CN106354867A - Multimedia resource recommendation method and device - Google Patents
Multimedia resource recommendation method and device Download PDFInfo
- Publication number
- CN106354867A CN106354867A CN201610817968.2A CN201610817968A CN106354867A CN 106354867 A CN106354867 A CN 106354867A CN 201610817968 A CN201610817968 A CN 201610817968A CN 106354867 A CN106354867 A CN 106354867A
- Authority
- CN
- China
- Prior art keywords
- multimedia resource
- recommended
- feature
- heading message
- classification
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/40—Information retrieval; Database structures therefor; File system structures therefor of multimedia data, e.g. slideshows comprising image and additional audio data
- G06F16/43—Querying
- G06F16/435—Filtering based on additional data, e.g. user or group profiles
Abstract
The invention relates to a multimedia resource recommendation method and device. The multimedia resource recommendation method comprises the steps that characteristics of multimedia resources to be recommended are respectively determined according to header information of multimedia resources to be recommended; the categories to which the multimedia resources to be recommended belong to are respectively determined; multimedia resource recommendation results are generated according to the categories to which the multimedia resources to be recommended belong to. By the adoption of the multimedia resource recommendation method, the categories of the multimedia resources in recommended information can be enriched, and the diversity of the multimedia resources is improved.
Description
Technical field
The present invention relates to MultiMedia Field, the recommendation method and device of more particularly, to a kind of multimedia resource.
Background technology
Internet era, in the particularly mobile Internet epoch, how to provide the user timely and valuable information is
The focus of numerous Internet firms research.In recent years, the development with machine learning system is it is recommended that system starts to support personalization
Generalization bounds.
At present, the recommendation to multimedia resource, is devoted to improving the forecasting accuracy of commending system, thus improve pushing away mostly
Recommend the clicking rate of information.In correlation technique, the multiformity of multimedia resource is often considered seldom.Taking video recommendations as a example, depending on
The multiformity of frequency often relies on the information such as the mark of video channel, interest tags and uploader, for example, controls each video frequency
The recommendation number of the video under road.
Using above-mentioned recommendation method, the multiformity of video derives from manually regular (video channel, interest tags and upload
The information such as the mark of person), there is more noise and various Sexual behavior mode does not have adaptivity.
Content of the invention
Technical problem
In view of this, the technical problem to be solved in the present invention is to provide a kind of recommendation method of multimedia resource, enriches and pushes away
Recommend the classification of multimedia resource in information, improve the multiformity of multimedia resource.
Solution
In order to solve above-mentioned technical problem, according to one embodiment of the invention, there is provided a kind of recommendation of multimedia resource
Method, comprising:
According to the heading message of each multimedia resource to be recommended, determine the spy of each described multimedia resource to be recommended respectively
Levy；
According to the feature of each described multimedia resource to be recommended, determine that each described multimedia resource to be recommended is belonged to respectively
Classification；
The classification being belonged to according to each described multimedia resource to be recommended, generates multimedia resource recommendation results.
For said method, in a kind of possible implementation, according to the heading message of each multimedia resource to be recommended,
Determine the feature of each described multimedia resource to be recommended respectively, comprising:
Obtain the heading message of each described multimedia resource to be recommended；
Participle is carried out to each described heading message, obtains the corresponding word of each described heading message；
According to the corresponding word of each described heading message, determine the feature of each described multimedia resource to be recommended respectively.
For said method, in a kind of possible implementation, according to the feature of each described multimedia resource to be recommended,
Determine the classification that each described multimedia resource to be recommended is belonged to respectively, comprising:
Feature according to each described multimedia resource to be recommended and center vector of all categories, calculate respectively and each described wait to push away
Recommend multimedia resource and distance of all categories；
According to each described multimedia resource to be recommended and distance of all categories, determine each described multimedia money to be recommended respectively
The classification that source is belonged to.
For said method, in a kind of possible implementation, according to the corresponding word of each described heading message, respectively really
The feature of fixed each described multimedia resource to be recommended, comprising:
According to the corresponding word of each described heading message, it is respectively adopted the fisrt feature that formula 1 calculates each described heading message,
Wherein, the corresponding word of described heading message is title (v)={ w1, w2 ..., wk }；K represents described heading message
The total number of corresponding word；niRepresent the number of times that i-th word wi occurs in described heading message；njRepresent j-th word wj in institute
State the number of times occurring in heading message；I, j represent the label of the corresponding word of described heading message, the span of i, j be [1,
k]；tfiRepresent the corresponding fisrt feature of i-th word wi；
According to the sum of the corresponding word of each described heading message and multimedia resource, it is respectively adopted formula 2 and calculates each described mark
The second feature of topic information,
Wherein, | d | represents the total number of multimedia resource；|di| represent that heading message includes many matchmakers of i-th word wi
The number of body resource；idfiRepresent the corresponding second feature of i-th word wi；
According to described fisrt feature and described second feature, it is respectively adopted the 3rd spy that formula 3 calculates each described heading message
Levy,
tfidfi=tfi×idfiFormula 3,
Wherein, tfidfiRepresent the corresponding third feature of i-th word wi.
For said method, in a kind of possible implementation, according to the feature of each described multimedia resource to be recommended
With center vector of all categories, calculate each described multimedia resource to be recommended and distance of all categories respectively, comprising:
Feature according to each described multimedia resource to be recommended and center vector of all categories, are respectively adopted formula 4 and calculate respectively
Described multimedia resource to be recommended and distance of all categories,
s(vn, cm)=cosin (f (vn), qm) formula 4,
Wherein, n represents multimedia resource v to be recommendednCorresponding label, n is integer；M represents classification cmCorresponding label, m
For integer；qmRepresent classification cmCenter vector；f(vn) represent multimedia resource v to be recommendednFeature；s(vn, cm) represent and treat
Recommend multimedia resource vnWith classification cmDistance.
For said method, in a kind of possible implementation, according to each described multimedia resource to be recommended with all kinds of
Other distance, determines the classification that each described multimedia resource to be recommended is belonged to respectively, comprising:
According to each described multimedia resource to be recommended and distance of all categories, be respectively adopted formula 5 determine each described to be recommended
The classification that multimedia resource is belonged to,
Wherein, h (vn) represent s (vn, cm) take the value of m during maximum, the class that each described multimedia resource to be recommended is belonged to
It is not the classification closest with each described multimedia resource.
For said method, in a kind of possible implementation, belonged to according to each described multimedia resource to be recommended
Classification, generate multimedia resource recommendation results, comprising:
The classification being belonged to according to each described multimedia resource to be recommended, determine of all categories in included each described wait to push away
Recommend the ranking results of multimedia resource；
According to described ranking results, generate described multimedia resource recommendation results.
In order to solve above-mentioned technical problem, according to another embodiment of the present invention, there is provided the pushing away of a kind of multimedia resource
Recommend device, comprising:
Characteristic determination module, for the heading message according to each multimedia resource to be recommended, determines respectively and each described waits to push away
Recommend the feature of multimedia resource；
Category determination module, is connected with described characteristic determination module, for according to each described multimedia resource to be recommended
Feature, determines the classification that each described multimedia resource to be recommended is belonged to respectively；
Recommendation results generation module, is connected with described category determination module, for according to each described multimedia money to be recommended
The classification that source is belonged to, generates multimedia resource recommendation results.
For said apparatus, in a kind of possible implementation, described characteristic determination module includes:
Heading message acquiring unit, for obtaining the heading message of each described multimedia resource to be recommended；
Heading message participle unit, is connected with described heading message acquiring unit, for carrying out to each described heading message
Participle, obtains the corresponding word of each described heading message；
Characteristics determining unit, is connected with described heading message participle unit, for corresponding according to each described heading message
Word, determines the feature of each described multimedia resource to be recommended respectively.
For said apparatus, in a kind of possible implementation, described category determination module includes:
Metrics calculation unit, for the feature according to each described multimedia resource to be recommended and center vector of all categories,
Calculate each described multimedia resource to be recommended and distance of all categories respectively；
Classification determination unit, is connected with described metrics calculation unit, for according to each described multimedia resource to be recommended with
Distance of all categories, determines the classification that each described multimedia resource to be recommended is belonged to respectively.
For said apparatus, in a kind of possible implementation,
Described characteristics determining unit, for according to the corresponding word of each described heading message, be respectively adopted formula 1 calculate each described
The fisrt feature of heading message,
Wherein, the corresponding word of described heading message is title (v)={ w1, w2 ..., wk }；K represents described heading message
The total number of corresponding word；niRepresent the number of times that i-th word wi occurs in described heading message；njRepresent j-th word wj in institute
State the number of times occurring in heading message；I, j represent the label of the corresponding word of described heading message, the span of i, j be [1,
k]；tfiRepresent the corresponding fisrt feature of i-th word wi；
Described characteristics determining unit, is additionally operable to the sum according to the corresponding word of each described heading message and multimedia resource,
It is respectively adopted the second feature that formula 2 calculates each described heading message,
Wherein, | d | represents the total number of multimedia resource；|di| represent that heading message includes many matchmakers of i-th word wi
The number of body resource；idfiRepresent the corresponding second feature of i-th word wi；
Described characteristics determining unit, is additionally operable to, according to described fisrt feature and described second feature, be respectively adopted formula 3 and calculate
The third feature of each described heading message,
tfidfi=tfi×idfiFormula 3,
Wherein, tfidfiRepresent the corresponding third feature of i-th word wi.
For said apparatus, in a kind of possible implementation,
Described metrics calculation unit, for the feature according to each described multimedia resource to be recommended and center of all categories to
Amount, is respectively adopted formula 4 and calculates each described multimedia resource to be recommended and distance of all categories,
s(vn, cm)=cosin (f (vn), qm) formula 4,
Wherein, n represents multimedia resource v to be recommendednCorresponding label, n is integer；M represents classification cmCorresponding label, m
For integer；qmRepresent classification cmCenter vector；f(vn) represent multimedia resource v to be recommendednFeature；s(vn, cm) represent and treat
Recommend multimedia resource vnWith classification cmDistance.
For said apparatus, in a kind of possible implementation,
Described classification determination unit, for according to each described multimedia resource to be recommended and distance of all categories, adopting respectively
Determine the classification that each described multimedia resource to be recommended is belonged to formula 5,
Wherein, h (vn) represent s (vn, cm) take the value of m during maximum, the class that each described multimedia resource to be recommended is belonged to
It is not the classification closest with each described multimedia resource.
For said apparatus, in a kind of possible implementation, described recommendation results generation module includes:
Ranking results determining unit, for the classification being belonged to according to each described multimedia resource to be recommended, determines all kinds of
The ranking results of included each described multimedia resource to be recommended in not；
Recommendation results signal generating unit, is connected with described ranking results determining unit, for according to described ranking results, generating
Described multimedia resource recommendation results.
Beneficial effect
The recommendation method of the multimedia resource of the embodiment of the present invention, can believe according to the title of each multimedia resource to be recommended
Breath, determines the classification that each multimedia resource to be recommended is belonged to, and the class being belonged to according to each multimedia resource to be recommended respectively
Not, multimedia resource recommendation results are generated.The recommendation method of the multimedia resource of the embodiment of the present invention, can enrich recommendation information
The classification of middle multimedia resource, improves the multiformity of multimedia resource.
According to below with reference to the accompanying drawings, to detailed description of illustrative embodiments, the further feature of the present invention and aspect will become
Clear.
Brief description
Including in the description and accompanying drawing and the description of the part that constitutes description together illustrates the present invention's
Exemplary embodiment, feature and aspect, and for explaining the principle of the present invention.
The flow chart that Fig. 1 illustrates the recommendation method of multimedia resource according to an embodiment of the invention；
Fig. 2 illustrates another flow chart of the recommendation method of multimedia resource according to an embodiment of the invention；
Fig. 3 illustrates another flow chart of the recommendation method of multimedia resource according to an embodiment of the invention；
Fig. 4 illustrates another flow chart of the recommendation method of multimedia resource according to an embodiment of the invention；
Fig. 5 illustrates the structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention；
Fig. 6 illustrates another structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention；
Fig. 7 illustrates another structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention；
Fig. 8 illustrates another structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention；
Fig. 9 illustrates another structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention.
Specific embodiment
Describe various exemplary embodiments, feature and the aspect of the present invention below with reference to accompanying drawing in detail.Identical in accompanying drawing
Reference represent the same or analogous element of function.Although the various aspects of embodiment shown in the drawings, remove
Non-specifically points out it is not necessary to accompanying drawing drawn to scale.
Special word " exemplary " means " as example, embodiment or illustrative " here.Here as " exemplary "
Illustrated any embodiment should not necessarily be construed as preferred or advantageous over other embodiments.
In addition, in order to better illustrate the present invention, giving numerous details in specific embodiment below.
It will be appreciated by those skilled in the art that not having some details, the present invention equally can be implemented.In some instances, for
Method well known to those skilled in the art, means, element and circuit are not described in detail, in order to highlight the purport of the present invention.
Embodiment 1
The flow chart that Fig. 1 illustrates the recommendation method of multimedia resource according to an embodiment of the invention.As shown in figure 1, should
The recommendation method of multimedia resource, may include that
In step s101, according to the heading message of each multimedia resource to be recommended, determine each described to be recommended many respectively
The feature of media resource.
In step s102, according to the feature of each described multimedia resource to be recommended, determine each described to be recommended many respectively
The classification that media resource is belonged to.
In step s103, the classification that belonged to according to each described multimedia resource to be recommended, generate multimedia resource and push away
Recommend result.
The multimedia resource (multimedia) of the embodiment of the present invention, can include such as text, sound, video and image
Etc. various media formats.Wherein, multimedia resource to be recommended can include from multimedia resources database obtain can be used in give birth to
Become the multimedia resource of multimedia resource recommendation results.The embodiment of the present invention does not limit the concrete application of multimedia resource to be recommended
Scene.For example, in video website, in the case of user input search keyword, can obtain and search keyword
Related video to be recommended；In the case that user's request plays target video, can obtain and related to target video wait to push away
Recommend video.
It should be noted that those skilled in the art it should be understood that can adopt obtain in various manners each to be recommended
Multimedia resource, is not construed as limiting to this.For example, recommended for the target video that user is watching, can according to mesh
The degree of association (for example belong to serial, have identical performer or director etc.) of mark video obtains video to be recommended it is also possible to root
Obtain video to be recommended according to the temperature that browses in a period of time (such as 3 days or 1 week), this is not construed as limiting.
Wherein, the heading message of multimedia resource to be recommended could be for indicating the brief sentence of multimedia resource.This
Inventive embodiments do not limit the concrete form of heading message.For example, heading message can be to reflect multimedia resource
The heading message of content, for example, " Chinese Women's Volleyball Team took Olympic champion by force again every 12 years ".Heading message can also be that multimedia resource is normal
Title, for example, " Journey to the West ", " apostle passerby ".The feature of the multimedia resource to be recommended of the embodiment of the present invention can be wrapped
Include the parameter related to the heading message of multimedia resource to be recommended.For example, feature can be tf (term frequency, word
Frequently), idf (inverse document frequency, reverse document-frequency) etc., is not construed as limiting to this.
Further, the feature according to multimedia resource to be recommended is it may be determined that what multimedia resource to be recommended was belonged to
Classification.Wherein, classification can be used for multimedia resource is classified.The embodiment of the present invention does not limit multimedia resource to be recommended
Classification determination mode.For example, it is possible to be carried out to the heading message of multimedia resource by k-mens (k average) clustering algorithm
Cluster, finds the center vector of heading message cluster, so that it is determined that the classification that multimedia resource is belonged to.
The embodiment of the present invention does not limit the classification that multimedia resource to be recommended belonged to and really fixes time.Real as the present invention
Apply an example of example, the various multimedia resources that can include for multimedia resources database, lower each many matchmaker of determination online
Classification that body resource is belonged to simultaneously is stored.In recommendation process, the multimedia resource to be recommended obtain can be directed on line,
The classification that each multimedia resource to be recommended is belonged to is obtained from the corresponding class library of stored multimedia resource.As
Another example of the embodiment of the present invention, in recommendation process, can be for the multimedia resource to be recommended obtaining on line, online
Above or determine, under line, the classification that each multimedia resource to be recommended is belonged to.
In a kind of possible implementation, as shown in Fig. 2 according to the heading message of each multimedia resource to be recommended, point
Do not determine the feature (step 101) of each described multimedia resource to be recommended, may include that
In step s201, obtain the heading message of each described multimedia resource to be recommended.
In step s202, participle is carried out to each described heading message, obtain the corresponding word of each described heading message.
In step s203, according to the corresponding word of each described heading message, determine each described multimedia money to be recommended respectively
The feature in source.
In the embodiment of the present invention, heading message is carried out with participle can be that the Chinese character sequence in heading message is cut into one
Each and every one single word.Participle is typically the basis of text mining, and the passage for input carries out rational participle, can make
Obtain the effect that equipment (such as computer, mobile phone, server etc.) reaches automatic identification sentence implication.The method of participle and accurately
Degree generally can directly influence the relevancy ranking to recommendation results.The embodiment of the present invention does not limit the concrete grammar of participle,
For example, it is possible to include the segmenting method based on string matching, the segmenting method based on the segmenting method understanding or based on statistics
Deng.
An example as the embodiment of the present invention is it is assumed that the multimedia resource each to be recommended obtaining is for example
Multimedia={ v1, v2..., vk}.Wherein, k represents the total number of multimedia resource to be recommended, v1Represent the 1st to be recommended
Multimedia resource, v2Represent the 2nd multimedia resource to be recommended, the like, vkRepresent k-th multimedia resource.Further
Ground, for n-th multimedia resource v to be recommendednHeading message row participle, obtain the corresponding word of heading message be title (vn)
={ w1, w2 ..., wk }.Wherein, k represents the total number of the corresponding word of heading message, and w1 represents the 1st word, and w2 represents the 2nd
Word, the like, wk represents k-th word.
In a kind of possible implementation, according to the corresponding word of each described heading message, determine respectively and each described wait to push away
Recommend the feature (step 203) of multimedia resource, may include that
According to the corresponding word of each described heading message, it is respectively adopted the fisrt feature that formula 1 calculates each described heading message,
Wherein, the corresponding word of described heading message is title (vn)={ w1, w2 ..., wk }；K represents described heading message
The total number of corresponding word, generally greater than or equal to 1 positive integer；niRepresent that i-th word wi goes out in described heading message
Existing number of times；njRepresent the number of times that j-th word wj occurs in described heading message；I, j represent that described heading message is corresponding
The label of word, the span of i, j is [1, k]；tfiRepresent the corresponding fisrt feature of i-th word wi；
According to the sum of the corresponding word of each described heading message and multimedia resource, it is respectively adopted formula 2 and calculates each described mark
The second feature of topic information,
Wherein, | d | represents the total number of multimedia resource, generally greater than or equal to 1 positive integer；|diRepresent title
Information includes the number of the multimedia resource of i-th word wi；idfiRepresent the corresponding second feature of i-th word wi；
According to described fisrt feature and described second feature, it is respectively adopted the 3rd spy that formula 3 calculates each described heading message
Levy,
tfidfi=tfi×idfiFormula 3,
Wherein, tfidfiRepresent the corresponding third feature of i-th word wi.
The embodiment of the present invention can using tf-idf (term frequency inverse document frequency,
Word frequency -- reverse document-frequency) algorithm obtains the feature of multimedia resource to be recommended.Wherein, tf-idf algorithm can be used for assessing
One word is for the significance level of an one of file or corpus field file set.It is understood that word
The number of times that occurs hereof with it of importance be directly proportional increase, but the frequency simultaneously occurring in corpus as well as it
Rate is inversely proportional to decline.The various forms of tf-idf algorithm is generally applied to information retrieval and text mining field, as file
The tolerance of degree of correlation or grading and user's inquiry between.
The feature of the multimedia resource of the embodiment of the present invention can assess the weight for multimedia resources database for the multimedia resource
Want degree.Further, the feature of multimedia resource can include fisrt feature, second feature and third feature.Wherein, first
Feature can be tf feature, and second feature can be idf feature, and third feature can be tf-idf feature.Given in portion
In file, tf feature can be the frequency that some given word occurs in this document.This numeral can be to word number
The normalization of (term count), to prevent it to be partial to long file.Idf feature can be the degree of a word general importance
Amount.The idf feature of a certain specific word, then will be able to obtain by the total number of file divided by the number of the file including this word
Business takes the logarithm and obtains.If because this word is not in corpus, may result in dividend is zero, therefore, as shown in Equation 2, typically
In the case of can by include this word file number add 1.Tf-idf feature can be tf feature and the product of idf feature.
In embodiments of the present invention, for n-th multimedia resource v to be recommendednHeading message carry out participle, marked
The corresponding word of topic information is title (vn)={ w1, w2 ..., wk }.It is possible to further the 1st word is calculated by formula 1
Tf feature tf (w1) of w1, can be calculated idf feature idf (w1) of the 1st word w1, can be calculated by formula 3 by formula 2
Obtain tf-idf feature tfidf (w1) of the 1st word w1.The like, tf feature tf of i-th word wi can be calculated
(wi), idf feature idf (wi) and tf-idf feature tfidf (wi) etc..Wherein, i represents the mark of heading message corresponding word wi
Number, the span of i is [1, k].
Further, by above derivation step, multimedia resource v to be recommended can be calculatednFeature f (vn)=
{ tf (w1), idf (w1), tfidf (w1) ..., tf (wk), idf (wk), tfidf (wk) }.Subsequent arithmetic for convenience, permissible
The identifier of uniform characteristics, for example, represents multimedia resource v to be recommended with xnFeature, obtain f (vn)={ f (v1)=
{ x1, x2, x3 ..., x (3k) }.Wherein, x1, x2, x3 can correspond to tf (w1) respectively, idf (w1), tfidf (w1) it is also possible to
Correspond to tf (w1), tf (w2), tf (w3) respectively, this is not construed as limiting.
As an example of the embodiment of the present invention, to the heading message of video to be recommended, for example " Chinese Women's Volleyball Team was every 12 years
Take Olympic champion again by force " carry out participle, can obtain for example " China, women's volleyball, every 12 years, take by force again, Olympic champion " word segmentation result.
Wherein, the total number of the corresponding word of heading message is 5, and word " Olympic champion " occurs 1 time, then the tf feature of word " Olympic champion "
For 0.20 (1/5).Assume that video library includes 4 videos, for example, " Chinese Women's Volleyball Team took Olympic champion by force again every 12 years ", " Journey to the West ",
" passerby apostle ", " Chen Long victory Li Zongwei takes Olympic champion by force ".Wherein, the total number of the video that video library includes is 4, heading message
The number including word " Olympic champion " is 2, then the idf of word " Olympic champion " is characterized as 0.50 (2/4).Further, word
The tf-idf of " Olympic champion " is characterized as 0.10 (0.20 × 0.50).
It should be noted that those skilled in the art are it should be understood that the title for multimedia resource to be recommended is believed
Breath, using different segmenting methods, may affect the result of calculation of the feature of multimedia resource to be recommended.In practical application mistake
Cheng Zhong, can select to adopt coarseness segmenting method or employing according to various index parameters (such as degree of accuracy, recall result etc.)
Fine granularity segmenting method, is not construed as limiting to this.
In a kind of possible implementation, as shown in figure 3, according to the feature of each described multimedia resource to be recommended, point
Do not determine the classification (step 102) that each described multimedia resource to be recommended is belonged to, may include that
In step s301, the feature according to each described multimedia resource to be recommended and center vector of all categories, respectively
Calculate each described multimedia resource to be recommended and distance of all categories.
In step s302, according to each described multimedia resource to be recommended and distance of all categories, determine each described respectively
The classification that multimedia resource to be recommended is belonged to.
In embodiments of the present invention, the center vector that k-mens clustering algorithm obtains m cluster, q can be first passed through1,
q2..., qm.Wherein, the center vector of each cluster corresponds to a classification.For example, q1Represent classification c1Center vector, q2Represent
Classification c2Center vector, the rest may be inferred, qmRepresent classification cmCenter vector.It should be noted that k-means algorithm is base
In the clustering algorithm of distance, mainly adopt distance as the evaluation index of similarity, that is, think that the distance of two objects is nearer, its
Similarity is bigger.K-means algorithm thinks that cluster can be formed by apart from close object, therefore, compact and independent obtaining
Cluster as final goal.
Further, in k-means algorithm, the selection of the center vector of m initial clustering has relatively to cluster result
Big impact.Specifically, the first step of k-means algorithm can be random any m object of selection as initial clustering
Center, initially represents a cluster.K-means algorithm concentrates each object remaining to data in each iteration, according to it
With the distance at each cluster center, each object is assigned to nearest cluster again.After having investigated all data objects, an iteration
Computing completes, and the middle vector of new cluster is computed.If before and after an iteration, the value of evaluation index does not become
Change, illustrate that algorithm has been restrained.
, details are provided below taking video as a example: the first step, randomly selects the title letter of m video from p video
Breath is as the center vector of initial clustering.Second step, the heading message to each video remaining, measure it and initially gather to each
The distance of the center vector of class, and it is grouped into the classification of nearest center vector.3rd step, recalculates obtained each
The center vector of the new cluster of individual classification.Iteration second step and the 3rd step, until the new center vector of cluster and former cluster
Center vector is equal or is less than specified threshold, terminates iteration.Thus, the center of m cluster is obtained by k-mens clustering algorithm
Vector, q1, q2..., qm.
In a kind of possible implementation, the feature according to each described multimedia resource to be recommended and center of all categories
Vector, is respectively adopted formula 4 and calculates each described multimedia resource to be recommended and distance of all categories,
s(vn, cm)=cosin (f (vn), qm) formula 4,
Wherein, n represents multimedia resource v to be recommendednCorresponding label, n is integer；M represents classification cmCorresponding label, m
For integer；qmRepresent classification cmCenter vector；f(vn) represent multimedia resource v to be recommendednFeature；s(vn, cm) represent and treat
Recommend multimedia resource vnWith classification cmDistance.
In a kind of possible implementation, according to each described multimedia resource to be recommended and distance of all categories, difference
The classification that each described multimedia resource to be recommended is belonged to is determined using formula 5,
Wherein, h (vn) represent s (vn, cm) take the value of m during maximum, the class that each described multimedia resource to be recommended is belonged to
It is not the classification closest with each described multimedia resource.
, it is assumed that user viewing video v, the corresponding video to be recommended of video v can for an example as the embodiment of the present invention
To include l={ v1, v2..., vn, then the corresponding word of the corresponding heading message of video to be recommended can include lt={ title
(v1), title (v2) ..., title (vn), further, the corresponding feature of video to be recommended can include flt={ f (v1),
f(v2) ..., f (vn)}.The corresponding classification of each video to be recommended, cl={ h (v can be calculated according to formula 4 and formula 51), h
(v2) ..., h (vn)}.
It should be noted that s (vn, cm)=cosin (f (vn), qm), wherein, cosine0 ° of value is 1,
Cosine90 ° of value is 0.Thus work as s (vn, cm) take maximum in the case of, video vnWith classification cmSimilarity highest.
In a kind of possible implementation, as shown in figure 4, the class being belonged to according to each described multimedia resource to be recommended
Not, generate multimedia resource recommendation results (step 103), comprising:
In step s401, the classification that belonged to according to each described multimedia resource to be recommended, determine of all categories middle wrapped
The ranking results of each described multimedia resource to be recommended including.
In step s402, according to described ranking results, generate described multimedia resource recommendation results.
An example as the embodiment of the present invention it is assumed that user viewing video v, obtains that video v is corresponding to be recommended to be regarded
Frequency includes l={ v1, v2..., v8, it is calculated the corresponding classification of each video to be recommended, cl={ h (v1-c2), h (v2-c4),
h(v3-c4), h (v4-c2), h (v5-c1), h (v6-c2), h (v7-c1), h (v8-c3)}.Obtain after being sorted out by arrangement, belong to
Classification c1Video to be recommended have v5, v7, belong to classification c2Video to be recommended have v1, v4, v6, belong to classification c3Wait push away
Recommending video has v8, belong to classification c4Video to be recommended have v2, v3.
Further, the embodiment of the present invention can also be to video l={ v to be recommended1, v2..., v8Be ranked up.For example,
Before determining the classification that belonged to of video to be recommended, video to be recommended can be ranked up, obtain ranking results (according to pushing away
Recommend descending) it is lp={ v5, v2, v8, v6, v3, v1, v7, v4}.Accordingly, it is determined that included video each to be recommended in of all categories
Ranking results, can obtain belonging to classification c1Video to be recommended ranking results be v5, v7, belong to classification c2Treat
The ranking results recommending video are v6, v1, v4, belong to classification c3Video to be recommended ranking results be v8, belong to classification
c4Video to be recommended ranking results be v2, v3.
In embodiments of the present invention, can be from video l={ v to be recommended1, v2..., v8In selected part video (such as 4
Individual etc.) generate video recommendations result.The embodiment of the present invention does not limit to choose from video to be recommended and generates video recommendations result
The concrete mode of video.For example, can be using the method for intersection selecting video.Specifically, can distinguish and obtain often successively
The forward video of sequence under individual classification, for generating video recommendations result.For example, from belonging to classification c1Video to be recommended
Middle selecting video v5, then from belonging to classification c2Video to be recommended in selecting video v6, the like, finally give video and push away
Recommending result is v5, v6, v8, v2.What the embodiment of the present invention did not limit video recommendations result represents form, for example, it is possible to by row
The forms such as table, form represent.
It should be noted that the embodiment of the present invention does not limit the concrete time point to video to be recommended sequence.For example, it is possible to
Before determining the classification that video to be recommended is belonged to, video to be recommended is ranked up；Video to be recommended can also determined
After the classification being belonged to, video to be recommended is ranked up.Additionally, the embodiment of the present invention does not limit sorting to video to be recommended
Method, for example can be ranked up it is also possible to related to video v according to video to be recommended according to the temperature of video to be recommended
Degree is ranked up, and can also be ranked up according to the comprehensive parameters (such as behavior characteristicss, attribute character etc.) of video to be recommended.Its
In, behavior characteristicss can be used to indicate that the situation to the behavior that video to be recommended is made for the user, for example viewing duration, comment number,
Number etc. is stepped on scoring, top.Resource characteristic can be used to indicate that the situation of the attribute of video to be recommended, such as video channel, interest mark
Sign etc..
When user is carried out with the personalized recommendation of video, not only need to predict the interest video of user, it is also contemplated that regarding
Whether frequency the status information such as can play.Based on above principle, the video of video sequence prediction can be screened, generate and meet
The video recommendations list requiring.In addition, the list of videos being unsatisfactory for recommendation list length is carried out hot video supplement, and return
To request user.Hot video can be for example that the interior video of preset time period (such as a week) clicks on row in embodiments of the present invention
The forward video of sequence.
The recommendation method of the multimedia resource of the embodiment of the present invention, the heading message using multimedia resource is gathered automatically
Class, and the multimedia resource classification of unartificial rule, the performance that there is self adaptation and excavate video recessiveness classification, effectively carry out
Multimedia resource category label.It is ensured that customer multi-media resource recommendation list while ensureing Accurate Prediction user interest
Multiformity, improves Consumer's Experience, excavates long-tail multimedia resource.
The recommendation method of the multimedia resource of the embodiment of the present invention, can believe according to the title of each multimedia resource to be recommended
Breath, determines the classification that each multimedia resource to be recommended is belonged to, and the class being belonged to according to each multimedia resource to be recommended respectively
Not, multimedia resource recommendation results are generated.The recommendation method of the multimedia resource of the embodiment of the present invention, can enrich recommendation information
The classification of middle multimedia resource, improves the multiformity of multimedia resource.
Embodiment 2
Fig. 5 illustrates the structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention.Fig. 5 can use
In the video broadcasting method shown in operation Fig. 1 to Fig. 4.
As shown in figure 5, the recommendation apparatus of described multimedia resource, may include that characteristic determination module 11, for according to each
The heading message of multimedia resource to be recommended, determines the feature of each described multimedia resource to be recommended respectively；Category determination module
13, it is connected with described characteristic determination module 11, for the feature according to each described multimedia resource to be recommended, determine each institute respectively
State the classification that multimedia resource to be recommended is belonged to；Recommendation results generation module 15, is connected with described category determination module 13, uses
In the classification being belonged to according to each described multimedia resource to be recommended, generate multimedia resource recommendation results.Concrete principle and showing
Example may refer to embodiment 1 and the associated description of Fig. 1.
In a kind of possible implementation, as shown in fig. 6, described characteristic determination module 11 includes: heading message obtains
Unit 111, for obtaining the heading message of each described multimedia resource to be recommended；Heading message participle unit 113, with described mark
Topic information acquisition unit 111 connects, and for carrying out participle to each described heading message, obtains each described heading message corresponding
Word；Characteristics determining unit 115, is connected with described heading message participle unit 113, for corresponding according to each described heading message
Word, determines the feature of each described multimedia resource to be recommended respectively.Concrete principle and example may refer to embodiment 1 and Fig. 2
Associated description.
In a kind of possible implementation, as shown in fig. 7, described category determination module 13 includes: metrics calculation unit
131, for the feature according to each described multimedia resource to be recommended and center vector of all categories, calculate respectively and each described wait to push away
Recommend multimedia resource and distance of all categories；Classification determination unit 133, is connected with described metrics calculation unit 131, for basis
Each described multimedia resource to be recommended and distance of all categories, determine the class that each described multimedia resource to be recommended is belonged to respectively
Not.Concrete principle and example may refer to embodiment 1 and the associated description of Fig. 3.
In a kind of possible implementation, described characteristics determining unit 115, for corresponding to according to each described heading message
Word, be respectively adopted the fisrt feature that formula 1 calculates each described heading message,
Wherein, the corresponding word of described heading message is titile (v)={ w1, w2 ..., wk }；K represents described heading message
The total number of corresponding word；niRepresent the number of times that i-th word wi occurs in described heading message；njRepresent j-th word wj in institute
State the number of times occurring in heading message；I, j represent the label of the corresponding word of described heading message, the span of i, j be [1,
k]；tfiRepresent the corresponding fisrt feature of i-th word wi；
Described characteristics determining unit 115, is additionally operable to total according to the corresponding word of each described heading message and multimedia resource
Number, is respectively adopted the second feature that formula 2 calculates each described heading message,
Wherein, | d | represents the total number of multimedia resource；|di| represent that heading message includes many matchmakers of i-th word wi
The number of body resource；idfiRepresent the corresponding second feature of i-th word wi；
Described characteristics determining unit 115, is additionally operable to, according to described fisrt feature and described second feature, be respectively adopted formula 3
Calculate the third feature of each described heading message,
tfidfi=tfi×idfiFormula 3,
Wherein, tfidfiRepresent the corresponding third feature of i-th word wi.
In a kind of possible implementation, described metrics calculation unit 131, for according to each described multimedia to be recommended
The feature of resource and center vector of all categories, be respectively adopted formula 4 calculate each described multimedia resource to be recommended with of all categories
Distance,
s(vn, cm)=cosin (f (vn), qm) formula 4,
Wherein, n represents multimedia resource v to be recommendednCorresponding label, n is integer；M represents classification cmCorresponding label, m
For integer；qmRepresent classification cmCenter vector；f(vn) represent multimedia resource v to be recommendednFeature；s(vn, cm) represent and treat
Recommend multimedia resource vnWith classification cmDistance.
In a kind of possible implementation, described classification determination unit 133, for according to each described multimedia to be recommended
Resource and distance of all categories, are respectively adopted the classification that formula 5 determines that each described multimedia resource to be recommended is belonged to,
Wherein, h (vn) represent s (vn, cm) take the value of m during maximum, the class that each described multimedia resource to be recommended is belonged to
It is not the classification closest with each described multimedia resource.
In a kind of possible implementation, as shown in figure 8, described recommendation results generation module 15 includes: ranking results
Determining unit 151, for the classification being belonged to according to each described multimedia resource to be recommended, determine of all categories in included each
The ranking results of described multimedia resource to be recommended；Recommendation results signal generating unit 153, with described ranking results determining unit even
Connect, for according to described ranking results, generating described multimedia resource recommendation results.Concrete principle and example may refer to implement
Example 1 and the associated description of Fig. 4.
The recommendation apparatus of the multimedia resource of the embodiment of the present invention, can believe according to the title of each multimedia resource to be recommended
Breath, determines the classification that each multimedia resource to be recommended is belonged to, and the class being belonged to according to each multimedia resource to be recommended respectively
Not, multimedia resource recommendation results are generated.The recommendation apparatus of the multimedia resource of the embodiment of the present invention, can enrich recommendation information
The classification of middle multimedia resource, improves the multiformity of multimedia resource.
Embodiment 3
Fig. 9 illustrates another structured flowchart of the recommendation apparatus of multimedia resource according to another embodiment of the present invention.Described
The recommendation apparatus 1100 of multimedia resource can be possessed the host server of computing capability, personal computer pc or can take
The portable computer of band or terminal etc..The specific embodiment of the invention does not limit to implementing of calculate node.
The recommendation apparatus 1100 of described multimedia resource include processor (processor) 1110, communication interface
(communications interface) 1120, memorizer (memory) 1130 and bus 1140.Wherein, processor 1110,
Communication interface 1120 and memorizer 1130 complete mutual communication by bus 1140.
Communication interface 1120 is used for and network device communications, and wherein the network equipment includes such as Virtual Machine Manager center, is total to
Enjoy storage etc..
Processor 1110 is used for configuration processor.Processor 1110 is probably a central processing unit cpu, or special collection
Become circuit asic (application specific integrated circuit), or be arranged to implement the present invention
One or more integrated circuits of embodiment.
Memorizer 1130 is used for depositing file.Memorizer 1130 potentially includes high speed ram memorizer it is also possible to also include non-
Volatile memory (non-volatile memory), for example, at least one disk memory.Memorizer 1130 can also be deposited
Memory array.Memorizer 1130 is also possible to by piecemeal, and described piece can be combined into virtual volume by certain rule.
In a kind of possible embodiment, said procedure can be the program code including computer-managed instruction.This journey
Sequence is particularly used in: realizes the operation of each step in embodiment 1.
Those of ordinary skill in the art are it is to be appreciated that each exemplary cell in embodiment described herein and algorithm
Step, being capable of being implemented in combination in electronic hardware or computer software and electronic hardware.These functions are actually with hardware also
Being software form to realize, the application-specific depending on technical scheme and design constraint.Professional and technical personnel can be directed to
Specifically application selects different methods to realize described function, but this realization is it is not considered that exceed the model of the present invention
Enclose.
If to be realized using in the form of computer software described function and as independent production marketing or use when,
To a certain extent it is believed that all or part (part for example prior art being contributed) of technical scheme is
Embody in form of a computer software product.This computer software product is generally stored inside the non-volatile of embodied on computer readable
In storage medium, including some instructions with so that computer equipment (can be that personal computer, server or network set
Standby etc.) all or part of step of execution various embodiments of the present invention method.And aforesaid storage medium include u disk, portable hard drive,
Read only memory (rom, read-only memory), random access memory (ram, random access memory), magnetic
Dish or CD etc. are various can be with the medium of store program codes.
The above, the only specific embodiment of the present invention, but protection scope of the present invention is not limited thereto, and any
Those familiar with the art the invention discloses technical scope in, change or replacement can be readily occurred in, all should contain
Cover within protection scope of the present invention.Therefore, protection scope of the present invention should be defined by described scope of the claims.
Claims (14)
1. a kind of recommendation method of multimedia resource is it is characterised in that include:
According to the heading message of each multimedia resource to be recommended, determine the feature of each described multimedia resource to be recommended respectively；
According to the feature of each described multimedia resource to be recommended, determine the class that each described multimedia resource to be recommended is belonged to respectively
Not；
The classification being belonged to according to each described multimedia resource to be recommended, generates multimedia resource recommendation results.
2. method according to claim 1 is it is characterised in that according to the heading message of each multimedia resource to be recommended, point
Do not determine the feature of each described multimedia resource to be recommended, comprising:
Obtain the heading message of each described multimedia resource to be recommended；
Participle is carried out to each described heading message, obtains the corresponding word of each described heading message；
According to the corresponding word of each described heading message, determine the feature of each described multimedia resource to be recommended respectively.
3. method according to claim 1 is it is characterised in that according to the feature of each described multimedia resource to be recommended, point
Do not determine the classification that each described multimedia resource to be recommended is belonged to, comprising:
Feature according to each described multimedia resource to be recommended and center vector of all categories, calculate each described to be recommended many respectively
Media resource and distance of all categories；
According to each described multimedia resource to be recommended and distance of all categories, determine each described multimedia resource institute to be recommended respectively
The classification of ownership.
4. method according to claim 2 is it is characterised in that according to the corresponding word of each described heading message, determine respectively
The feature of each described multimedia resource to be recommended, comprising:
According to the corresponding word of each described heading message, it is respectively adopted the fisrt feature that formula 1 calculates each described heading message,
Wherein, the corresponding word of described heading message is title (v)={ w1, w2 ..., wk }；K represents that described heading message corresponds to
Word total number；niRepresent the number of times that i-th word wi occurs in described heading message；njRepresent j-th word wj in described mark
The number of times occurring in topic information；I, j represent the label of the corresponding word of described heading message, and the span of i, j is [1, k]；tfi
Represent the corresponding fisrt feature of i-th word wi；
According to the sum of the corresponding word of each described heading message and multimedia resource, it is respectively adopted formula 2 and calculates each described title letter
The second feature of breath,
Wherein, | d | represents the total number of multimedia resource；|di| represent that heading message includes the multimedia resource of i-th word wi
Number；idfiRepresent the corresponding second feature of i-th word wi；
According to described fisrt feature and described second feature, it is respectively adopted the third feature that formula 3 calculates each described heading message,
tfidfi=tfi×idfiFormula 3,
Wherein, tfidfiRepresent the corresponding third feature of i-th word wi.
5. method according to claim 3 is it is characterised in that according to the feature of each described multimedia resource to be recommended and each
The center vector of classification, calculates each described multimedia resource to be recommended and distance of all categories respectively, comprising:
Feature according to each described multimedia resource to be recommended and center vector of all categories, be respectively adopted formula 4 calculate each described
Multimedia resource to be recommended and distance of all categories,
s(vn, cm)=cosin (f (vn), qm) formula 4,
Wherein, n represents multimedia resource v to be recommendednCorresponding label, n is integer；M represents classification cmCorresponding label, m is whole
Number；qmRepresent classification cmCenter vector；f(vn) represent multimedia resource v to be recommendednFeature；s(vn, cm) represent to be recommended
Multimedia resource vnWith classification cmDistance.
6. method according to claim 5 it is characterised in that according to each described multimedia resource to be recommended with of all categories
Distance, determines the classification that each described multimedia resource to be recommended is belonged to respectively, comprising:
According to each described multimedia resource to be recommended and distance of all categories, it is respectively adopted formula 5 and determines each described many matchmakers to be recommended
The classification that body resource is belonged to,
Wherein, h (vn) represent s (vn, cm) taking the value of m during maximum, the classification that each described multimedia resource to be recommended is belonged to is
The classification closest with each described multimedia resource.
7. method according to any one of claim 1 to 6 is it is characterised in that provide according to each described multimedia to be recommended
The classification that source is belonged to, generates multimedia resource recommendation results, comprising:
The classification being belonged to according to each described multimedia resource to be recommended, determine of all categories in included each described to be recommended many
The ranking results of media resource；
According to described ranking results, generate described multimedia resource recommendation results.
8. a kind of recommendation apparatus of multimedia resource are it is characterised in that include:
Characteristic determination module, for the heading message according to each multimedia resource to be recommended, determines each described to be recommended many respectively
The feature of media resource；
Category determination module, is connected with described characteristic determination module, for the feature according to each described multimedia resource to be recommended,
Determine the classification that each described multimedia resource to be recommended is belonged to respectively；
Recommendation results generation module, is connected with described category determination module, for according to each described multimedia resource institute to be recommended
The classification of ownership, generates multimedia resource recommendation results.
9. device according to claim 8 is it is characterised in that described characteristic determination module includes:
Heading message acquiring unit, for obtaining the heading message of each described multimedia resource to be recommended；
Heading message participle unit, is connected with described heading message acquiring unit, for participle is carried out to each described heading message,
Obtain the corresponding word of each described heading message；
Characteristics determining unit, is connected with described heading message participle unit, for according to the corresponding word of each described heading message, point
Do not determine the feature of each described multimedia resource to be recommended.
10. device according to claim 8 is it is characterised in that described category determination module includes:
Metrics calculation unit, for the feature according to each described multimedia resource to be recommended and center vector of all categories, difference
Calculate each described multimedia resource to be recommended and distance of all categories；
Classification determination unit, is connected with described metrics calculation unit, for according to each described multimedia resource to be recommended with all kinds of
Other distance, determines the classification that each described multimedia resource to be recommended is belonged to respectively.
11. devices according to claim 9 it is characterised in that
Described characteristics determining unit, for according to the corresponding word of each described heading message, being respectively adopted formula 1 and calculating each described title
The fisrt feature of information,
Wherein, the corresponding word of described heading message is title (v)={ w1, w2 ..., wk }；K represents that described heading message corresponds to
Word total number；niRepresent the number of times that i-th word wi occurs in described heading message；njRepresent j-th word wj in described mark
The number of times occurring in topic information；I, j represent the label of the corresponding word of described heading message, and the span of i, j is [1, k]；tfi
Represent the corresponding fisrt feature of i-th word wi；
Described characteristics determining unit, is additionally operable to the sum according to the corresponding word of each described heading message and multimedia resource, respectively
Calculate the second feature of each described heading message using formula 2,
Wherein, | d | represents the total number of multimedia resource；|di| represent that heading message includes the multimedia resource of i-th word wi
Number；idfiRepresent the corresponding second feature of i-th word wi；
Described characteristics determining unit, is additionally operable to according to described fisrt feature and described second feature, is respectively adopted formula 3 and calculates each institute
State the third feature of heading message,
tfidfi=tfi×idfiFormula 3,
Wherein, tfidfiRepresent the corresponding third feature of i-th word wi.
12. devices according to claim 10 it is characterised in that
Described metrics calculation unit, for the feature according to each described multimedia resource to be recommended and center vector of all categories,
It is respectively adopted formula 4 and calculate each described multimedia resource to be recommended and distance of all categories,
s(vn, cm)=cosin (f (vn), qm) formula 4,
Wherein, n represents multimedia resource v to be recommendednCorresponding label, n is integer；M represents classification cmCorresponding label, m is whole
Number；qmRepresent classification cmCenter vector；f(vn) represent multimedia resource v to be recommendednFeature；s(vn, cm) represent to be recommended
Multimedia resource vnWith classification cmDistance.
13. devices according to claim 12 it is characterised in that
Described classification determination unit, for according to each described multimedia resource to be recommended and distance of all categories, being respectively adopted formula 5
Determine the classification that each described multimedia resource to be recommended is belonged to,
Wherein, h (vn) represent s (vn, cm) taking the value of m during maximum, the classification that each described multimedia resource to be recommended is belonged to is
The classification closest with each described multimedia resource.
14. devices any one of according to Claim 8 to 13 are it is characterised in that described recommendation results generation module bag
Include:
Ranking results determining unit, for the classification being belonged to according to each described multimedia resource to be recommended, determine of all categories in
The ranking results of included each described multimedia resource to be recommended；
Recommendation results signal generating unit, is connected with described ranking results determining unit, described for according to described ranking results, generating
Multimedia resource recommendation results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610817968.2A CN106354867A (en) | 2016-09-12 | 2016-09-12 | Multimedia resource recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610817968.2A CN106354867A (en) | 2016-09-12 | 2016-09-12 | Multimedia resource recommendation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106354867A true CN106354867A (en) | 2017-01-25 |
Family
ID=57859751
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610817968.2A Pending CN106354867A (en) | 2016-09-12 | 2016-09-12 | Multimedia resource recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106354867A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107688587A (en) * | 2017-02-15 | 2018-02-13 | 腾讯科技（深圳）有限公司 | A kind of media information methods of exhibiting and device |
| CN109492216A (en) * | 2018-09-19 | 2019-03-19 | 平安科技（深圳）有限公司 | Water note identifies automatically and the measures and procedures for the examination and approval, device and computer readable storage medium |
| CN110659404A (en) * | 2018-06-13 | 2020-01-07 | 腾讯科技（深圳）有限公司 | Information recommendation method and device and storage medium |
| CN110807127A (en) * | 2018-08-01 | 2020-02-18 | 北京优酷科技有限公司 | Video recommendation method and device |
| CN111400615A (en) * | 2020-03-19 | 2020-07-10 | 腾讯科技（深圳）有限公司 | Resource recommendation method, device, equipment and storage medium |
| CN112000820A (en) * | 2020-08-10 | 2020-11-27 | 海信电子科技(武汉)有限公司 | Media asset recommendation method and display device |
| WO2021068610A1 (en) * | 2019-10-12 | 2021-04-15 | 平安国际智慧城市科技股份有限公司 | Resource recommendation method and apparatus, electronic device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8063295B2 (en) * | 2002-10-03 | 2011-11-22 | Polyphonic Human Media Interface, S.L. | Method and system for video and film recommendation |
| CN103164471A (en) * | 2011-12-15 | 2013-06-19 | 盛乐信息技术（上海）有限公司 | Recommendation method and system of video text labels |
| CN103744928A (en) * | 2013-12-30 | 2014-04-23 | 北京理工大学 | Network video classification method based on historical access records |
| CN104021140A (en) * | 2014-05-08 | 2014-09-03 | 北京奇艺世纪科技有限公司 | Network video processing method and device |
-
2016
- 2016-09-12 CN CN201610817968.2A patent/CN106354867A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8063295B2 (en) * | 2002-10-03 | 2011-11-22 | Polyphonic Human Media Interface, S.L. | Method and system for video and film recommendation |
| CN103164471A (en) * | 2011-12-15 | 2013-06-19 | 盛乐信息技术（上海）有限公司 | Recommendation method and system of video text labels |
| CN103744928A (en) * | 2013-12-30 | 2014-04-23 | 北京理工大学 | Network video classification method based on historical access records |
| CN104021140A (en) * | 2014-05-08 | 2014-09-03 | 北京奇艺世纪科技有限公司 | Network video processing method and device |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107688587A (en) * | 2017-02-15 | 2018-02-13 | 腾讯科技（深圳）有限公司 | A kind of media information methods of exhibiting and device |
| CN107688587B (en) * | 2017-02-15 | 2023-02-10 | 腾讯科技（深圳）有限公司 | Media information display method and device |
| CN110659404A (en) * | 2018-06-13 | 2020-01-07 | 腾讯科技（深圳）有限公司 | Information recommendation method and device and storage medium |
| CN110659404B (en) * | 2018-06-13 | 2024-03-15 | 深圳市雅阅科技有限公司 | Information recommendation method, device and storage medium |
| CN110807127A (en) * | 2018-08-01 | 2020-02-18 | 北京优酷科技有限公司 | Video recommendation method and device |
| CN109492216A (en) * | 2018-09-19 | 2019-03-19 | 平安科技（深圳）有限公司 | Water note identifies automatically and the measures and procedures for the examination and approval, device and computer readable storage medium |
| WO2021068610A1 (en) * | 2019-10-12 | 2021-04-15 | 平安国际智慧城市科技股份有限公司 | Resource recommendation method and apparatus, electronic device and storage medium |
| CN111400615A (en) * | 2020-03-19 | 2020-07-10 | 腾讯科技（深圳）有限公司 | Resource recommendation method, device, equipment and storage medium |
| CN112000820A (en) * | 2020-08-10 | 2020-11-27 | 海信电子科技(武汉)有限公司 | Media asset recommendation method and display device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106354867A (en) | Multimedia resource recommendation method and device | |
| US20220114199A1 (en) | System and method for information recommendation | |
| Wang et al. | Latent aspect rating analysis without aspect keyword supervision | |
| US10204163B2 (en) | Active prediction of diverse search intent based upon user browsing behavior | |
| TWI636416B (en) | Method and system for multi-phase ranking for content personalization | |
| Chen et al. | Sheepdog: group and tag recommendation for flickr photos by automatic search-based learning | |
| WO2017181612A1 (en) | Personalized video recommendation method and device | |
| CN104428768B (en) | Knowledge panel | |
| Agarwal et al. | Statistical methods for recommender systems | |
| CN109511015B (en) | Multimedia resource recommendation method, device, storage medium and equipment | |
| CN106326391A (en) | Method and device for recommending multimedia resources | |
| CN106294830A (en) | The recommendation method and device of multimedia resource | |
| CN109753601B (en) | Method and device for determining click rate of recommended information and electronic equipment | |
| CN103455545A (en) | Location estimation of social network users | |
| CN105426514A (en) | Personalized mobile APP recommendation method | |
| CN104077415A (en) | Searching method and device | |
| CN103714063A (en) | Data analysis method and data analysis system | |
| US20120117043A1 (en) | Measuring Duplication in Search Results | |
| CN111344696B (en) | System and method for evaluating advertisements | |
| CN103262079B (en) | Search device and search method | |
| CN113961823B (en) | News recommendation method, system, storage medium and equipment | |
| CN107944026A (en) | A kind of method, apparatus, server and the storage medium of atlas personalized recommendation | |
| CN114330329A (en) | Service content searching method and device, electronic equipment and storage medium | |
| CN106156351A (en) | Multimedia resource recommendation information generates method and device | |
| Bhatnagar et al. | A novel aspect based framework for tourism sector with improvised aspect and opinion mining algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170125 |
|
| RJ01 | Rejection of invention patent application after publication |