Background technology
Microblogging is a kind of social network-i i-platform sharing the broadcast type of brief real-time information by paying close attention to mechanism.User can other people receive content by paying close attention to (follow), also can forward, comment on and collect the content that other people issue.Also to be foremost microblogging be the earliest U.S. pushes away spy (twitter), and domestic main flow microblog comprises Sina's microblogging and Tengxun's microblogging etc.
In the prior art, microblogging has two large shortcomings.One is that the number that user pays close attention to limits by Dunbar number, and another is the accurate input being difficult to realize advertisement in microblog.
Dunbar number is proposed by the guest anthropologist sieve Dunbar (Robin Dunbar) of Regius professor.One of Dunbar mathematics opinion basic theory being considered to social networks, namely the social number upper limit of the mankind is 150 people.This theory makes microblog users sink into awkward predicament.User wishes to pay close attention to more people to obtain more internet information on the one hand, and when the user number paid close attention to as user on the other hand exceedes Dunbar number, user cannot bear a large amount of information loads.If cause the reason of this predicament to be that you have paid close attention to someone in microblogging, you just have to receive all the elements that he issues, and can not receive certain class content selectively.The content that such as certain microblogging famous person issues contains the fields such as IT, education, business administration and life & amusement, even if you are only interested in the content in certain field that he issues, you also have to receive its all the elements issued.If publisher can send content according to classification, and each user can receive content according to publisher and classification, the quantity that so user pays close attention to other people will surmount the restriction of Dunbar number widely, and the content that microblog users obtains is also more accurate and personalized.
Delivery accurate advertisement on microblogging is a difficult problem for puzzlement industry always.Microblogging is thrown in advertisement and cannot obtain the huge income of throwing in advertisement on a search engine and producing, its reason is that advertisement thrown in the keyword can submitted to according to user on a search engine, and the keyword that user submits to usually can show the demand of user, therefore on a search engine can delivery accurate advertisement.Such as time user search keyword " smart mobile phone ", just can throw in the advertisement relevant to smart mobile phone to user.But microblogging does not but have such advantage.The general data that microblogging can utilize is user social contact relation data, and the demand of user is difficult to be obtained by user social contact relation, is therefore difficult to delivery accurate advertisement on microblogging.If can improve existing microblogging, make user can come content distributed according to classification, content can be received according to publisher and classification again, just can use for reference the successful practice of search engine, delivery accurate advertisement on microblogging.According to statistics, the microblogging registered user sum of the company such as Sina and Tengxun breaks through 600,000,000, and the number of users logged in every day has exceeded 40,000,000.Because microblog users group is most active colony in China Internet, they are crowds the most responsive to fangle, are also the crowds that purchasing power is the highest, on microblogging, therefore throw in advertising potential huge.
Except microblogging, also there are the problems referred to above in other social network-i i-platform.Such as, if group's number of micro-letter or QQ is more, will produce information overload situation, this greatly inhibits user to issue and receive the demand of content.Also have, some businessmans wish to be marketed by micro-telecommunications services number, but businessman has multiclass product needed to promote usually, the bulk information that therefore businessman issues usually causes interference to user, and micro-telecommunications services business also has to appearance measure to limit the content distributed quantity of businessman.If can arrange multiple content type channel in each micro-letter or QQ group, user just can receive content by category as required, and then reduces the generation of information overload situation.Equally, if multiple marketing categories channel can be offered under each micro-telecommunications services number, user just can receive market content by category as required, so both decreases bulk information bothering user, can provide personalized precision marketing service again for user.
Embodiment
By reference to the accompanying drawings the inventive method is described in further detail.
Fig. 1 is that user collects, the explanation of classification collection and set of advertisements.In social networks, composition user is numbered to each user ID and collects U={1,2 ..., M}, wherein M is total number of users.At least one content type of each curriculum offering that content publisher issues for it on social networks, is numbered composition classification collection C={1,2 by each content type ..., N}, wherein N is classification sum.Composition set of advertisements A={1 is numbered, 2 to the advertisement that advertiser throws in ..., Q}, wherein Q is advertisement sum.
Fig. 2 is the method to set up in distributing content data storehouse.When user is content distributed on social networks, need the classification that this content is set simultaneously, and store a content record in the distributing content data storehouse of server, this record comprises the user ID of context number, the content of issue, content type and content publisher.Wherein, content publisher belongs to described user and collects U, and content type belongs to described classification collection C.Such as in fig. 2 the 1st article record represents that user 7342 has issued " content 1 ", and user 7342 is the issue classification that " content 1 " is arranged is 653.Each content at least comprises a classification.In addition, the content issuing classification cannot be defined, one can be included into and be called in the public classification of " other ".
Fig. 3 is the Content Selection set R of user nnmethod to set up.Content Selection set Rnrepresent that user n (n ∈ U) receives the content and content distributed classification which publisher issues.Content Selection set Rncan be expressed as formally ..., (a, b) ...), wherein (a, b) ∈ Rn, a ∈ U, b ∈ C, element (a, b) represents that user n receives the content of the classification b that user a issues.Content Selection set Rnin each element in a database with one record form store.Fig. 3 illustrates the Content Selection set R of the user being numbered 958958storage means, wherein, every a line represents Content Selection set R958an element.It is the content of 32 that the data representation user 958 of Fig. 3 is provided with the class number receiving user 256 issue respectively, the class number receiving user 3746 issue is the content of 8765, the classification receiving user 55365 issue is the content of 32, the classification receiving user 3647 issue is the content of 768, receive user 3456 issue classification be the content of 95, and receive user 37480 issue classification be 32 and classification be the content of 692.Therefore, the Content Selection set R of user 958958={ (256,32), (3746,8765), (55365,32), (3647,768), (3456,95), (37480,32), (37480,692) }.When user receives multiple classifications of a publisher content distributed, by the different elements in each classification in publisher's mark and described multiple classification respectively component content screening set, such as go up (37480,32) and (37480,692) in example.
Fig. 4 is the issue classification collection of user and the method to set up of reception classification collection.If certain user has issued the content that classification is c, we have just said that the issue classification collection of this user comprises classification c.The issue classification collection formalization of user i is expressed as s (i, 1), s (i, 2) ..., s (i, j) ..., s (i, SPi), wherein s (i, j) represents that the jth of user i issues the title of classification, SPirepresent the classification number of the issue classification collection of user i.The issue classification collection of each user can be obtained according to the distributing content data storehouse of Fig. 2.The issue classification collection that such as can obtain user 1 according to Fig. 2 is { 86,323,587}.Suppose that classification 86,323 and 587 represents " management " respectively, " education ", and " IT ", then have SP for user 11=3, s (1,1)=management, s (1,2)=education, s (1,3)=IT.
If certain user have received the content that classification is d, we just say that the reception classification collection of this user comprises classification d.The reception classification collection formalization of user i is expressed as r (i, 1), r (i, 2) ..., r (i, j) ..., r (i, RPi), wherein r (i, j) represents that the jth of user i receives the title of classification, RPirepresent the classification number of the reception classification collection of user i.The reception classification collection that such as can obtain user 958 according to Fig. 3 is { 32,95,692,768,8765}.Suppose that classification 32,95,692,768 and 8765 represents " literature " respectively, " history ", " music ", " physical culture " and " shopping ", then have RP for user 958958=5, r (958,1)=literature, r (958,2)=history, r (958,3)=music, r (958,4)=physical culture, r (958,5)=shopping.Described issue classification and described reception classification are all the elements of classification collection C.
Fig. 5 is a kind of method issued and receive social network content of classifying.Described method is included in the server of accessing Internet and performs following steps:
Step S1. obtains and stores user and collects U={1,2 ..., M} and classification collection C={1,2 ..., N};
Step S2. receives and stores the content distributed signal of arbitrary user m (m ∈ U), at least one issue classification that described signal comprises a content of described user m issue, described user m is this curriculum offering and the mark of described user m, described issue classification belongs to described classification collection C;
Step S3. receives and stores the signal that arbitrary user n (n ∈ U) receives content, described signal comprises the mark of described user n, the mark of content publisher user p (p ∈ U) and at least one reception classification, and described reception classification belongs to the issue classification collection of described user p;
Step S4., according to the signal of described reception content, by the mark of described user p and at least one reception classification described, adds the Content Selection set R of described user n tonin;
Step S5. when described user n asks update content, according to described Content Selection set Rn, at least one up-to-date is content distributedly sent to described user n;
Return described step S2.
In an application example of method described in Fig. 5, described user m, user n and user p represent described user respectively and collect any one user in U, instead of refer in particular to certain user.Such as, when certain performs described step S2 to step S6, the mark of user m, user n and user p is respectively 353,5764 and 2333.And upper once perform described step S2 to step S6 time, the mark of user m, user n and user p is respectively 6543,84763 and 67.
In an application example of method described in Fig. 5, described method also comprises described user n and inquires about content distributed step according to specific classification of issuing.Such as user n will search the content of microblog that class number is 653, then according to the distributing content data storehouse of Fig. 2, find classification be 653 content comprise content 1, content 5 and content 9.
In an application example of method described in Fig. 5, described method also comprises described user n and inquires about other users and also set up according to specific classification of issuing the step associated with the social activity of other users described.Such as user n will search and issue the publisher that class number is 653, then according to the distributing content data storehouse of Fig. 2 find issue classification be 653 content publisher be user 7342, user 9735 and user 4167 respectively.Then described user n pays close attention to these content publishers respectively.
In an application example of method described in Fig. 5, if described user n wishes all the elements paying close attention to described user p issue, then each classification that described user n pays close attention to described user p is set.
In an application example of method described in Fig. 5, described user m forwards a content, identical with the treatment scheme that user issues a content.Such as after user m forwards a content, can be that this forwarding curriculum offering one issues classification, this issue classification can be identical with the issue classification being forwarded content, also can be different.When category searches the content of social networks, comprise the content of searching and being forwarded.
Fig. 6 is the method for expressing of knowledge mapping.In described server, store a knowledge mapping, therefore, the described issue classification of step S2 described in method described in Fig. 5 is selected from described knowledge mapping by described user m.In an application example, knowledge mapping is the tree structure of Fig. 6, and each upperseat concept comprises at least one subordinate concept.Such as upperseat concept is science, and subordinate concept comprises natural science, social science and cognitive science.And when natural science is upperseat concept, subordinate concept comprises physics, chemistry, biology, uranology, geology, meteorology, architecture, medical science, agronomy, psychology, electronics, system science, mathematics and computer science etc.The effect of knowledge mapping is content type standardization and structuring that user is selected.If user-defined content type is not in knowledge mapping, then new classification is added in knowledge mapping.
Fig. 7 is the method to set up of the advertisement subset under particular category.If the content type that each user on social networks is arranged is regarded as a keyword, the business model that so advertiser buys keyword in a search engine also can be applied to social networks.In this patent method, advertiser carrys out releasing advertisements by buying content type, and is that the advertisement of issuing under this classification arranges a bid.The concrete production method of advertisement subset is described for advertisement a (a ∈ A) below.Suppose that advertiser be the content type that its advertisement a buys is " smart mobile phone ", and be provided with bid under " smart mobile phone " for each click 0.5 yuan for advertisement a.If the class number of " smart mobile phone " is 2340, then advertisement a is just assigned to advertisement subset A2340in.Each advertisement in set of advertisements A is assigned at least one advertisement subset.
Fig. 8 is the estimation method of advertisement.First set of advertisements A={1 is set, 2 ..., Q}, and advertisement subset A is set under each classification k ∈ Ck, whereinand be each described advertisement subset Akin each advertisement a bid is set.Therefore, before returning described step S2, described in Fig. 5, method is further comprising the steps of:
Step S61. is that described user n generates one group of classification;
Step S62. searches the advertisement subset corresponding to each classification in described one group of classification;
Step S63. carries out valuation to each advertisement under each described advertisement subset found;
At least one maximum for described valuation advertisement pushing is given described user n by step S64..
In an application example of method described in Fig. 8, described one group of classification comprises each classification of the reception classification collection of described user n.The definition of described reception classification collection is see Fig. 4.
In an application example of method described in Fig. 8, described one group of classification comprises each classification of the issue classification collection of described user n.The definition of described issue classification collection is see Fig. 4.
In an application example of method described in Fig. 8, described one group of classification comprises the issue classification of each content that described user n forwards, comments on or collects in microblog.
In an application example of method described in Fig. 8, the valuation of each advertisement is determined by the clicking rate of the bid of advertisement and advertisement.Such as when user 958 asks to upgrade content of microblog, server is except according to Content Selection set R958outside user 958 rendering content, also throw in advertisement to user 958.The reception classification collection supposing user 958 for 32,95,692,768,8765}, therefore respectively to advertisement subset A32, A95, A692, A768and A8765in each advertisement carry out valuation.It is the bid that this advertisement is arranged that the clicking rate that the valuation of advertisement equals advertisement is multiplied by advertiser.The clicking rate of advertisement obtains according to the statistics to user's click logs, or obtains according to ad click prediction model.The bid of advertisement is arranged by advertiser.Then according to advertisement valuation, to advertisement subset A32, A95, A692, A768and A8765in each advertisement sort, and by least one maximum for valuation advertisement pushing give described user 958.Equally, one group of advertisement subset can also be determined according to the issue classification collection of user 958, and bid ranking and input are carried out to each advertisement wherein.
In an application example of method described in Fig. 8, the valuation of each advertisement is determined by the occurrence number of the bid of advertisement, the clicking rate of advertisement and advertisement place classification.The clicking rate of advertisement obtains according to the statistics to user's click logs, and the bid of advertisement is arranged by advertiser.The occurrence number of classification obtains according to the Content Selection set statistics of user.Such as in figure 3, the reception classification collection of user 958 is that { 32,95,692,768,8765}, the occurrence number of classification is that classification 32 occurs 3 times, and classification 95,692,768 and 8765 occurs 1 time respectively.Therefore according to above-mentioned data respectively to advertisement subset A32, A95, A692, A768and A8765in each advertisement carry out valuation.The modified value f (x) of this advertisement is multiplied by the bid that the clicking rate that the valuation of advertisement equals this advertisement is multiplied by this advertisement again.Wherein x is the occurrence number of this advertisement place classification, and described f (x) is the increasing function of x, and f (x)>=1, such asp>=0.Then according to the size of advertisement valuation to advertisement subset A32, A95, A692, A768and A8765in each advertisement sort, and by least one maximum for valuation advertisement pushing give described user 958.
Fig. 9 is the method at microblogging promulgating advertisement.D10 is the reception content of microblog page of user n, and wherein D11 and D12 is respectively advertising area, and each advertising area comprises multiple advertisement position, and each advertisement position throws in advertisement according to advertisement valuation size.D13 is content of microblog viewing area, such as, according to the Content Selection set R of user nnregion D13 throws in content of microblog to user n.
The above application example is only preferably application example of the present invention, and is not used to limit protection scope of the present invention.