CN104750789B - The recommendation method and device of label - Google Patents
The recommendation method and device of label Download PDFInfo
- Publication number
- CN104750789B CN104750789B CN201510107973.XA CN201510107973A CN104750789B CN 104750789 B CN104750789 B CN 104750789B CN 201510107973 A CN201510107973 A CN 201510107973A CN 104750789 B CN104750789 B CN 104750789B
- Authority
- CN
- China
- Prior art keywords
- label
- candidate
- user
- data
- application
- 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
Abstract
The present invention provides a kind of recommendation method and device of label.The embodiment of the present invention passes through the description data according to candidate label, and user's generated behavioral data and the user generated behavioral data in the other application other than the application in identified application, the description data of the user obtained, obtain at least one candidate label, using as recommend label, make it possible to show obtained recommendation label to user, the recommendation label obtained by being then based on user's generated behavioral data in the whole network, so that these recommend label to be probably the label interested to user, to guide user according to these recommendation labels, execute relevant operation, such as, label customization etc., in this way, the operating efficiency of label can be effectively improved.
Description
【Technical field】
The present invention relates to the recommendation method and devices of the recommended technology of label more particularly to a kind of label.
【Background technology】
Social Label (Social tagging) is referred to as label, is a kind of more flexible, interesting mode classification, it permits
Family allowable freely marks the resources such as various resources, such as webpage, scientific paper and multimedia.Social Label can help user
Taxonomic revision and inquiry various information, be widely used in Social Label website (for example, Flickr, Picassa, YouTube,
Plaxo etc.), blog (for example, Blogger, WordPress, LiveJournal etc.), encyclopaedia (for example, Wikipedia,
PBWiki etc.), the systems such as microblogging (for example, Twitter, Jaiku etc.).
Recommend interested label for user, becomes a current research hotspot.
【Invention content】
The many aspects of the present invention provide a kind of recommendation method and device of label, to recommend interested mark for user
Label.
An aspect of of the present present invention provides a kind of recommendation method of label, including：
Determine the currently used application of user；
According to the description data of the description data of candidate label and the user, at least one candidate label is obtained, to make
To recommend label；The description data of the user are to be obtained according to the historical behavior data of the user, the history of the user
Behavioral data include the user in the application caused by behavioral data and the user other than the application
Other application in generated behavioral data；
Show the recommendation label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The description data of label include the behavior of the semantic description data of candidate label, the distribution description data and candidate label of candidate label
At least one in data is described.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The semantic description data of label include at least one in following data：
The bookmark name of candidate label；
Keyword under candidate label；
With the bookmark name of the relevant extension tag of semanteme of candidate label；And
With the relevant expanded keyword of semanteme of the keyword under candidate label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, it is described according to time
It selects the description data of label and the description data of the user, obtains at least one candidate label, using as before recommending label,
Further include：
According to a candidate label, text collection is obtained；
According to the text collection, the keyword under candidate's label is obtained.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The distribution description data of label include at least one in following data：
The category distribution of candidate label；And
The theme distribution of candidate label, the theme distribution include the distributed intelligence of M specific subject, and M is to be more than or wait
In 1 integer.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The behavior description data of label include at least one in following data：
With the relevant correlation tag of behavior of candidate label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The description data of label further include the application characteristic of candidate label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
Label using characteristic include in following data at least one of：
The temperature of candidate label in the application；And
The Annual distribution of candidate label in the application.
Another aspect of the present invention provides a kind of recommendation apparatus of label, including：
Determination unit, for determining the currently used application of user；
Obtaining unit, for according to the description data of candidate label and the description data of the user, obtaining at least one
Candidate label, using as recommend label；The description data of the user are to be obtained according to the historical behavior data of the user, institute
The historical behavior data for stating user include the user in the application caused by behavioral data and the user in addition to
Generated behavioral data in other application except the application；
Show unit, for showing the recommendation label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The description data of label include the behavior of the semantic description data of candidate label, the distribution description data and candidate label of candidate label
At least one in data is described.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The semantic description data of label include at least one in following data：
The bookmark name of candidate label；
Keyword under candidate label；
With the bookmark name of the relevant extension tag of semanteme of candidate label；And
With the relevant expanded keyword of semanteme of the keyword under candidate label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, described device is also
Including excavating unit, it is used for
According to a candidate label, text collection is obtained；And
According to the text collection, the keyword under candidate's label is obtained.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The distribution description data of label include at least one in following data：
The category distribution of candidate label；And
The theme distribution of candidate label, the theme distribution include the distributed intelligence of M specific subject, and M is to be more than or wait
In 1 integer.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The behavior description data of label include at least one in following data：
With the relevant correlation tag of behavior of candidate label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
The description data of label further include the application characteristic of candidate label.
The aspect and any possible implementation manners as described above, it is further provided a kind of realization method, the candidate mark
Label using characteristic include in following data at least one of：
The temperature of candidate label in the application；And
The Annual distribution of candidate label in the application.
As shown from the above technical solution, the embodiment of the present invention according to the description data of candidate label and user by existing
Generated behavioral data and the user are produced in the other application other than the application in identified application
Behavioral data, the description data of the user of acquisition obtain at least one candidate label, using as recommending label so that
Obtained recommendation label can be showed to user, obtained by being then based on user's generated behavioral data in the whole network
Recommend label so that these recommend label to be probably the label interested to user, to guide user to be pushed away according to these
Label is recommended, relevant operation is executed, for example, label customization etc., in such manner, it is possible to effectively improve the operating efficiency of label.
In addition, using technical solution provided by the invention, since the description data of candidate label include the language of candidate label
At least one of in the behavior description data of justice description data, the distribution description data of candidate label and candidate label so that energy
It is enough that candidate label is described from multiple dimensions, therefore, it is possible to effectively improve the reliability of label recommendations.
In addition, using technical solution provided by the invention, by the way that the application characteristic of candidate label is increased to candidate
In the description data of label, enabling candidate label is described from using dimension, without being directed to each application distribution pair
Candidate label is described, therefore, it is possible to effectively improve the efficiency of label recommendations.
In addition, using technical solution provided by the invention, due to considering user other than identified application
Generated behavioral data in other application, enabling the dimension of the other application except the application carries out the user
Description, the label recommendations in the case of cold start-up therefore, it is possible to realize the application, simultaneously, additionally it is possible to effectively improve label recommendations
Coverage.
【Description of the drawings】
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to embodiment or description of the prior art
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description be the present invention some realities
Example is applied, it for those of ordinary skill in the art, without having to pay creative labor, can also be attached according to these
Figure obtains other attached drawings.
Fig. 1 is the flow diagram of the recommendation method for the label that one embodiment of the invention provides；
Fig. 2 is the structural schematic diagram of the recommendation apparatus for the label that another embodiment of the present invention provides；
Fig. 3 is the structural schematic diagram of the recommendation apparatus for the label that another embodiment of the present invention provides.
【Specific implementation mode】
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
The whole other embodiments obtained without creative efforts, shall fall within the protection scope of the present invention.
It should be noted that terminal involved in the embodiment of the present invention can include but is not limited to mobile phone, individual digital
Assistant (Personal Digital Assistant, PDA), radio hand-held equipment, tablet computer (Tablet Computer),
PC (Personal Computer, PC), MP3 player, MP4 players, wearable device (for example, intelligent glasses,
Smartwatch, Intelligent bracelet etc.) etc..
In addition, the terms "and/or", only a kind of incidence relation of description affiliated partner, indicates may exist
Three kinds of relationships, for example, A and/or B, can indicate：Individualism A exists simultaneously A and B, these three situations of individualism B.Separately
Outside, character "/" herein, it is a kind of relationship of "or" to typically represent forward-backward correlation object.
Fig. 1 is the flow diagram of the recommendation method for the label that one embodiment of the invention provides, as shown in Figure 1.
101, the currently used application of user is determined.
102, according to the description data of the description data of candidate label and the user, at least one candidate label is obtained,
Using as recommend label；The description data of the user are to be obtained according to the historical behavior data of the user, the user's
Historical behavior data include the user in the application caused by behavioral data and the user in addition to the application
Except other application in generated behavioral data.
The currently used application of user is different with the other application other than the currently used application of user
Application, installed in different types of terminal with the Taobao's application etc. installed on mobile phone for example, the Taobao installed on PC applies
Different application, alternatively, for another example the Baidu map installed on mobile phone apply with installed on mobile phone Baidu search application etc. phases
The different application, etc. installed in same type terminal, the present embodiment is to this without being particularly limited to.
It is understood that the other application other than the currently used application of user, can also be approximately considered is
Whole applications in entire internet including currently used application.
103, show the recommendation label.
It should be noted that some or all of 101~103 executive agent can be the application for being located locally terminal,
Or can also be the plug-in unit being located locally in the application of terminal or Software Development Kit (Software Development
Kit, SDK) etc. functional units, can also be either processing engine in the server of network side or can also be position
In the distributed system of network side, the present embodiment is to this without being particularly limited to, and the present embodiment is to this without being particularly limited to.
It is understood that the application can be mounted in the local program (nativeApp) in terminal, or may be used also
To be a web page program (webApp) of browser in terminal, the present embodiment is to this without being particularly limited to.
In this way, passing through description data and the user generated behavior in identified application according to candidate label
Data and the user generated behavioral data in the other application other than the application, the user's of acquisition
Data are described, at least one candidate label is obtained, using as recommending label, enabling show obtained recommendation mark to user
Label, the recommendation label obtained by being then based on user's generated behavioral data in the whole network so that these recommend label very
It may be exactly the label interested to user, to guide user according to these recommendation labels, relevant operation be executed, for example, mark
Signing system etc., in such manner, it is possible to effectively improve the operating efficiency of label.
Optionally, in a possible realization method of the present embodiment, in 102, the used candidate label
The distributions of the description data semantic description data, candidate label that can include but is not limited to candidate label data and candidate are described
At least one of in the behavior description data of label, the present embodiment is to this without being particularly limited to.
The semantic description data of so-called candidate's label, from itself semantic angle of candidate label, in particulate
The semanteme that the candidate label of description of degree is included.In addition, also further contemplate itself semantic information content of candidate label compared with
It is few, it therefore, can be by the provided content-data of the currently used application of user, utilizing unsupervised machine learning mould
Type, for example, Word2Vec, RandomWalk etc., semantic related expanding is carried out to itself semantic description of candidate label.
During a concrete implementation, it is described candidate label semantic description data can include but is not limited to it is following
At least one of in data：
The bookmark name of candidate label；
Keyword under candidate label；
With the bookmark name of the relevant extension tag of semanteme of candidate label；And
With the relevant expanded keyword of semanteme of the keyword under candidate label.
Keyword under candidate's label, refers in word or phrase of the subject content for expressing candidate label etc.
Hold.Specifically, before 102, text collection can also be obtained, for example, question and answer class application further according to a candidate label
In question and answer content-data etc. corresponding to candidate's label, in turn, then can obtain candidate's label according to the text collection
Under keyword.
The distribution of so-called candidate's label describes data, from the system angle of candidate label, in coarseness
The concept and classification system of the candidate label of description.
During a concrete implementation, it is described candidate label distribution describe data can include but is not limited to it is following
At least one of in data：
The category distribution of candidate label；And
The theme distribution of candidate label, the theme distribution include the distributed intelligence of M specific subject, and M is to be more than or wait
In 1 integer.
Specifically, in one example, the theme distribution of the category distribution of the candidate label and the candidate label,
It can set manually.
Specifically, in another example, machine excavation method can be utilized, point for specifying classification about several is generated
Class device.For example, can specifically utilize history tab and the associated content-data of the history tab, training is specified about several
The history tab grader of classification.In turn, using the history tab grader, the category distribution of candidate label is generated.
Specifically, in another example, measure of supervision can be utilized, the grader about M specific subject is generated.Example
Such as, it can specifically determine the definition of the quantity and each specific subject of specific subject, collect training data, using training data,
History tab grader of the training about M specific subject, to obtain the grader about M specific subject.In turn, this is utilized
About the grader of M specific subject, the theme distribution of candidate label is generated.This method, precision is high, but recall rate is low.
Specifically, in another example, non-supervisory method can be utilized, the theme mould about M specific subject is established
Type.For example, specifically the topic in history tab and the associated content-data of the history tab can be combined into a segment,
Word segmentation processing is carried out to the segment, to generate training data.Utilize training data, training topic model (Topic Model).So
Afterwards, operation is optimized to topic model, for example, semantic-based delete operation, semantic-based deduplication operation etc., to obtain
Topic model about M specific subject.In turn, the topic model using this about M specific subject generates candidate label
Theme distribution.This method, precision is slightly lower, but recall rate is high.
It is understood that in another example, it specifically can also be special about M to being generated using measure of supervision
Determine the grader of theme, and the topic model about M specific subject is established using non-supervisory method, carries out at integration
Reason, to obtain model of the relatively reliable generation about M specific subject.
Specifically, the grader about M specific subject generated using measure of supervision may be used, generate candidate mark
The theme distribution of label.For example, the history tab in the historical behavior data of the whole users of acquisition, using as candidate label, in turn
Using the grader, the theme distribution of candidate label is generated.
Specifically, it may be used and establish topic model about M specific subject using non-supervisory method, generate candidate
The theme distribution of label.For example, the history tab in the historical behavior data of the whole users of acquisition, using as candidate label, with
And the topic in the associated content-data of candidate's label, by the topic in candidate label and the associated content-data of candidate's label
Mesh is combined into a segment, and word segmentation processing is carried out to the segment, to generate word segmentation result, and then utilizes the topic model, generates
The theme distribution of candidate label.
It is understood that the theme distribution progress for the candidate label that can also be specifically generated to above two method is whole
It closes, to obtain the theme distribution of relatively reliable candidate label.
The behavior description data of so-called candidate's label, from user to label recommendations resulting subsequent feedback behavior angle
Degree sets out, to describe the dynamic behaviour feature of candidate label.
During a concrete implementation, it is described candidate label behavior description data can include but is not limited to it is following
At least one of in data：
With the relevant correlation tag of behavior of candidate label.
Specifically, in one example, the behavior of candidate label can be the customization behavior of candidate label, or can be with
For the behavior of applying of candidate label, such as the question and answer behavior in the application of question and answer class, the present embodiment is to this without being particularly limited to.
For example, can specifically utilize the history question and answer class behavior data corresponding to some candidate label, candidate mark is calculated
The co-occurrence probabilities of label and other labels, will meet other labels corresponding to the co-occurrence probabilities of pre-set Correlation Criteria, really
It is set to and the relevant correlation tag of the behavior of candidate's label.
Optionally, in a possible realization method of the present embodiment, in 102, the used candidate label
Description data can further include but be not limited to the application characteristic of candidate label, the present embodiment is to this without spy
It does not limit.
The application characteristic of so-called candidate's label, and from user to the resulting subsequent feedback row of label recommendations
It sets out for angle, to describe the dynamic behaviour feature of candidate label.
During a concrete implementation, it is described candidate label application characteristic can include but is not limited to it is following
At least one of in data：
The temperature of candidate label in the application；And
The Annual distribution of candidate label in the application.
Optionally, in a possible realization method of the present embodiment, in 102, the used user's retouches
It states data and can include but is not limited in the currently used application of user as described in the user description data and in addition to user works as
It is preceding used in application except other application in as described in the user description data, the present embodiment is to this without especially limiting
It is fixed.
Wherein, in the currently used application of user as described in the user description data, specifically can be according to the user
Generated behavioral data obtains in the application.
During a concrete implementation, in the currently used application of user as described in the user description data, tool
Body can include but is not limited in the behavior description data of the semantic description data of user, the distribution description data of user and user
At least one of, the present embodiment is to this without being particularly limited to.
The semantic description data of so-called user, it is emerging in currently used application to can include but is not limited to user
At least one of in interest description and historical conventions tag set.
Specifically, in one example, log-on message that specifically can be according to user in currently used application, is obtained
Interesting measure of the user in currently used application is obtained, for example, interest set of words.
Specifically, in another example, historical behavior that specifically can be according to user in currently used application
Data obtain interesting measure of the user in currently used application, for example, interest set of words.
The distribution of so-called user describes data, and it is emerging in currently used application to can include but is not limited to user
At least one of in the category distribution of interest and the theme distribution of interest.
Specifically, in one example, machine excavation method can be utilized, the classification for specifying classification about several is generated
Device.For example, can be specifically associated with the user interest using user interest (user interests of the whole users acquired)
Content-data, training specifies the user interest grader of classification about several.In turn, using the user interest grader,
Generate the category distribution of interest of the user in currently used application.
Specifically, in another example, measure of supervision can be utilized, the grader about M specific subject is generated.Example
Such as, it can specifically determine the definition of the quantity and each specific subject of specific subject, collect training data, using training data,
User interest label grader of the training about M specific subject, to obtain the grader about M specific subject.In turn, it utilizes
The grader about M specific subject generates the theme distribution of interest of the user in currently used application.This side
Method, precision is high, but recall rate is low.
Specifically, in another example, non-supervisory method can be utilized, the theme mould about M specific subject is established
Type.For example, specifically the topic in user interest and the associated content-data of the user interest can be combined into a segment,
Word segmentation processing is carried out to the segment, to generate training data.Utilize training data, training topic model (Topic Model).So
Afterwards, operation is optimized to topic model, for example, semantic-based delete operation, semantic-based deduplication operation etc., to obtain
Topic model about M specific subject.In turn, the topic model using this about M specific subject generates user current
The theme distribution of interest in used application.This method, precision is slightly lower, but recall rate is high.
It is understood that in another example, it specifically can also be special about M to being generated using measure of supervision
Determine the grader of theme, and the topic model about M specific subject is established using non-supervisory method, carries out at integration
Reason, to obtain model of the relatively reliable generation about M specific subject.
Specifically, the grader about M specific subject generated using measure of supervision may be used, generate user and exist
The theme distribution of interest in currently used application.For example, history row of the acquisition user in currently used application
For the interest in data, using the interest as user in currently used application, and then the grader is utilized, generate user
The theme distribution of interest in currently used application.
Specifically, it may be used and establish topic model about M specific subject using non-supervisory method, generate user
The theme distribution of interest in currently used application.For example, the interest in the historical behavior data of the whole users of acquisition,
Topic using in the content-data of interest and the interest relationship as user in currently used application, by user
Interest in currently used application and the topic in the content-data of the interest relationship are combined into a segment, to the piece
Duan Jinhang word segmentation processings to generate word segmentation result, and then utilize the topic model, generate user in currently used application
Interest theme distribution.
It is understood that specifically can also be to user that above two method is generated in currently used application
The theme distribution of interest integrated, to obtain the theme of interest of the relatively reliable user in currently used application
Distribution.
The behavior description data of so-called user, the behavior that can include but is not limited to association user related to user are retouched
State at least one in data.
Specifically, in one example, the behavior description data of association user, to describe the behavior of association user, tool
Body can be the customization behavior of association user, or can also apply behavior for association user, asking in being applied such as question and answer class
Behavior etc. is answered, the present embodiment is to this without being particularly limited to.
During another concrete implementation, in the currently used application of user as described in the user description data,
The specific application characteristic that can further include user, the present embodiment is to this without being particularly limited to.
The application characteristic of so-called user can include but is not limited at least one in following data：
The Annual distribution of interest of the user in currently used application in this application.
Wherein, in the other application other than the currently used application of user as described in the user description data, tool
Body can generated behavioral data obtains in the other application other than the application according to the user.
During a concrete implementation, about this in the other application other than the currently used application of user
The description data of user can specifically include but be not limited in the semantic description data of user and the distribution description data of user
At least one of, the present embodiment is to this without being particularly limited to.
The semantic description data of so-called user can include but is not limited to user and be answered in addition to user is currently used
The interesting measure in other application except and at least one in historical conventions tag set.
Specifically, in one example, it can specifically obtain user according to log-on message of the user in other application and exist
Interesting measure in other application, for example, interest set of words.
Specifically, in another example, it can specifically be obtained according to historical behavior data of the user in other application
Interesting measure of the user in other application, for example, interest set of words.
Optionally, in a possible realization method of the present embodiment, in 102, candidate label can specifically be calculated
The relevance parameter of description data of description data and the user can will then meet pre-set recommendation item in turn
Candidate label corresponding to the relevance parameter of part, as recommendation label.
Wherein, so-called relevance parameter can include but is not limited at least one in following data：
The relevance score of the semantic description data of candidate label and the semantic description data of user；
The relevance score of the distribution description data of candidate label and the distribution description data of user；
The relevance score of the behavior description data of candidate label and the behavior description data of user；And
The relevance score using characteristic using characteristic and user of candidate label.
During a concrete implementation, the mode of crossing dependency specifically may be used, respectively from user current
Historical conventions tag set in currently used application of interesting measure, user in used application, user except
Interesting measure in other application and user except user currently used application in addition to user it is currently used
Application except other application in the semantic description data of users such as historical conventions tag set sub- dimension, calculate its with
The bookmark name of candidate label, the keyword under candidate label, the tag name with the relevant extension tag of semanteme of candidate label
Claim and the son with the semantic description data of the candidate label such as the relevant expanded keyword of semanteme of the keyword under candidate label
The degree of cross-correlation of dimension using the weight of every sub- dimension, in a manner of Weighted Fusion, obtains the language of candidate label in turn
The relevance score of justice description data and the semantic description data of user.
During another concrete implementation, the mode of crossing dependency specifically may be used, working as respectively from user
The theme distribution of interest of the category distribution and user of interest in application used in preceding in currently used application
The sub- dimension of the distribution description data of equal users calculates itself and the category distribution of candidate label and the theme point of candidate label
The degree of cross-correlation of the sub- dimension of the distribution description data of candidate's label such as cloth, in turn, using the weight of every sub- dimension, to add
The mode of fusion is weighed, the relevance score of the distribution description data of candidate label and the distribution description data of user is obtained.
It, specifically can be according to the behavior description data of candidate label and the row of user during another concrete implementation
To describe data, using matrix decomposition (Matrix Factorization, MF) algorithm, for example, collaborative filtering
(Collaborative Filtering, CF) algorithm, singular value decomposition (Singular Value Decomposition, SVD)
Deng the relevance score of the behavior description data of the candidate label of acquisition and the behavior description data of user.
Another concrete implementation during, specifically can according to candidate label answering using characteristic and user
With characteristic, the composition of the two is inquired into state, is inquired in heuristics rule base, by query result, as candidate
The relevance score using characteristic using characteristic and user of label.
After obtaining these relevance scores, specifically each relevance score can all be marked respectively as candidate
The relevance parameter of the description data of label and the description data of the user in turn then can be according to the relevance parameter, respectively
It obtains and recommends label, or candidate label can also be obtained according to the weight of each relevance score, in a manner of Weighted Fusion
The relevance parameter of description data of description data and the user can then be pushed away in turn according to the relevance parameter
Label is recommended, the present embodiment is to this without being particularly limited to.
During another concrete implementation, relevance parameter can specifically be arranged according to sequence from big to small
Sequence, then, selection come the candidate label corresponding to preceding P relevance parameters, as recommendation label.
During another concrete implementation, a relevance threshold can be specifically pre-set, then, selection is more than
Or the candidate label corresponding to the relevance parameter equal to the relevance threshold, as recommendation label.
The technical solution that the present embodiment is provided, can be abstracted as metadata layer, characteristic layer and model layer these three
Details.Wherein, metadata layer includes mainly the currently used application of user, and in addition to user is currently used
The user behavior data in other application except, for example, the problem of user in answer platform scene issues, answer are used
The data etc. that answer of the family on platform generates；Characteristic layer is user's row in the currently used application accumulation of user
In data basis, to complete the description to user in the description and application of candidate label to be recommended, and work as in addition to user
On the basis of the user behavior data of other application accumulation except application used in preceding, user in other application is retouched in completion
It states, the two describe constituted intermediate data；Model layer can be divided into several recommendation subsystems, each to recommend subsystem
According to respective Generalization bounds, relevance parameter is generated, finally, then recommends correlation ginseng caused by subsystem according to each
Number completes the determination for recommending label.
In the present embodiment, produced by according to the description data and user of candidate label in identified application
Behavioral data and the user generated behavioral data in the other application other than the application, acquisition it is described
The description data of user obtain at least one candidate label, using as recommending label, enabling show to user and obtained
Recommend label, the recommendation label obtained by being then based on user's generated behavioral data in the whole network so that these recommendations
Label is probably the label interested to user, to guide user according to these recommendation labels, executes relevant operation, example
Such as, label customization etc., in such manner, it is possible to effectively improve the operating efficiency of label.
In addition, using technical solution provided by the invention, since the description data of candidate label include the language of candidate label
At least one of in the behavior description data of justice description data, the distribution description data of candidate label and candidate label so that energy
It is enough that candidate label is described from multiple dimensions, therefore, it is possible to effectively improve the reliability of label recommendations.
In addition, using technical solution provided by the invention, by the way that the application characteristic of candidate label is increased to candidate
In the description data of label, enabling candidate label is described from using dimension, without being directed to each application distribution pair
Candidate label is described, therefore, it is possible to effectively improve the efficiency of label recommendations.
In addition, using technical solution provided by the invention, due to considering user other than identified application
Generated behavioral data in other application, enabling the dimension of the other application except the application carries out the user
Description, the label recommendations in the case of cold start-up therefore, it is possible to realize the application, simultaneously, additionally it is possible to effectively improve label recommendations
Coverage.
It should be noted that for each method embodiment above-mentioned, for simple description, therefore it is all expressed as a series of
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the described action sequence because
According to the present invention, certain steps can be performed in other orders or simultaneously.Secondly, those skilled in the art should also know
It knows, embodiment described in this description belongs to preferred embodiment, and involved action and module are not necessarily of the invention
It is necessary.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiment.
Fig. 2 is the structural schematic diagram of the recommendation apparatus for the label that another embodiment of the present invention provides, as shown in Figure 2.This reality
The recommendation apparatus for applying the label of example can include determining that unit 21, obtaining unit 22 and show unit 23.Wherein it is determined that unit
21, for determining the currently used application of user；Obtaining unit 22 is used for the description data according to candidate label and the use
The description data at family obtain at least one candidate label, using as recommending label；The description data of the user are according to
The historical behavior data of user obtain, and the historical behavior data of the user include that the user is generated in the application
Behavioral data and the user generated behavioral data in the other application other than the application；Show unit 23,
For showing the recommendation label.
It should be noted that some or all of recommendation apparatus of label that the present embodiment is provided can be to be located locally
The application of terminal, or can also be the plug-in unit being located locally in the application of terminal or Software Development Kit (Software
Development Kit, SDK) etc. functional units, or can also be processing engine in the server of network side, or
Person can also be positioned at network side distributed system, the present embodiment to this without being particularly limited to, the present embodiment to this not into
Row is particularly limited to.
It is understood that the application can be mounted in the local program (nativeApp) in terminal, or may be used also
To be a web page program (webApp) of browser in terminal, the present embodiment is to this without being particularly limited to.
Optionally, in a possible realization method of the present embodiment, the time used by the obtaining unit 22
The distribution that the description data of label can include but is not limited to the semantic description data, candidate label of candidate label is selected to describe data
With at least one in the behavior description data of candidate label, the present embodiment is to this without being particularly limited to.
During a concrete implementation, it is described candidate label semantic description data can include but is not limited to it is following
At least one of in data：
The bookmark name of candidate label；
Keyword under candidate label；
With the bookmark name of the relevant extension tag of semanteme of candidate label；And
With the relevant expanded keyword of semanteme of the keyword under candidate label.
Specifically, as shown in figure 3, the recommendation apparatus for the label that the present embodiment is provided can further include excavation list
Member 31, for according to a candidate label, obtaining text collection；And according to the text collection, obtain under candidate's label
Keyword.
During a concrete implementation, it is described candidate label distribution describe data can include but is not limited to it is following
At least one of in data：
The category distribution of candidate label；And
The theme distribution of candidate label, the theme distribution include the distributed intelligence of M specific subject, and M is to be more than or wait
In 1 integer.
During a concrete implementation, it is described candidate label behavior description data can include but is not limited to it is following
At least one of in data：
With the relevant correlation tag of behavior of candidate label.
Optionally, in a possible realization method of the present embodiment, the time used by the obtaining unit 22
Select the description data of label can further include but be not limited to the application characteristic of candidate label, the present embodiment to this not
It is particularly limited.
During a concrete implementation, it is described candidate label application characteristic can include but is not limited to it is following
At least one of in data：
The temperature of candidate label in the application；And
The Annual distribution of candidate label in the application.
It should be noted that method in the corresponding embodiments of Fig. 1, it can be by the recommendation apparatus of label provided in this embodiment
It realizes.Detailed description may refer to the correlated resources in the corresponding embodiments of Fig. 1, and details are not described herein again.
It is true in determination unit institute according to the description data and user of candidate label by obtaining unit in the present embodiment
Generated behavioral data and the user generated row in the other application other than the application in fixed application
For data, the description data of the user of acquisition obtain at least one candidate label, using as recommending label so that show
Unit can show obtained recommendation label to user, be obtained by being then based on user's generated behavioral data in the whole network
The recommendation label obtained so that these recommend labels to be probably the label interested to user, to guide user according to this
It is a little to recommend label, relevant operation is executed, for example, label customization etc., in such manner, it is possible to effectively improve the operating efficiency of label.
In addition, using technical solution provided by the invention, since the description data of candidate label include the language of candidate label
At least one of in the behavior description data of justice description data, the distribution description data of candidate label and candidate label so that energy
It is enough that candidate label is described from multiple dimensions, therefore, it is possible to effectively improve the reliability of label recommendations.
In addition, using technical solution provided by the invention, by the way that the application characteristic of candidate label is increased to candidate
In the description data of label, enabling candidate label is described from using dimension, without being directed to each application distribution pair
Candidate label is described, therefore, it is possible to effectively improve the efficiency of label recommendations.
In addition, using technical solution provided by the invention, due to considering user other than identified application
Generated behavioral data in other application, enabling the dimension of the other application except the application carries out the user
Description, the label recommendations in the case of cold start-up therefore, it is possible to realize the application, simultaneously, additionally it is possible to effectively improve label recommendations
Coverage.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided by the present invention, it should be understood that disclosed system, device and method can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of division of logic function, formula that in actual implementation, there may be another division manner, such as multiple units or component
It can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, it is shown or
The mutual coupling, direct-coupling or communication connection discussed can be the indirect coupling by some interfaces, device or unit
It closes or communicates to connect, can be electrical, machinery or other forms.
The unit illustrated as separating component may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, you can be located at a place, or may be distributed over multiple
In network element.Some or all of unit therein can be selected according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, it can also
It is that each unit physically exists alone, it can also be during two or more units be integrated in one unit.Above-mentioned integrated list
The form that hardware had both may be used in member is realized, can also be realized in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit being realized in the form of SFU software functional unit can be stored in one and computer-readable deposit
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are used so that a computer
It is each that equipment (can be personal computer, server or the network equipment etc.) or processor (processor) execute the present invention
The part steps of embodiment the method.And storage medium above-mentioned includes：USB flash disk, mobile hard disk, read-only memory (Read-
Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disc or CD etc. it is various
The medium of program code can be stored.
Finally it should be noted that：The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, it will be understood by those of ordinary skill in the art that：It still may be used
With technical scheme described in the above embodiments is modified or equivalent replacement of some of the technical features；
And these modifications or replacements, various embodiments of the present invention technical solution that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (14)
1. a kind of recommendation method of label, which is characterized in that including：
Determine the currently used application of user；
According to the description data of the description data of candidate label and the user, at least one candidate label is obtained, using as pushing away
Recommend label；The description data of the user are to be obtained according to the historical behavior data of the user, the historical behavior of the user
Data include the user in the application caused by behavioral data and the user in its other than the application
Generated behavioral data during he applies；The description data of candidate's label include the semantic description data of candidate label, wait
Select at least one in the distribution description data of label and the behavior description data of candidate label；
Show the recommendation label.
2. according to the method described in claim 1, it is characterized in that, the semantic description data of candidate's label include following number
At least one of in：
The bookmark name of candidate label；
Keyword under candidate label；
With the bookmark name of the relevant extension tag of semanteme of candidate label；And
With the relevant expanded keyword of semanteme of the keyword under candidate label.
3. according to the method described in claim 2, it is characterized in that, the description data according to candidate label and the user
Description data, at least one candidate label is obtained, as before recommending label, to further include：
According to a candidate label, text collection is obtained；
According to the text collection, the keyword under candidate's label is obtained.
4. according to the method described in claims 1 to 3 any claim, which is characterized in that the distribution of candidate's label is retouched
It includes at least one in following data to state data：
The category distribution of candidate label；And
The theme distribution of candidate label, the theme distribution include the distributed intelligence of M specific subject, and M is more than or equal to 1
Integer.
5. according to the method described in claims 1 to 3 any claim, which is characterized in that the behavior of candidate's label is retouched
It includes at least one in following data to state data：
With the relevant correlation tag of behavior of candidate label.
6. according to the method described in claims 1 to 3 any claim, which is characterized in that the description number of candidate's label
According to the application characteristic for further including candidate label.
7. according to the method described in claim 6, it is characterized in that, the application characteristic of candidate's label includes lower columns
At least one of in：
The temperature of candidate label in the application；And
The Annual distribution of candidate label in the application.
8. a kind of recommendation apparatus of label, which is characterized in that including：
Determination unit, for determining the currently used application of user；
Obtaining unit, for according to the description data of candidate label and the description data of the user, obtaining at least one candidate
Label, using as recommend label；The description data of the user are to be obtained according to the historical behavior data of the user, the use
The historical behavior data at family include the user in the application caused by behavioral data and the user in addition to described
Generated behavioral data in other application except；The description data of candidate's label include the semanteme of candidate label
The distribution of description data, candidate label describes at least one in data and the behavior description data of candidate label；
Show unit, for showing the recommendation label.
9. device according to claim 8, which is characterized in that the semantic description data of candidate's label include following number
At least one of in：
The bookmark name of candidate label；
Keyword under candidate label；
With the bookmark name of the relevant extension tag of semanteme of candidate label；And
With the relevant expanded keyword of semanteme of the keyword under candidate label.
10. device according to claim 9, which is characterized in that described device further includes excavating unit, is used for
According to a candidate label, text collection is obtained；And
According to the text collection, the keyword under candidate's label is obtained.
11. according to the device described in claim 8~10 any claim, which is characterized in that the distribution of candidate's label
Description data include at least one in following data：
The category distribution of candidate label；And
The theme distribution of candidate label, the theme distribution include the distributed intelligence of M specific subject, and M is more than or equal to 1
Integer.
12. according to the device described in claim 8~10 any claim, which is characterized in that the behavior of candidate's label
Description data include at least one in following data：
With the relevant correlation tag of behavior of candidate label.
13. according to the device described in claim 8~10 any claim, which is characterized in that the description of candidate's label
Data further include the application characteristic of candidate label.
14. device according to claim 13, which is characterized in that the application characteristic of candidate's label includes following
At least one of in data：
The temperature of candidate label in the application；And
The Annual distribution of candidate label in the application.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510107973.XA CN104750789B (en) | 2015-03-12 | 2015-03-12 | The recommendation method and device of label |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510107973.XA CN104750789B (en) | 2015-03-12 | 2015-03-12 | The recommendation method and device of label |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104750789A CN104750789A (en) | 2015-07-01 |
| CN104750789B true CN104750789B (en) | 2018-10-16 |
Family
ID=53590473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510107973.XA Active CN104750789B (en) | 2015-03-12 | 2015-03-12 | The recommendation method and device of label |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104750789B (en) |
Families Citing this family (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105138572B (en) * | 2015-07-27 | 2019-12-10 | 百度在线网络技术（北京）有限公司 | Method and device for acquiring relevance weight of user tag |
| CN105117449B (en) * | 2015-08-14 | 2019-08-16 | 百度在线网络技术（北京）有限公司 | A kind of method and apparatus for generating the label of content item |
| CN105701139A (en) * | 2015-11-26 | 2016-06-22 | 中国传媒大学 | Holographic video material indexing method |
| CN105975472A (en) * | 2015-12-09 | 2016-09-28 | 乐视网信息技术（北京）股份有限公司 | Method and device for recommendation |
| CN105701182A (en) * | 2016-01-07 | 2016-06-22 | 百度在线网络技术（北京）有限公司 | Information pushing method and apparatus |
| CN105808642B (en) * | 2016-02-24 | 2019-12-24 | 北京百度网讯科技有限公司 | Recommendation method and device |
| CN107301188B (en) * | 2016-04-15 | 2020-11-10 | 北京搜狗科技发展有限公司 | Method for acquiring user interest and electronic equipment |
| CN105976161A (en) * | 2016-04-29 | 2016-09-28 | 随身云（北京）信息技术有限公司 | Time axis-based intelligent recommendation calendar and user-based presentation method |
| CN107436875B (en) * | 2016-05-25 | 2020-12-04 | 华为技术有限公司 | Text classification method and device |
| CN107526741B (en) * | 2016-06-21 | 2021-05-18 | 华为技术有限公司 | User label generation method and device |
| CN106850780A (en) * | 2017-01-16 | 2017-06-13 | 北京奇虎科技有限公司 | System-level application information recommends method, device and mobile terminal |
| CN106960033B (en) * | 2017-03-22 | 2021-09-14 | 阿里巴巴（中国）有限公司 | Method and device for labeling information stream |
| CN110555155B (en) * | 2017-08-30 | 2023-04-07 | 腾讯科技（北京）有限公司 | Article information recommendation method, device and storage medium |
| CN109636430A (en) * | 2017-10-09 | 2019-04-16 | 北京京东尚科信息技术有限公司 | Object identifying method and its system |
| CN108170665B (en) * | 2017-11-29 | 2021-06-04 | 有米科技股份有限公司 | Keyword expansion method and device based on comprehensive similarity |
| CN108363550A (en) * | 2017-12-28 | 2018-08-03 | 中兴智能交通股份有限公司 | A kind of method and apparatus of data cached update and storage |
| CN110134783B (en) * | 2018-02-09 | 2023-11-10 | 阿里巴巴集团控股有限公司 | Personalized recommendation method, device, equipment and medium |
| CN108897592A (en) * | 2018-06-22 | 2018-11-27 | 珠海市君天电子科技有限公司 | A kind of software methods of exhibiting and relevant device |
| CN110737824B (en) * | 2018-07-03 | 2022-08-09 | 百度在线网络技术（北京）有限公司 | Content query method and device |
| CN109801101A (en) * | 2019-01-03 | 2019-05-24 | 深圳壹账通智能科技有限公司 | Label determines method, apparatus, computer equipment and storage medium |
| CN111046224B (en) * | 2019-12-02 | 2023-04-07 | 上海麦克风文化传媒有限公司 | Real-time recall method for audio products |
| CN111899047A (en) * | 2020-07-14 | 2020-11-06 | 拉扎斯网络科技（上海）有限公司 | Resource recommendation method and device, computer equipment and computer-readable storage medium |
| CN112182098B (en) * | 2020-09-15 | 2021-08-31 | 国网江苏省电力有限公司信息通信分公司 | Information push method and information push server based on cloud computing and big data |
| CN117235586B (en) * | 2023-11-16 | 2024-02-09 | 青岛小帅智能科技股份有限公司 | Hotel customer portrait construction method, system, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102760163A (en) * | 2012-06-12 | 2012-10-31 | 奇智软件（北京）有限公司 | Personalized recommendation method and device of characteristic information |
| CN104123360A (en) * | 2014-07-18 | 2014-10-29 | 腾讯科技（深圳）有限公司 | Application recommendation data acquisition method, device and system and electronic device |
| CN104216881A (en) * | 2013-05-29 | 2014-12-17 | 腾讯科技（深圳）有限公司 | Method and device for recommending individual labels |
-
2015
- 2015-03-12 CN CN201510107973.XA patent/CN104750789B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102760163A (en) * | 2012-06-12 | 2012-10-31 | 奇智软件（北京）有限公司 | Personalized recommendation method and device of characteristic information |
| CN104216881A (en) * | 2013-05-29 | 2014-12-17 | 腾讯科技（深圳）有限公司 | Method and device for recommending individual labels |
| CN104123360A (en) * | 2014-07-18 | 2014-10-29 | 腾讯科技（深圳）有限公司 | Application recommendation data acquisition method, device and system and electronic device |
Non-Patent Citations (1)
| Title |
|---|
| 社交网络中的用户标签推荐;陆子龙;《中国优秀硕士学位论文全文数据库 信息科技辑》;20140315;I139-110第21-32页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104750789A (en) | 2015-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104750789B (en) | The recommendation method and device of label | |
| Ozsoy | From word embeddings to item recommendation | |
| Abdel-Hafez et al. | A survey of user modelling in social media websites | |
| US9910930B2 (en) | Scalable user intent mining using a multimodal restricted boltzmann machine | |
| US9864803B2 (en) | Method and system for multimodal clue based personalized app function recommendation | |
| CN103164521B (en) | A kind ofly to browse and the keyword calculation method of search behavior and device based on user | |
| CN102054003B (en) | Methods and systems for recommending network information and creating network resource index | |
| US11487946B2 (en) | Content editing using content modeling and semantic relevancy scoring | |
| US11423234B2 (en) | Content generation using target content derived modeling and unsupervised language modeling | |
| CN103136188A (en) | Method and system used for sentiment estimation of web browsing user | |
| Kanwal et al. | A review of text-based recommendation systems | |
| US10176260B2 (en) | Measuring semantic incongruity within text data | |
| CN103955464A (en) | Recommendation method based on situation fusion sensing | |
| CN103678304A (en) | Method and device for pushing specific content for predetermined webpage | |
| CN105518661A (en) | Browsing images via mined hyperlinked text snippets | |
| CN105389329A (en) | Open source software recommendation method based on group comments | |
| CN104077415A (en) | Searching method and device | |
| CN104751354A (en) | Advertisement cluster screening method | |
| CN110069713B (en) | Personalized recommendation method based on user context perception | |
| CN104142990A (en) | Search method and device | |
| CN103823849A (en) | Method and device for acquiring entries | |
| JP2014106661A (en) | User state prediction device, method and program | |
| Zhu et al. | Cyber-physical-social-thinking modeling and computing for geological information service system | |
| CN104199836A (en) | Annotation user model construction method based on child interest division | |
| Sajeev et al. | Effective web personalization system based on time and semantic relatedness |
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 |