CN103136224A - Recommendation method and device for keywords - Google Patents
Recommendation method and device for keywords Download PDFInfo
- Publication number
- CN103136224A CN103136224A CN201110379470XA CN201110379470A CN103136224A CN 103136224 A CN103136224 A CN 103136224A CN 201110379470X A CN201110379470X A CN 201110379470XA CN 201110379470 A CN201110379470 A CN 201110379470A CN 103136224 A CN103136224 A CN 103136224A
- Authority
- CN
- China
- Prior art keywords
- word
- keyword
- unexpected rival
- utilization factor
- recommendation
- 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.)
- Pending
Links
Images
Abstract
The invention provides a recommendation method and a device for keywords. The recommendation method comprises the followings steps: digging out search terms as black-horse words, wherein an average popularization result number of the search terms is less than a .preset maximum popularization result number, and a predicted use ratio is larger than a preset use ratio threshold value, and adding the black-horse words into a recommendation word stock; and enabling the words in the recommendation word stock to be intended to be keywords which are recommended to a user, wherein the relevancy degree of the words and business of the user reaches a preset relevancy degree threshold value. According to the recommendation method and the device for the keywords, use ratio of popularization resources can be improved, waste of the popularization resources is reduced, income of providers who provide search engines is improved, and needs of clients and users can be better met.
Description
[technical field]
The present invention relates to technical field of the computer network, particularly a kind of recommend method of keyword and device.
[background technology]
Search is promoted as a kind of successful network promotion form, and commercial value is outstanding, and searched engine supplier adopts widely.Promote or marketing in order to do on network, client's (client who relates in the present invention is for doing the enterprise that promotes or market by network) buys keyword to the search engine supplier, when the user common user of network (user who relates in the present invention for) uses this keyword to start on search engine to search for, when representing the large search result, can represent to the user client's who buys this keyword popularization.Usually above the large search result or the right side, sequence can change according to other clients' that buy this keyword purchase payment status in the position of this popularization.The form of promoting typically refers to the link to this client web site.
The existing search way of promotion is recommended other keywords relevant to this keyword to the user when normally the user inputs certain keyword, and the confession user selects purchase.This keyword way of recommendation is merely based on the degree of correlation of keyword and candidate keywords; yet; in the actual application that search is promoted; the discontented situation in popularization position that some keyword usually can occur; namely when this keyword of user search; the situation of vacancy appears in the popularization position of this keyword, and this situation has following defective:
The waste of one, popularization resource.When the search that occurs based on this keyword, promote resource and be not fully utilized, when causing the wasting of resources, also affected search engine supplier's income.
Two, customer demand does not well satisfy.The client is when buying keyword, and hope can gain the initiative usually, buys and use keyword prior to other colleagues, and the keyword that vacancy appears in this popularization position is client " the value depression " that strive to find just.
Three, user's request is not well satisfied.When the user used search engine to search for certain keyword, the deficiency of Extended Results can cause some more relevant client not show in promoting the position, and user's actual demand may be not being met.
Obviously, existing keyword disseminate technology can not be found the discontented keyword in above-mentioned popularization position, thereby addresses the aforementioned drawbacks.
[summary of the invention]
The invention provides a kind of recommend method and device of keyword, so that improve the utilization factor of promoting resource, better satisfy client and user's demand.
Concrete technical scheme is as follows:
A kind of recommend method of keyword, the method comprises:
S1, from search excavate daily record average Extended Results number less than default maximum Extended Results number and estimate utilization factor greater than the search word of default utilization factor threshold value as unexpected rival's word, and described unexpected rival's word is added the recommendation dictionary;
S2, the word that reaches default degree of correlation threshold value with client's traffic aided degree in described recommendation dictionary is defined as recommending described client's keyword.
In described step S1, the excavation of described unexpected rival's word specifically comprises:
S11, obtain the search daily record in the setting-up time section;
S12, calculate the average Extended Results number of each search word in the search daily record, the ratio that the average Extended Results number of search word the searching times of Extended Results occurs for Extended Results number and this search word of this search word appearance;
S13, determine average Extended Results number less than the search word of default maximum Extended Results number as candidate unexpected rival's word;
S14, determine the expectation utilization factor of each candidate unexpected rival's word, select to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value as unexpected rival's word.
More preferably, also comprise between described step S11 and S12: S16, determine that from described search daily record searching times is greater than the search word of default searching times threshold value;
The average Extended Results number that calculates each search word in the search daily record in described step S12 is: the average Extended Results number that calculates each definite search word of described step S16.
Particularly, described in step S14, the expectation utilization factor of definite each candidate unexpected rival's word is:
According to existing utilization factor parameter value and the candidate unexpected rival's word and the degree of correlation of recommending dictionary of recommending keyword in dictionary, determine the expectation utilization factor of each candidate unexpected rival's word, described utilization factor parameter value comprises at least: number of clicks or purchase number of times.
According to
The expectation utilization factor score (w) of calculated candidate unexpected rival's word w, wherein, P
iBe i utilization factor parameter, M is the number of utilization factor parameter, α
iBe the weighted value of i utilization factor parameter, avg (P
i) be the existing P that recommends to satisfy in dictionary and between candidate unexpected rival's word w the keyword of presetting the correlativity requirement
iMean value.
Preferably, select described in step S14 to estimate that utilization factor comprises as unexpected rival's word greater than candidate unexpected rival's word of default utilization factor threshold value:
Select to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value, and obtain described unexpected rival's word after candidate unexpected rival's word of selecting is filtered according to default filtering policy.
Wherein, described step S2 specifically comprises:
Determine in described recommendation dictionary to reach with the degree of correlation of keyword that described client inputs the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Determine in described recommendation dictionary to reach with the degree of correlation of keyword that described client buys the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Extract proper vector from described client's business data, determine in described recommendation dictionary to reach with the degree of correlation of described proper vector the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client.
Further, the method also comprises: the keyword that will recommend described client sorts;
Wherein the mode of sequence comprises: will recommend that in described client's keyword, unexpected rival's word comes the front; Perhaps, consider unexpected rival's word weight in the sort algorithm of the keyword of recommending described client.
Preferably, identify the Regional Property of word in described recommendation dictionary, extract described client's Regional Property in described step S2, with reach default degree of correlation threshold value with client's traffic aided degree in described recommendation dictionary and have a keyword that the word of Domain Properties in the same manner is defined as recommending described client.
Further, the method also comprises: unexpected rival's word is provided unexpected rival's word sign or rationale for the recommendation in recommending described client's keyword.
A kind of recommendation apparatus of keyword, this device comprises:
Unexpected rival's word excavates the unit, be used for from the search daily record excavate average Extended Results number less than default maximum Extended Results number and estimate utilization factor greater than the search word of default utilization factor threshold value as unexpected rival's word, and described unexpected rival's word is added the recommendation dictionary;
The keyword recommendation unit is used for the keyword that word that traffic aided degree with described recommendation dictionary and client reaches default degree of correlation threshold value is defined as recommending described client.
Wherein, described unexpected rival's word excavates the unit and specifically comprises:
The log acquisition subelement is used for obtaining the search daily record in the setting-up time section;
Computation subunit be used for to be calculated the average Extended Results number of each search word of search daily record, the ratio that the average Extended Results number of search word the searching times of Extended Results occurs for Extended Results number and this search word of this search word appearance;
The first chooser unit, be used for determining average Extended Results number less than the search word of default maximum Extended Results number as candidate unexpected rival's word;
The second chooser unit is used for determining the expectation utilization factor of each candidate unexpected rival's word, selects to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value, and offers dictionary and add subelement;
Dictionary adds subelement, is used for the candidate unexpected rival's word that receives is added the recommendation dictionary as unexpected rival's word.
Preferably, described computation subunit determines that from described search daily record searching times is greater than the search word of default searching times threshold value, the average Extended Results number of each search word of calculative determination.
Particularly, described the second existing utilization factor parameter value and candidate unexpected rival's word and the degree of correlation of recommending dictionary of recommending keyword in dictionary of chooser unit basis, determine the expectation utilization factor of each candidate unexpected rival's word, described utilization factor parameter value comprises at least: number of clicks or purchase number of times.
Described the second chooser unit according to
The expectation utilization factor score (w) of calculated candidate unexpected rival's word w, wherein, P
iBe i utilization factor parameter, M is the number of utilization factor parameter, α
iBe the weighted value of i utilization factor parameter, avg (P
i) be the existing P that recommends to satisfy in dictionary and between candidate unexpected rival's word w the keyword of presetting the correlativity requirement
iMean value.
Further, described unexpected rival's word excavates the unit and also comprises:
Filter subelement, be used for described the second chooser unit is offered and offer again described dictionary after candidate unexpected rival's word that dictionary adds subelement filters according to default filtering policy and add subelement.
Wherein, described keyword recommendation unit determines in described recommendation dictionary to reach with the degree of correlation of keyword that described client inputs the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Determine in described recommendation dictionary to reach with the degree of correlation of keyword that described client buys the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Extract proper vector from described client's business data, determine in described recommendation dictionary to reach with the degree of correlation of described proper vector the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client.
Further, this device also comprises: the keyword sequencing unit is used for the keyword that described keyword recommendation unit is recommended is sorted;
Wherein the mode of sequence comprises: will recommend that in described client's keyword, unexpected rival's word comes the front; Perhaps, consider unexpected rival's word weight in the sort algorithm of the keyword of recommending described client.
Preferably, described keyword recommendation unit also is used for identifying the Regional Property of described recommendation dictionary word, when determining to recommend described client's keyword, with reach default degree of correlation threshold value with client's traffic aided degree in described recommendation dictionary and have a keyword that the word of Domain Properties in the same manner is defined as recommending described client.
In addition, described keyword recommendation unit also is used for providing unexpected rival's word sign or rationale for the recommendation at the keyword of recommending described client with unexpected rival's word.
as can be seen from the above technical solutions, the present invention is by excavating average Extended Results number less than default maximum Extended Results number and estimating utilization factor greater than the search word of default utilization factor threshold value and add the recommendation dictionary from the search daily record, make it possible to promoting the position discontented and estimate that the higher keyword of utilization factor recommends the client, thereby improve the utilization factor of promoting resource, reduced the waste of promoting resource, make the client can obtain fast " value depression ", gain the initiative, simultaneously also make the user when searching key word, can offer the abundant Extended Results of user, satisfied user's actual demand.
[description of drawings]
The main method process flow diagram that Fig. 1 provides for the embodiment of the present invention one;
The method for digging process flow diagram of unexpected rival's word that Fig. 2 provides for the embodiment of the present invention two;
The instance graph to user's recommended keywords that Fig. 3 provides for the embodiment of the present invention three;
The apparatus structure schematic diagram that Fig. 4 provides for the embodiment of the present invention four.
[embodiment]
In order to make the purpose, technical solutions and advantages of the present invention clearer, describe the present invention below in conjunction with the drawings and specific embodiments.
Embodiment one,
The main method process flow diagram that Fig. 1 provides for the embodiment of the present invention one, as shown in Figure 1, the method can comprise the following steps:
Step 101: from search excavate daily record average Extended Results number less than default maximum Extended Results number and estimate utilization factor greater than the search word of default utilization factor threshold value as unexpected rival's word, and unexpected rival's word of determining is added the recommendation dictionary.
This step is the excavation step of recommended keywords, soon promotes the position and exists vacancy and higher keyword (being called unexpected rival's word in the embodiment of the present invention) excavation of expectation utilization factor out to add in the recommendation dictionary, makes these unexpected rival's words can recommend the client.The concrete method for digging of unexpected rival's word will be described in detail in embodiment two.
Can only comprise unexpected rival's word in the recommendation dictionary that relates in the embodiment of the present invention, can also comprise except unexpected rival's word that also all clients' purchases keyword.
Step 102: the word that will recommend in dictionary traffic aided degree with the client to reach default degree of correlation threshold value is defined as recommending this client's keyword.
This step is the recommendation step of keyword, namely find the keyword that is fit to promote to the client in recommending dictionary, adopt the mode of active or passive triggering to recommend the client, in addition, also can relate to the sequence of keyword in recommendation process, concrete recommendation and sortord will be described in detail in embodiment three.
Embodiment two,
The method for digging process flow diagram of unexpected rival's word that Fig. 2 provides for the embodiment of the present invention two, i.e. the specific implementation of above-mentioned steps 101, as shown in Figure 2, this method for digging can comprise the following steps:
Step 201: obtain the search daily record in the setting-up time section.
Step 202: determine that from the search daily record of obtaining searching times (PV) is greater than the search word of default searching times threshold value (PV_th).
Usually the searching times of search word can embody the value of this search word, and lower its value of searching times is lower, and the present invention excludes choosing of unexpected rival's word with the search word that searching times is less than or equal to default searching times threshold value.Certainly, if also execution in step 202 not of the factor of not considering searching times.
Step 203: the average Extended Results number (ASN) of each search word that calculation procedure 202 is determined.
In this step, can adopt following formula to calculate:
Wherein PV_show is the Extended Results number that search word occurs, and epv is the searching times that Extended Results appears in search word.
For example, " fresh flower express delivery " Extended Results occurs as search word in the search daily record searching times is 200 times, the Extended Results number that occurs in these searching times is total up to 800, the average Extended Results number of this search word is 4 so, and this average Extended Results number has embodied the popularization position behaviour in service of search word.
Step 204: determine average Extended Results number less than the search word of default maximum Extended Results number as candidate unexpected rival's word.
Upper example continues, if search word " fresh flower express delivery " is 5 as the maximum Extended Results number of keyword, that is to say, average popularizations that arranges for " fresh flower express delivery " is 5, just illustrate that may there be the situation of promoting the position vacancy in this search word " fresh flower express delivery ", this with it as candidate unexpected rival's word.
So far in fact realize the excavation of vacancy popularization resource, this part has been promoted resource as candidate unexpected rival's word of recommending the client, can reduce the waste situation of the popularization resource that has vacancy.
Step 205: determine the expectation utilization factor of each candidate unexpected rival's word, select to estimate that utilization factor is greater than candidate unexpected rival's word of default utilization factor threshold value.
In this step, can according to existing utilization factor parameter value and the candidate unexpected rival's word and the degree of correlation of recommending dictionary of recommending keyword in dictionary, determine the expectation utilization factor of each candidate unexpected rival's word.Can adopt the method for machine learning, keyword in existing recommendation dictionary as training set, is recorded the utilization factor parameter of each keyword in this training set, such as number of clicks, purchase number of times, what this training set embodied is client and user's practice result, and can in time upgrade.Based on this training set, candidate unexpected rival's word is trained, the expectation utilization factor of each candidate unexpected rival's word of the relatedness computation of calculated candidate unexpected rival's word and training set, particularly, calculation training concentrate with candidate unexpected rival's word between satisfy the keyword of default correlativity requirement, utilize the expectation utilization factor of the utilization factor calculation of parameter candidate unexpected rival word of these keywords.
Wherein, the degree of correlation of candidate unexpected rival's word and training set can adopt such as existing correlation calculations methods such as BM25 algorithms and calculate.
For example, when adopting the degree of correlation of BM25 algorithm calculated candidate unexpected rival's word and training set, formula is:
Wherein, q is candidate unexpected rival's word, d is the keyword in training set, N is the keyword sum in training set, and c (w, d) is the occurrence number of word w in training set, df (w) is for comprising the number of files of word w in extensive document sets, | d| is the length of training set, and avgdl is the average length of each keyword in training set, b, k
1And k
3Be default parameter, b is used for controlling the punishment degree of training set length, k
1Be used for controlling word w at the contribution degree of training set occurrence number, k
3Be used for controlling word w in the contribution degree of candidate unexpected rival's word occurrence number.
When the expectation utilization factor of calculated candidate unexpected rival word, can adopt following formula:
Wherein, score (w) is the expectation utilization factor of candidate unexpected rival's word w, P
iBe i utilization factor parameter, M is the number of utilization factor parameter, α
iBe the weighted value of i utilization factor parameter, avg (P
i) be in training set and candidate unexpected rival's word w between satisfy the P of the keyword that default correlativity requires
iMean value.
Suppose to exist two utilization factor parameters: clicks and purchase number, if satisfying the keyword of default correlativity requirement in training set and between candidate unexpected rival's word w is A, B and C, the value that obtains after the value that obtains after the clicks of A, B and C in training set being averaging and the purchase number of A, B and C are averaging is weighted summation, namely obtains the expectation utilization factor of candidate unexpected rival's word w.
The method of this machine learning obtains the expectation utilization factor of candidate unexpected rival's word according to the known utilization factor of existing keyword, this training patterns that adopts can be excavated out and estimate the higher unexpected rival's word of utilization factor in candidate unexpected rival's word fast, thereby improve as far as possible the utilization factor of promoting resource, can bring more value to the client on the one hand, can offer on the other hand the more Extended Results of user.
Before this step, can at first carry out the url decoding to unexpected rival's word, usually unexpected rival's word is all url coding forms in the search daily record, before machine learning, in order to make unexpected rival's word consistent with the keyword form of the recommendation dictionary of server stores, at first each candidate unexpected rival's word is carried out the url decoding, this part is current techique, does not repeat them here.
Step 206: according to default filtering policy, obtain unexpected rival's word after candidate unexpected rival's word that step 205 is selected filters.
Filtering policy in this step normally filters out the keyword of violating national legislation, for example relates to the keyword of Huang or reaction.
Embodiment three,
In above-mentioned steps 102, when determining to recommend client's keyword, can adopt following three kinds of modes:
First kind of way: the way of recommendation of passive triggering, when the client inputs keyword, trigger the recommendation of keyword: determine to recommend in dictionary to reach with the degree of correlation of keyword that the client inputs the word of presetting degree of correlation threshold value, with this word as the keyword of recommending this client.
The second way: the way of recommendation initiatively, determine to recommend in dictionary to reach with the degree of correlation of keyword that the client buys the word of presetting degree of correlation threshold value, with this word as the keyword of recommending this client.
The third mode: the way of recommendation initiatively, extract proper vector from client's business data, determine to recommend in dictionary to reach with the degree of correlation of client's proper vector the word of default degree of correlation threshold value.
In this kind mode, can extract proper vector from the business data such as client's the popularization page, customer information (such as client's register name) or business information (business tine of filling in such as the client, industry scope etc.), then by in the calculated recommendation dictionary between word and proper vector the mode of the degree of correlation determine to recommend this client's keyword.The mode of wherein extracting proper vector is prior art, does not repeat them here.
The above-mentioned second way and the third mode can log in or client when inputting keyword the client, are initiatively the lead referral keyword.
In addition, above-mentioned three kinds of modes can be used in combination in any way, for example, and when the client inputs keyword, can adopt first kind of way and the second way to determine the keyword of recommending the client, jointly recommend the client after union got in the keyword that dual mode is determined.
In addition, the relatedness computation method that relates in above-mentioned three kinds of modes all can adopt existing relatedness computation method, and the present invention is not limited.
After adopting aforesaid way to determine the keyword of recommending the client, when recommending to the client, can adopt following keyword sortord:
Sortord 1, will come the front to unexpected rival's word in the keyword of lead referral.
If have unexpected rival's word in the keyword of lead referral, preferentially to lead referral unexpected rival word, so that the client can preferably see unexpected rival's word, thereby improve the purchased probability of unexpected rival's word.
Sortord 2, consider unexpected rival's word weight in the sort algorithm of keyword.
In existing keyword sort algorithm, what usually consider is the degree of correlation, user search number of times etc. of keyword and client's business, in embodiments of the present invention, whether be that the calculating that a weight participates in the keyword sequence given in unexpected rival's word with keyword, thereby make unexpected rival's word can appear at the keyword prostatitis of recommendation as far as possible.
As a kind of preferred embodiment, to the lead referral keyword time, can unexpected rival's word wherein be marked, this mark can include but not limited to: unexpected rival's word sign is provided, the rationale for the recommendation of this unexpected rival's word etc. is provided.
Wherein, unexpected rival's word sign can include but not limited to: with unexpected rival's word add black, unexpected rival's word (is for example adopted special color, mark pattern identification
NEW) etc.Rationale for the recommendation is for example: " the most emerging search word is wished your one step of straightforward man, seizes commercial opportunity ".This rationale for the recommendation can show all the time when recommending unexpected rival's word, also can show when mouse-over unexpected rival word.
Cite an actual example, as shown in Figure 3, suppose that the client inputs keyword " smart mobile phone ", the triggering keyword is recommended, adopt at least a keyword of recommending this client of determining in above-mentioned first kind of way, the second way and the third mode to be: smart mobile phone, business intelligence mobile phone, to buy smart mobile phone, second-hand smart mobile phone, domestic smart mobile phone, smart mobile phone wholesale etc., unexpected rival's word can be stood out when sequence, and unexpected rival's word is adopted diagrammatic representation
NEWMark.In addition, the dog-eat-dog degree of can further mark the average daily volumes of searches (as shown in Figure 3) of institute's recommended keywords, determining according to average Extended Results number etc.
Preferably, consider that client's input demand has notable feature usually on the region, namely the client may preferably promote the input of content in same region, therefore can introduce regional characteristic in above-mentioned recommendation process.Particularly, when recommending the formation of dictionary, due to based on be the search daily record, can get the Regional Property (obtaining of this Regional Property can be adopted existing mode) of word, to user's recommended keywords the time, will recommend in dictionary that traffic aided degree with the client reaches default degree of correlation threshold value and have the keyword that the word of Domain Properties in the same manner is defined as recommending the client.
Be more than the detailed description that the method that the embodiment of the present invention provides is carried out, be described in detail below by four pairs of devices provided by the present invention of embodiment.
Embodiment four,
The structure drawing of device that Fig. 4 provides for the embodiment of the present invention four, as shown in Figure 4, this device comprises: unexpected rival's word excavates unit 400 and keyword recommendation unit 410.
Unexpected rival's word excavate unit 400 excavate from the search daily record average Extended Results number less than default maximum Extended Results number and estimate utilization factor greater than the search word of default utilization factor threshold value as unexpected rival's word, and unexpected rival's word is added the recommendation dictionary.
The word that keyword recommendation unit 410 will recommend in dictionary the traffic aided degree with the client to reach default degree of correlation threshold value is defined as recommending client's keyword.
Wherein, unexpected rival's word excavates the unit and can specifically comprise: log acquisition subelement 401, computation subunit 402, the first chooser unit 403, the second chooser unit 404 and dictionary add subelement 405.
Usually the searching times of search word can embody the value of this search word, lower its value of searching times is lower, therefore, in order to improve the digging efficiency of unexpected rival's word, the search word that can at first searching times be less than or equal to default searching times threshold value excludes choosing of unexpected rival's word.At this moment, computation subunit 402 determines that from the search daily record searching times is greater than the search word of default searching times threshold value, the average Extended Results number of each search word of calculative determination.
The first chooser unit 403 determine average Extended Results numbers less than the search word of default maximum Extended Results number as candidate unexpected rival's word.
The expectation utilization factor of each candidate unexpected rival's words is determined in the second chooser unit 404, selects to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value, and offers dictionary and add subelement 405.
Particularly, the expectation utilization factor of each candidate unexpected rival's word is determined according to existing utilization factor parameter value and the candidate unexpected rival's words and the degree of correlation of recommending dictionary of recommending keyword in dictionaries in the second chooser unit 404.Can adopt the method for machine learning, with the keyword in existing recommendation dictionary as training set, the utilization factor parameter of each keyword can comprise number of clicks, buy number of times etc. in this training set, and what this training set embodied is client and user's practice result, and can in time upgrade.Based on this training set, candidate unexpected rival's word is trained, the degree of correlation according to candidate unexpected rival's word and training set is given the expectation utilization factor for each candidate unexpected rival's word, particularly, calculation training concentrate with candidate unexpected rival's word between satisfy the keyword of default correlativity requirement, utilize the expectation utilization factor of the utilization factor parameter value calculation candidate unexpected rival word of these keywords.
For example can according to
The expectation utilization factor score (w) of calculated candidate unexpected rival's word w, wherein, P
iBe i utilization factor parameter, M is the number of utilization factor parameter, α
iBe the weighted value of i utilization factor parameter, avg (P
i) be the existing P that recommends to satisfy in dictionary and between candidate unexpected rival's word w the keyword of presetting the correlativity requirement
iMean value.。
Dictionary adds subelement 405 the candidate unexpected rival's word that receives is added the recommendation dictionary as unexpected rival's word.
Further, unexpected rival's word excavation unit 400 can also comprise: filter subelement 406.
Filtering 406 pairs, subelement the second chooser unit 404 offers and offers dictionary after candidate unexpected rival's word that dictionary adds subelement 405 filters according to default filtering policy again and add subelement 405.
When determining to recommend client's keyword, keyword recommendation unit 410 can adopt following three kinds of modes:
First kind of way: determine to recommend to reach with the degree of correlation of keyword that the client inputs in dictionary the word of default degree of correlation threshold value, with definite word as the keyword of recommending the client.
The second way: determine to recommend to reach with the degree of correlation of keyword that the client buys in dictionary the word of default degree of correlation threshold value, with definite word as the keyword of recommending the client.
The third mode: extract proper vector from client's business data, determine to recommend to reach with the degree of correlation of proper vector in dictionary the word of default degree of correlation threshold value, with the word determined as the keyword of recommending the client.
Further, this device also comprises: keyword sequencing unit 420 is used for the keyword that keyword recommendation unit 410 is recommended is sorted.Wherein the mode of sequence comprises: will recommend that in client's keyword, unexpected rival's word comes the front; Perhaps, consider unexpected rival's word weight in the sort algorithm of the keyword of recommending the client.
The input demand of considering the client has notable feature usually on the region, namely the client may preferably promote the input of content in same region, therefore can introduce regional characteristic in above-mentioned recommendation process.At this moment, keyword recommendation unit 410 also is used for the Regional Property that the dictionary word is recommended in identification, when determining to recommend client's keyword, will recommend in dictionary that traffic aided degree with the client reaches default degree of correlation threshold value and have the keyword that the word of Domain Properties in the same manner is defined as recommending the client.
The above is only preferred embodiment of the present invention, and is in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of making, is equal to replacement, improvement etc., within all should being included in the scope of protection of the invention.
Claims (20)
1. the recommend method of a keyword, is characterized in that, the method comprises:
S1, from search excavate daily record average Extended Results number less than default maximum Extended Results number and estimate utilization factor greater than the search word of default utilization factor threshold value as unexpected rival's word, and described unexpected rival's word is added the recommendation dictionary;
S2, the word that reaches default degree of correlation threshold value with client's traffic aided degree in described recommendation dictionary is defined as recommending described client's keyword.
2. recommend method according to claim 1, is characterized in that, in described step S1, the excavation of described unexpected rival's word specifically comprises:
S11, obtain the search daily record in the setting-up time section;
S12, calculate the average Extended Results number of each search word in the search daily record, the ratio that the average Extended Results number of search word the searching times of Extended Results occurs for Extended Results number and this search word of this search word appearance;
S13, determine average Extended Results number less than the search word of default maximum Extended Results number as candidate unexpected rival's word;
S14, determine the expectation utilization factor of each candidate unexpected rival's word, select to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value as unexpected rival's word.
3. recommend method according to claim 2, is characterized in that, also comprises between described step S11 and S12: S16, determine that from described search daily record searching times is greater than the search word of default searching times threshold value;
The average Extended Results number that calculates each search word in the search daily record in described step S12 is: the average Extended Results number that calculates each definite search word of described step S16.
4. recommend method according to claim 2, is characterized in that, determines described in step S14 that the expectation utilization factor of each candidate unexpected rival's word is:
According to existing utilization factor parameter value and the candidate unexpected rival's word and the degree of correlation of recommending dictionary of recommending keyword in dictionary, determine the expectation utilization factor of each candidate unexpected rival's word, described utilization factor parameter value comprises at least: number of clicks or purchase number of times.
5. recommend method according to claim 4, is characterized in that, according to
The expectation utilization factor score (w) of calculated candidate unexpected rival's word w, wherein, P
iBe i utilization factor parameter, M is the number of utilization factor parameter, α
iBe the weighted value of i utilization factor parameter, avg (P
i) be the existing P that recommends to satisfy in dictionary and between candidate unexpected rival's word w the keyword of presetting the correlativity requirement
iMean value.
6. recommend method according to claim 2, is characterized in that, selects described in step S14 to estimate that utilization factor comprises as unexpected rival's word greater than candidate unexpected rival's word of default utilization factor threshold value:
Select to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value, and obtain described unexpected rival's word after candidate unexpected rival's word of selecting is filtered according to default filtering policy.
7. recommend method according to claim 1, is characterized in that, described step S2 specifically comprises:
Determine in described recommendation dictionary to reach with the degree of correlation of keyword that described client inputs the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Determine in described recommendation dictionary to reach with the degree of correlation of keyword that described client buys the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Extract proper vector from described client's business data, determine in described recommendation dictionary to reach with the degree of correlation of described proper vector the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client.
8. according to claim 1 or 7 described recommend methods, is characterized in that, the method also comprises: the keyword that will recommend described client sorts;
Wherein the mode of sequence comprises: will recommend that in described client's keyword, unexpected rival's word comes the front; Perhaps, consider unexpected rival's word weight in the sort algorithm of the keyword of recommending described client.
9. recommend method according to claim 1, it is characterized in that, identify the Regional Property of word in described recommendation dictionary, extract described client's Regional Property in described step S2, with reach default degree of correlation threshold value with client's traffic aided degree in described recommendation dictionary and have a keyword that the word of Domain Properties in the same manner is defined as recommending described client.
10. the described recommend method of arbitrary claim of according to claim 1 to 7, is characterized in that, the method also comprises: unexpected rival's word is provided unexpected rival's word sign or rationale for the recommendation in recommending described client's keyword.
11. the recommendation apparatus of a keyword is characterized in that, this device comprises:
Unexpected rival's word excavates the unit, be used for from the search daily record excavate average Extended Results number less than default maximum Extended Results number and estimate utilization factor greater than the search word of default utilization factor threshold value as unexpected rival's word, and described unexpected rival's word is added the recommendation dictionary;
The keyword recommendation unit is used for the keyword that word that traffic aided degree with described recommendation dictionary and client reaches default degree of correlation threshold value is defined as recommending described client.
12. recommendation apparatus according to claim 11 is characterized in that, described unexpected rival's word excavates the unit and specifically comprises:
The log acquisition subelement is used for obtaining the search daily record in the setting-up time section;
Computation subunit be used for to be calculated the average Extended Results number of each search word of search daily record, the ratio that the average Extended Results number of search word the searching times of Extended Results occurs for Extended Results number and this search word of this search word appearance;
The first chooser unit, be used for determining average Extended Results number less than the search word of default maximum Extended Results number as candidate unexpected rival's word;
The second chooser unit is used for determining the expectation utilization factor of each candidate unexpected rival's word, selects to estimate utilization factor greater than candidate unexpected rival's word of default utilization factor threshold value, and offers dictionary and add subelement;
Dictionary adds subelement, is used for the candidate unexpected rival's word that receives is added the recommendation dictionary as unexpected rival's word.
13. recommendation apparatus according to claim 12 is characterized in that, described computation subunit determines that from described search daily record searching times is greater than the search word of default searching times threshold value, the average Extended Results number of each search word of calculative determination.
14. recommendation apparatus according to claim 12, it is characterized in that, described the second chooser unit concrete utilization factor parameter value and candidate unexpected rival's word and the degree of correlation of recommending dictionary according to keyword in existing recommendation dictionary, determine the expectation utilization factor of each candidate unexpected rival's word, described utilization factor parameter value comprises at least: number of clicks or purchase number of times.
15. recommendation apparatus according to claim 14 is characterized in that, described the second chooser unit according to
The expectation utilization factor score (w) of calculated candidate unexpected rival's word w, wherein, P
iBe i utilization factor parameter, M is the number of utilization factor parameter, α
iBe the weighted value of i utilization factor parameter, avg (P
i) be the existing P that recommends to satisfy in dictionary and between candidate unexpected rival's word w the keyword of presetting the correlativity requirement
iMean value.
16. recommendation apparatus according to claim 12 is characterized in that, described unexpected rival's word excavates the unit and also comprises:
Filter subelement, be used for described the second chooser unit is offered and offer again described dictionary after candidate unexpected rival's word that dictionary adds subelement filters according to default filtering policy and add subelement.
17. recommendation apparatus according to claim 11, it is characterized in that, described keyword recommendation unit determines in described recommendation dictionary to reach with the degree of correlation of keyword that described client inputs the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Determine in described recommendation dictionary to reach with the degree of correlation of keyword that described client buys the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client; Perhaps,
Extract proper vector from described client's business data, determine in described recommendation dictionary to reach with the degree of correlation of described proper vector the word of default degree of correlation threshold value, with the word determined as the keyword of recommending described client.
18. according to claim 11 or 17 described recommendation apparatus is characterized in that, this device also comprises: the keyword sequencing unit is used for the keyword that described keyword recommendation unit is recommended is sorted;
Wherein the mode of sequence comprises: will recommend that in described client's keyword, unexpected rival's word comes the front; Perhaps, consider unexpected rival's word weight in the sort algorithm of the keyword of recommending described client.
19. recommendation apparatus according to claim 11, it is characterized in that, described keyword recommendation unit also is used for identifying the Regional Property of described recommendation dictionary word, when determining to recommend described client's keyword, with reach default degree of correlation threshold value with client's traffic aided degree in described recommendation dictionary and have a keyword that the word of Domain Properties in the same manner is defined as recommending described client.
20. according to claim 11 to the 17 described recommendation apparatus of arbitrary claim, it is characterized in that, described keyword recommendation unit also is used for providing unexpected rival's word sign or rationale for the recommendation at the keyword of recommending described client with unexpected rival's word.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110379470XA CN103136224A (en) | 2011-11-24 | 2011-11-24 | Recommendation method and device for keywords |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110379470XA CN103136224A (en) | 2011-11-24 | 2011-11-24 | Recommendation method and device for keywords |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103136224A true CN103136224A (en) | 2013-06-05 |
Family
ID=48496060
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110379470XA Pending CN103136224A (en) | 2011-11-24 | 2011-11-24 | Recommendation method and device for keywords |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103136224A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103412926A (en) * | 2013-08-14 | 2013-11-27 | 上海兆民云计算科技有限公司 | Method for building cloud storage retrieval index based on network access point characteristics |
| CN103559284A (en) * | 2013-11-07 | 2014-02-05 | 北京国双科技有限公司 | Word expansion method and device for webpage keywords |
| CN103902714A (en) * | 2014-04-03 | 2014-07-02 | 北京国双科技有限公司 | Method and device for filtering keywords |
| CN104091058A (en) * | 2014-06-27 | 2014-10-08 | 北京君和信达科技有限公司 | Safety inspection conclusion submitting method and device |
| CN104424342A (en) * | 2013-09-11 | 2015-03-18 | 携程计算机技术（上海）有限公司 | Method for keyword matching, and device, server and system of method |
| CN104965842A (en) * | 2014-11-27 | 2015-10-07 | 深圳市腾讯计算机系统有限公司 | Search recommending method and apparatus |
| CN105183710A (en) * | 2015-06-23 | 2015-12-23 | 武汉传神信息技术有限公司 | Method for automatically generating document summary |
| CN106484698A (en) * | 2015-08-25 | 2017-03-08 | 北京奇虎科技有限公司 | A kind of method for pushing of search keyword and device |
| CN106557480A (en) * | 2015-09-25 | 2017-04-05 | 阿里巴巴集团控股有限公司 | Implementation method and device that inquiry is rewritten |
| CN106598976A (en) * | 2015-10-15 | 2017-04-26 | 百度在线网络技术（北京）有限公司 | Internet-based information promotion method and apparatus |
| CN106649323A (en) * | 2015-10-29 | 2017-05-10 | 北京国双科技有限公司 | Method and device for recommending keyword |
| CN106919693A (en) * | 2017-03-07 | 2017-07-04 | 广州优视网络科技有限公司 | It is a kind of to improve the method and apparatus that hot word exposes coverage rate |
| CN107066497A (en) * | 2016-12-29 | 2017-08-18 | 努比亚技术有限公司 | A kind of searching method and device |
| WO2017143703A1 (en) * | 2016-02-24 | 2017-08-31 | 百度在线网络技术（北京）有限公司 | Offline resource mining method and device |
| CN107871259A (en) * | 2016-09-26 | 2018-04-03 | 阿里巴巴集团控股有限公司 | A kind of processing method of information recommendation, device and client |
| CN108304533A (en) * | 2018-01-29 | 2018-07-20 | 上海名轩软件科技有限公司 | Keyword recommendation method and equipment |
| CN110069676A (en) * | 2017-09-28 | 2019-07-30 | 北京国双科技有限公司 | Keyword recommendation method and device |
| CN113836379A (en) * | 2021-09-26 | 2021-12-24 | 北京百炼智能科技有限公司 | Intelligent recommendation method and system based on customer image |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101331487A (en) * | 2005-12-15 | 2008-12-24 | 微软公司 | Advertising keyword cross-selling |
| CN101650731A (en) * | 2009-08-31 | 2010-02-17 | 浙江大学 | Method for generating suggested keywords of sponsored search advertisement based on user feedback |
| US20100076991A1 (en) * | 2008-09-09 | 2010-03-25 | Kabushiki Kaisha Toshiba | Apparatus and method product for presenting recommended information |
-
2011
- 2011-11-24 CN CN201110379470XA patent/CN103136224A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101331487A (en) * | 2005-12-15 | 2008-12-24 | 微软公司 | Advertising keyword cross-selling |
| US20100076991A1 (en) * | 2008-09-09 | 2010-03-25 | Kabushiki Kaisha Toshiba | Apparatus and method product for presenting recommended information |
| CN101650731A (en) * | 2009-08-31 | 2010-02-17 | 浙江大学 | Method for generating suggested keywords of sponsored search advertisement based on user feedback |
Non-Patent Citations (1)
| Title |
|---|
| 唐卫东等: "《基于关键词效能的搜索引擎优化策略分析》", 《现代情报》, vol. 31, no. 10, 31 October 2011 (2011-10-31) * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103412926A (en) * | 2013-08-14 | 2013-11-27 | 上海兆民云计算科技有限公司 | Method for building cloud storage retrieval index based on network access point characteristics |
| CN104424342A (en) * | 2013-09-11 | 2015-03-18 | 携程计算机技术（上海）有限公司 | Method for keyword matching, and device, server and system of method |
| CN103559284A (en) * | 2013-11-07 | 2014-02-05 | 北京国双科技有限公司 | Word expansion method and device for webpage keywords |
| CN103559284B (en) * | 2013-11-07 | 2017-08-01 | 北京国双科技有限公司 | Web Page Key Words open up word method and apparatus |
| CN103902714B (en) * | 2014-04-03 | 2017-08-01 | 北京国双科技有限公司 | Keyword filter method and device |
| CN103902714A (en) * | 2014-04-03 | 2014-07-02 | 北京国双科技有限公司 | Method and device for filtering keywords |
| CN104091058A (en) * | 2014-06-27 | 2014-10-08 | 北京君和信达科技有限公司 | Safety inspection conclusion submitting method and device |
| CN104965842A (en) * | 2014-11-27 | 2015-10-07 | 深圳市腾讯计算机系统有限公司 | Search recommending method and apparatus |
| CN105183710A (en) * | 2015-06-23 | 2015-12-23 | 武汉传神信息技术有限公司 | Method for automatically generating document summary |
| CN106484698A (en) * | 2015-08-25 | 2017-03-08 | 北京奇虎科技有限公司 | A kind of method for pushing of search keyword and device |
| CN106557480A (en) * | 2015-09-25 | 2017-04-05 | 阿里巴巴集团控股有限公司 | Implementation method and device that inquiry is rewritten |
| CN106557480B (en) * | 2015-09-25 | 2020-07-07 | 阿里巴巴集团控股有限公司 | Method and device for realizing query rewriting |
| CN106598976A (en) * | 2015-10-15 | 2017-04-26 | 百度在线网络技术（北京）有限公司 | Internet-based information promotion method and apparatus |
| CN106649323B (en) * | 2015-10-29 | 2020-07-03 | 北京国双科技有限公司 | Method and device for recommending keywords |
| CN106649323A (en) * | 2015-10-29 | 2017-05-10 | 北京国双科技有限公司 | Method and device for recommending keyword |
| US11416502B2 (en) | 2016-02-24 | 2022-08-16 | Baidu Online Network Technology (Beijing) Co., Ltd. | Method and apparatus for mining offline resources |
| WO2017143703A1 (en) * | 2016-02-24 | 2017-08-31 | 百度在线网络技术（北京）有限公司 | Offline resource mining method and device |
| CN107871259A (en) * | 2016-09-26 | 2018-04-03 | 阿里巴巴集团控股有限公司 | A kind of processing method of information recommendation, device and client |
| CN107066497A (en) * | 2016-12-29 | 2017-08-18 | 努比亚技术有限公司 | A kind of searching method and device |
| CN106919693B (en) * | 2017-03-07 | 2020-12-01 | 阿里巴巴（中国）有限公司 | Method and device for improving hot word exposure coverage rate |
| CN106919693A (en) * | 2017-03-07 | 2017-07-04 | 广州优视网络科技有限公司 | It is a kind of to improve the method and apparatus that hot word exposes coverage rate |
| CN110069676A (en) * | 2017-09-28 | 2019-07-30 | 北京国双科技有限公司 | Keyword recommendation method and device |
| CN108304533A (en) * | 2018-01-29 | 2018-07-20 | 上海名轩软件科技有限公司 | Keyword recommendation method and equipment |
| CN113836379A (en) * | 2021-09-26 | 2021-12-24 | 北京百炼智能科技有限公司 | Intelligent recommendation method and system based on customer image |
| CN113836379B (en) * | 2021-09-26 | 2023-08-25 | 北京百炼智能科技有限公司 | Intelligent recommendation method and system based on client image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103136224A (en) | Recommendation method and device for keywords | |
| CN106557513B (en) | Event information pushing method and event information pushing device | |
| CN104750789B (en) | The recommendation method and device of label | |
| CN105005582B (en) | The recommendation method and device of multimedia messages | |
| US10049132B2 (en) | Personalizing query rewrites for ad matching | |
| Liu et al. | Who is. com? Learning to parse WHOIS records | |
| CN103118111B (en) | Information push method based on data from a plurality of data interaction centers | |
| CN101957834B (en) | Content recommending method and device based on user characteristics | |
| CN102693271B (en) | A kind of network information recommending method and system | |
| CN103207881B (en) | Querying method and device | |
| CN105808685A (en) | Promotion information pushing method and device | |
| US20180102947A1 (en) | Network based data traffic latency reduction | |
| US9584589B2 (en) | Friend recommendation method, apparatus and storage medium | |
| CN103778151A (en) | Method and device for identifying characteristic group and search method and device | |
| US20200065337A1 (en) | Gain adjustment component for computer network routing infrastructure | |
| CN104750760A (en) | Application software recommending method and device | |
| CN103365904A (en) | Advertising information searching method and system | |
| JP2011227721A (en) | Interest extraction device, interest extraction method, and interest extraction program | |
| CN103944935A (en) | Method and device for releasing web site information | |
| US20110055229A1 (en) | System and method for generating a valuation of revenue opportunity for a keyword from a valuation of online sessions on a website from user activities following a keyword search | |
| CN112131259A (en) | Similar malware recommendation method, device, medium and equipment | |
| CN106528066B (en) | Business process operation method and device | |
| WO2017050991A1 (en) | Aggregating profile information | |
| CN104391982A (en) | Information recommendation method and information recommendation system | |
| CN104657882B (en) | The acquisition methods and device of promotion effect data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20130605 |