keywords of the first user selection
To one or more keyword categories, the keyword for belonging to the same keyword categories is considered as the same keyword, when second
When user scans for, the matched keyword of query information institute of second user, keyword generic, corresponding wait throw is determined
Information is put, using operational feedback information of the existing second user in the corresponding information to be put of the keyword categories, is adjusted
The target ranking parameter value of whole matched keyword, so adjust the corresponding information to be put of matched keyword row
Name.
To keep the purposes, technical schemes and advantages of the application clearer, below in conjunction with the application specific embodiment and
Technical scheme is clearly and completely described in corresponding attached drawing.Obviously, described embodiment is only the application one
Section Example, instead of all the embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not doing
Every other embodiment obtained under the premise of creative work out, shall fall in the protection scope of this application.
According to an embodiment of the present application, it provides a kind of ranking parameter value based on keyword clustering and determines method.
It is that the ranking parameter value based on keyword clustering that one embodiment of the application is related to determines method with reference to Fig. 1, Fig. 1
Flow chart.
At step S101, multiple key word informations of the first user selection and the ranking parameter of each keyword are received
Value.Wherein, which can be used for adjusting ranking of the corresponding information to be put of the keyword in the page.
First user can choose multiple keywords and select the ranking parameter value to each keyword.Each keyword
Ranking parameter value can be used for adjusting ranking of the information to be put provided by corresponding first user of the keyword in the page.
Example is made as with Adwords triggering advertisement machine, and the first user can be advertiser i.e. advertisement and launch user, which can be with
It is purchasing price of the advertiser to keyword.Generally, an advertiser is higher to the purchasing price of a certain keyword, the key
Ranking of the advertisement that the corresponding advertiser of word launches in the corresponding all advertisements of the keyword is higher.The key of Adwords
Word triggering advertisement mechanism can bid according to keyword of the historical data to selection, it is preferable that can make all keys
Word price having the same, to facilitate subsequent be adjusted.
At step S102, multiple keywords of the first user selection are clustered, to obtain one or more keys
Word class.
Specifically, same category of keyword can will be belonged in multiple keywords and aggregates into a keyword categories.Root
It, can query information or search information based on keyword itself and the corresponding keyword according to one embodiment of the application
(such as query word query) determines the degree of correlation in the multiple keyword between each keyword, thus to the multiple pass
Keyword is clustered.
In one embodiment, it can refer to and correlation be based on to multiple keywords according to one embodiment of the application shown in Fig. 2
Degree is clustered, flow chart the step of to obtain one or more keyword categories.
Step S201, by keyword each in the multiple keyword and the corresponding one or more for triggering the keyword
Query word carries out word segmentation processing, to obtain multiple word information set corresponding with the multiple keyword respectively.
Wherein, include in the corresponding word information set of each keyword in the multiple keyword: with the key
The weight of word corresponding one or more lexical items and one or more of lexical items in the keyword.
It is according to one embodiment of the application by keyword each in the multiple keyword and correspondence with reference to Fig. 3, Fig. 3
The one or more query words for triggering the keyword carry out word segmentation processing, corresponding with the multiple keyword more respectively to obtain
The flow chart of the step of a word information set (step S201).
Step S301, by keyword each in the multiple keyword and the corresponding one or more for triggering the keyword
Query word carries out word segmentation processing, to obtain one or more lexical items corresponding with the keyword.
Wherein, lexical item (Term) is exactly the text segment in natural language, with basic meaning of one's words feature.The text of user's input
What word, the title of webpage, the content of article, song title etc. were construed as being made of lexical item.In each language
Lexical item is metastable set, and the text quantity that single lexical item includes searches for the query word (query), the sentence that use than user
Generally want small etc. the text quantity for including.
It specifically, can be to one of each keyword itself and the corresponding triggering keyword in multiple keyword
Or multiple queries word carries out word segmentation processing, thus by the keyword and the corresponding one or more query words for triggering the keyword
It is divided into multiple lexical items.Wherein, it is corresponding to may be constructed the keyword for the corresponding one or more query words for triggering the keyword
Inquire word list.
According to an embodiment of the present application, predetermined quantity (predetermined number can be extracted from multiple lexical items that segmentation obtains
Amount can determine as the case may be) lexical item, as one or more lexical items corresponding with the keyword, for example, being mentioned
(i.e. the frequency of occurrences is higher than all lexical items or the higher lexical item of the part frequency of occurrences that the lexical item of taking-up can be expression practical significance
The lexical item of given threshold).
Step S302 determines weight of the one or more lexical items corresponding with the keyword in the keyword.
It, can be according to each lexical item in the keyword and the corresponding triggering pass according to one embodiment of the application
The number of the keyword of the number and the appearance lexical item that occur in one or more query words of keyword, determines the lexical item
Weight in the keyword.
Specifically, the corresponding each lexical item of a keyword can be calculated using formula (1) by TF-IDF statistical method
Weight in the keyword；
Wherein, the M query word (query) of keyword d itself and corresponding triggering keyword d segmented
Lexical item b1、b2、…、bi、…、bM, in formula (1), biIndicate i-th corresponding of lexical item of keyword d；wiIndicate lexical item biIt is closing
Weight in keyword b；TF(bi) it is lexical item biIn one or more query words of keyword d itself and corresponding triggering keyword d
The number occurred in (inquiry word list)；DF(bi) it is lexical item b occur in whole keywordsiKeyword number, wherein word
Item biNo matter occur from some keyword itself also occurs from the corresponding one or more query words for triggering the keyword
In (inquiry word list), it can be seen as occurring lexical item b in the keywordi。
Step S303 exists according to one or more lexical items corresponding with the keyword and one or more of lexical items
Weight in the keyword constructs the corresponding word information set of the keyword.
Wherein, include in the corresponding word information set of each keyword in the multiple keyword: with the key
The weight of word corresponding one or more lexical items and one or more of lexical items in the keyword.
It specifically, can be by the corresponding one or more lexical items of each keyword and each lexical item in the keyword
Weight is expressed as a word information vector:
d={(b1,w1),(b2,w2)…(bi,wi)…(bM,wM) (2)
Wherein, wiFor one or more query words of keyword d itself and corresponding triggering keyword d are carried out at participle
Manage the obtained corresponding M lexical item b of keyword d1、b2、…、bi、…、bMIn, i-th of lexical item biCorresponding weight.
By executing above-mentioned step S301~S303, each keyword is corresponding in available the multiple keyword
Word information set, followed by step S202.
Step S202, according to obtained multiple word information set corresponding with the multiple keyword respectively, determine described in
The degree of correlation in multiple keywords between each keyword.
According to one embodiment of the application, one or more potential classes can be introduced between each keyword, and are counted
Weight of each potential class in each keyword is calculated, so that it is determined that in the multiple keyword between each keyword
The degree of correlation.
It is obtained with reference to the basis that Fig. 4, Fig. 4 are one embodiments of the application corresponding with the multiple keyword respectively multiple
Word information set, the specific flow chart for the step of determining the degree of correlation between each keyword.
As shown in figure 4, obtaining the potential class of one or more relevant to the multiple keyword, often at step S401
One potential class is the identical theme that several keywords have in the multiple keyword.
Specifically, a potential class can be construed to a theme, for example, " joyful down toy ", " teddy bear public affairs
The keywords such as son " and " teddy bear down toy " can indicate that then " toy " can be used as one and dive with theme " toy "
In class.Potential class can be the theme that can generally represent the keyword hidden after keyword literal message, also, each pass
Keyword can indicate with multiple and different themes, and e.g., the hiding theme of " AC Milan ", " Real Madrid ", " Barcelona " these words can be with
It is football, is also possible to sport.
Step S402 is calculated each potential based on the corresponding word information set of keyword each in the multiple keyword
Weight of the class in the multiple keyword in each keyword.
It is according to an embodiment of the present application corresponding based on each keyword in the multiple keyword with reference to Fig. 5, Fig. 5
Word information set, the step of calculating the weight of each potential class in the multiple keyword in each keyword (step S402)
Flow chart.
As shown in figure 5, at step S501, according to including with the pass in the corresponding word information set of each keyword
Weight of the corresponding one or more lexical items of keyword in the keyword, calculates each potential class in the multiple keyword
The probability of appearance and the probability that each keyword occurs under each potential class.
In one embodiment, probability latent semantic analysis model PLSA(Probabilistic Latent can be used
Semantic Analysis), and counted according to weight of the corresponding one or more lexical items of each keyword in the keyword
Calculate probability that each potential class occurs in the multiple keyword and under each potential class each keyword occur it is general
Rate.
Specifically, N number of potential class t is introduced1、t2、…、tk、…、tN, tkK-th of potential class is indicated, with P (tk) indicate potential
Class tkThe probability of appearance, that is, potential class tkThe probability occurred in all keywords, and P (d | tk) indicate in potential class tkLower key
The probability that word d occurs, that is, given theme tkThe probability that lower keyword d occurs, and P (b | tk) indicate in potential class tkLower lexical item b goes out
Existing probability, that is, given theme tkThe probability that lower lexical item b occurs.
P (t can be calculated by executing EM (Expectation Maximization) iteration of certain numberk)、P(d
|tk) and P (b | tk):
E step: estimating the desired value of unknown parameter, estimates p (t in the present embodimentk)、p(d|tk) and p (b | tk) expectation
Value, and utilize p (tk)、p(d|tk) and p (b | tk) desired value or the last resulting p (t of iterationk)、p(d|tk) and p (b | tk)
Calculate P (tk| d, b) desired value, such as formula (3).
Wherein, P (tk| d, b) it indicates in potential class tkThe probability that lower keyword d and lexical item b occur jointly, that is, use theme tk
Come the probability for explaining keyword d and lexical item b occurs jointly, w (d, b) is power of the corresponding lexical item b of keyword d in the keyword
Weight.
M step: P (t obtained in the previous step is utilizedk| d, b) desired value maximizes current parameter Estimation.Namely according to E
P (t obtained in stepk| d, b), it calculates in potential class tkProbability P that lower lexical item b occurs (b | tk), in potential class tkLower keyword d
Appearance probability P (d | tk) and each potential applications class tkProbability P (the t of appearancek)。
E step is alternately performed with after the certain number of M step, can will finally obtain P (tk) and P (d | tk) value is as each potential
Class tkProbability P (the t of appearancek) and in each potential class tkUnder each keyword d occur probability P (d | tk), wherein it executes
Depending on the number of EM iteration is by concrete condition.
Step S502, the probability occurred in the multiple keyword according to each potential class and under each potential class
The probability that each keyword occurs calculates weight of each potential class in the multiple keyword in each keyword.
Specifically, formula (7) be can use and calculate potential class tkWeight in keyword d
Weight of the available each potential class in each keyword through the above steps, next can be according to every
The degree of correlation of weight calculation each keyword of a potential class in each keyword.
Step S403 is determined each in the multiple keyword according to weight of each potential class in each keyword
The degree of correlation between a keyword.
Each potential class t according to obtained in step S4021、t2、…、tk、…、tNWeight in keyword d…、…、It can be write as the corresponding potential class distribution characteristics vector of keyword d:
According to the application
One embodiment, determine the degree of correlation between each keyword, can by calculate any two keyword between it is similar
Degree is to embody, wherein it can be calculated according to weight of each potential class in the two keywords, specifically, Ke Yigen
According to keyword d1、d2Theme distribution feature vector, pass through KL distance calculate the two keywords between similarity, such as formula
(10) shown in:
Wherein,(d1) with(d2) it is respectively keyword d1And d2It is tieed up in the kth of respective theme distribution vector
Weight.
Step S203 clusters the multiple keyword according to the degree of correlation between each keyword, with
To one or more keyword categories.
It specifically, can be more than the same keyword of keyword cut-in of predetermined threshold by the degree of correlation in the multiple keyword
Classification, to obtain one or more keyword categories.Such as: by determining similarity sim2 (d1,d2) whether result be more than or equal to one
Predetermined threshold 0.65, if it is larger than or equal to 0.65 keyword d1、d2Between the degree of correlation both be sufficient to make cluster to one
Under keyword categories.
Above step, by multiple keyword clusterings at one or more keyword categories, next, can use keyword
The target ranking parameter value of the result adjustment keyword categories of cluster.
At step S103, according to keyword and its ranking parameter value that the first user selects, the letter to be put is shown
Breath, to obtain operational feedback information of the second user in the corresponding information to be put of each keyword categories.
Wherein, the first user selects one or more keywords, and has corresponding ranking parameter value to each keyword.The
One user provides the information to be put for corresponding to the keyword of its selection, and showing in search display platform (page) should be to
Impression information.
Wherein, second user operates the information to be put of displaying, generates the keyword of the corresponding information to be put
Operational feedback information based on the keyword categories that aforementioned cluster generates, second user can be collected in each crucial part of speech in turn
It Shang not operational feedback information in corresponding information to be put.
Further, operational feedback information can recorde in operation log, and operational feedback information may include such as: input
Second user that query information is inquired (such as: terminal user, search user) to the first user show on webpage wait throw
Put the operation (as clicked, browsing) of information, the keyword that the result of operation generation and query information match, by operating
Information, the corresponding keyword of information to be put and the first affiliated user to be put, etc..
It specifically, can be according to the multiple keywords and advertiser that advertiser selects along the example being used at step S101
The advertisement that the corresponding advertiser of multiple keywords launches is shown to the purchasing price of each keyword, that is, according to the advertiser
To the purchasing price of each keyword, bidding for the advertisement that the advertiser launches in the corresponding all advertisements of each keyword is determined
Ranking, and when needing to show the corresponding advertisement of each keyword (the inquiry query as being directed to user) according to the bid ranking
Show the advertisement of the corresponding advertiser of each keyword.Advertiser launch advertising display for a period of time after, terminal user exists
Matching keywords and the advertisement that the corresponding keyword is shown can be viewed when input inquiry message reference webpage, and then to displaying
Advertisement occur to click, the browsing operation such as details, and generate operational feedback information (as clicked, clicking the ad revenue generated
Deng).Operational feedback information such as to ad click operation and feedback information relevant to the operation, will be recorded in operation log
In.By using by every time click charge cpc(cost perclick) advertisement search release platform for, obtain each key
The click click(of the corresponding advertisement of word such as bring flowing of access, ad revenue), using aforementioned cluster, obtain each key
The corresponding click click(income of word class).
At step S104, second user query information currently entered is received, determines the inquiry letter of second user input
Cease matched keyword, the corresponding information to be put of matched keyword and keyword belonging to matched keyword
Classification.
Second user such as terminal user or search user, when being scanned for data such as search shopping website, input
Query information can determine the matched keyword of query information institute.The matched keyword of the query information can be their ability to
The keyword of triggering, the keyword of the corresponding triggering of the one or more query words for e.g., in the query information including.
By the matched keyword, its corresponding impression information can be determined.It in turn, can by the cluster of aforementioned keyword
To determine matched keyword keyword categories affiliated in one or more keyword categories, or even determine pass belonging to its
Keyword classification information to be put (one or more information to be put) accordingly.
For example, terminal user has input query information " red one-piece dress autumn and winter ", then query information energy can be determined
The keyword matched has " red one-piece dress ", corresponds to the keyword matched, there is the advertisement of corresponding " red one-piece dress ", and
According to the cluster of front it is found that the keyword belongs to " one-piece dress " classification.
At step S105, behaviour of the second user in the determining corresponding information to be put of the keyword categories is utilized
Make feedback information, adjustment matched keyword target ranking parameter value.
Wherein, second user treats impression information when accessing webpage and is operated (click, browsing etc.), and it is anti-to generate operation
Feedforward information can recorde the operational feedback information in the corresponding information to be put of keywords multiple under determining keyword categories
Into operation log.Thus, each keyword categories obtained by cluster of the correspondence in operation log are recorded in based on these
Operational feedback information, the first user can be obtained to the current target value of each keyword categories, such as the first user is to current
The input and output value of keyword categories or input-output ratio Roi, etc..
By taking the application of aforementioned cpc advertisement charging as an example, according to second user in a series of of first user's keyword clustering
Operational feedback information on keyword categories, such as the number such as the click of terminal user in the advertisement that keyword categories triggered generation
According on statistics cpc advertisement search display platform, the cost that advertiser generates on each keyword categories, i.e., advertiser is every
The investment cost and each keyword categories of a keyword categories be the advertiser's bring output/income (such as click volume,
Flowing of access) Income.Then, the first user is calculated in the Roi of each keyword categories.
Specifically, advertiser, can be corresponding to the keyword categories for advertiser to total investment of a certain keyword categories
Total click investment, that is, because the corresponding advertisement of keyword each in the keyword categories is clicked the sum of the expense of generation, that is,
The corresponding advertisement of the keyword categories that advertiser launches because pay required for being clicked advertisement platform system expense it
With, the corresponding advertisement of keyword each in the keyword categories can be clicked generation expense be added.A certain key part of speech
Not Wei Gross Output Income brought by advertiser, it is obtained the corresponding advertisement of the keyword categories can be launched for advertiser
Click income, that is, the sum of click income obtained of the corresponding advertisement of each keyword in the keyword categories (or for should
Total click income of the corresponding advertisement of keyword categories).Advertiser is advertiser to the input and output value of the keyword categories
Advertisement corresponding with the keyword categories is put into total click of the corresponding advertisement of the keyword categories to be brought by advertiser
Total click income ratio.
Such as: click income (10) can be calculated according to the following formula:
Income=click*rpc (10)
Wherein, click expression is clicked number, rpc(return per click) it indicates to click average yield every time.It can
The corresponding advertisement point obtained of each keyword in the keyword categories is launched first to calculate advertiser according to formula (10)
Income is hit, then is added the corresponding advertisement of each keyword click income obtained to obtain the corresponding advertisement of the keyword categories
Income is always clicked to advertiser's bring.
It is the advertiser to total investment (cost) of each keyword categories and each keyword categories by above-mentioned advertiser
Bring Gross Output Income, Roi of the available advertiser to the keyword categories.
For example, in the Adwords account of a certain advertiser, each keyword of a certain keyword categories (one-piece dress classification)
Ranking parameter value (price), investment (cost) and output (income) can be as shown in table 1:
Table 1
Thus, it is possible to which the total investment and Gross Output according to keywords all in this keyword categories calculate the key part of speech
Other input and output value Roi, available according to table 1, total investment of this keyword categories is (by the input of each keyword
It is added) it is cost Cost=43, Gross Output (being added the output value of each keyword) is Incom=60, according to formula (11):
Input and output value Roi=1.395 of this keyword categories can be calculated.
Further, the first user of acquisition is to the current target value of the keyword categories and the first user to determination
The keyword categories in each keyword ranking parameter value, determine (in other words adjust) first user couple and the inquiry
The target ranking parameter value of the keyword of information matches.
The adjustment target ranking parameter value may include: the to be achieved of the keyword categories for obtaining the first user to the determination
Target value (such as: target input and output value to be achieved)；And according to the target value to be achieved and the first user to determination
The keyword categories current target value (such as: the input and output value being calculated by the record in operation log), with
And first user to the ranking parameter value of each keyword in the determining keyword categories, determine the first user to the pass
The unified ranking parameter value of keyword classification；Unified ranking parameter value of first user to the keyword categories is determined as first
Target ranking parameter value of the user to the matched keyword of the query information.
Previous example is continued to use, since the advertiser by obtaining is to the Roi of keyword categories, can be determined by the Roi
The rise in price or the range of price decrease of all keywords in the category can guarantee that the Adwords account of advertiser is whole and reach one rapidly
A target Roi, when being not in price reduction, long-time interative computation can be only achieved a target Roi, improve data processing performance
And efficiency.
In one embodiment, the application as shown in Figure 7 is corresponding wait throw in the keyword categories according to second user
It puts the operational feedback information in information and determines the first user couple and the target ranking parameter value of the matched keyword of query information
The flow chart of one embodiment.
At step S701, the target value to be achieved of the first user is obtained.
Previous example, target value to be achieved are continued to use, i.e., target input and output value to be achieved can be the first user
Desired input and output value.For example, can be the total click for the corresponding advertisement of the keyword categories that advertiser demand communication obtains
Putting into advertisement corresponding with the keyword categories is that the desired value of the ratio of income, i.e. advertiser are always clicked brought by advertiser
Want the rate of return on investment obtained.
At step S702, according to the target value to be achieved of the first user, the first user to the keyword categories
Current target value and the first user determine the first user couple to the ranking parameter value of each keyword in the keyword categories
The unified ranking parameter value of the keyword categories.
Above example is continued to use, for example, advertiser target input and output value targetRoi to be achieved is 1.5, is calculated
The advertiser arrived is 1.395 to the current input and output value Roi of the determining keyword categories, and the keyword categories determined
Average price (average value of the price of each keyword) i.e. ranking parameter value be OldPrice=2.250, then this keyword
The new price of the unification of classification, that is, the unified ranking parameter value newPrice for belonging to all keywords of this keyword categories, it can
To be calculated according to formula (12):
The new price of unification of this keyword categories can be calculated (for Adwords system, system according to formula (12)
One ranking parameter value is exactly the new price of unification of all keywords of this keyword categories) it is newPrice=2.25*
(1.395/1.5)=2.09。
At step S703, unified ranking parameter value of first user to the keyword categories is determined as the first user
To the target ranking parameter value with the matched keyword of the query information.That is, by the first user to the query information
The unified ranking parameter value of matched keyword categories is determined as the first user to the target ranking parameter value of the keyword.
According to an embodiment of the present application, the target ranking parameter value for belonging to the keyword of same keyword categories can phase
Together, that is to say, that the keyword for belonging to same keyword categories can be considered as the same keyword, belong to the same keyword
The target ranking parameter value of all keywords of classification is identical.
At step S106, corresponding first user of the matched keyword is adjusted using the target ranking parameter value
The ranking of information to be put, to show corresponding first user of the matched keyword to second user according to the ranking
Information to be put.
That is, adjusting the pass according to target ranking parameter value of first user to the matched keyword of the query information
Ranking of the corresponding first user information to be put of keyword in the corresponding all information to be put of the key copula, so as to basis
The ranking shows information to be put corresponding to the corresponding keyword of query information of its input to second user, to make first
User obtains desired target input and output value.
Previous example is continued to use, by the obtained i.e. new bid of keyword of target ranking parameter value, keyword is corresponding wide
The sequence for accusing the advertisement of main offer also changes, and it is successive thus to influence the sequence that advertisement shows on the page of webpage.
The ranking parameter value determining device based on keyword clustering that present invention also provides a kind of.
Fig. 6 is schematically shown to be determined according to the ranking parameter value based on keyword clustering of the application one embodiment
The structural block diagram of device.
As shown in fig. 6, device 600 includes: the first receiving module 601, cluster module 602, display module 603, second connects
Receive module 604, determining module 605 and adjustment module 606.
First receiving module 601 can be used for receiving the first user (such as advertiser) selection multiple key word informations and
The ranking parameter value (such as purchasing price of keyword) of each keyword.Wherein, which can be used for adjusting and be somebody's turn to do
Ranking of the corresponding information (such as advertisement) to be put of keyword in the page.Concrete function processing is referring to step S101.
Cluster module 602 can be used for clustering multiple keywords that the first user selects, to obtain one or more
A keyword categories.Concrete function processing is referring to step S102.
Display module 603 can be used for the keyword and its ranking parameter value selected according to the first user, show it is described to
Impression information, to obtain operational feedback information of the second user in the corresponding information to be put of each keyword categories.Specifically
Function treatment is referring to step S103.
Second receiving module 604 can be used for receiving that second user (such as terminal user, search user) is currently entered looks into
Ask information, determine the corresponding information to be put of the matched keyword of query information of second user input, the matched keyword of institute,
And keyword categories belonging to matched keyword.Concrete function processing is referring to step S104.
Determining module 605 can be used for using second user in the determining corresponding information to be put of the keyword categories
On operational feedback information, adjustment matched keyword target ranking parameter value.Concrete function processing is referring to step
S105。
Adjustment module 606 can be used for adjusting the matched keyword corresponding the using the target ranking parameter value
The ranking of one user information to be put, to show the matched keyword corresponding the to second user according to the ranking
One user information to be put.Concrete function processing is referring to step S106.
According to one embodiment of the application, cluster module 601 be may further include: word segmentation processing module, the degree of correlation
Determining module and keyword clustering module (not shown).
Wherein, word segmentation processing module, can be used for should by keyword each in the multiple keyword and corresponding triggering
One or more query words of keyword carry out word segmentation processing, to obtain multiple word letters corresponding with the multiple keyword respectively
Breath set.
Degree of correlation determining module can be used for according to obtained multiple word informations corresponding with the multiple keyword respectively
Set, determines the degree of correlation in the multiple keyword between each keyword；
Keyword clustering module can be used for according to the degree of correlation between each keyword to the multiple keyword
It is clustered, to obtain one or more keyword categories.
According to one embodiment of the application, the corresponding word information set of each keyword in the multiple keyword
In may include: it is corresponding with the keyword one or more lexical items and one or more of lexical items in the keyword
In weight.
According to one embodiment of the application, word segmentation processing module be may further include: word segmentation module, weight determine mould
Block, building module (not shown).
Word segmentation module can be used for one by keyword each in the multiple keyword and the corresponding triggering keyword
A or multiple queries word carries out word segmentation processing, to obtain one or more lexical items corresponding with the keyword.
Weight determination module is determined for one or more lexical items corresponding with the keyword in the keyword
In weight.
Module is constructed, can be used for according to one or more lexical items corresponding with the keyword and one or more
Weight of a lexical item in the keyword constructs the corresponding word information set of the keyword.
According to one embodiment of the application, weight determination module can be further used for: according to each lexical item described
The number that occurs in keyword and the corresponding one or more query words for triggering the keyword and there is the lexical item
The number of keyword determines weight of the lexical item in the keyword.
According to one embodiment of the application, degree of correlation determining module be may further include: potential class obtains module, power
Re-computation module and the first determining module (not shown).
Wherein, potential class obtains module, can be used for obtaining one or more relevant to the multiple keyword potential
Class；
Weight calculation module can be used for based on the corresponding word information set of keyword each in the multiple keyword,
Calculate weight of each potential class in the multiple keyword in each keyword；
First determining module can be used for being determined according to weight of each potential class in each keyword described more
The degree of correlation in a keyword between each keyword.
According to one embodiment of the application, weight calculation module be may further include: the first computational submodule, second
Computational submodule (not shown).
Wherein, the first computational submodule, can be used for according in the corresponding word information set of each keyword include with
Weight of the corresponding one or more lexical items of the keyword in the keyword, calculates each potential class in the multiple pass
The probability that occurs in keyword and the probability that each keyword occurs under each potential class
Second computational submodule can be used for the probability occurred in each keyword according to each potential class and every
The each potential class of probability calculation that each keyword occurs under a potential class is in the multiple keyword in each keyword
Weight.
According to one embodiment of the application, keyword clustering module can be further used for: by the multiple keyword
The middle degree of correlation is more than that the keyword of predetermined threshold is divided into same keyword categories, to obtain one or more keyword categories.
According to one embodiment of the application, determining module be may further include: obtain module, the first determining module with
And the second determining module (not shown).
Wherein, module is obtained, can be used for obtaining the target value to be achieved of the first user.
First determining module can be used for the target value to be achieved according to the first user, the first user to the keyword
The current target value of classification and the first user determine first to the ranking parameter value of each keyword in the keyword categories
Unified ranking parameter value of the user to the keyword categories.
Second determining module can be used for for unified ranking parameter value of first user to the keyword categories being determined as
Target ranking parameter value of first user to the matched keyword of the query information.
Since the function that the device of the present embodiment is realized essentially corresponds to aforementioned FIG. 1 to FIG. 5, method shown in Fig. 7 reality
Example is applied, therefore not detailed place in the description of the present embodiment, it may refer to the related description in previous embodiment, do not do herein superfluous
It states.
In a typical configuration, calculating equipment includes one or more processors (CPU), input/output interface, net
Network interface and memory.
Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/or
The forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flashRAM).Memory is showing for computer-readable medium
Example.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer readable instructions, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase change memory (PRAM), static random access memory (SRAM), moves
State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electric erasable
Programmable read only memory (EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM),
Digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetic storage devices
Or any other non-transmission medium, can be used for storage can be accessed by a computing device information.As defined in this article, it calculates
Machine readable medium does not include temporary computer readable media (transitory media), such as the data-signal and carrier wave of modulation.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want
There is also other identical elements in the process, method of element, commodity or equipment.
It will be understood by those skilled in the art that embodiments herein can provide as method, system or computer program product.
Therefore, complete hardware embodiment, complete software embodiment or embodiment combining software and hardware aspects can be used in the application
Form.It is deposited moreover, the application can be used to can be used in the computer that one or more wherein includes computer usable program code
The shape for the computer program product implemented on storage media (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
Formula.
The above description is only an example of the present application, is not intended to limit this application.For those skilled in the art
For, various changes and changes are possible in this application.All any modifications made within the spirit and principles of the present application are equal
Replacement, improvement etc., should be included within the scope of the claims of this application.
Claims (8)
1. a kind of ranking parameter value based on keyword clustering determines method characterized by comprising
It receives multiple key word informations of the first user selection and the ranking parameter value of each keyword, the ranking parameter value is used
In adjusting ranking of the corresponding information to be put of the keyword in the page；
Multiple keywords of first user selection are clustered, to obtain one or more keyword categories；
According to keyword and its ranking parameter value that the first user selects, the information to be put is shown, to obtain second user
Operational feedback information in the corresponding information to be put of each keyword categories；
Receive second user query information currently entered, determining and the matched keyword of the query information, the matched pass
Keyword categories belonging to the corresponding information to be put of keyword and the matched keyword；
Using operational feedback information of the second user in the determining corresponding information to be put of the keyword categories, institute is adjusted
The target ranking parameter value of matched keyword；
Using the target ranking parameter value, adjustment the corresponding information to be put of matched keyword ranking；
Wherein, multiple keywords of the first user selection are clustered, to obtain one or more keyword categories, comprising:
Keyword each in the multiple keyword and the corresponding one or more query words for triggering the keyword are divided
Word processing, to obtain multiple word information set corresponding with the multiple keyword respectively；
According to obtained multiple word information set corresponding with the multiple keyword respectively, determine each in the multiple keyword
The degree of correlation between a keyword；
The multiple keyword is clustered according to the degree of correlation between each keyword, to obtain one or more passes
Keyword classification.
2. the method according to claim 1, wherein each keyword is corresponding described in the multiple keyword
Include in word information set:
The power of corresponding with the keyword one or more lexical item and one or more of lexical items in the keyword
Weight.
3. according to the method described in claim 2, it is characterized in that, by keyword each in the multiple keyword and correspondence
The one or more query words for triggering the keyword carry out word segmentation processing, corresponding with the multiple keyword more respectively to obtain
A word information set, comprising:
Keyword each in the multiple keyword and the corresponding one or more query words for triggering the keyword are divided
Word processing, to obtain one or more lexical items corresponding with the keyword；
Determine weight of the one or more lexical items corresponding with the keyword in the keyword；
According to one or more lexical items corresponding with the keyword and one or more of lexical items in the keyword
Weight, construct the corresponding word information set of the keyword.
4. according to the method described in claim 3, it is characterized in that, determining one or more lexical items corresponding with the keyword
Weight in the keyword further comprises:
Occurred in the keyword and the corresponding one or more query words for triggering the keyword according to each lexical item
Number and the number for the keyword of the lexical item occur, determine weight of the lexical item in the keyword.
5. according to the method described in claim 2, it is characterized in that, corresponding with the multiple keyword respectively according to what is obtained
Multiple word information set determine the degree of correlation between each keyword, further comprise:
Obtain the potential class of one or more relevant to the multiple keyword；
Based on the corresponding word information set of keyword each in the multiple keyword, each potential class is calculated in the multiple pass
Weight in keyword in each keyword；
It is determined in the multiple keyword between each keyword according to weight of each potential class in each keyword
The degree of correlation.
6. according to the method described in claim 5, it is characterized in that, corresponding based on each keyword in the multiple keyword
Word information set calculates weight of each potential class in the multiple keyword in each keyword, comprising:
Existed according to the one or more lexical items corresponding with the keyword for including in the corresponding word information set of each keyword
Weight in the keyword calculates probability that each potential class occurs in the multiple keyword and in each potential class
Under the probability that occurs of each keyword；
The probability occurred in each keyword according to each potential class and each keyword appearance under each potential class
Probability calculates weight of each potential class in the multiple keyword in each keyword.
7. the method according to claim 1, wherein according to the degree of correlation between each keyword to described
Multiple keywords are clustered, and to obtain one or more keyword categories, further comprise:
The keyword that the degree of correlation in the multiple keyword is more than predetermined threshold is divided into same keyword categories, to obtain one
Or multiple keyword categories.
8. a kind of ranking parameter value determining device based on keyword clustering characterized by comprising
First receiving module, the multiple key word informations and the first user for receiving the first user selection are to each keyword
Ranking parameter value, the ranking parameter value is for adjusting row of the corresponding first user information to be put of the keyword in the page
Name；
Cluster module, multiple keywords for selecting the first user cluster, to obtain one or more crucial parts of speech
Not；
Display module, the ranking parameter of multiple keywords and the first user for being selected according to the first user to each keyword
Value, shows the information to be put of corresponding first user of the multiple keyword, to obtain second user in each crucial part of speech
Operational feedback information in not corresponding information to be put；
Second receiving module determines the matched pass of the query information for receiving second user query information currently entered
Keyword, the corresponding information to be put of matched keyword and keyword categories belonging to matched keyword；
Determining module, for anti-using operation of the second user in the determining corresponding information to be put of the keyword categories
Feedforward information, adjustment matched keyword target ranking parameter value；
Adjust module, using the target ranking parameter value, adjustment the corresponding information to be put of matched keyword ranking；
Wherein, the cluster module further comprises:
Word segmentation processing module, for by keyword each in the multiple keyword and corresponding one for triggering the keyword or
Multiple queries word carries out word segmentation processing, to obtain multiple word information set corresponding with the multiple keyword respectively；
Degree of correlation determining module, for multiple word information set corresponding with the multiple keyword respectively that basis obtains, really
The degree of correlation in fixed the multiple keyword between each keyword；
Keyword clustering module, for being gathered according to the degree of correlation between each keyword to the multiple keyword
Class, to obtain one or more keyword categories.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410119451.7A CN104951460B (en) | 2014-03-27 | 2014-03-27 | Method and device is determined based on the ranking parameter value of keyword clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410119451.7A CN104951460B (en) | 2014-03-27 | 2014-03-27 | Method and device is determined based on the ranking parameter value of keyword clustering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104951460A CN104951460A (en) | 2015-09-30 |
| CN104951460B true CN104951460B (en) | 2019-02-05 |
Family
ID=54166123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410119451.7A Active CN104951460B (en) | 2014-03-27 | 2014-03-27 | Method and device is determined based on the ranking parameter value of keyword clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104951460B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844379A (en) * | 2015-12-04 | 2017-06-13 | 北京国双科技有限公司 | The processing method and processing device that keyword is delivered |
| CN105528728A (en) * | 2015-12-09 | 2016-04-27 | 江苏易销电子商务有限公司 | Mall e-commerce service platform based on cloud computing and method thereof |
| CN105930442B (en) * | 2016-04-19 | 2019-06-18 | 广州天趣网络科技有限公司 | A kind of site search engine optimization method and system based on statistics calculating |
| CN106874500B (en) * | 2017-02-24 | 2020-06-09 | 百度在线网络技术（北京）有限公司 | Method, device, equipment and storage medium for diagnosing material display |
| CN107481031A (en) * | 2017-07-19 | 2017-12-15 | 安徽拓通信科技集团股份有限公司 | Advertisement ROI detection methods based on big data |
| CN108537599B (en) * | 2018-04-17 | 2023-11-07 | 北京三快在线科技有限公司 | Query feedback method and device based on keyword aggregation and storage medium |
| CN110544110A (en) * | 2019-07-30 | 2019-12-06 | 中至数据集团股份有限公司 | SEM keyword release abnormity monitoring method and device and readable storage medium |
| CN110706021A (en) * | 2019-09-12 | 2020-01-17 | 微梦创科网络科技（中国）有限公司 | Advertisement putting method and system |
| CN113010636A (en) * | 2021-02-23 | 2021-06-22 | 玉米社(深圳)网络科技有限公司 | Method for rapidly detecting ranking of all keywords of website |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101482926A (en) * | 2009-02-19 | 2009-07-15 | 北京大学 | Extensible self-adapting multi-core classification method |
| CN102902671A (en) * | 2011-07-25 | 2013-01-30 | 腾讯科技（深圳）有限公司 | Search method and device for advertising system |
| CN103164454A (en) * | 2011-12-15 | 2013-06-19 | 百度在线网络技术（北京）有限公司 | Keyword grouping method and keyword grouping system |
| CN103514191A (en) * | 2012-06-20 | 2014-01-15 | 百度在线网络技术（北京）有限公司 | Method and device for determining keyword matching mode of target popularization information |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7792858B2 (en) * | 2005-12-21 | 2010-09-07 | Ebay Inc. | Computer-implemented method and system for combining keywords into logical clusters that share similar behavior with respect to a considered dimension |
-
2014
- 2014-03-27 CN CN201410119451.7A patent/CN104951460B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101482926A (en) * | 2009-02-19 | 2009-07-15 | 北京大学 | Extensible self-adapting multi-core classification method |
| CN102902671A (en) * | 2011-07-25 | 2013-01-30 | 腾讯科技（深圳）有限公司 | Search method and device for advertising system |
| CN103164454A (en) * | 2011-12-15 | 2013-06-19 | 百度在线网络技术（北京）有限公司 | Keyword grouping method and keyword grouping system |
| CN103514191A (en) * | 2012-06-20 | 2014-01-15 | 百度在线网络技术（北京）有限公司 | Method and device for determining keyword matching mode of target popularization information |
Non-Patent Citations (1)
| Title |
|---|
| 基于数据挖掘的SEM投放模型研究;邹佳君;《中国优秀硕士学位论文全文数据库信息科技辑》;20111215(第12期);I138-638,正文第31页5段至第32页第2段，第47页第6段至第48页第1段 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104951460A (en) | 2015-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104951460B (en) | Method and device is determined based on the ranking parameter value of keyword clustering | |
| US11941660B1 (en) | Conversion path performance measures and reports | |
| Grbovic et al. | Scalable semantic matching of queries to ads in sponsored search advertising | |
| CA2803552C (en) | System and method for evaluating search queries to identify titles for content production | |
| US8655912B2 (en) | Computer-implemented method and system for combining keywords into logical clusters that share similar behavior with respect to a considered dimension | |
| CN105765573B (en) | Improvements in website traffic optimization | |
| CN103914468A (en) | Method and device for searching for released information | |
| Karande et al. | Optimizing budget constrained spend in search advertising | |
| CN101685521A (en) | Method for showing advertisements in webpage and system | |
| CN103729359A (en) | Method and system for recommending search terms | |
| CN102194015B (en) | Retrieval information heat statistical method | |
| CN104679771A (en) | Individual data searching method and device | |
| US20140258001A1 (en) | Systems and Methods for Determining Net-New Keywords in Expanding Live Advertising Campaigns in Targeted Advertising Systems | |
| CN108765042A (en) | A kind of retail shop's proposed algorithm based on user Yu retail shop's network-physical space interbehavior | |
| CN106484698A (en) | A kind of method for pushing of search keyword and device | |
| CN108537599B (en) | Query feedback method and device based on keyword aggregation and storage medium | |
| US20160335678A1 (en) | Collective Expansion of Bid-Terms for Campaign Optimization | |
| Raghavan et al. | Probabilistic first pass retrieval for search advertising: from theory to practice | |
| US20080256037A1 (en) | Method and system for generating an ordered list | |
| Katukuri et al. | Large-scale recommendations in a dynamic marketplace | |
| CN116051196A (en) | Advertisement recommendation method, advertisement recommendation device, advertisement recommendation equipment, advertisement recommendation storage medium and advertisement recommendation computer product | |
| TWI486799B (en) | A method and a device for determining a weight value of a search word, a search result generating method, and a device | |
| CN102902671B (en) | Searching method in ad system and device | |
| Delfianto et al. | Content-targeted advertising using genetic programming | |
| AU2015203826A1 (en) | Conversion path performance measures and reports |
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 |