CN108965916A - A kind of method, the method, device and equipment of model foundation of live video assessment - Google Patents
A kind of method, the method, device and equipment of model foundation of live video assessment Download PDFInfo
- Publication number
- CN108965916A CN108965916A CN201710380359.XA CN201710380359A CN108965916A CN 108965916 A CN108965916 A CN 108965916A CN 201710380359 A CN201710380359 A CN 201710380359A CN 108965916 A CN108965916 A CN 108965916A
- Authority
- CN
- China
- Prior art keywords
- live
- live video
- video
- barrage
- barrage data
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 81
- 230000035945 sensitivity Effects 0.000 claims abstract description 57
- 230000005484 gravity Effects 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims description 57
- 239000011159 matrix material Substances 0.000 claims description 43
- 206010020751 Hypersensitivity Diseases 0.000 claims description 20
- 208000026935 allergic disease Diseases 0.000 claims description 20
- 230000007815 allergy Effects 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 20
- 238000003860 storage Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 14
- 238000007726 management method Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 238000007477 logistic regression Methods 0.000 description 4
- 238000012544 monitoring process Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 230000036651 mood Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 208000001613 Gambling Diseases 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 235000012054 meals Nutrition 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 210000000689 upper leg Anatomy 0.000 description 1
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
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/234—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs
- H04N21/23418—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics
-
- 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
- H04N21/266—Channel or content management, e.g. generation and management of keys and entitlement messages in a conditional access system, merging a VOD unicast channel into a multicast channel
-
- 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
- H04N21/47—End-user applications
- H04N21/478—Supplemental services, e.g. displaying phone caller identification, shopping application
- H04N21/4788—Supplemental services, e.g. displaying phone caller identification, shopping application communicating with other users, e.g. chatting
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The invention discloses a kind of methods of live video assessment, comprising: obtains the barrage data of target live video；Determine to characterize the feature of each sensitive dimension from barrage data；The feature of each sensitive dimension is input in live streaming assessment models, determines that the Sensitivity Index of target live video, Sensitivity Index reflect the specific gravity of the included sensitive content of target live video；When Sensitivity Index is greater than sensitivity threshold, determine that target live video is objectionable video.The method of live video assessment provided by the embodiments of the present application, whether the video content that can effectively analyze network direct broadcasting is sensitive, to effectively be managed to network direct broadcasting.
Description
Technical field
The present invention relates to Internet technical fields, and in particular to method, the live streaming assessment models of a kind of live video assessment
The method, device and equipment of foundation.
Background technique
As the rise that network direct broadcasting and network video play earns more flows, net to attract more users
The Pornograph being broadcast live in network live streaming also increases significantly therewith.
How effectively to identify whether the content of program request or live streaming is Pornograph to user in a network, and to including color
It is major internet platform provider problems faced that the live streaming of feelings content, which is managed,.
It is all currently to identify whether intercepted picture includes Pornograph by carrying out screenshot to video content.It is this
Identification method, the more difficult identification if when there was only a bit of Pornograph inside prolonged video, and the party
Method is since complexity cannot be used for monitoring live content, so as to cause the video content that can not effectively manage network direct broadcasting.
Summary of the invention
In order to solve the problems, such as can not effectively to manage the video content of network direct broadcasting in the prior art, the embodiment of the present invention is mentioned
For a kind of method of live video assessment, whether the video content that can effectively analyze network direct broadcasting is sensitive, thus effectively
Network direct broadcasting is managed.The embodiment of the present application also provides the methods that live streaming assessment models are established, and and live video
The corresponding device of method and equipment that method, the live streaming assessment models of assessment are established.
The application first aspect provides a kind of method of live video assessment, comprising:
Obtain the barrage data of target live video；
Determine to characterize the feature of each sensitive dimension from the barrage data；
The feature of each sensitive dimension is input in live streaming assessment models, determines the sensitivity of the target live video
Index, the Sensitivity Index reflect the specific gravity of the included sensitive content of target live video；
When the Sensitivity Index is greater than sensitivity threshold, determine that the target live video is objectionable video.
The application second aspect provides a kind of method that live streaming assessment models are established, comprising:
Obtain the barrage data for the first kind live streaming for being chosen as sample and the barrage data of the second class live streaming, the first kind
Live streaming is sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
According to the barrage data of the barrage data of the first kind live video and the second class live video, training is initially commented
Model is estimated, to obtain the live streaming assessment models.
The application third aspect provides a kind of device of live video assessment, comprising:
Program module is obtained, for obtaining the barrage data of target live video；
First determines program module, for determining to characterize from the barrage data that the acquisition program module obtains
The feature of each sensitivity dimension；
Second determines program module, for determining the feature of the determining each sensitive dimension of program module by described first
It is input in live streaming assessment models, determines that the Sensitivity Index of the target live video, the Sensitivity Index reflect the target
The specific gravity of the included sensitive content of live video；
Third determines program module, and the Sensitivity Index for determining when the described second determining program module is greater than sensitivity
When threshold value, determine that the target live video is objectionable video.
The application fourth aspect provides a kind of device that live streaming assessment models are established, comprising:
Program module is obtained, for obtaining the barrage data for the first kind live streaming for being chosen as sample and the bullet of the second class live streaming
Curtain data, the first kind live streaming are sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
Model training program module, the bullet of the first kind live video for being obtained according to the acquisition program module
The barrage data of curtain data and the second class live video, training initial assessment model, to obtain the live streaming assessment models.
The 5th aspect of the application provides a kind of computer equipment, comprising: input/output (I/O) interface, processor and storage
Device is stored with the instruction of the assessment of live video described in first aspect in the memory；
The processor is used to execute the instruction of the live video assessment stored in memory, executes as described in relation to the first aspect
Live video assessment method the step of.
The 6th aspect of the application provides a kind of computer equipment, comprising: input/output (I/O) interface, processor and storage
Device is stored with the instruction of the foundation of live streaming assessment models described in second aspect in the memory；
The processor is used to execute the instruction that the live streaming assessment models stored in memory are established, and executes such as second aspect
The step of method that the live streaming assessment models are established.
The another aspect of the application provides a kind of computer readable storage medium, in the computer readable storage medium
It is stored with instruction, when run on a computer, so that computer executes method described in above-mentioned various aspects.
The another aspect of the application provides a kind of computer program product comprising instruction, when it runs on computers
When, so that computer executes method described in above-mentioned various aspects.
Compared with it can not preferably manage live video content in the prior art, live video provided by the embodiments of the present application
The method of assessment, whether the video content that can effectively analyze network direct broadcasting is sensitive, to effectively carry out to network direct broadcasting
Management.
Detailed description of the invention
Fig. 1 is one network architecture schematic diagram of distributed system in the embodiment of the present application；
Fig. 2 is the structural schematic diagram of distributed system under quasi-ization scene in the embodiment of the present application；
Fig. 3 is the embodiment schematic diagram that the method that assessment models are established is broadcast live in the embodiment of the present application；
Fig. 4 is another embodiment schematic diagram for the method that live streaming assessment models provided by the embodiments of the present application are established；
Fig. 5 is an embodiment schematic diagram of the method that live video is assessed in the embodiment of the present application；
Fig. 6 is a Sample Scenario schematic diagram of the method that live video is assessed in the embodiment of the present application；
Fig. 7 is another embodiment schematic diagram for the method that live video is assessed in the embodiment of the present application；
Fig. 8 is an embodiment schematic diagram of the device that live video is assessed in the embodiment of the present application；
Fig. 9 is another embodiment schematic diagram for the device that live video is assessed in the embodiment of the present application；
Figure 10 is the embodiment schematic diagram that the device that assessment models are established is broadcast live in the embodiment of the present application；
Figure 11 is an embodiment schematic diagram of computer equipment in the embodiment of the present application.
Specific embodiment
With reference to the accompanying drawing, the embodiment of the present invention is described, it is clear that described embodiment is only the present invention
The embodiment of a part, instead of all the embodiments.Those of ordinary skill in the art are it is found that with the development of technology, this hair
The technical solution that bright embodiment provides is equally applicable for similar technical problem.
The embodiment of the present invention provides a kind of method of live video assessment, can effectively analyze in the video of network direct broadcasting
Whether appearance is sensitive, to effectively be managed to network direct broadcasting.The embodiment of the present application also provides corresponding devices.Divide below
It is not described in detail.
Network direct broadcasting refers to be broadcast live online by internet platform, now with many network direct broadcasting platforms, Yong Huan
After filling corresponding live streaming platform software, so that it may the live streaming room on selection live streaming platform, into the live streaming in the viewing room
Hold.
Because being required without what the main broadcaster being broadcast live in the network platform at present, what main broadcaster was broadcast live out
Content is also multifarious, has live streaming to have a meal, and has live streaming to make up, has live streaming to play games, some sensitive contents are also broadcast live
, the sensitive content being broadcast live in the network platform is often referred to Pornograph, it is of course also possible to refer to otherwise sensitive content, such as
Gambling etc..
Platform provider will be managed these sensitive contents, it is necessary to the video content of direct broadcasting room each on platform
Screenshot is carried out, whether include sensitive content, but this way to manage, not only treating capacity is big if then analyzing in screenshot, but also to cutting
The analysis of figure may be also not accurate enough, and the video content of network direct broadcasting can not effectively be managed by leading to platform provider still.
Live streaming is to be interacted by barrage with spectators.And be broadcast live decadent content necessarily in order to attract spectators,
Flow of the people.Therefore, unsound barrage has already appeared during interaction, while unsound barrage can also be through the entire mistake of live streaming
Journey.Therefore, text analyzing is carried out for barrage in the embodiment of the present application carry out different degrees of monitoring, such as based on the analysis results
There are a large amount of unhealthy vocabulary in fruit, while direct broadcasting room flow increases sharply, then such needs key monitoring, can be effectively to live streaming
Video content is managed.
Video is managed according to the barrage in live video, needs first to establish live streaming assessment models.
It is related to the foundation and use to live streaming assessment models in the embodiment of the present application, live streaming assessment models refer to and can calculate
The Sensitivity Index of live video out, the Sensitivity Index can reflect the specific gravity of the included sensitive content of target live video.
Either the foundation of live streaming assessment models is still used, requires to realize based on hind computation equipment, backstage
It calculates equipment to exist usually in the form of cluster, it is of course also possible to be independent calculating equipment.Below in the form of cluster
It is introduced.
As shown in FIG. 1, FIG. 1 is one network architecture schematic diagrams of distributed system in the embodiment of the present application；
Distributed system provided by the embodiment of the present application includes main controlled node (Master) 10, network 20 and multiple work
Node (Worker) 30, main controlled node 10 and multiple working nodes 30 can be communicated by network 20, in the distributed system
Multiple working nodes are responsible for storing various live videos, which can assess in the embodiment of the present application for live video
Device or the device established of live streaming assessment models, be responsible for the live video that is stored according to each working node and establish live streaming to comment
Estimate model or determine the Sensitivity Index for the video being broadcast live, which reflects that the target live video includes quick
Feel the specific gravity of content.When it is determined that alarm prompt can be sent to backstage manager when objectionable video.The application is real
The main controlled node 10 applied in example can have one, can also have multiple.For example, the reliability in order to guarantee system, can dispose
One spare main controlled node, for sharing a part of load, or working as in currently running main controlled node load too high
When the main controlled node of preceding operation breaks down, the work of the main controlled node is taken over.Main controlled node 10 and working node 30 are ok
It is physical host.
The system that the distributed system can also be virtualization, the form of expression of the distributed system in the case where virtualizing scene
As shown in Fig. 2, the distributed system under the virtualization scene includes hardware layer and operates in the virtual machine monitoring on hardware layer
Device (VMM) 1001 and multiple virtual machines 1002.One or more virtual machine be can choose as main controlled node, Yi Jiduo
A virtual machine is as working node.
Specifically, virtual machine 1002: one or more simulated in common hardware resource by software virtual machine
Virtual computer, and these virtual machines work just as real computer, and operation system can be installed on virtual machine
System and application program, virtual machine may also access Internet resources.For the application program run in virtual machine, virtual machine is just
It seem to work in real computer.
Hardware layer: the hardware platform of virtualized environment operation can be taken out by the hardware resource of one or more physical hosts
As obtaining.Wherein, hardware layer may include multiple hardwares, for example including processor 1004 (such as CPU) and memory 1005, also
It may include network interface card 1003 (such as RDMA network interface card), the input/output of high speed/low speed (I/O, Input/Output) equipment, and tool
There are other equipment of particular procedure function.
In addition, the distributed system under the virtualization scene can also include host (Host): as management level, to
Complete management, the distribution of hardware resource；Virtual hardware platform is presented for virtual machine；Realize the scheduling and isolation of virtual machine.Wherein,
Host may be monitor of virtual machine (VMM)；In addition, VMM and 1 privileged virtual machine cooperation, the two combine composition sometimes
Host.Wherein, virtual hardware platform provides various hardware resources to each virtual machine run thereon, such as provides virtual processor
(such as VCPU), virtual memory, virtual disk, Microsoft Loopback Adapter.Wherein, the virtual disk can correspond to Host a file or
One logic block device of person.Virtual machine operates on virtual hardware platform of the Host for its preparation, and one or more is run on Host
A virtual machine.
Privileged virtual machine: a kind of special virtual machine, also referred to as driving domain, such as this special virtual machine is in Xen
It is referred to as Dom0 on Hypervisor platform, the true physical equipment such as network interface card, SCSI disk is mounted in the virtual machine
Driver, can detect and directly access these true physical equipments.The phase that other virtual machines utilize Hypervisor to provide
Mechanism is answered to access true physical equipment by privileged virtual machine.
It is that first describe the embodiment of the present application below with reference to the network architecture in the description of the network architecture of the application above
In live streaming assessment models establish method.
As shown in figure 3, an embodiment of the method that live streaming assessment models provided by the embodiments of the present application are established includes:
101, the barrage data for the first kind live streaming for being chosen as sample and the barrage data of the second class live streaming are obtained, described the
One kind live streaming is sensitive live streaming, and the second class live streaming is non-sensitive live streaming.
When establishing live streaming assessment models, need to filter out the live streaming comprising sensitive content from history live video in advance
Video, such as: the live video comprising Pornograph will also filter out the normal live video not comprising sensitive content.
102, according to the barrage data of the barrage data of the first kind live video and the second class live video, training is just
Beginning assessment models, to obtain the live streaming assessment models.
Compared with it can not effectively analyze live video content in the prior art, live streaming assessment provided by the embodiments of the present application
The method of model foundation can establish live streaming assessment models, thus effectively to being broadcast live in direct broadcasting room by barrage data
Video analyzed, be conducive to platform provider and timely manage the live video comprising sensitive content.
It is wherein, described according to the barrage data of the first kind live video and the barrage data of the second class live video,
Training initial assessment model to obtain the live streaming assessment models may include:
It is determined from the barrage data of the first kind live video and the barrage data of the second class live video each
The feature of each sensitive dimension is characterized in live video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, to obtain
The live streaming assessment models.
Wherein, in the barrage data from the first kind live video and the barrage data of the second class live video really
Make the feature that each sensitive dimension is characterized in each live video, comprising:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described every
Training barrage text and training barrage data on flows in a live video；
The words-frequency feature for determining sensitive word included in the trained barrage text, from the trained barrage data on flows
Determine the increment and click each live streaming view that the barrage quantity for training pattern, each live video are clicked
At least one of the ratio of allergy sense video is clicked in the object of frequency.
In the embodiment of the present application, barrage data include barrage text and barrage data on flows two parts, training barrage text
With the barrage text and barrage data on flows that training barrage data on flows is exactly for training live streaming assessment models.
The text that barrage text is inputted by direct broadcasting room where spectators, barrage data on flows includes barrage quantity, described every
In the ratio for clicking allergy sense video in the object of increment and click each live video that a live video is clicked
At least one.
Wherein, the increment that each live video is clicked is spectators' increment for the live video.Described in click
The ratio of allergy sense video is clicked in the object of each live video: referring to having in all spectators for the live video
Cross the spectator attendance of viewing objectionable video historical record ratio shared in all spectators again.
The value of feature about each sensitive dimension, if various situations include sensitive content, the value of the dimension can
Think 1, if not including sensitive content, the value of the dimension can be 0.
The value of each dimension is the value of each x in x matrix, because being all sample number used in the model training process
According to so hθ(x) value be it is known, the value of θ can be trained by multiple samples, thus obtain live streaming assessment mould
Type.
By taking pornographic live streaming as an example, the training process that assessment models are broadcast live in the embodiment of the present application is introduced below with reference to Fig. 4.
Barrage full dose data and pre-stored historical data, model training personnel have been directed to the barrage data of each video
It is marked, if in the barrage data of a certain video including Pornograph, 0 is labeled as, if in the barrage data of a certain video
Do not include Pornograph, is normal video barrage data, is then labeled as 1.
A: pornographic barrage data are marked off from barrage full dose data.
It can be 0 according to the label of each video barrage data, such as label, then can mark off pornographic barrage data.
B: normal barrage data are marked off from barrage full dose data.
It can be 1 according to the label of each video barrage data, such as label, then can mark off normal barrage data.
C: from pornographic barrage data and normal barrage data, bullet is extracted respectively for the barrage data of each video
Curtain text.
It after obtaining barrage text, needs barrage text being converted to the data that computer can identify, uses text here
The most common dictionary segmenting method in identification the inside, by common preposition ', obtain ', auxiliary words of mood ', oh ' etc. removes；So
The number that dictionary finds out the word occurred inside text is corresponded to afterwards.
D: from pornographic barrage data and normal barrage data, for each video barrage data extract respectively it is each quick
Feel the sensitive features of dimension.
In the embodiment, sensitive dimension is pornographic dimension, and sensitive features are pornographic feature.
The embodiment sensitive features may include barrage number, viewing live streaming number growth rate per second, in viewing live streaming user
Watched user's accounting etc. of pornographic live streaming.
E: the word frequency of sensitive word is calculated from barrage text.
Wherein molecule is the frequency of occurrence of sensitive word, and denominator is the total degree that all words occur.
In the embodiment, molecule is the number that pornographic word occurs, and denominator is the sum of all words in the barrage of the video.
F: it is modeled using sorting algorithm.
The process is using logistic regression algorithm (Logistic Regression, LR) algorithm, using each in Sample video
The feature of the pornographic word frequency of video and pornographic dimension, to initial assessment modelIt is instructed
Practice, obtains the process of weight matrix θ.
Because being all sample data, h used in the model training processθ(x) value be it is known, by more
A sample can train the value of θ, to obtain live streaming assessment models.
G: live streaming assessment models.
It is 5 to be described in the embodiment of the present application with reference to the accompanying drawing to the description of the establishment process of live streaming assessment models above
The process that target live video is assessed using live streaming assessment models.
Refering to Fig. 5, an embodiment of the method for live video assessment provided by the embodiments of the present application includes:
201, the barrage data of target live video are obtained.
Target live video can be the video that is being broadcast live in the direct broadcasting room of the network platform.Barrage data include it is all with
The relevant data of barrage, including barrage text, the user for sending barrage, barrage quantity and the user's increment for sending barrage etc.
Data.
202, determine to characterize the feature of each sensitive dimension from the barrage data.
The step may include:
Barrage text and barrage data on flows are extracted from the barrage data；
The words-frequency feature for determining sensitive word included in the barrage text determines barrage from the barrage data on flows
Allergy sense view is clicked in the object of increment and the click target live video that quantity, the target live video are clicked
At least one of ratio of frequency.
The text that barrage text is inputted by direct broadcasting room where spectators, barrage data on flows includes barrage quantity, described every
In the ratio for clicking allergy sense video in the object of increment and click each live video that a live video is clicked
At least one.
Wherein, the increment that target live video is clicked is spectators' increment for the target live video.It clicks
The ratio of allergy sense video is clicked in the object of the target live video: referring to all sights for the target live video
In crowd, there is the spectator attendance of viewing objectionable video historical record ratio shared in all spectators again.
203, the feature of each sensitive dimension is input in live streaming assessment models, determines the target live video
Sensitivity Index, the Sensitivity Index reflect the specific gravity of the included sensitive content of target live video.
The step may include:
The Sensitivity Index of the target live video is determined according to following live streaming assessment models；
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, the x include the increment that is clicked of the words-frequency feature and the target live video and
Click at least one of the ratio that allergy sense video is clicked in the object of the target live video；
θ is weight matrix, each weight in the weight matrix obtained in the live streaming assessment models training process
Value, the number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ.
In the calculating process, because the value of x and θ is all given value, it is possible to calculate hθ(x) value, such as: hθ(x)=
0.7。
204, when the Sensitivity Index is greater than sensitivity threshold, determine that the target live video is objectionable video.
If sensitivity threshold is 0.5, h can be determinedθ(x) when being greater than 0.5, then the target live video is sensitive view
Frequently, if hθ(x) less than 0.5, then the target live video is non-sensitive video.Then alarm can be sent to live streaming management backstage
Prompt information.
As shown in fig. 6, in a direct broadcasting room of network direct broadcasting platform one " main broadcaster " wear skirt in fanatic dance, fanatic dance process
In can also expose thigh, skirt is also to seethe up and down, and the number of direct broadcasting room spectators is continuously increased during fanatic dance, Hen Duoguan
Crowd also leaves a message in barrage, such as: spectators one: barrage content 1, spectators two: barrage content 22, spectators three: barrage content 3
Deng.There are many spectators also constantly to beat reward, the main broadcaster that please be broadcast live jumps hot again more hot.
Background devices can obtain the barrage data of this direct broadcasting room, including barrage text during the main broadcaster is broadcast live,
Such as: spectators one: barrage content 1, spectators two: barrage content 22, spectators three: barrage content 3 etc..Spectators' number can also be obtained
Spectators' ratio etc. of pornographic video was watched in amount, spectators' increment and spectators.
These sensitive dimensional characteristics are input in live streaming assessment models by background devices, calculate Sensitivity Index hθ(x),
If hθ(x)=0.7 it, then can determine that the video that the direct broadcasting room is broadcast live for pornographic video, sends alarm prompt to the direct broadcasting room
Information, such as: " paying attention to live content, otherwise close room ".
Compared with it can not preferably manage live video content in the prior art, live video provided by the embodiments of the present application
The method of assessment, whether the video content that can effectively analyze network direct broadcasting is sensitive, to effectively carry out to network direct broadcasting
Management.
The establishment process of live streaming assessment models as shown in connection with fig. 4, live video assessment provided by the embodiments of the present application
Method can also be understood refering to the process of Fig. 7.In fact, the process of model training can be off-line training, it is also possible to
The process of on-line training.
A: pornographic barrage data are marked off from barrage full dose data.
It can be 0 according to the label of each video barrage data, such as label, then can mark off pornographic barrage data.
B: normal barrage data are marked off from barrage full dose data.
It can be 1 according to the label of each video barrage data, such as label, then can mark off normal barrage data.
C: from pornographic barrage data and normal barrage data, bullet is extracted respectively for the barrage data of each video
Curtain text.
It after obtaining barrage text, needs barrage text being converted to the data that computer can identify, uses text here
The most common dictionary segmenting method in identification the inside, by common preposition ', obtain ', auxiliary words of mood ', oh ' etc. removes；So
The number that dictionary finds out the word occurred inside text is corresponded to afterwards.
D: from pornographic barrage data and normal barrage data, for each video barrage data extract respectively it is each quick
Feel the sensitive features of dimension.
In the embodiment, sensitive dimension is pornographic dimension, and sensitive features are pornographic feature.
The embodiment sensitive features may include barrage number, viewing live streaming number growth rate per second, in viewing live streaming user
Watched user's accounting etc. of pornographic live streaming.
E: the word frequency of sensitive word is calculated from barrage text.
Wherein molecule is the frequency of occurrence of sensitive word, and denominator is the total degree that all words occur.
In the embodiment, molecule is the number that pornographic word occurs, and denominator is the sum of all words in the barrage of the video.
F: it is modeled using sorting algorithm.
The process is using logistic regression algorithm (Logistic Regression, LR) algorithm, using each in Sample video
The feature of the pornographic word frequency of video and pornographic dimension, to initial assessment modelIt is instructed
Practice, obtains the process of weight matrix θ.
Because being all sample data, h used in the model training processθ(x) value be it is known, by more
A sample can train the value of θ, to obtain live streaming assessment models.
G: live streaming assessment models.
H: the direct broadcasting room barrage data being broadcast live are obtained.
The barrage text and sensitive features for extracting the direct broadcasting room barrage data carry out word frequency analysis to barrage text.
Word frequency and sensitive features are all input in live streaming assessment models, are predicted.
I: prediction result includes normal and malice.
If malice, then refering to Fig. 6, alarm prompt first can be sent to backstage manager, then by back-stage management people
Member pays close attention to the direct broadcasting room, then sends alarm prompt to the direct broadcasting room, directly can also send alarm prompt to the direct broadcasting room.
Retouching for the process of assessment is broadcast live the above are the foundation to live streaming assessment models and using live streaming assessment models
It states, the dress that the device and live streaming assessment models for introducing the live video assessment in the embodiment of the present application with reference to the accompanying drawing are established
It sets.
Refering to Fig. 8, an embodiment of the device 40 of live video assessment provided by the embodiments of the present application includes:
Program module 401 is obtained, for obtaining the barrage data of target live video；
First determines program module 402, for determining from the barrage data that the acquisition program module 401 obtains
The feature of each sensitive dimension is characterized out；
Second determines program module 403, for determining the determining each sensitive dimension of program module 402 for described first
Feature be input in live streaming assessment models, determine the Sensitivity Index of the target live video, the Sensitivity Index reflects institute
State the specific gravity of the included sensitive content of target live video；
Third determines program module 404, and the Sensitivity Index for determining that program module 403 determines when described second is big
When sensitivity threshold, determine that the target live video is objectionable video.
Compared with it can not preferably manage live video content in the prior art, live video provided by the embodiments of the present application
The device of assessment, whether the video content that can effectively analyze network direct broadcasting is sensitive, to effectively carry out to network direct broadcasting
Management.
Optionally, in another embodiment of the device 40 of live video assessment provided by the embodiments of the present application,
First determines that program module 402 is used for:
Barrage text and barrage data on flows are extracted from the barrage data；
The words-frequency feature for determining sensitive word included in the barrage text determines barrage from the barrage data on flows
Allergy sense view is clicked in the object of increment and the click target live video that quantity, the target live video are clicked
At least one of ratio of frequency.
Optionally, in another embodiment of the device 40 of live video assessment provided by the embodiments of the present application,
Described second determines that program module 403 is used for:
The Sensitivity Index of the target live video is determined according to following live streaming assessment models；
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, the x include the increment that is clicked of the words-frequency feature and the target live video and
Click at least one of the ratio that allergy sense video is clicked in the object of the target live video；
θ is weight matrix, each weight in the weight matrix obtained in the live streaming assessment models training process
Value, the number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ.
Optionally, it refering in another embodiment of Fig. 9, live video provided by the embodiments of the present application device assessed, fills
Setting 40 further includes model training program module 405,
Program module 401 is obtained, is also used to obtain the barrage data and second for the first kind live video for being chosen as sample
The barrage data of class live video, the first kind live streaming are sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
View is broadcast live in model training program module 405, the first kind for being obtained according to the acquisition program module 401
The barrage data of the barrage data of frequency and the second class live video, training initial assessment model, to obtain the live streaming assessment mould
Type.
Optionally, in another embodiment of the device 40 of live video assessment provided by the embodiments of the present application,
Model training program module 405 is used for:
It is determined from the barrage data of the first kind live video and the barrage data of the second class live video each
The feature of each sensitive dimension is characterized in live video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, to obtain
The live streaming assessment models.
Optionally, in another embodiment of the device 40 of live video assessment provided by the embodiments of the present application,
Model training program module 405 is used for:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described every
Training barrage text and training barrage data on flows in a live video；
The words-frequency feature for determining sensitive word included in the trained barrage text, from the trained barrage data on flows
Determine the increment and click each live streaming view that the barrage quantity for training pattern, each live video are clicked
At least one of the ratio of allergy sense video is clicked in the object of frequency.
It is the description to the device of live video assessment above, can be managed in conjunction with the correlation technique of Fig. 1 to the part Fig. 7
Solution, it is no longer repeated at this place.
Refering to fig. 10, an embodiment of the device 50 that live streaming assessment models provided by the embodiments of the present application are established includes:
Program module 501 is obtained, for obtaining the barrage data and the live streaming of the second class of the first kind live streaming for being chosen as sample
Barrage data, first kind live streaming is sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
View is broadcast live in model training program module 502, the first kind for being obtained according to the acquisition program module 501
The barrage data of the barrage data of frequency and the second class live video, training initial assessment model, to obtain the live streaming assessment mould
Type.
Compared with it can not effectively analyze live video content in the prior art, live streaming assessment provided by the embodiments of the present application
The device of model foundation can establish live streaming assessment models, thus effectively to being broadcast live in direct broadcasting room by barrage data
Video analyzed, be conducive to platform provider and timely manage the live video comprising sensitive content.
Optionally, in another embodiment for the device 50 that live streaming assessment models provided by the embodiments of the present application are established,
The model training program module 502 is used for:
It is determined from the barrage data of the first kind live video and the barrage data of the second class live video each
The feature of each sensitive dimension is characterized in live video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, to obtain
The live streaming assessment models.
Optionally, in another embodiment for the device 50 that live streaming assessment models provided by the embodiments of the present application are established,
The model training program module 502 is used for:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described every
Training barrage text and training barrage data on flows in a live video；
The words-frequency feature for determining sensitive word included in the trained barrage text, from the trained barrage data on flows
Determine the increment and click each live streaming view that the barrage quantity for training pattern, each live video are clicked
At least one of the ratio of allergy sense video is clicked in the object of frequency.
The device 50 that live streaming assessment models provided by the embodiments of the present application are established can be corresponding to the part Fig. 7 refering to fig. 1
Description is understood that this place no longer does repetition and repeats.
Figure 11 is the structural schematic diagram of computer equipment 60 provided in an embodiment of the present invention.The computer equipment 60 is wrapped
Including processor 610, memory 650 and transceiver 630, memory 650 may include read-only memory and random access memory,
And operational order and data are provided to processor 610.The a part of of memory 650 can also include that non-volatile random is deposited
Reservoir (NVRAM).
In some embodiments, memory 650 stores following element, executable modules or data structures, or
Their subset of person or their superset:
In embodiments of the present invention, by calling the operational order of the storage of memory 650, (operational order is storable in behaviour
Make in system),
Obtain the barrage data of target live video；
Determine to characterize the feature of each sensitive dimension from the barrage data；
The feature of each sensitive dimension is input in live streaming assessment models, determines the sensitivity of the target live video
Index, the Sensitivity Index reflect the specific gravity of the included sensitive content of target live video；
When the Sensitivity Index is greater than sensitivity threshold, determine that the target live video is objectionable video.
Compared with it can not preferably manage live video content in the prior art, computer provided by the embodiments of the present application is set
Standby, whether the video content that can effectively analyze network direct broadcasting is sensitive, to effectively be managed to network direct broadcasting.
Processor 610 controls the operation of computer equipment 60, and processor 610 can also be known as CPU (Central
Processing Unit, central processing unit).Memory 650 may include read-only memory and random access memory, and
Instruction and data is provided to processor 610.The a part of of memory 650 can also include nonvolatile RAM
(NVRAM).The various components of computer equipment 60 are coupled by bus system 620 in specific application, wherein bus
System 620 can also include power bus, control bus and status signal bus in addition etc. in addition to including data/address bus.But it is
For the sake of clear explanation, in figure various buses are all designated as bus system 620.
The method that the embodiments of the present invention disclose can be applied in processor 610, or be realized by processor 610.
Processor 610 may be a kind of IC chip, the processing capacity with signal.During realization, the above method it is each
Step can be completed by the integrated logic circuit of the hardware in processor 610 or the instruction of software form.Above-mentioned processing
Device 610 can be general processor, digital signal processor (DSP), specific integrated circuit (ASIC), ready-made programmable gate array
Arrange (FPGA) either other programmable logic device, discrete gate or transistor logic, discrete hardware components.It may be implemented
Or disclosed each method, step and logic diagram in the execution embodiment of the present invention.General processor can be microprocessor
Or the processor is also possible to any conventional processor etc..The step of method in conjunction with disclosed in the embodiment of the present invention, can be with
Be embodied directly in hardware decoding processor and execute completion, or in decoding processor hardware and software module combination executed
At.Software module can be located at random access memory, and flash memory, read-only memory, programmable read only memory or electrically-erasable can
In the storage medium of this fields such as programmable memory, register maturation.The storage medium is located at memory 650, and processor 610 is read
Information in access to memory 650, in conjunction with the step of its hardware completion above method.
Optionally, processor 610 is used for:
Barrage text and barrage data on flows are extracted from the barrage data；
The words-frequency feature for determining sensitive word included in the barrage text determines barrage from the barrage data on flows
Allergy sense view is clicked in the object of increment and the click target live video that quantity, the target live video are clicked
At least one of ratio of frequency.
Optionally, processor 610 is used for:
The Sensitivity Index of the target live video is determined according to following live streaming assessment models；
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, the x include the increment that is clicked of the words-frequency feature and the target live video and
Click at least one of the ratio that allergy sense video is clicked in the object of the target live video；
θ is weight matrix, each weight in the weight matrix obtained in the live streaming assessment models training process
Value, the number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ.
Optionally, processor 610 is also used to:
Obtain the barrage data of the barrage data and the second class live video that are chosen as the first kind live video of sample, institute
First kind live streaming is stated as sensitive live streaming, the second class live streaming is non-sensitive live streaming；
According to the barrage data of the barrage data of the first kind live video and the second class live video, training is initially commented
Model is estimated, to obtain the live streaming assessment models.
Optionally, processor 610 is used for:
It is determined from the barrage data of the first kind live video and the barrage data of the second class live video each
The feature of each sensitive dimension is characterized in live video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, to obtain
The live streaming assessment models.
Optionally, processor 610 is used for:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described every
Training barrage text and training barrage data on flows in a live video；
The words-frequency feature for determining sensitive word included in the trained barrage text, from the trained barrage data on flows
Determine the increment and click each live streaming view that the barrage quantity for training pattern, each live video are clicked
At least one of the ratio of allergy sense video is clicked in the object of frequency.
Above to computer equipment 60 description can the description refering to fig. 1 to the part Fig. 7 understand, this place no longer weigh
It repeats again.
In addition, the computer equipment in the embodiment of the present application can also realize the function for the device that live streaming assessment models are established
It can, that is to say, that the actual physics form for the device that live streaming assessment models are established can also be with the hardware configuration of 1 part refering to fig. 1
Understood, corresponding hardware capability is for completing following steps:
Processor 610 is used for:
Obtain the barrage data for the first kind live streaming for being chosen as sample and the barrage data of the second class live streaming, the first kind
Live streaming is sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
According to the barrage data of the barrage data of the first kind live video and the second class live video, training is initially commented
Model is estimated, to obtain the live streaming assessment models.
Optionally, processor 610 is used for:
It is determined from the barrage data of the first kind live video and the barrage data of the second class live video each
The feature of each sensitive dimension is characterized in live video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, to obtain
The live streaming assessment models.
Optionally, processor 610 is used for:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described every
Training barrage text and training barrage data on flows in a live video；
The words-frequency feature for determining sensitive word included in the trained barrage text, from the trained barrage data on flows
Determine the increment and click each live streaming view that the barrage quantity for training pattern, each live video are clicked
At least one of the ratio of allergy sense video is clicked in the object of frequency.
Above to the description of corresponding function in computer equipment, can describe to be managed to the corresponding of the part Fig. 7 refering to fig. 1
Solution, it is no longer repeated at this place.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.
The computer program product includes one or more computer instructions.Load and execute on computers the meter
When calculation machine program instruction, entirely or partly generate according to process or function described in the embodiment of the present invention.The computer can
To be general purpose computer, special purpose computer, computer network or other programmable devices.The computer instruction can be deposited
Storage in a computer-readable storage medium, or from a computer readable storage medium to another computer readable storage medium
Transmission, for example, the computer instruction can pass through wired (example from a web-site, computer, server or data center
Such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (such as infrared, wireless, microwave) mode to another website
Website, computer, server or data center are transmitted.The computer readable storage medium can be computer and can deposit
Any usable medium of storage either includes that the data storages such as one or more usable mediums integrated server, data center are set
It is standby.The usable medium can be magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or partly lead
Body medium (such as solid state hard disk Solid State Disk (SSD)) etc..
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium may include: ROM, RAM, disk or CD etc..
Be provided for the embodiments of the invention above live video assessment method, live streaming assessment models establish method,
Device and equipment are described in detail, and specific case used herein carries out the principle of the present invention and embodiment
It illustrates, the above description of the embodiment is only used to help understand the method for the present invention and its core ideas；Meanwhile for this field
Those skilled in the art, according to the thought of the present invention, there will be changes in the specific implementation manner and application range, to sum up
Described, the contents of this specification are not to be construed as limiting the invention.
Claims (15)
1. a kind of method of live video assessment characterized by comprising
Obtain the barrage data of target live video；
Determine to characterize the feature of each sensitive dimension from the barrage data；
The feature of each sensitive dimension is input in live streaming assessment models, determines that the sensitivity of the target live video refers to
Number, the Sensitivity Index reflect the specific gravity of the included sensitive content of target live video；
When the Sensitivity Index is greater than sensitivity threshold, determine that the target live video is objectionable video.
2. the method according to claim 1, wherein described determine to characterize each sensitivity from the barrage data
The feature of dimension, comprising:
Barrage text and barrage data on flows are extracted from the barrage data；
The words-frequency feature for determining sensitive word included in the barrage text determines barrage number from the barrage data on flows
Allergy sense video is clicked in the object of increment and the click target live video that amount, the target live video are clicked
At least one of ratio.
3. according to the method described in claim 2, it is characterized in that, the feature by each sensitive dimension is input to live streaming
In assessment models, the Sensitivity Index of the target live video is determined, comprising:
The Sensitivity Index of the target live video is determined according to following live streaming assessment models；
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, the increment and click of the x being clicked including the words-frequency feature and the target live video
At least one of the ratio of allergy sense video is clicked in the object of the target live video；
θ is weight matrix, and each weight in the weight matrix obtained in the live streaming assessment models training process takes
Value, the number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ.
4. method according to claim 1 to 3, which is characterized in that the method also includes:
Obtain the barrage data of the barrage data and the second class live video that are chosen as the first kind live video of sample, described the
One kind live streaming is sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
According to the barrage data of the barrage data of the first kind live video and the second class live video, training initial assessment mould
Type, to obtain the live streaming assessment models.
5. according to the method described in claim 4, it is characterized in that, the barrage data according to the first kind live video
With the barrage data of the second class live video, training initial assessment model, to obtain the live streaming assessment models, comprising:
Each live streaming is determined from the barrage data of the first kind live video and the barrage data of the second class live video
The feature of each sensitive dimension is characterized in video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, described to obtain
Assessment models are broadcast live.
6. according to the method described in claim 5, it is characterized in that, the barrage data from the first kind live video and
The feature that each sensitive dimension is characterized in each live video is determined in the barrage data of second class live video, comprising:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described each straight
Broadcast the training barrage text and training barrage data on flows in video；
The words-frequency feature for determining sensitive word included in the trained barrage text is determined from the trained barrage data on flows
The increment that is clicked for the barrage quantity of training pattern, each live video and click each live video
At least one of the ratio of allergy sense video is clicked in object.
7. a kind of method that live streaming assessment models are established characterized by comprising
Obtain the barrage data for the first kind live streaming for being chosen as sample and the barrage data of the second class live streaming, the first kind live streaming
For sensitivity live streaming, the second class live streaming is non-sensitive live streaming；
According to the barrage data of the barrage data of the first kind live video and the second class live video, training initial assessment mould
Type, to obtain the live streaming assessment models.
8. the method according to the description of claim 7 is characterized in that the barrage data according to the first kind live video
With the barrage data of the second class live video, training initial assessment model, to obtain the live streaming assessment models, comprising:
Each live streaming is determined from the barrage data of the first kind live video and the barrage data of the second class live video
The feature of each sensitive dimension is characterized in video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, described to obtain
Assessment models are broadcast live.
9. according to the method described in claim 8, it is characterized in that, the barrage data from the first kind live video and
The feature that each sensitive dimension is characterized in each live video is determined in the barrage data of second class live video, comprising:
From the barrage data that barrage data and second class that the first kind is broadcast live are broadcast live, extract respectively described each straight
Broadcast the training barrage text and training barrage data on flows in video；
The words-frequency feature for determining sensitive word included in the trained barrage text is determined from the trained barrage data on flows
The increment that is clicked for the barrage quantity of training pattern, each live video and click each live video
At least one of the ratio of allergy sense video is clicked in object.
10. a kind of device of live video assessment characterized by comprising
Program module is obtained, for obtaining the barrage data of target live video；
First determines program module, each quick for determining to characterize from the barrage data that the acquisition program module obtains
Feel the feature of dimension；
Second determines program module, for determining that the feature of the determining each sensitive dimension of program module inputs for described first
Into live streaming assessment models, determine that the Sensitivity Index of the target live video, the Sensitivity Index reflect the target live streaming
The specific gravity of the included sensitive content of video；
Third determines program module, and the Sensitivity Index for determining when the described second determining program module is greater than sensitivity threshold
When, determine that the target live video is objectionable video.
11. device according to claim 10, which is characterized in that
First determines that program module is used for:
Barrage text and barrage data on flows are extracted from the barrage data；
The words-frequency feature for determining sensitive word included in the barrage text determines barrage number from the barrage data on flows
Allergy sense video is clicked in the object of increment and the click target live video that amount, the target live video are clicked
At least one of ratio.
12. a kind of device that live streaming assessment models are established characterized by comprising
Program module is obtained, for obtaining the barrage data for the first kind live streaming for being chosen as sample and the barrage number of the second class live streaming
According to the first kind live streaming is sensitive live streaming, and the second class live streaming is non-sensitive live streaming；
Model training program module, the barrage number of the first kind live video for being obtained according to the acquisition program module
According to the barrage data with the second class live video, training initial assessment model, to obtain the live streaming assessment models.
13. device according to claim 12, which is characterized in that
The model training program module is used for:
Each live streaming is determined from the barrage data of the first kind live video and the barrage data of the second class live video
The feature of each sensitive dimension is characterized in video；
Following initial assessment model is trained using the feature for characterizing each sensitive dimension in each live video；
The initial assessment model are as follows:
Wherein,
hθ(x) and g (θTIt x) is Sensitivity Index；
X is matrix, and the x's includes the feature that each sensitive dimension is characterized in each live video；
θ is weight matrix；The number of the θ is corresponding with the number of the x, the θTIndicate the transposed matrix of θ；
Feature training by characterizing each sensitive dimension in each live video obtains the value of the θ, described to obtain
Assessment models are broadcast live.
14. a kind of computer equipment characterized by comprising input/output (I/O) interface, processor and memory, it is described
The instruction of any live video assessment of claim 1-6 is stored in memory；
The processor is used to execute the instruction of the live video assessment stored in memory, executes as claim 1-6 is any
The step of method of the live video assessment.
15. a kind of computer equipment characterized by comprising input/output (I/O) interface, processor and memory, it is described
The instruction that any live streaming assessment models of claim 7-9 are established is stored in memory；
The processor is used to execute the instruction that the live streaming assessment models stored in memory are established, and executes such as claim 7-9
The step of method that any live streaming assessment models are established.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710380359.XA CN108965916B (en) | 2017-05-25 | 2017-05-25 | Live video evaluation method, model establishment method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710380359.XA CN108965916B (en) | 2017-05-25 | 2017-05-25 | Live video evaluation method, model establishment method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108965916A true CN108965916A (en) | 2018-12-07 |
| CN108965916B CN108965916B (en) | 2021-05-25 |
Family
ID=64494458
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710380359.XA Active CN108965916B (en) | 2017-05-25 | 2017-05-25 | Live video evaluation method, model establishment method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108965916B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109327715A (en) * | 2018-08-01 | 2019-02-12 | 阿里巴巴集团控股有限公司 | A kind of video Risk Identification Method, device and equipment |
| CN111147876A (en) * | 2019-12-26 | 2020-05-12 | 山东爱城市网信息技术有限公司 | Live broadcast content monitoring method, equipment, storage medium and platform based on block chain |
| CN111382383A (en) * | 2018-12-28 | 2020-07-07 | 广州市百果园信息技术有限公司 | Method, device, medium and computer equipment for determining sensitive type of webpage content |
| CN112055230A (en) * | 2020-09-03 | 2020-12-08 | 北京中润互联信息技术有限公司 | Live broadcast monitoring method and device, computer equipment and readable storage medium |
| CN114598899A (en) * | 2022-03-15 | 2022-06-07 | 中科大数据研究院 | Bullet screen broadcast analysis method based on crawler |
| CN114840477A (en) * | 2022-06-30 | 2022-08-02 | 深圳乐播科技有限公司 | File sensitivity index determining method based on cloud conference and related product |
| CN116228320A (en) * | 2023-03-01 | 2023-06-06 | 深圳市快美妆科技有限公司 | Live advertisement putting effect analysis system and method |
| CN117395470A (en) * | 2023-08-31 | 2024-01-12 | 江苏初辰文化发展有限公司 | Live broadcast content evaluation detection method based on barrage sharing |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8191152B1 (en) * | 2009-01-23 | 2012-05-29 | Intuit Inc. | Methods systems and articles of manufacture for generating and displaying video related to financial application |
| CN103218608A (en) * | 2013-04-19 | 2013-07-24 | 中国科学院自动化研究所 | Network violent video identification method |
| CN104918066A (en) * | 2014-03-11 | 2015-09-16 | 上海数字电视国家工程研究中心有限公司 | Video content censoring method and system |
| CN105574003A (en) * | 2014-10-10 | 2016-05-11 | 华东师范大学 | Comment text and score analysis-based information recommendation method |
| CN105872773A (en) * | 2016-06-01 | 2016-08-17 | 北京奇虎科技有限公司 | Video broadcast monitoring method and device |
| CN105930411A (en) * | 2016-04-18 | 2016-09-07 | 苏州大学 | Classifier training method, classifier and sentiment classification system |
| CN105956550A (en) * | 2016-04-29 | 2016-09-21 | 浪潮电子信息产业股份有限公司 | Video discriminating method and device |
| CN106250837A (en) * | 2016-07-27 | 2016-12-21 | 腾讯科技（深圳）有限公司 | The recognition methods of a kind of video, device and system |
-
2017
- 2017-05-25 CN CN201710380359.XA patent/CN108965916B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8191152B1 (en) * | 2009-01-23 | 2012-05-29 | Intuit Inc. | Methods systems and articles of manufacture for generating and displaying video related to financial application |
| CN103218608A (en) * | 2013-04-19 | 2013-07-24 | 中国科学院自动化研究所 | Network violent video identification method |
| CN104918066A (en) * | 2014-03-11 | 2015-09-16 | 上海数字电视国家工程研究中心有限公司 | Video content censoring method and system |
| CN105574003A (en) * | 2014-10-10 | 2016-05-11 | 华东师范大学 | Comment text and score analysis-based information recommendation method |
| CN105930411A (en) * | 2016-04-18 | 2016-09-07 | 苏州大学 | Classifier training method, classifier and sentiment classification system |
| CN105956550A (en) * | 2016-04-29 | 2016-09-21 | 浪潮电子信息产业股份有限公司 | Video discriminating method and device |
| CN105872773A (en) * | 2016-06-01 | 2016-08-17 | 北京奇虎科技有限公司 | Video broadcast monitoring method and device |
| CN106250837A (en) * | 2016-07-27 | 2016-12-21 | 腾讯科技（深圳）有限公司 | The recognition methods of a kind of video, device and system |
Non-Patent Citations (1)
| Title |
|---|
| 龚勋: "《重磅：286页一天搞懂深度学习ppt》", 《豆丁网》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109327715A (en) * | 2018-08-01 | 2019-02-12 | 阿里巴巴集团控股有限公司 | A kind of video Risk Identification Method, device and equipment |
| CN109327715B (en) * | 2018-08-01 | 2021-06-04 | 创新先进技术有限公司 | Video risk identification method, device and equipment |
| CN111382383A (en) * | 2018-12-28 | 2020-07-07 | 广州市百果园信息技术有限公司 | Method, device, medium and computer equipment for determining sensitive type of webpage content |
| CN111147876A (en) * | 2019-12-26 | 2020-05-12 | 山东爱城市网信息技术有限公司 | Live broadcast content monitoring method, equipment, storage medium and platform based on block chain |
| CN112055230A (en) * | 2020-09-03 | 2020-12-08 | 北京中润互联信息技术有限公司 | Live broadcast monitoring method and device, computer equipment and readable storage medium |
| CN114598899A (en) * | 2022-03-15 | 2022-06-07 | 中科大数据研究院 | Bullet screen broadcast analysis method based on crawler |
| CN114840477A (en) * | 2022-06-30 | 2022-08-02 | 深圳乐播科技有限公司 | File sensitivity index determining method based on cloud conference and related product |
| CN116228320A (en) * | 2023-03-01 | 2023-06-06 | 深圳市快美妆科技有限公司 | Live advertisement putting effect analysis system and method |
| CN116228320B (en) * | 2023-03-01 | 2024-02-06 | 广州网优优数据技术股份有限公司 | Live advertisement putting effect analysis system and method |
| CN117395470A (en) * | 2023-08-31 | 2024-01-12 | 江苏初辰文化发展有限公司 | Live broadcast content evaluation detection method based on barrage sharing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108965916B (en) | 2021-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108965916A (en) | A kind of method, the method, device and equipment of model foundation of live video assessment | |
| JP6826586B2 (en) | Dependency-based container deployment methods, systems, and programs | |
| US9635101B2 (en) | Proposed storage system solution selection for service level objective management | |
| CN103645957B (en) | A kind of resources of virtual machine management-control method and device | |
| CN102759979B (en) | A kind of energy consumption of virtual machine method of estimation and device | |
| US10756976B2 (en) | Data network and execution environment replication for network automation and network applications | |
| US20200028935A1 (en) | Workload rebalancing in heterogeneous resource environments | |
| US8635484B2 (en) | Event based correlation of power events | |
| US10063427B1 (en) | Visualizing and interacting with resources of an infrastructure provisioned in a network | |
| US20060235664A1 (en) | Model-based capacity planning | |
| US10841173B2 (en) | System and method for determining resources utilization in a virtual network | |
| US20180152392A1 (en) | Hybrid cloud management | |
| US11924117B2 (en) | Automated local scaling of compute instances | |
| WO2009029496A1 (en) | Virtualization planning system | |
| EP2766810A1 (en) | Methods and systems for planning execution of an application in a cloud computing system | |
| US10616078B1 (en) | Detecting deviating resources in a virtual environment | |
| KR102062576B1 (en) | Vnf monitoring system for monitoring both virtual network function based on virtual machine and virtual network function based on container | |
| Guzek et al. | A holistic model of the performance and the energy efficiency of hypervisors in a high‐performance computing environment | |
| US11055568B2 (en) | Method and system that measure application response time | |
| US11546224B2 (en) | Virtual network layer for distributed systems | |
| US9164746B2 (en) | Automatic topology extraction and plotting with correlation to real time analytic data | |
| US20140359365A1 (en) | Integrated Configuration Management and Monitoring for Computer Systems | |
| US11886939B2 (en) | System, device, method and datastack for managing applications that manage operation of assets | |
| US20180067839A1 (en) | Using workload profiling and analytics to understand and score complexity of test environments and workloads | |
| US10805180B2 (en) | Enterprise cloud usage and alerting system |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |