CN109325146A - A kind of video recommendation method, device, storage medium and server - Google Patents
A kind of video recommendation method, device, storage medium and server Download PDFInfo
- Publication number
- CN109325146A CN109325146A CN201811339832.0A CN201811339832A CN109325146A CN 109325146 A CN109325146 A CN 109325146A CN 201811339832 A CN201811339832 A CN 201811339832A CN 109325146 A CN109325146 A CN 109325146A
- Authority
- CN
- China
- Prior art keywords
- video
- candidate
- user
- list
- theme
- 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
Abstract
The present invention provides a kind of video recommendation method, device, storage medium and servers, comprising: obtains the corresponding text information of video that the user that user pays close attention to or described is currently played；Model is generated according to the corresponding text information of the video and preparatory trained theme, determines the theme of the video；Video identical with the theme is searched in video database as candidate video, constructs video candidate list；Obtain the user comment information of candidate video described in the video candidate list；According to the user comment information of the candidate video, the candidate video in the candidate video list is screened, obtains the recommendation list of videos of the video；Recommend video to the user according to the recommendation list of videos.The present invention can carry out individualized video recommendation, effectively meet user demand, enhance user experience.
Description
Technical field
The present invention relates to technical field of information processing more particularly to a kind of video recommendation method, device, storage medium kimonos
Business device.
Background technique
Currently, many video applications (Application, APP) all can recommend video content to user.The prior art
In, the video content that most of APP recommend is based on the selected theme of APP oneself, that is, the video content recommended to all users
It is all identical.But since the user group of APP would generally be different, identical recommendation does not have specific aim to user, thus
Cause the video content recommended that cannot meet the needs of users, expends the search time of user, poor user experience.
Summary of the invention
The embodiment of the invention provides a kind of video recommendation method, device, storage medium and servers, to solve existing skill
In art, since the user group of APP would generally be different, identical recommendation does not have specific aim to user, so as to cause recommendation
Video content cannot meet the needs of users, the search time of user is expended, the problem of poor user experience.
The first aspect of the embodiment of the present invention provides a kind of video recommendation method, comprising:
Obtain the corresponding text information of video that the user that user pays close attention to or described is currently played；
Model is generated according to the corresponding text information of the video and preparatory trained theme, determines the master of the video
Topic；
Video identical with the theme is searched in video database as candidate video, constructs video candidate list；
Obtain the user comment information of candidate video described in the video candidate list；
According to the user comment information of the candidate video, the candidate video in the candidate video list is carried out
Screening, obtains the recommendation list of videos of the video；
Recommend video to the user according to the recommendation list of videos.
The second aspect of the embodiment of the present invention provides a kind of video recommendations device, comprising:
Information acquisition unit, the corresponding text of video being currently played for obtaining the user that user pays close attention to or described
This information；
Theme determination unit, for generating mould according to the corresponding text information of the video and preparatory trained theme
Type determines the theme of the video；
Candidate list construction unit, for searching video identical with the theme in video database as candidate view
Frequently, video candidate list is constructed；
Comment information acquiring unit, the user comment for obtaining candidate video described in the video candidate list are believed
Breath；
Recommendation list determination unit arranges the candidate video for the user comment information according to the candidate video
The candidate video in table is screened, and the recommendation list of videos of the video is obtained；
Video recommendations unit, for recommending video to the user according to the recommendation list of videos.
The third aspect of the embodiment of the present invention provides a kind of server, including memory and processor, the storage
Device is stored with the computer program that can be run on the processor, and the processor is realized such as when executing the computer program
Lower step:
Obtain the corresponding text information of video that the user that user pays close attention to or described is currently played；
Model is generated according to the corresponding text information of the video and preparatory trained theme, determines the master of the video
Topic；
Video identical with the theme is searched in video database as candidate video, constructs video candidate list；
Obtain the user comment information of candidate video described in the video candidate list；
According to the user comment information of the candidate video, the candidate video in the candidate video list is carried out
Screening, obtains the recommendation list of videos of the video；
Recommend video to the user according to the recommendation list of videos.
The fourth aspect of the embodiment of the present invention provides a kind of computer readable storage medium, the computer-readable storage
Media storage has computer program, and the computer program realizes following steps when being executed by processor:
Obtain the corresponding text information of video that the user that user pays close attention to or described is currently played；
Model is generated according to the corresponding text information of the video and preparatory trained theme, determines the master of the video
Topic；
Video identical with the theme is searched in video database as candidate video, constructs video candidate list；
Obtain the user comment information of candidate video described in the video candidate list；
According to the user comment information of the candidate video, the candidate video in the candidate video list is carried out
Screening, obtains the recommendation list of videos of the video；
Recommend video to the user according to the recommendation list of videos.
It is corresponding by obtaining the video that the user that user pays close attention to or described is currently played in the embodiment of the present invention
Text information generates model according to the corresponding text information of the video and preparatory trained theme, determines the video
Then theme searches video identical with the theme as candidate video in video database, constructs video candidate list,
The user comment information for obtaining candidate video described in the video candidate list again, according to the user comment of the candidate video
Information screens the candidate video in the candidate video list, obtains the recommendation list of videos of the video, most
Video is recommended to the user according to the recommendation list of videos afterwards, personalized video recommendations are realized, for each user's
Demand recommends video, and recommendation is not single and can effectively meet the needs of users, so that the time that user voluntarily searches for is saved,
Effectively enhancing user experience improves user's viscosity.
Detailed description of the invention
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to embodiment or description of the prior art
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description is only of the invention some
Embodiment for those of ordinary skill in the art without any creative labor, can also be according to these
Attached drawing obtains other attached drawings.
Fig. 1 is the implementation flow chart of video recommendation method provided in an embodiment of the present invention；
Fig. 2 is the specific implementation flow chart of video recommendation method S102 provided in an embodiment of the present invention；
Fig. 3 is a kind of specific implementation flow chart of video recommendation method S105 provided in an embodiment of the present invention；
Fig. 4 is another specific implementation flow chart of video recommendation method S105 provided in an embodiment of the present invention；
Fig. 5 is the structural block diagram of video recommendations device provided in an embodiment of the present invention；
Fig. 6 is the schematic diagram of server provided in an embodiment of the present invention.
Specific embodiment
In order to make the invention's purpose, features and advantages of the invention more obvious and easy to understand, below in conjunction with the present invention
Attached drawing in embodiment, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that disclosed below
Embodiment be only a part of the embodiment of the present invention, and not all embodiment.Based on the embodiments of the present invention, this field
Those of ordinary skill's all other embodiment obtained without making creative work, belongs to protection of the present invention
Range.
Fig. 1 shows the implementation process of video recommendation method provided in an embodiment of the present invention, and this method process includes step
S101 to S106.The specific implementation principle of each step is as follows:
S101: the corresponding text information of video that the user that user pays close attention to or described is currently played is obtained.
Specifically, the corresponding text information of the video includes but is not limited to that the heading message of video and video profile are believed
Breath.In embodiments of the present invention, the video that user is currently played in client is obtained, alternatively, obtaining the account that user logs in
Family obtains the video of user's concern according to the account.
Optionally, if it is corresponding to obtain the video that user's the last time plays currently without video being played on by user
Text information.Specifically, the history for obtaining user in client plays record, obtains user according to play time and broadcasts the last time
The text information for the video put.
S102: model is generated according to the corresponding text information of the video and preparatory trained theme, determines the view
The theme of frequency.
In the present embodiment, the theme, which generates model, can be LDA model.LDA(Latent Dirichlet
Allocation be) that a kind of document subject matter generates model, also referred to as three layers of bayesian probability model, comprising word, theme and
Document three-decker.Document obeys multinomial distribution to theme, and theme to word obeys multinomial distribution.LDA model (the document
Theme generates model) for determining the theme probability of the video according to the keyword of the video.Specifically, by the video
Corresponding text information is input in preparatory trained LDA model, exports institute according to the text information by the LDA model
State the theme of video.
As an embodiment of the present invention, as shown in Fig. 2, video recommendation method step provided in an embodiment of the present invention
The specific implementation flow of S102, details are as follows:
A1: word segmentation processing is carried out to the text information of the video, obtains each participle for constituting the text information.Tool
Body, stammerer participle can be used, word segmentation processing is carried out to the text information.To the text information carry out word segmentation processing it
Before, to save memory space and improving participle efficiency, the text information is pre-processed, stop words, stop words packet are removed
Fullstop, comma, branch etc. are included, that is, screens out the punctuation marks such as fullstop, comma, the branch in the text information.
A2: word vocabulary table is constructed according to each participle.
A3: the extracting keywords from the word vocabulary table.Specifically, institute can be determined according to preset keyword thesaurus
State the quantity of the keyword in participle.
A4: the keyword is input to trained theme in advance and generates model, exports the theme of the video.
In embodiments of the present invention, the corresponding text information of the video is subjected to word segmentation processing, and obtains the text
Keyword in information obtains and the corresponding text information of the video is input in preparatory trained LDA model, by described
LDA model exports the theme probability of the video according to the keyword, according to the master of video described in the theme determine the probability
Topic, the theme of the video can rapidly and accurately be determined by generating model using trained theme.
Optionally, in embodiments of the present invention, the training step of the trained topic model in advance includes:
(1), the corresponding text information of Sample video of known video theme is obtained, text information includes video title and view
Frequency brief introduction；
(2), word segmentation processing is carried out to the text information using stammerer participle, and removes stop words, screened out fullstop, tease
Number, the punctuation marks such as branch, obtain participle vocabulary；
(3), it according to the TF_IDF matrix based on the training of wikipedia news corpus, is extracted from the participle vocabulary
Kernel keyword establishes lists of keywords；
(4) according to the lists of keywords of the Sample video of multiple known video themes training LDA model, the LDA is determined
The optimized parameter of model, until the theme weight of the theme video corresponding with the lists of keywords of input of LDA model output
It closes.
As an embodiment of the present invention, the specific implementation of video recommendation method step A3 provided in an embodiment of the present invention
Process, details are as follows:
A31: the key coefficient of each participle in the word vocabulary table is calculated according to the following formula:
Key-ratio=TF × IDF (1)
Wherein, key-ratio is the key coefficient, and TF is word frequency of the participle in the word vocabulary table, and IDF is indicated
The word frequency segmented in other texts, the IDF are determined according to trained word frequency inverse file frequency matrix.Specifically, it obtains
The news corpus for taking appointed website such as wikipedia website carries out stammerer participle to news corpus by text, removes stop words, obtain
To vocabulary, frequency matrix is generated based on the vocabulary, wherein horizontal axis is document, and the longitudinal axis is keyword, according to the word frequency
Matrix determines the IDF.
A32: each participle in the word vocabulary table is arranged from high to low by the key coefficient.
A33: the participle that specified quantity is successively chosen according to rank results is determined as the keyword of the word vocabulary table.
In embodiments of the present invention, it is determined in the participle for constituting the corresponding text information of the video according to keyword coefficient
Keyword.
S103: video identical with the theme is searched in video database as candidate video, it is candidate to construct video
List.
Specifically, according to the theme of the video determined in step S102, the video of video database is screened,
Video identical with the theme of the video is determined as candidate video, constructs video candidate list.The video candidate list
In include candidate video video number.In the embodiment of the present invention, by building video candidate list in video database
Video carries out primary screening, to go out to meet the video of user demand from Effective selection in video in the sea, improves video recommendations
Efficiency.
S104: the user comment information of candidate video described in the video candidate list is obtained.
Specifically, the video in video database has corresponding user comment information, and the user comment information of video includes
At least one watched the comment of the user of the video, and the user comment information includes favorable comment, difference is commented and neutral comment.?
It include user comment information collection in the video database in the embodiment of the present invention, the user comment information of same video is stored
It is concentrated in same user comment information, uses video number as the label of the user comment information collection of same video.According to institute
The video number of candidate video in video candidate list is stated, the user for obtaining candidate video described in the video candidate list comments
By information,
S105: according to the user comment information of the candidate video, to the candidate view in the candidate video list
Frequency is screened, and the recommendation list of videos of the video is obtained.
In embodiments of the present invention, the user comment information for obtaining same candidate video, according to the use of the candidate video
Family comment information determines the recommendation list of videos of the video.Optionally, count the user comment information favorable comment quantity and
Neutral number of reviews determines the positive rating of the candidate video, according to the positive rating of the candidate video to the candidate video
The candidate video in list is screened, and the candidate video that positive rating is not reached to default positive rating screens out, according to favorable comment
The candidate video that rate reaches default positive rating, which is established, recommends list of videos.
As an embodiment of the present invention, Fig. 3 shows video recommendation method S105's provided in an embodiment of the present invention
Specific implementation flow, details are as follows:
B1: the user comment information of the video is obtained.
B2: according to the user comment information of the candidate video and the user comment information of the video, described in calculating
The similarity factor of candidate video described in candidate video list and the video.
It is optionally, described according to the user comment information of the candidate video and the user comment information of the video,
Calculate the similarity factor of candidate video and the video described in the candidate video list, comprising:
The similarity factor of candidate video described in the candidate list Yu the video is calculated according to the following formula:
Wherein, r indicates similarity factor, constructs commenting for the candidate video according to the user comment information of the candidate video
By vector, C indicates the comment vector of the candidate video, constructs commenting for the video according to the user comment information of the video
By vector, V indicates the comment vector of the video, and i indicates i-th of user, and the vector dimension of C and V are N, and N indicates user
Number, CiIndicate i-th of user to the comment vector of the candidate video, ViIndicate i-th of user to the comment of the video to
Amount.
Optionally, quantify user comment information, user comment information includes favorable comment, difference is commented and neutral comment, foundation are commented
Quantify table by score value, the comment score value quantization table includes the mapping relations of user comment and score value, quantifies table from the comment
The middle corresponding user comment score value of user comment information for searching the candidate video, according to the user comment of the candidate video
The corresponding user comment score value of information constructs the comment vector C of the candidate video；Described in being searched from comment quantization table
The corresponding user comment score value of the user comment information of video, according to the corresponding user comment of the user comment information of the video
Score value constructs the comment vector V of the video.The dimension of the comment vector of the candidate video and the quantity of user comment information
Correlation, for example, the dimension of the comment vector V of the candidate video is if the candidate video has 100 user comment informations
1*100。
Optionally, according to the word frequency building comment vector of keyword in the user comment information.Specifically, i-th is searched
A user calculates i-th of user to the user comment of the candidate video to the user comment information of the candidate video
The word frequency that keyword is specified in information, constructs the candidate video of the user i according to the word frequency of the designated key word
Comment on vector Ci.Similar, i-th of user is searched to the user comment information of the video, calculates i-th of user to institute
The word frequency that keyword is specified in the user comment information of video is stated, according to described i-th of the word frequency of designated key word building
The comment vector V of the candidate video of useri。
B3: according to the similarity factor, the candidate video in the candidate video list is screened, the view is obtained
The recommendation list of videos of frequency.
In embodiments of the present invention, the similarity factor is for identifying candidate video journey similar to the video
Degree.Similarity factor and similarity degree are positively correlated, and similarity factor is bigger, and the similarity degree of video is higher.According to the similarity factor,
Candidate video in the candidate video list is screened, the candidate video that similarity factor is lower than default similarity factor is sieved
It removes, retains the candidate video that similarity factor is greater than or equal to default similarity factor, establish the recommendation list of videos of the video.
As an embodiment of the present invention, Fig. 4 shows video recommendation method S105's provided in an embodiment of the present invention
Specific implementation flow, details are as follows:
C1: it according to the user comment information and trained neural network model, obtains in the candidate video list
Candidate video user scoring.
C2: it is scored according to the user, the candidate video in the candidate video list is screened, the view is obtained
The recommendation list of videos of frequency.
In embodiments of the present invention, the candidate video in the candidate video list is screened according to user's scoring,
Candidate video by user's scoring lower than default scoring is screened out from the candidate video list, and user's scoring is greater than or equal to
The candidate video of default scoring retains, to obtain the recommendation list of videos of the video.
Specifically, the rating matrix of user comment information is constructed, horizontal axis is user, and the longitudinal axis is candidate video list.It is described
User's scoring is obtained by trained neural network model DNN.In embodiments of the present invention, the neural network model
Training step is as follows:
(1), sample of users comment data collection is obtained, the sample of users that the sample of users comment data is concentrated comments on label
There is the corresponding user's scoring manually marked；
(2), data prediction is carried out to the sample of users comment that the sample of users comment data is concentrated.Specifically, right
Every sample user comment carries out a point processing, removes stop words, and by TF_IDF matrix extracting keywords, obtains each sample
The lists of keywords of user comment；
(3), input vector is generated according to the lists of keywords, specifically, based on the specified network platform such as Wiki hundred
The term vector model that section is trained, term vector model include specified dimension, carry out word insertion to the lists of keywords to get arriving
The term vector of same dimension, as input vector.
(4), neural network model is constructed, is input with the input vector, is scored using the user manually marked as defeated
Out, model training is carried out, until determining the optimized parameter of the neural network model, obtains trained neural network model.
Optionally, parameter learning is carried out using trellis traversal method, determines the optimized parameter of the neural network model, mainly traverses
Index has: learning rate r, frequency of training epoch_num, lot number amount batch_size, termination error expect_loss etc.；Model
The condition that training terminates has following: 1, frequency of training reaches certain number；2, error has had arrived at specified index.
Illustratively, 4 layers of neural network model are constructed in the embodiment of the present invention, including input layer, two hidden layers and defeated
Layer out, in which:
Input layer a: node, the vector of 10 dimension *, 256 dimension, the vector as output layer；(256 dimensions are as word
The dimension of vector)；
1:100 node of hidden layer, 1 dimension *, 100 dimension, activation primitive are relu function；
2:200 node of hidden layer, 1*200 dimension, activation primitive are relu function；
Output layer: 10 nodes (assuming that having 10 sample of users scorings), dimension is 1 dimension, and activation primitive is
Logistics function.
Optionally, in embodiments of the present invention, the candidate in the candidate video list is regarded according to the similarity factor
Frequency carries out the first minor sort, carries out first time sieve to the candidate video in the candidate video list according to first time ranking results
Choosing will screen the candidate video retained for the first time and be determined as the first candidate video list, and first candidate video includes not
An only candidate video, and the candidate video in the first candidate video list is arranged from high to low by similarity factor
Sequence.Further according to user scoring to by screening the progress of the candidate video in the first obtained candidate video list for the first time
Postsearch screening constructs the recommendation list of videos of the video according to the candidate video that programmed screening retains, wherein described to push away
The candidate video recommended in list of videos is ranked up from high to low according to user scoring.
In the embodiment of the present invention, by carrying out screening and sequencing twice to the candidate video in the candidate video list, obtain
Recommendation list of videos is taken, so that the candidate video recommended in list of videos is high-quality and meets user demand.
S106: video is recommended to the user according to the recommendation list of videos.
In embodiments of the present invention, it is carried out according to the candidate video to have sorted in the recommendation list of videos to the user
Recommend.
It is corresponding by obtaining the video that the user that user pays close attention to or described is currently played in the embodiment of the present invention
Text information generates model according to the corresponding text information of the video and preparatory trained theme, determines the video
Then theme searches video identical with the theme as candidate video in video database, constructs video candidate list,
The user comment information for obtaining candidate video described in the video candidate list again, according to the user comment of the candidate video
Information screens the candidate video in the candidate video list, obtains the recommendation list of videos of the video, most
Video is recommended to the user according to the recommendation list of videos afterwards, personalized video recommendations are realized, for each user's
Demand recommends video, and recommendation is not single and can effectively meet the needs of users, so that the time that user voluntarily searches for is saved,
Effectively enhancing user experience improves user's viscosity.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process
Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit
It is fixed.
Corresponding to video recommendation method described in foregoing embodiments, Fig. 5 shows video provided by the embodiments of the present application and pushes away
The structural block diagram of device is recommended, for ease of description, illustrates only part relevant to the embodiment of the present application.
Referring to Fig. 5, which includes: information acquisition unit 51, theme determination unit 52, candidate list building
Unit 53, comment information acquiring unit 54, recommendation list determination unit 55, video recommendations unit 56, in which:
Information acquisition unit 51, it is corresponding for obtaining the video that the user that user pays close attention to or described is currently played
Text information；
Theme determination unit 52, for generating mould according to the corresponding text information of the video and preparatory trained theme
Type determines the theme of the video；
Candidate list construction unit 53, for searching video identical with the theme in video database as candidate
Video constructs video candidate list；
Comment information acquiring unit 54, the user comment for obtaining candidate video described in the video candidate list are believed
Breath；
Recommendation list determination unit 55, for the user comment information according to the candidate video, to the candidate video
The candidate video in list is screened, and the recommendation list of videos of the video is obtained；
Video recommendations unit 56, for recommending video to the user according to the recommendation list of videos.
Optionally, the theme determination unit 52 includes:
Determining module is segmented, word segmentation processing is carried out for the text information to the video, obtains and constitute the text envelope
Each participle of breath；
Vocabulary constructs module, for constructing word vocabulary table according to each participle；
Keyword abstraction module, for the extracting keywords from the word vocabulary table；
Theme determining module generates model for the keyword to be input to trained theme in advance, described in output
The theme of video.
Optionally, the keyword abstraction module includes:
Key coefficient computational submodule, for calculating the key of each participle in the word vocabulary table according to the following formula
Coefficient:
Key-ratio=TF × IDF,
Wherein, key-ratio is the key coefficient, and TF is word frequency of the participle in the word vocabulary table, and IDF is indicated
The word frequency segmented in other texts, the IDF are determined according to trained word frequency inverse file frequency matrix；
Sorting sub-module, for arranging each participle in the word vocabulary table from high to low by the key coefficient
Column；
Keyword specifies submodule, and the participle for successively choosing specified quantity according to rank results is determined as the word
The keyword of vocabulary.
Optionally, the recommendation list determination unit 55 includes:
User comment information obtains module, for obtaining the user comment information of the video.
Similarity factor computing module, for according to the user comment information of the candidate video and the user of the video
Comment information calculates the similarity factor of candidate video and the video described in the candidate video list；
First recommendation list determining module is used for according to the similarity factor, to the candidate in the candidate video list
Video is screened, and the recommendation list of videos of the video is obtained.
Optionally, the similarity factor computing module includes:
Similarity factor computational submodule, for calculating candidate video described in the candidate list and institute according to the following formula
State the similarity factor of video:
Wherein, r indicates similarity factor, constructs commenting for the candidate video according to the user comment information of the candidate video
By vector, the comment vector of candidate video described in C table constructs the comment of the video according to the user comment information of the video
Vector, V indicate the comment vector of the video, and i indicates i-th of user, and the vector dimension of C and V are N, and N indicates number of users,
CiIndicate i-th of user to the comment vector of the candidate video, ViIndicate i-th of user to the comment vector of the video.
Optionally, the recommendation list determination unit 55 includes:
User, which scores, obtains module, for obtaining according to the user comment information and trained neural network model
The user of candidate video in the candidate video list scores；
Second recommendation list determining module, for being scored according to the user, to the candidate in the candidate video list
Video is screened, and the recommendation list of videos of the video is obtained.
It is corresponding by obtaining the video that the user that user pays close attention to or described is currently played in the embodiment of the present invention
Text information generates model according to the corresponding text information of the video and preparatory trained theme, determines the video
Then theme searches video identical with the theme as candidate video in video database, constructs video candidate list,
The user comment information for obtaining candidate video described in the video candidate list again, according to the user comment of the candidate video
Information screens the candidate video in the candidate video list, obtains the recommendation list of videos of the video, most
Video is recommended to the user according to the recommendation list of videos afterwards, personalized video recommendations are realized, for each user's
Demand recommends video, and recommendation is not single and can effectively meet the needs of users, so that the time that user voluntarily searches for is saved,
Effectively enhancing user experience improves user's viscosity.
Fig. 6 is the schematic diagram for the server that one embodiment of the invention provides.As shown in fig. 6, the server 6 of the embodiment wraps
It includes: processor 60, memory 61 and being stored in the computer that can be run in the memory 61 and on the processor 60
Program 62, such as video recommendations program.The processor 60 realizes that above-mentioned each video pushes away when executing the computer program 62
Recommend the step in embodiment of the method, such as step 101 shown in FIG. 1 is to 106.Alternatively, the processor 60 executes the calculating
The function of each module/unit in above-mentioned each Installation practice, such as the function of unit 51 to 56 shown in Fig. 5 are realized when machine program 62
Energy.
Illustratively, the computer program 62 can be divided into one or more module/units, it is one or
Multiple module/units are stored in the memory 61, and are executed by the processor 60, to complete the present invention.Described one
A or multiple module/units can be the series of computation machine program instruction section that can complete specific function, which is used for
Implementation procedure of the computer program 62 in the server 6 is described.
The server 6 can be desktop PC, notebook, palm PC and cloud server etc. and calculate equipment.
The server may include, but be not limited only to, processor 60, memory 61.It will be understood by those skilled in the art that Fig. 6 is only
It is the example of server 6, does not constitute the restriction to server 6, may include than illustrating more or fewer components or group
Close certain components or different components, for example, the server can also include input-output equipment, network access equipment,
Bus etc..
The processor 60 can be central processing unit (Central Processing Unit, CPU), can also be
Other general processors, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor
Deng.
The memory 61 can be the internal storage unit of the server 6, such as the hard disk or memory of server 6.
The memory 61 is also possible to the External memory equipment of the server 6, such as the plug-in type being equipped on the server 6 is hard
Disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card
(Flash Card) etc..Further, the memory 61 can also both include the internal storage unit of the server 6 or wrap
Include External memory equipment.The memory 61 is for other programs needed for storing the computer program and the server
And data.The memory 61 can be also used for temporarily storing the data that has exported or will export.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated module/unit be realized in the form of SFU software functional unit and as independent product sale or
In use, can store in a computer readable storage medium.Based on this understanding, the present invention realizes above-mentioned implementation
All or part of the process in example method, can also instruct relevant hardware to complete, the meter by computer program
Calculation machine program can be stored in a computer readable storage medium, the computer program when being executed by processor, it can be achieved that on
The step of stating each embodiment of the method.Wherein, the computer program includes computer program code, the computer program generation
Code can be source code form, object identification code form, executable file or certain intermediate forms etc..The computer-readable medium
It may include: any entity or device, recording medium, USB flash disk, mobile hard disk, magnetic that can carry the computer program code
Dish, CD, computer storage, read-only memory (ROM, Read-Only Memory), random access memory (RAM,
Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..It should be noted that described
The content that computer-readable medium includes can carry out increasing appropriate according to the requirement made laws in jurisdiction with patent practice
Subtract, such as does not include electric carrier signal and electricity according to legislation and patent practice, computer-readable medium in certain jurisdictions
Believe signal.
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations；Although referring to aforementioned reality
Applying example, invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each
Technical solution documented by embodiment is modified or equivalent replacement of some of the technical features；And these are modified
Or replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all
It is included within protection scope of the present invention.
Claims (10)
1. a kind of video recommendation method characterized by comprising
Obtain the corresponding text information of video that the user that user pays close attention to or described is currently played；
Model is generated according to the corresponding text information of the video and preparatory trained theme, determines the theme of the video；
Video identical with the theme is searched in video database as candidate video, constructs video candidate list；
Obtain the user comment information of candidate video described in the video candidate list；
According to the user comment information of the candidate video, the candidate video in the candidate video list is sieved
Choosing, obtains the recommendation list of videos of the video；
Recommend video to the user according to the recommendation list of videos.
2. video recommendation method according to claim 1, which is characterized in that described according to the corresponding text envelope of the video
Breath generates model with trained theme in advance, determines the theme of the video, comprising:
Word segmentation processing is carried out to the text information of the video, obtains each participle for constituting the text information；
Word vocabulary table is constructed according to each participle；
The extracting keywords from the word vocabulary table；
The keyword is input to trained theme in advance and generates model, exports the theme of the video.
3. video recommendation method according to claim 2, which is characterized in that described extract from the word vocabulary table is closed
Keyword, comprising:
The key coefficient of each participle in the word vocabulary table is calculated according to the following formula:
Key-ratio=TF × IDF,
Wherein, key-ratio is the key coefficient, and TF be word frequency of the participle in the word vocabulary table, described in IDF is indicated
The word frequency in other texts is segmented, the IDF is determined according to trained word frequency inverse file frequency matrix；
Each participle in the word vocabulary table is arranged from high to low by the key coefficient；
The participle that specified quantity is successively chosen according to rank results is determined as the keyword of the word vocabulary table.
4. video recommendation method according to claim 1, which is characterized in that the user according to the candidate video comments
By information, the candidate video in the candidate video list is screened, obtains the recommendation list of videos of the video,
Include:
Obtain the user comment information of the video；
According to the user comment information of the candidate video and the user comment information of the video, the candidate video is calculated
The similarity factor of candidate video described in list and the video；
According to the similarity factor, the candidate video in the candidate video list is screened, pushing away for the video is obtained
Recommend list of videos.
5. video recommendation method according to claim 4, which is characterized in that the user according to the candidate video comments
By information and the user comment information of the video, candidate video described in the candidate video list and the video are calculated
Similarity factor, comprising:
The similarity factor of candidate video described in the candidate list Yu the video is calculated according to the following formula:
Wherein, r indicates similarity factor, according to the user comment information of the candidate video construct the comment of the candidate video to
Amount, the comment vector of candidate video described in C table, according to the user comment information of the video construct the comment of the video to
Amount, V indicate the comment vector of the video, and i indicates i-th of user, and the vector dimension of C and V are N, and N indicates number of users, Ci
Indicate i-th of user to the comment vector of the candidate video, ViIndicate i-th of user to the comment vector of the video.
6. video recommendation method according to claim 1, which is characterized in that the user according to the candidate video comments
By information, the candidate video in the candidate video list is screened, obtains the recommendation list of videos of the video,
Include:
According to the user comment information and trained neural network model, the candidate view in the candidate video list is obtained
The user of frequency scores；
It is scored according to the user, the candidate video in the candidate video list is screened, pushing away for the video is obtained
Recommend list of videos.
7. a kind of video recommendations device, which is characterized in that the video recommendations device includes:
Information acquisition unit, the corresponding text envelope of video being currently played for obtaining the user that user pays close attention to or described
Breath；
Theme determination unit, for generating model according to the corresponding text information of the video and preparatory trained theme, really
The theme of the fixed video；
Candidate list construction unit, for searching identical with theme video in video database as candidate video,
Construct video candidate list；
Comment information acquiring unit, for obtaining the user comment information of candidate video described in the video candidate list；
Recommendation list determination unit, for the user comment information according to the candidate video, in the candidate video list
The candidate video screened, obtain the recommendation list of videos of the video；
Video recommendations unit, for recommending video to the user according to the recommendation list of videos.
8. video recommendations device according to claim 6, which is characterized in that the recommendation list determination unit includes:
User, which scores, obtains module, for according to the user comment information and trained neural network model, described in acquisition
The user of candidate video in candidate video list scores；
Second recommendation list determining module, for being scored according to the user, to the candidate video in the candidate video list
It is screened, obtains the recommendation list of videos of the video.
9. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In the step of realization video recommendation method as described in any one of claims 1 to 6 when the computer program is executed by processor
Suddenly.
10. a kind of server, including memory, processor and storage can transport in the memory and on the processor
Capable computer program, which is characterized in that the processor is realized when executing the computer program as in claim 1 to 6
The step of any one video recommendation method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811339832.0A CN109325146A (en) | 2018-11-12 | 2018-11-12 | A kind of video recommendation method, device, storage medium and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811339832.0A CN109325146A (en) | 2018-11-12 | 2018-11-12 | A kind of video recommendation method, device, storage medium and server |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109325146A true CN109325146A (en) | 2019-02-12 |
Family
ID=65259523
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811339832.0A Pending CN109325146A (en) | 2018-11-12 | 2018-11-12 | A kind of video recommendation method, device, storage medium and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109325146A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109996102A (en) * | 2019-04-11 | 2019-07-09 | 北京字节跳动网络技术有限公司 | Video information synchronous display method, device, equipment and storage medium |
| CN110059219A (en) * | 2019-05-24 | 2019-07-26 | 广东工业大学 | A kind of video preference prediction technique, device, equipment and readable storage medium storing program for executing |
| CN110245261A (en) * | 2019-05-24 | 2019-09-17 | 中山大学 | A kind of latent structure method and system in multi-modal short video recommendation system |
| CN110347869A (en) * | 2019-06-05 | 2019-10-18 | 北京达佳互联信息技术有限公司 | A kind of video generation method, device, electronic equipment and storage medium |
| CN111597469A (en) * | 2020-05-19 | 2020-08-28 | 北京字节跳动网络技术有限公司 | Display position determining method and device, electronic equipment and storage medium |
| CN112417207A (en) * | 2020-11-24 | 2021-02-26 | 未来电视有限公司 | Video recommendation method, device, equipment and storage medium |
| CN113495942A (en) * | 2020-04-01 | 2021-10-12 | 百度在线网络技术（北京）有限公司 | Method and device for pushing information |
| CN113590804A (en) * | 2021-06-23 | 2021-11-02 | 北京百度网讯科技有限公司 | Video theme generation method and device and electronic equipment |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103136275A (en) * | 2011-12-02 | 2013-06-05 | 盛乐信息技术（上海）有限公司 | System and method for recommending personalized video |
| CN103455613A (en) * | 2013-09-06 | 2013-12-18 | 南京大学 | Interest aware service recommendation method based on MapReduce model |
| CN104219575A (en) * | 2013-05-29 | 2014-12-17 | 酷盛(天津）科技有限公司 | Related video recommending method and system |
| CN105915949A (en) * | 2015-12-23 | 2016-08-31 | 乐视网信息技术（北京）股份有限公司 | Video content recommending method, device and system |
| CN106570008A (en) * | 2015-10-09 | 2017-04-19 | 阿里巴巴集团控股有限公司 | Recommendation method and device |
| CN106951547A (en) * | 2017-03-27 | 2017-07-14 | 西安电子科技大学 | A kind of cross-domain recommendation method based on intersection user |
| CN107332905A (en) * | 2017-06-30 | 2017-11-07 | 广州优视网络科技有限公司 | Information-pushing method, device and server |
| CN107818491A (en) * | 2017-09-30 | 2018-03-20 | 平安科技（深圳）有限公司 | Electronic installation, Products Show method and storage medium based on user's Internet data |
| CN108491529A (en) * | 2018-03-28 | 2018-09-04 | 百度在线网络技术（北京）有限公司 | Information recommendation method and device |
-
2018
- 2018-11-12 CN CN201811339832.0A patent/CN109325146A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103136275A (en) * | 2011-12-02 | 2013-06-05 | 盛乐信息技术（上海）有限公司 | System and method for recommending personalized video |
| CN104219575A (en) * | 2013-05-29 | 2014-12-17 | 酷盛(天津）科技有限公司 | Related video recommending method and system |
| CN103455613A (en) * | 2013-09-06 | 2013-12-18 | 南京大学 | Interest aware service recommendation method based on MapReduce model |
| CN106570008A (en) * | 2015-10-09 | 2017-04-19 | 阿里巴巴集团控股有限公司 | Recommendation method and device |
| CN105915949A (en) * | 2015-12-23 | 2016-08-31 | 乐视网信息技术（北京）股份有限公司 | Video content recommending method, device and system |
| CN106951547A (en) * | 2017-03-27 | 2017-07-14 | 西安电子科技大学 | A kind of cross-domain recommendation method based on intersection user |
| CN107332905A (en) * | 2017-06-30 | 2017-11-07 | 广州优视网络科技有限公司 | Information-pushing method, device and server |
| CN107818491A (en) * | 2017-09-30 | 2018-03-20 | 平安科技（深圳）有限公司 | Electronic installation, Products Show method and storage medium based on user's Internet data |
| CN108491529A (en) * | 2018-03-28 | 2018-09-04 | 百度在线网络技术（北京）有限公司 | Information recommendation method and device |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109996102B (en) * | 2019-04-11 | 2021-01-05 | 北京字节跳动网络技术有限公司 | Video information synchronous display method, device, equipment and storage medium |
| CN109996102A (en) * | 2019-04-11 | 2019-07-09 | 北京字节跳动网络技术有限公司 | Video information synchronous display method, device, equipment and storage medium |
| CN110059219A (en) * | 2019-05-24 | 2019-07-26 | 广东工业大学 | A kind of video preference prediction technique, device, equipment and readable storage medium storing program for executing |
| CN110245261A (en) * | 2019-05-24 | 2019-09-17 | 中山大学 | A kind of latent structure method and system in multi-modal short video recommendation system |
| CN110245261B (en) * | 2019-05-24 | 2022-09-09 | 中山大学 | Feature construction method and system in multi-modal short video recommendation system |
| CN110347869A (en) * | 2019-06-05 | 2019-10-18 | 北京达佳互联信息技术有限公司 | A kind of video generation method, device, electronic equipment and storage medium |
| CN110347869B (en) * | 2019-06-05 | 2021-07-09 | 北京达佳互联信息技术有限公司 | Video generation method and device, electronic equipment and storage medium |
| CN113495942B (en) * | 2020-04-01 | 2022-07-05 | 百度在线网络技术（北京）有限公司 | Method and device for pushing information |
| CN113495942A (en) * | 2020-04-01 | 2021-10-12 | 百度在线网络技术（北京）有限公司 | Method and device for pushing information |
| CN111597469A (en) * | 2020-05-19 | 2020-08-28 | 北京字节跳动网络技术有限公司 | Display position determining method and device, electronic equipment and storage medium |
| CN112417207A (en) * | 2020-11-24 | 2021-02-26 | 未来电视有限公司 | Video recommendation method, device, equipment and storage medium |
| CN112417207B (en) * | 2020-11-24 | 2023-02-21 | 未来电视有限公司 | Video recommendation method, device, equipment and storage medium |
| CN113590804A (en) * | 2021-06-23 | 2021-11-02 | 北京百度网讯科技有限公司 | Video theme generation method and device and electronic equipment |
| CN113590804B (en) * | 2021-06-23 | 2023-08-04 | 北京百度网讯科技有限公司 | Video theme generation method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109325146A (en) | A kind of video recommendation method, device, storage medium and server | |
| CN108573411B (en) | Mixed recommendation method based on deep emotion analysis and multi-source recommendation view fusion of user comments | |
| CN106709040B (en) | Application search method and server | |
| CN111324728A (en) | Text event abstract generation method and device, electronic equipment and storage medium | |
| CN103678329B (en) | Recommend method and device | |
| US20130060769A1 (en) | System and method for identifying social media interactions | |
| CN110909164A (en) | Text enhancement semantic classification method and system based on convolutional neural network | |
| CN103092943B (en) | A kind of method of advertisement scheduling and advertisement scheduling server | |
| CN105786977A (en) | Mobile search method and device based on artificial intelligence | |
| CN104978314B (en) | Media content recommendations method and device | |
| EP2842060A1 (en) | Recommending keywords | |
| WO2014107801A1 (en) | Methods and apparatus for identifying concepts corresponding to input information | |
| CN112836487B (en) | Automatic comment method and device, computer equipment and storage medium | |
| EP3732592A1 (en) | Intelligent routing services and systems | |
| CN109918662A (en) | A kind of label of e-sourcing determines method, apparatus and readable medium | |
| CN113254711B (en) | Interactive image display method and device, computer equipment and storage medium | |
| CN109598586A (en) | A kind of recommended method based on attention model | |
| CN109255012A (en) | A kind of machine reads the implementation method and device of understanding | |
| Yang et al. | A topic model for co-occurring normal documents and short texts | |
| CN110222260A (en) | A kind of searching method, device and storage medium | |
| CN111444304A (en) | Search ranking method and device | |
| CN111400584A (en) | Association word recommendation method and device, computer equipment and storage medium | |
| Kiran et al. | User specific product recommendation and rating system by performing sentiment analysis on product reviews | |
| Moya et al. | Integrating web feed opinions into a corporate data warehouse | |
| CN113961823B (en) | News recommendation method, system, storage medium and equipment |
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 |