CN108737859A - Video recommendation method based on barrage and device - Google Patents
Video recommendation method based on barrage and device Download PDFInfo
- Publication number
- CN108737859A CN108737859A CN201810426715.1A CN201810426715A CN108737859A CN 108737859 A CN108737859 A CN 108737859A CN 201810426715 A CN201810426715 A CN 201810426715A CN 108737859 A CN108737859 A CN 108737859A
- Authority
- CN
- China
- Prior art keywords
- user
- video
- barrage
- vector
- similarity
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
- H04N21/25866—Management of end-user data
- H04N21/25891—Management of end-user data being end-user preferences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/235—Processing of additional data, e.g. scrambling of additional data or processing content descriptors
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/251—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/252—Processing of multiple end-users' preferences to derive collaborative data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/435—Processing of additional data, e.g. decrypting of additional data, reconstructing software from modules extracted from the transport stream
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4662—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms
- H04N21/4666—Learning process for intelligent management, e.g. learning user preferences for recommending movies characterized by learning algorithms using neural networks, e.g. processing the feedback provided by the user
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Computer Graphics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Social Psychology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a kind of video recommendation method and device based on barrage, and this method includes：According to the language material of barrage, model training is carried out to the word vector sum affective tag of language material by neural network, and predict the emotion of barrage；User is configured to indicate user to the user vector of the fancy grade of all kinds of videos by the barrage issued in video by user, and wherein every one-dimensional representation user of user vector is to fancy grade a kind of in all kinds of videos；The primary vector that the character introduction of video is obtained by PV-DM models, the secondary vector of the distribution of each implicit theme of the barrage of video is obtained by theme generation model, and primary vector and secondary vector build video vector jointly；It is scored the prediction of video according to user's similarity and video similarity prediction user, video recommendations by prediction scoring not less than predetermined threshold are to user, wherein by the similarity respectively between user and the similarity calculation score between video after, score is weighted to obtain prediction scoring.
Description
Technical field
The present invention relates to video services technology field more particularly to a kind of video recommendation methods and device based on barrage.
Background technology
With the development of Online Video website, barrage has become a kind of exchange way of hot topic.In Online Video website
Number of videos is various, and commending system has been used to video website to promote user experience.It is presently recommended that mode is generally divided into two kinds：
One is according to keyword input by user, is matched with the title of video or introduction, return to the higher video of degree of correlation；Separately
One is the data such as the history of analysis user viewing record, scoring and comment, analyze the hobby of user, recommending user may like
Joyous video.
First way is matched by keyword input by user, when user can not accurately describe oneself demand or
When the keyword of input is wrong, then accurate recommendation results cannot be provided.And the second way relies primarily on user's rating matrix
Recommend with the keyword of extraction user's barrage.But in the video website of mainstream, user can not directly make video and comment
Point, therefore user's rating matrix can not be directly obtained.And it extracts the method for keyword there is problems with：(1) user is to this
The emotion of keyword is not necessarily positive, if user does not like this keyword, that is recommended his video and can not carry
Rise user experience；(2) keyword can not represent the interest of user completely, it may be possible to the word that user mentions once in a while, in this way
The video of recommendation can not promote user experience.
Invention content
In view of the above-mentioned problems, the present invention establishes the model of user and predicts user's by the barrage content of analysis user
Emotion recommends the video that user may like.
In a first aspect, an embodiment of the present invention provides a kind of video recommendation methods based on barrage, including：According to barrage
Language material carries out model training to the word vector sum affective tag of language material by neural network, and predicts the emotion of barrage；Pass through use
The barrage that family is issued in video, the user vector for being configured to user that can indicate user to the fancy grade of all kinds of videos,
Wherein every one-dimensional representation user of user vector is to fancy grade a kind of in all kinds of videos；Video is obtained by PV-DM models
Character introduction primary vector, by theme generate model obtain video barrage each implicit theme distribution second to
Amount, primary vector and secondary vector build video vector jointly；Predict user to regarding according to user's similarity and video similarity
The prediction of frequency is scored, will prediction scoring not less than predetermined threshold video recommendations to user, wherein by respectively between user
Similarity and video between similarity calculation score after, score is weighted to obtain prediction scoring.
Optionally, it using Skip-gram model trainings word vector, is sent in and is regarded using shot and long term memory network prediction user
The feeling polarities of barrage in frequency.
Optionally, the often one-dimensional of user vector includes the score obtained according to the emotion of barrage, and score is liked for indicating
Good degree.
Optionally, the similarity between user and the similarity between video are obtained by cosine similarity respectively.
Optionally, in recommendation step, occurrence number most multiple labels when choosing viewing video in user's predetermined time,
The video collection for needing to predict is obtained according to label.
Second aspect, the present invention disclose a kind of video recommendations device based on barrage, including：Sentiment analysis model training mould
Block, for the language material according to barrage, model training is carried out to the word vector sum affective tag of the language material by neural network, and
Predict the emotion of the barrage；User model builds module, the barrage for being issued in video by user, will be described
User is configured to the user vector that can indicate the user to the fancy grade of all kinds of videos, wherein the user vector is every
User described in one-dimensional representation is to fancy grade a kind of in all kinds of videos；Video Model builds module, for passing through PV-DM
Model obtains the primary vector of the character introduction of the video, and each hidden of the barrage of the video is obtained by theme generation model
The secondary vector of distribution containing theme, the primary vector and the secondary vector build video vector jointly；Recommending module, use
In predicting that the user scores to the prediction of video according to user's similitude and Video similarity, prediction scoring is not less than
The video recommendations of predetermined threshold give the user, wherein passing through the similarity respectively between user and the similarity between video
After calculating score, the score is weighted to obtain the prediction scoring.
Optionally, it using Skip-gram model trainings word vector, is sent in and is regarded using shot and long term memory network prediction user
The feeling polarities of barrage in frequency.
Optionally, the often one-dimensional of user vector includes the score obtained according to the emotion of barrage, and score is liked for indicating
Good degree.
Optionally, the similarity between user and the similarity between video are obtained by cosine similarity respectively.
Optionally, in recommendation step, occurrence number most multiple labels when choosing viewing video in user's predetermined time,
The video collection for needing to predict is obtained according to label.
The third aspect, the embodiment of the present invention provide a kind of computing device, including memory and one or more at
Manage device；Wherein, computing device further includes：One or more units, one or more units be stored in memory and by with
It is set to and is executed by one or more processors, one or more units include the instruction for executing following steps：According to barrage
Language material, model training is carried out to the word vector sum affective tag of language material by neural network, and predict the emotion of barrage；Pass through
The barrage that user issues in video, by user be configured to can to indicate user to the user of the fancy grade of all kinds of videos to
Amount, wherein every one-dimensional representation user of user vector is to fancy grade a kind of in all kinds of videos；It is regarded by PV-DM models
The primary vector of the character introduction of frequency obtains the second of the distribution of each implicit theme of the barrage of video by theme generation model
Vector, primary vector and secondary vector build video vector jointly；User couple is predicted according to user's similarity and video similarity
The prediction of video is scored, will prediction scoring not less than predetermined threshold video recommendations to user, wherein by respectively to user it
Between similarity and video between similarity calculation score after, score is weighted to obtain prediction scoring.
Further, the embodiment of the present invention provides the computer program product being used in combination with computing device, and feature exists
In, including computer-readable storage medium and it is embedded in computer program mechanism therein；Wherein, computer program mechanism packet
Include the instruction for executing following steps：According to the language material of barrage, the word vector sum affective tag of language material is carried out by neural network
Model training, and predict the emotion of barrage；User is configured to indicate user by the barrage issued in video by user
To the user vector of the fancy grade of all kinds of videos, wherein every one-dimensional representation user of user vector is to a kind of in all kinds of videos
Fancy grade；The primary vector that the character introduction of video is obtained by PV-DM models generates model by theme and obtains video
The secondary vector of the distribution of each implicit theme of barrage, primary vector and secondary vector build video vector jointly；According to user
Similarity and video similarity prediction user score to the prediction of video, the video recommendations by prediction scoring not less than predetermined threshold
To user, wherein after by the similarity respectively between user and the similarity calculation score between video, score is weighted
Obtain prediction scoring.
Compared with prior art, the main distinction and its effect are embodiment of the present invention：Technical scheme of the present invention energy
It is enough to predict that the real feelings of user are expressed according to the barrage of user, the video that recommended user likes.
Description of the drawings
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, other are can also be obtained according to these attached drawings
Attached drawing.
Fig. 1 is the flow chart of the recommendation method 100 according to a first embodiment of the present invention based on barrage.
Fig. 2 is the schematic diagram of the feeling polarities of LSTM model predictions barrage according to the ... of the embodiment of the present invention.
Fig. 3 is the structure diagram of the recommendation apparatus 300 based on barrage for third embodiment of the invention.
Specific implementation mode
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.Illustrative system and method embodiment described herein is not intended to be limited.
First embodiment
Fig. 1 is the flow chart of the recommendation method 100 according to the ... of the embodiment of the present invention based on barrage.As shown in Figure 1, this method
Specific process flow is as described below：
S110, the language material according to barrage carry out model by neural network to the word vector sum affective tag of the language material
Training, and predict the emotion of the barrage.
According to an embodiment of the invention, it will be understood that it needs to collect the relevant various data of video, such as：Video information,
User information, user's barrage and affective tag etc..Wherein, video information include but not limited to video number, video type label,
Video text introduction etc.；User information includes but not limited to Customs Assigned Number etc.；User's barrage may include Customs Assigned Number, video volume
Number, barrage sending time, barrage reproduction time, barrage content etc..Affective tag is the table marked for the barrage content of user
Show the forward direction (positive) of orientation of emotion, the label of negative sense (negative) and neutral (neutral).
Since barrage is usually shorter and has very strong randomness, existing segmentation methods are difficult to play satisfactory effect
Fruit.Therefore the vector of the word of the embodiment of the present invention study barrage indicates, and is used for sentiment analysis.It, can be with as an example
Using existing skip-gram model trainings word vector.
According to an embodiment of the invention, using LSTM (Long Short-Term Memory, shot and long term memory network) mould
Type predicts the feeling polarities of barrage.As shown in Fig. 2, the barrage of extraction user is the sequence of word, by the corresponding word vector w of wordi(i
It=1,2 ..., n) is sequentially inputted in LSTM units, each LSTM units can export a hi(i=1,2 ..., n), it will
The hidden layer output vector h of the last onenAs sentence expression it is defeated finally obtain be input to softmax layers.Softmax layers will
Vector is divided into three labels：Positive (positive), negative sense (negative) and neutrality (neutral).This method is using intersection
Entropy is as loss function：
Wherein, DTIndicate that all training barrages, C are affective tag number.gc(d) value is 1 or 0, respectively represents bullet
Curtain d belongs to or is not belonging to this current affective tag.Pc(d) the softmax layers of probability divided barrage d for classification c are represented.It can be with
Understand, model training carries out parameter update using backpropagation and stochastic gradient descent.
The user is configured to indicate the user by S120, the barrage issued in video by user
To the user vector of the fancy grade of all kinds of videos, wherein user is to described all kinds of described in every one-dimensional representation of the user vector
A kind of fancy grade in video.
In order to weigh the similarity of user, the embodiment of the present invention is by calculating hobby journey of the user to each video tab
It spends to calculate the similarity between user.Assuming that the video tab collection of entire video website is combined into L, the character representation of user i is |
L | the vectorial u of dimensioni, per one-dimensional j (j=1 .., | L |) user is represented to label ljFancy grade.It can structure in the following manner
It makes user vector and weighs the similarity between user：
(1) video that user is transmitted across barrage is obtained, V is denoted asi；
(2) to each video v ∈ Vi, obtain its tag set Lv, to the barrage d that user i is sent in this video, more
New its user vector ui:
ui[Lv]=ui[Lv]+scoresentiment(d)
Its meaning is that the emotion for the barrage d that user sends in video v is emotion of the user to the affiliated labels of video v, to
Measure uiThe corresponding dimension of label of middle video v can add the score score by being obtained according to the emotion of barrage dsentiment(d).This
Barrage emotion is divided into positive (positive) in the embodiment of invention, negative sense (negative) and neutrality (neutral), thus
scoresentiment(d) it is defined as：
User vector u in this wayiJust represent fancy grades of the user i to all types of videos.
(3) after building user vector, the similarity between cosine similarity measurement user can be used：
S130, obtained by PV-DM models the video character introduction primary vector, by theme generate model obtain
To the secondary vector of the distribution of each implicit theme of the barrage of the video, the primary vector and the common structure of the secondary vector
Build video vector.
Further, the embodiment of the present invention carries out vectorization in terms of two to video：It the character introduction of video and regards
The barrage of frequency.As an example, the introduction of all videos and the barrage of video are collected, with PV-DM (Distributed
Memory Model of Paragraph Vectors, the distribution memory models of sentence vector) model training obtains character introduction portion
The vector dividedMeanwhile using the barrage of each video as document, using LDA (Latent Dirichlet
Allocation) document subject matter generates model and obtains the distribution of the implicit theme about the barrage document.Video v in this wayiBullet
Curtain is represented by the vector of T dimensionsT is the number of implicit theme.Splicing is available after two kinds of vectors are unitization
The vector v of video ii。
It is appreciated that it is similar with user's similarity is weighed, the phase between cosine similarity measurement video equally may be used
Like degree：
Further, S140, predict that according to user's similitude and Video similarity, the user scores to the prediction of video,
The prediction scoring is given to the user not less than the video recommendations of predetermined threshold, wherein by similar between user respectively
After similarity calculation score between degree and video, the score is weighted to obtain the prediction scoring.
If to all video estimation scores, efficiency is relatively low, therefore, in an embodiment of the present invention, Ke Yixian
A series of videos are selected as candidate.As an example, within user i nearest a period of times, such as several days or one week, viewing
M video in, collect the label of these videos, choose the most k labels of occurrence number, and find owning under these labels
Video is denoted as V as candidate collectioncan.According to an embodiment of the invention, from the similarity between user and the phase between video
Like the marking of two aspect prediction users of degree.To each video v ∈ Vcan, user i it is possible marking be：
Score (i, v)=k1·scoreuser+k2·scorevideo
Wherein scoreuserIndicate the prediction obtained from user's similarity marking, scorevideoExpression is obtained from video similarity
The prediction marking arrived, wherein k1And k2For parameter, k1+k2=1, k1And k2It respectively represents to the stressing property in terms of the two.
scoreuserAnd scorevideoIt is represented by：
Wherein DvIndicate the set of barrage in video v, | Dv| indicate the size of the set, udRepresent the user for sending barrage d
Vector.LvIndicate the tag set of video v, VlIndicate the video collection for having same label l with video v, | Vl| for the big of the set
It is small.s(ud, ui) and s (vj, v) and indicate the similarity between user and the similarity between video.
And sp (vj) represent viewing video vjUser to video vjOverall assessment, can be expressed as：
It is possible thereby to predict the marking of user i, and return to VcanN video of middle prediction highest scoring gives user i.
Therefore, the recommendation method based on barrage of the embodiment of the present invention can predict that user's is true according to the barrage of user
Emotional expression, the video that recommended user likes.
The each method embodiment of the present invention can be realized in a manner of software, hardware, firmware etc..No matter the present invention be with
Software, hardware or firmware mode realize that instruction code may be stored in any kind of computer-accessible memory
In (such as permanent either revisable volatibility is either non-volatile solid or non-solid, it is fixed or
The replaceable medium etc. of person).Equally, memory may, for example, be programmable logic array (Programmable Array
Logic, referred to as " PAL "), random access memory (Random Access Memory, referred to as " RAM "), programmable read-only deposit
Reservoir (Programmable Read Only Memory, referred to as " PROM "), read-only memory (Read-Only Memory, letter
Claim " ROM "), electrically erasable programmable read-only memory (Electrically Erasable Programmable ROM, referred to as
" EEPROM "), disk, CD, digital versatile disc (Digital Versatile Disc, referred to as " DVD ") etc..
Second embodiment
Fig. 3 is the schematic block diagram of the recommendation apparatus 300 according to the ... of the embodiment of the present invention based on barrage.The device is for holding
Row above method flow, including：
Sentiment analysis model training module 310, for the language material according to barrage, by neural network to the word of the language material
Vector sum affective tag carries out model training, and predicts the emotion of the barrage.
User model builds module 320, the barrage for being issued in video by user, and the user is built
For that can indicate user vector of the user to the fancy grade of all kinds of videos, wherein every one-dimensional representation of the user vector
The user is to fancy grade a kind of in all kinds of videos.
Video Model build module 330, character introduction for obtaining the video by PV-DM models first to
Amount, by theme generate model obtain the video barrage each implicit theme distribution secondary vector, described first to
Amount and the secondary vector build video vector jointly.
Recommending module 340, for predicting pre- test and appraisal of the user to video according to user's similitude and Video similarity
Point, the prediction scoring is given to the user not less than the video recommendations of predetermined threshold, wherein by respectively between user
After similarity calculation score between similarity and video, the score is weighted to obtain the prediction scoring.
Optionally, it using Skip-gram model trainings word vector, is sent in and is regarded using shot and long term memory network prediction user
The feeling polarities of barrage in frequency.
Optionally, the often one-dimensional of user vector includes the score obtained according to the emotion of barrage, and score is liked for indicating
Good degree.
Optionally, the similarity between user and the similarity between video are obtained by cosine similarity respectively.
Optionally, in recommendation step, occurrence number most multiple labels when choosing viewing video in user's predetermined time,
The video collection for needing to predict is obtained according to label.
First embodiment is method embodiment corresponding with present embodiment, and present embodiment can be with first embodiment
It works in coordination implementation.The relevant technical details mentioned in first embodiment are still effective in the present embodiment, in order to reduce weight
Multiple, which is not described herein again.Correspondingly, the relevant technical details mentioned in present embodiment are also applicable in first embodiment
In.
Therefore, the recommendation apparatus based on barrage of the embodiment of the present invention can predict that user's is true according to the barrage of user
Emotional expression, the video that recommended user likes.
Further, based on the same technical idea, the embodiment of the present invention provides a kind of computing device, including storage
Device and one or more processor；Wherein, computing device further includes：One or more units, one or more unit quilts
It stores in memory and is configured to be executed by one or more processors, one or more units include following for executing
The instruction of step：
According to the language material of barrage, model training is carried out to the word vector sum affective tag of language material by neural network, and pre-
Survey the emotion of barrage；
The barrage issued in video by user, the hobby journey for being configured to user that can indicate user to all kinds of videos
The user vector of degree, wherein every one-dimensional representation user of user vector is to fancy grade a kind of in all kinds of videos；
The primary vector that the character introduction of video is obtained by PV-DM models generates model by theme and obtains video
The secondary vector of the distribution of each implicit theme of barrage, primary vector and secondary vector build video vector jointly；
It is scored the prediction of video according to user's similarity and video similarity prediction user, by prediction scoring not less than pre-
The video recommendations of threshold value are determined to user, wherein being obtained by the similarity respectively between user and the similarity calculation between video
After point, score is weighted to obtain prediction scoring.
Based on the same technical idea, the computer journey being used in combination with computing device is provided according to an embodiment of the invention
Sequence product, which is characterized in that including computer-readable storage medium and be embedded in computer program mechanism therein；Wherein,
Computer program mechanism includes executing the instruction of following steps：
According to the language material of barrage, model training is carried out to the word vector sum affective tag of language material by neural network, and pre-
Survey the emotion of barrage；
The barrage issued in video by user, the hobby journey for being configured to user that can indicate user to all kinds of videos
The user vector of degree, wherein every one-dimensional representation user of user vector is to fancy grade a kind of in all kinds of videos；
The primary vector that the character introduction of video is obtained by PV-DM models generates model by theme and obtains video
The secondary vector of the distribution of each implicit theme of barrage, primary vector and secondary vector build video vector jointly；
It is scored the prediction of video according to user's similarity and video similarity prediction user, by prediction scoring not less than pre-
The video recommendations of threshold value are determined to user, wherein being obtained by the similarity respectively between user and the similarity calculation between video
After point, score is weighted to obtain prediction scoring.
It should be appreciated that in order to simplify the disclosure and help to understand one or more of each inventive aspect, it is right above
In the description of exemplary embodiment of the present invention, each feature of the invention be grouped together into sometimes single embodiment, figure or
In person's descriptions thereof.However, the method for the disclosure should be construed to reflect following intention：I.e. claimed hair
It is bright to require features more more than the feature being expressly recited in each claim.More precisely, such as claims institute
As reflection, inventive aspect is all features less than single embodiment disclosed above.Therefore, it then follows specific embodiment party
Thus claims of formula are expressly incorporated in the specific implementation mode, wherein each claim itself is as the present invention's
Separate embodiments.
In addition, it will be appreciated by those of skill in the art that although some embodiments described herein include other embodiments
In included certain features rather than other feature, but the combination of the feature of different embodiments means in of the invention
Within the scope of and form different embodiments.For example, in detail in the claims, embodiment claimed it is one of arbitrary
It mode can use in any combination.
Although disclosed herein various aspects and embodiment, other aspects and embodiment are for those skilled in the art
For will be apparent.Various aspects and embodiment disclosed herein are for illustrative purposes, and are not intended to be limited, very
Real range should be indicated by the full scope for the equivalent that appended claims and such claim are authorized to.Also
It is appreciated that term used herein is merely to describe the purpose of specific embodiment, and be not intended to be limited.
Claims (12)
1. a kind of recommendation method based on barrage, which is characterized in that including：
Sentiment analysis model training step, according to the language material of barrage, by neural network to the word vector sum emotion of the language material
Label carries out model training, and predicts the emotion of the barrage；
The user is configured to indicate by user model construction step, the barrage issued in video by user
The user is to the user vector of the fancy grade of all kinds of videos, wherein user couple described in every one-dimensional representation of the user vector
A kind of fancy grade in all kinds of videos；
Video Model construction step is obtained the primary vector of the character introduction of the video by PV-DM models, is given birth to by theme
Obtain the secondary vector of the distribution of each implicit theme of the barrage of the video at model, the primary vector and described second to
The common structure video vector of amount；
Recommendation step predicts that the user scores to the prediction of video according to user's similarity and video similarity, will be described pre-
Test and appraisal point give the user not less than the video recommendations of predetermined threshold, wherein passing through similarity and video respectively between user
Between similarity calculation score after, the score is weighted to obtain the prediction and is scored.
2. the recommendation method according to claim 1 based on barrage, which is characterized in that use Skip-gram model trainings
The word vector predicts that the user sends the feeling polarities of the barrage in video using shot and long term memory network.
3. the recommendation method according to claim 1 based on barrage, which is characterized in that user vector includes per one-dimensional
According to the score that the emotion of the barrage obtains, the score is for indicating the fancy grade.
4. the recommendation method according to claim 1 based on barrage, which is characterized in that similarity between the user and
Similarity between the video is obtained by cosine similarity respectively.
5. the recommendation method according to claim 1 based on barrage, which is characterized in that in the recommendation step, choose institute
The most multiple labels of occurrence number, obtain according to the label and need that predicts to regard when stating viewing video in user's predetermined time
Frequency is gathered.
6. a kind of recommendation apparatus based on barrage, which is characterized in that including,
Sentiment analysis model training module, for the language material according to barrage, by neural network to the word vector sum of the language material
Affective tag carries out model training, and predicts the emotion of the barrage；
User model builds module, and the barrage for being issued in video by user, the user is configured to can
User vector of the user to the fancy grade of all kinds of videos is indicated, wherein being used described in every one-dimensional representation of the user vector
Family is to fancy grade a kind of in all kinds of videos；
Video Model builds module, and the primary vector of the character introduction for obtaining the video by PV-DM models passes through master
Topic generate model obtain the video barrage each implicit theme distribution secondary vector, the primary vector and described the
The common structure video vector of two vectors；
Recommending module, for predicting that the user scores to the prediction of video according to user's similarity and video similarity, by institute
Video recommendations of the prediction scoring not less than predetermined threshold are stated to the user, wherein by similarity respectively between user and
After similarity calculation score between video, the score is weighted to obtain the prediction scoring.
7. the recommendation apparatus according to claim 6 based on barrage, which is characterized in that use Skip-gram model trainings
The word vector predicts that the user sends the feeling polarities of the barrage in video using shot and long term memory network.
8. the recommendation apparatus according to claim 6 based on barrage, which is characterized in that user vector includes per one-dimensional
According to the score that the emotion of the barrage obtains, the score is for indicating the fancy grade.
9. the recommendation apparatus according to claim 6 based on barrage, which is characterized in that similarity between the user and
Similarity between the video is obtained by cosine similarity respectively.
10. the recommendation apparatus according to claim 6 based on barrage, which is characterized in that in the recommending module, choose institute
The most multiple labels of occurrence number, obtain according to the label and need that predicts to regard when stating viewing video in user's predetermined time
Frequency is gathered.
11. a kind of computing device, including memory and one or more processor；Wherein, the computing device also wraps
It includes：
One or more units, one or more of units are stored in the memory and are configured to by one
Or multiple processors execute, one or more of units include the instruction for executing following steps：
According to the language material of barrage, model training is carried out to the word vector sum affective tag of the language material by neural network, and pre-
Survey the emotion of the barrage；
The user is configured to that the user can be indicated to all kinds of videos by the barrage issued in video by user
Fancy grade user vector, wherein user described in every one-dimensional representation of the user vector is to a kind of in all kinds of videos
Fancy grade；
The primary vector that the character introduction of the video is obtained by PV-DM models generates model by theme and obtains described regard
The secondary vector of the distribution of each implicit theme of the barrage of frequency, the primary vector and the secondary vector build jointly video to
Amount；
It predicts that the user scores to the prediction of video according to user's similarity and video similarity, the prediction is scored not small
The user is given in the video recommendations of predetermined threshold, wherein passing through the similarity respectively between user and similar between video
After degree calculates score, the score is weighted to obtain the prediction scoring.
12. a kind of computer program product being used in combination with computing device as claimed in claim 11, which is characterized in that packet
It includes computer-readable storage medium and is embedded in computer program mechanism therein；Wherein, the computer program mechanism packet
Include the instruction for executing following steps：
According to the language material of barrage, model training is carried out to the word vector sum affective tag of the language material by neural network, and pre-
Survey the emotion of the barrage；
The user is configured to that the user can be indicated to all kinds of videos by the barrage issued in video by user
Fancy grade user vector, wherein user described in every one-dimensional representation of the user vector is to a kind of in all kinds of videos
Fancy grade；
The primary vector that the character introduction of the video is obtained by PV-DM models generates model by theme and obtains described regard
The secondary vector of the distribution of each implicit theme of the barrage of frequency, the primary vector and the secondary vector build jointly video to
Amount；
It predicts that the user scores to the prediction of video according to user's similarity and video similarity, the prediction is scored not small
The user is given in the video recommendations of predetermined threshold, wherein passing through the similarity respectively between user and similar between video
After degree calculates score, the score is weighted to obtain the prediction scoring.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810426715.1A CN108737859A (en) | 2018-05-07 | 2018-05-07 | Video recommendation method based on barrage and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810426715.1A CN108737859A (en) | 2018-05-07 | 2018-05-07 | Video recommendation method based on barrage and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108737859A true CN108737859A (en) | 2018-11-02 |
Family
ID=63937000
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810426715.1A Pending CN108737859A (en) | 2018-05-07 | 2018-05-07 | Video recommendation method based on barrage and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108737859A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109361932A (en) * | 2018-11-23 | 2019-02-19 | 武汉斗鱼网络科技有限公司 | The method that temperature prediction is broadcast live, device, equipment and medium |
| CN109743617A (en) * | 2018-12-03 | 2019-05-10 | 清华大学 | A kind of video playing jumps air navigation aid and equipment |
| CN109862397A (en) * | 2019-02-02 | 2019-06-07 | 广州虎牙信息科技有限公司 | A kind of video analysis method, apparatus, equipment and storage medium |
| CN110113673A (en) * | 2019-04-30 | 2019-08-09 | 北京奇艺世纪科技有限公司 | A kind of barrage display methods, device and electronic equipment |
| CN110263188A (en) * | 2019-05-29 | 2019-09-20 | 深圳市元征科技股份有限公司 | Media data methods of marking, device and equipment |
| CN110267111A (en) * | 2019-05-24 | 2019-09-20 | 平安科技（深圳）有限公司 | Video barrage analysis method, device and storage medium, computer equipment |
| CN111050193A (en) * | 2019-11-12 | 2020-04-21 | 汉口北进出口服务有限公司 | User portrait construction method and device, computer equipment and storage medium |
| CN111368204A (en) * | 2020-03-09 | 2020-07-03 | 北京字节跳动网络技术有限公司 | Content pushing method and device, electronic equipment and computer readable medium |
| CN111708901A (en) * | 2020-06-19 | 2020-09-25 | 腾讯科技（深圳）有限公司 | Multimedia resource recommendation method and device, electronic equipment and storage medium |
| CN111860237A (en) * | 2020-07-07 | 2020-10-30 | 中国科学技术大学 | Video emotion fragment identification method and device |
| CN112231579A (en) * | 2019-12-30 | 2021-01-15 | 北京邮电大学 | Social video recommendation system and method based on implicit community discovery |
| CN112632277A (en) * | 2020-12-15 | 2021-04-09 | 五八同城信息技术有限公司 | Resource processing method and device for target content object |
| CN112749296A (en) * | 2019-10-31 | 2021-05-04 | 北京达佳互联信息技术有限公司 | Video recommendation method and device, server and storage medium |
| CN113766281A (en) * | 2021-09-10 | 2021-12-07 | 北京快来文化传播集团有限公司 | Short video recommendation method, electronic device and computer-readable storage medium |
| CN114064974A (en) * | 2021-11-15 | 2022-02-18 | 腾讯科技（深圳）有限公司 | Information processing method, information processing apparatus, electronic device, storage medium, and program product |
| CN116781949A (en) * | 2023-07-26 | 2023-09-19 | 新励成教育科技股份有限公司 | Recommendation method, system, equipment and storage medium for talent lecture live broadcast |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7685022B1 (en) * | 2007-06-15 | 2010-03-23 | Amazon Technologies, Inc. | System and method of offering media content |
| CN104504059A (en) * | 2014-12-22 | 2015-04-08 | 合一网络技术(北京)有限公司 | Multimedia resource recommending method |
| CN105095508A (en) * | 2015-08-31 | 2015-11-25 | 北京奇艺世纪科技有限公司 | Multimedia content recommendation method and multimedia content recommendation apparatus |
| CN106454422A (en) * | 2016-07-01 | 2017-02-22 | 江苏省公用信息有限公司 | Fingerprint recognition-based IPTV program recommending method and apparatus |
| CN106952111A (en) * | 2017-02-27 | 2017-07-14 | 东软集团股份有限公司 | Personalized recommendation method and device |
| CN107038480A (en) * | 2017-05-12 | 2017-08-11 | 东华大学 | A kind of text sentiment classification method based on convolutional neural networks |
-
2018
- 2018-05-07 CN CN201810426715.1A patent/CN108737859A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7685022B1 (en) * | 2007-06-15 | 2010-03-23 | Amazon Technologies, Inc. | System and method of offering media content |
| CN104504059A (en) * | 2014-12-22 | 2015-04-08 | 合一网络技术(北京)有限公司 | Multimedia resource recommending method |
| CN105095508A (en) * | 2015-08-31 | 2015-11-25 | 北京奇艺世纪科技有限公司 | Multimedia content recommendation method and multimedia content recommendation apparatus |
| CN106454422A (en) * | 2016-07-01 | 2017-02-22 | 江苏省公用信息有限公司 | Fingerprint recognition-based IPTV program recommending method and apparatus |
| CN106952111A (en) * | 2017-02-27 | 2017-07-14 | 东软集团股份有限公司 | Personalized recommendation method and device |
| CN107038480A (en) * | 2017-05-12 | 2017-08-11 | 东华大学 | A kind of text sentiment classification method based on convolutional neural networks |
Non-Patent Citations (2)
| Title |
|---|
| 邓扬 等: "基于弹幕情感分析的视频片段推荐模型", 《计算机应用》 * |
| 雷鸣 等: "情感分析在电影推荐系统中的应用", 《计算机工程与应用》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109361932A (en) * | 2018-11-23 | 2019-02-19 | 武汉斗鱼网络科技有限公司 | The method that temperature prediction is broadcast live, device, equipment and medium |
| CN109361932B (en) * | 2018-11-23 | 2021-01-01 | 武汉斗鱼网络科技有限公司 | Live broadcast heat prediction method, device, equipment and medium |
| CN109743617A (en) * | 2018-12-03 | 2019-05-10 | 清华大学 | A kind of video playing jumps air navigation aid and equipment |
| CN109743617B (en) * | 2018-12-03 | 2020-11-24 | 清华大学 | Skip navigation method and device for video playing |
| CN109862397A (en) * | 2019-02-02 | 2019-06-07 | 广州虎牙信息科技有限公司 | A kind of video analysis method, apparatus, equipment and storage medium |
| CN109862397B (en) * | 2019-02-02 | 2021-11-09 | 广州虎牙信息科技有限公司 | Video analysis method, device, equipment and storage medium |
| CN110113673A (en) * | 2019-04-30 | 2019-08-09 | 北京奇艺世纪科技有限公司 | A kind of barrage display methods, device and electronic equipment |
| CN110267111A (en) * | 2019-05-24 | 2019-09-20 | 平安科技（深圳）有限公司 | Video barrage analysis method, device and storage medium, computer equipment |
| CN110263188A (en) * | 2019-05-29 | 2019-09-20 | 深圳市元征科技股份有限公司 | Media data methods of marking, device and equipment |
| CN110263188B (en) * | 2019-05-29 | 2023-03-28 | 深圳市元征科技股份有限公司 | Media data scoring method, device and equipment |
| CN112749296A (en) * | 2019-10-31 | 2021-05-04 | 北京达佳互联信息技术有限公司 | Video recommendation method and device, server and storage medium |
| CN112749296B (en) * | 2019-10-31 | 2024-01-26 | 北京达佳互联信息技术有限公司 | Video recommendation method, device, server and storage medium |
| CN111050193B (en) * | 2019-11-12 | 2022-06-10 | 汉口北进出口服务有限公司 | User portrait construction method and device, computer equipment and storage medium |
| CN111050193A (en) * | 2019-11-12 | 2020-04-21 | 汉口北进出口服务有限公司 | User portrait construction method and device, computer equipment and storage medium |
| CN112231579A (en) * | 2019-12-30 | 2021-01-15 | 北京邮电大学 | Social video recommendation system and method based on implicit community discovery |
| CN112231579B (en) * | 2019-12-30 | 2022-10-28 | 北京邮电大学 | Social video recommendation system and method based on implicit community discovery |
| CN111368204A (en) * | 2020-03-09 | 2020-07-03 | 北京字节跳动网络技术有限公司 | Content pushing method and device, electronic equipment and computer readable medium |
| CN111708901A (en) * | 2020-06-19 | 2020-09-25 | 腾讯科技（深圳）有限公司 | Multimedia resource recommendation method and device, electronic equipment and storage medium |
| CN111708901B (en) * | 2020-06-19 | 2023-10-13 | 腾讯科技（深圳）有限公司 | Multimedia resource recommendation method and device, electronic equipment and storage medium |
| CN111860237A (en) * | 2020-07-07 | 2020-10-30 | 中国科学技术大学 | Video emotion fragment identification method and device |
| CN111860237B (en) * | 2020-07-07 | 2022-09-06 | 中国科学技术大学 | Video emotion fragment identification method and device |
| CN112632277A (en) * | 2020-12-15 | 2021-04-09 | 五八同城信息技术有限公司 | Resource processing method and device for target content object |
| CN113766281A (en) * | 2021-09-10 | 2021-12-07 | 北京快来文化传播集团有限公司 | Short video recommendation method, electronic device and computer-readable storage medium |
| CN114064974A (en) * | 2021-11-15 | 2022-02-18 | 腾讯科技（深圳）有限公司 | Information processing method, information processing apparatus, electronic device, storage medium, and program product |
| CN116781949A (en) * | 2023-07-26 | 2023-09-19 | 新励成教育科技股份有限公司 | Recommendation method, system, equipment and storage medium for talent lecture live broadcast |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108737859A (en) | Video recommendation method based on barrage and device | |
| Anastasopoulos et al. | Machine learning for public administration research, with application to organizational reputation | |
| Salminen et al. | Creating and detecting fake reviews of online products | |
| CN108363804B (en) | Local model weighted fusion Top-N movie recommendation method based on user clustering | |
| Bagheri et al. | ADM-LDA: An aspect detection model based on topic modelling using the structure of review sentences | |
| Baesens et al. | Neural network survival analysis for personal loan data | |
| Potha et al. | Cyberbullying detection using time series modeling | |
| CN103164463B (en) | Method and device for recommending labels | |
| CN111061946A (en) | Scenario content recommendation method and device, electronic equipment and storage medium | |
| US20190354887A1 (en) | Knowledge graph based learning content generation | |
| CN112163165A (en) | Information recommendation method, device, equipment and computer readable storage medium | |
| Li et al. | Identifying manipulated offerings on review portals | |
| KR102468827B1 (en) | Method, device and system for sourcing products based on artificial intelligence and providing category-customized review analysis service | |
| US11416884B2 (en) | Personality trait-based customer behavior prediction | |
| CN110851621B (en) | Method, device and storage medium for predicting video highlight level based on knowledge graph | |
| CN112288042A (en) | Updating method and device of behavior prediction system, storage medium and computing equipment | |
| CN110390052A (en) | Search for recommended method, the training method of CTR prediction model, device and equipment | |
| CN113742567B (en) | Recommendation method and device for multimedia resources, electronic equipment and storage medium | |
| CN110110372B (en) | Automatic segmentation prediction method for user time sequence behavior | |
| Phan et al. | An approach for a decision-making support system based on measuring the user satisfaction level on twitter | |
| CN116738066B (en) | Rural travel service recommendation method and device, electronic equipment and storage medium | |
| Kaghazgaran et al. | Wide-ranging review manipulation attacks: Model, empirical study, and countermeasures | |
| Bauman et al. | Recommending Items with Conditions Enhancing User Experiences Based on Sentiment Analysis of Reviews. | |
| CN115730125A (en) | Object identification method and device, computer equipment and storage medium | |
| Barbosa et al. | Predicting portuguese steam review helpfulness using artificial neural networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181102 |