A video recommendation method, device, equipment and medium

Abstract

Embodiments of the present disclosure relate to a video recommendation method, apparatus, device, and medium. The method includes: obtaining a target video and associated information of the target video; inputting the target video and the associated information of the target video into an interest probability model to obtain an interest bias elimination probability of a user for the target video, where the interest probability model is trained based on sample videos and corresponding interest bias elimination probability labels, and the interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probability of the user for the target video and the sample video to eliminate noise bias and duration bias; determining a recommended video based on the interest bias elimination probability of the target video. Since the interest bias elimination probability label considering the distribution of noise bias and duration bias of the sample video is obtained through model learning in the present disclosure, and the model is continuously updated dynamically to adapt to different distribution transformations, the accuracy of determining the interest bias elimination probability of the video is greatly improved, thereby improving the video recommendation effect.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular, to a video recommendation method, apparatus, device, and medium. Background Art

[0002] With the development of technology, watching videos has become an increasingly important form of entertainment and learning for people. The viewing duration of a user for a video is mainly affected by two factors: the length of the video itself and the noise viewing where the user loses interest after watching for a while. In related technologies, video recommendation mainly focuses on the influence of video duration, while the video recommendation method that simultaneously considers video duration and noise viewing is static and has low accuracy, affecting the video recommendation effect. Summary of the Invention

[0003] To solve the above technical problems, the present disclosure provides a video recommendation method, apparatus, device, and medium.

[0004] An embodiment of the present disclosure provides a video recommendation method, the method comprising:

[0005] Obtaining a target video and associated information of the target video;

[0006] Inputting the target video and the associated information of the target video into an interest probability model to obtain an interest deviation elimination probability of the user for the target video, where the interest probability model is trained based on a sample video and its corresponding interest deviation elimination probability label, and the interest deviation elimination probability and the interest deviation elimination probability label respectively represent the interest probabilities of the user for the target video and the sample video for eliminating noise deviation and duration deviation;

[0007] Determining a recommended video based on the interest deviation elimination probability of the target video.

[0008] An embodiment of the present disclosure further provides a video recommendation apparatus, the apparatus comprising:

[0009] An obtaining module, configured to obtain a target video and associated information of the target video;

[0010] A probability module, configured to input the target video and the associated information of the target video into an interest probability model to obtain an interest deviation elimination probability of the user for the target video, where the interest probability model is trained based on a sample video and its corresponding interest deviation elimination probability label, and the interest deviation elimination probability and the interest deviation elimination probability label respectively represent the interest probabilities of the user for the target video and the sample video for eliminating noise deviation and duration deviation;

[0011] A determining module, configured to determine a recommended video based on the interest deviation elimination probability of the target video.

[0012] An embodiment of the present disclosure also provides an electronic device, which includes: a processor; a memory for storing executable instructions executable by the processor; the processor is configured to read the executable instructions from the memory and execute the instructions to implement the video recommendation method provided by the embodiment of the present disclosure.

[0013] An embodiment of the present disclosure also provides a computer-readable storage medium storing a computer program for executing the video recommendation method provided by the embodiment of the present disclosure.

[0014] The technical solution provided by the embodiment of the present disclosure has the following advantages compared with the prior art: The video recommendation solution provided by the embodiment of the present disclosure obtains a target video and associated information of the target video; inputs the target video and the associated information of the target video into an interest probability model to obtain an interest bias elimination probability of the user for the target video, where the interest probability model is trained based on sample videos and their corresponding interest bias elimination probability labels, and the interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probabilities of the user for the target video and the sample video to eliminate noise bias and duration bias; determines a recommended video based on the interest bias elimination probability of the target video. By adopting the above technical solution, an interest probability model can be obtained through training using the interest bias elimination probability labels of sample videos that have learned to eliminate noise bias and duration bias through model learning. Using this interest probability model, the interest bias elimination probability of the target video can be determined and then the recommended video can be determined. Since the interest bias elimination probability labels of sample videos considering the distribution of noise bias and duration bias are obtained through model learning, and the model is continuously updated dynamically to adapt to different distribution transformations, compared with the static implementation in the related art, it can not only eliminate noise bias and duration bias, but also greatly improve the accuracy of determining the interest bias elimination probability of the video, thereby improving the video recommendation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In combination with the accompanying drawings and referring to the following specific embodiments, the above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the original elements and elements are not necessarily drawn to scale.

[0016] Figure 1 It is a flowchart of a video recommendation method provided by some embodiments of the present disclosure;

[0017] Figure 2 It is a flowchart of another video recommendation method provided by some embodiments of the present disclosure;

[0018] Figure 3 It is a viewing duration distribution diagram of sample videos provided by some embodiments of the present disclosure;

[0019] Figure 4 Structural schematic diagram of a hybrid distribution prediction model provided by some embodiments of the present disclosure;

[0020] Figure 5 Schematic diagram of the training process of an interest probability model provided by some embodiments of the present disclosure;

[0021] Figure 6 Schematic diagram of the prediction process of an interest probability model provided by some embodiments of the present disclosure;

[0022] Figure 7 Structural schematic diagram of a video recommendation device provided by some embodiments of the present disclosure;

[0023] Figure 8 Structural schematic diagram of an electronic device provided by some embodiments of the present disclosure. Detailed implementation manners

[0024] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0025] It should be understood that the steps recited in the method embodiments of the present disclosure can be executed in a different order and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0026] The term "including" and its variants used herein are open-ended, that is, "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0027] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0028] It should be noted that the modifications of "one" and "multiple" mentioned in the present disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly specified in the context, it should be understood as "one or more".

[0029] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are for illustrative purposes only and are not used to limit the scope of these messages or information.

[0030] In the related art, there are mainly the following several ways of video recommendation systems: One is the heuristic method, which divides the viewing duration of a user by the video duration and calls it the playback completion rate. However, the trend between the viewing duration and the duration is not a simple linear relationship, and the accuracy is relatively low; Another is to convert the duration prediction into the viewing duration quantiles grouped by video duration to reduce the negative impact of the length of the video itself; Still another is to standardize the viewing duration according to different video durations and use the standardized score as a supervision signal to train and evaluate the video recommendation model, but there is still much room for improvement. The above methods mainly focus on the influence of video duration, making the user preference signals they predict still inaccurate; While the interest debiasing method that simultaneously focuses on duration deviation and noisy viewing only has static methods at present, with relatively low accuracy, which affects the video recommendation effect.

[0031] To solve the above problems, the embodiments of the present disclosure provide a video recommendation method, which will be introduced below in combination with specific embodiments.

[0032] Figure 1 It is a schematic flowchart of the video recommendation method provided by some embodiments of the present disclosure. This method can be executed by a video recommendation device, where the device can be implemented by software and / or hardware and is generally integrated in an electronic device. As Figure 1 shown, this method includes:

[0033] Step 101, obtain a target video and the associated information of the target video.

[0034] Among them, the target video can be a video for which it is necessary to perform video analysis to determine the degree of interest of the user in the video. The number of target videos can be one or more. For example, the target video can be multiple short videos that need to be analyzed and processed in a video application. By analyzing multiple short videos, the recommended videos for subsequent recommendation are determined. The associated information of the target video includes user-related information and video-related information. The user-related information can include user historical video behavior data, user location, user gender, and other user-related information obtained after obtaining user authorization. The video-related information can include video identification, video classification, and other video-related information. The video classification can include, for example, movies and TV shows, food, sports, etc.

[0035] Specifically, the video recommendation device can obtain the target video that needs to be processed currently and the associated information of the target video. The specific acquisition source is not limited and can be determined according to business requirements. For example, it can be obtained from the service end of application programs such as video application programs and live broadcast application programs.

[0036] Step 102: Input the target video and its associated information into the interest probability model to obtain the interest bias elimination probability of the user for the target video. The interest probability model is trained based on sample videos and their corresponding interest bias elimination probability labels. The interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probabilities of the user for the target video and the sample video to eliminate noise bias and duration bias.

[0037] Among them, the interest probability model can be a model that predicts the degree of interest or the probability of interest of the user in other videos based on the viewing duration of the user for the sample videos. This interest probability model can eliminate noise bias and duration bias to obtain a more accurate prediction result. The interest bias elimination probability can be the interest probability of whether the user is interested in the video obtained through prediction, eliminating noise bias and duration bias. The viewing duration of the user for the video is mainly affected by two aspects: noise bias and duration bias. The noise bias can be the bias caused by the viewing duration of noise viewing. Noise viewing refers to the viewing when the user needs to watch the video for a period of time and then realizes that they are not interested. Fundamentally speaking, noise viewing stems from the user's trust in the recommendation system and the influence of the title, etc. At this time, the statistics of the viewing duration of the user for this video need to be eliminated to be accurate. The duration bias can be the influence of the original duration of the video on the viewing duration. Eliminating this duration bias can avoid the influence of the length of the video itself on the viewing of the video.

[0038] The interest bias elimination probability label can be the interest probability of the user for the sample video to eliminate noise bias and duration bias determined through analysis by the mixed distribution prediction model for the sample video.

[0039] In the embodiments of the present disclosure, the video recommendation device can first train an interest probability model based on the sample video and its interest bias elimination probability label, and then input the target video and its associated information into this interest probability model for analysis and processing, and output the interest bias elimination probability of the user for the target video. The value of the interest bias elimination probability is between 0 and 1. The greater the interest bias elimination probability, the greater the possibility that the user is interested in the target video.

[0040] Exemplarily, Figure 2 is a schematic flowchart of another video recommendation method provided by some embodiments of the present disclosure. As Figure 2 shown, in a feasible implementation manner, before the above step 101 or step 102, the video recommendation method of the embodiments of the present disclosure can also include training an interest probability model based on the interest bias elimination probability label obtained through model learning for the sample video. The specific steps are as follows:

[0041] Step 201: Determine the interest bias elimination probability label of the sample video based on the mixed distribution prediction model.

[0042] The mixture distribution prediction model may be a model constructed by a mixture density network that can estimate specific parameters of the potential mixture Gaussian distribution of the video. For example, Figure 3 The viewing time distribution diagram of the sample videos provided in some embodiments of the present disclosure is as follows: Figure 3 As shown in the figure, the viewing time distribution of a sample video with an original duration of 12 seconds is shown. The horizontal axis of the figure is the viewing time, and the vertical axis is the video exposure ratio. The video exposure ratio can be the ratio of the number of video viewing times to the total statistical number. The sum of the video exposure ratios of all video viewing times is 1. The viewing time due to noise is usually shorter. The peak close to the coordinate in the figure represents the noise deviation distribution, and the peak close to the right of the figure represents the duration deviation distribution, which basically presents a bimodal Gaussian distribution.

[0043] Because the noise bias and duration bias of a video exhibit a mixed Gaussian distribution within the distribution of video viewing duration, this mixed distribution prediction model can output the expected noise bias and duration bias of the video. These two expectations can then be used to generate interest-debiased probability labels for sample videos. The sample videos can be used to train the interest probability model, and their viewing duration is known. The expected noise bias represents the expected percentage of viewing time affected by noise bias when the user is not interested, while the expected duration bias represents the expected percentage of viewing time affected by duration bias when the user is interested.

[0044] In some embodiments, determining the interest debiasing probability label of a sample video based on a mixed distribution prediction model may include: taking the viewing time of the sample video as input data, training through a loss function, and obtaining the noise deviation expectation and duration deviation expectation determined by the mixed distribution prediction model for the sample video; generating the interest debiasing probability label of the sample video based on the viewing time, noise deviation expectation, and duration deviation expectation of the sample video. Optionally, generating the interest debiasing probability label of the sample video based on the viewing time, noise deviation expectation, and duration deviation expectation of the sample video may include: inputting the viewing time, noise deviation expectation, duration deviation expectation, and hyperparameters of the sample video into a preset formula for calculation to determine the interest debiasing probability label of the sample video, wherein the value of the interest debiasing probability label is between 0 and 1.

[0045] The loss function may be a loss function based on the negative log-likelihood definition error, which is only an example. Figure 4 A schematic diagram of the structure of a hybrid distribution prediction model provided in some embodiments of the present disclosure, such as Figure 4As shown, the mixed distribution prediction model may include a probability module, an expectation module, and a variance module. The probability module can be used to calculate the probability that a video belongs to a certain Gaussian distribution. The output probability of the probability module includes a noise deviation probability and a duration deviation probability. The expectation module can be used to calculate the expectation that a video belongs to a certain Gaussian distribution. The output expectation of the expectation module includes a noise deviation expectation and a duration deviation expectation. The variance module can be used to calculate the variance that a video belongs to a certain Gaussian distribution. The output variance of the variance module includes a noise deviation variance and a duration deviation variance. The loss function in the mixed distribution prediction model is related to the viewing duration of the sample video, the output probability of the probability module, the noise deviation expectation and duration deviation expectation output by the expectation module, and the output variance of the variance module.

[0046] Specifically, when determining the interest debiasing probability label of the sample video, the video recommendation device can transform the viewing duration of the sample video, such as logarithmic operation and square root operation, to obtain relevant features of the viewing duration. The viewing duration of the sample video and the relevant features are used as input data to be input into a mixture density network for training. The mixture density network includes a probability module, an expectation module, and a variance module. The parameters of the mixture density network are continuously optimized using the output of these modules with a loss function until the conditions are met. The mixture density network at this time is determined as the mixed distribution prediction model, and the noise deviation expectation and duration deviation expectation determined by the mixed distribution prediction model for the sample video are output at this time. Then, the viewing duration, noise deviation expectation, duration deviation expectation, and hyperparameters of the sample video are input into a preset formula for calculation to obtain the interest debiasing probability label.

[0047] Exemplarily, the viewing duration of the sample video can include the sum of the viewing duration affected by the duration deviation when the user is interested and the viewing duration affected by the noise deviation when the user is not interested. The viewing duration of the sample video can be represented as g, and can be expressed by the following formula y = debias_watch_label * duration_bias + (1 - debias_watch_label) * noise_bias. Then, by introducing a sensitivity factor α and an exp exponential function to adjust the debiasing distribution, a preset formula can be obtained, and the preset formula can be expressed as:

[0048]

[0049] The debias_watch_label represents the interest debiasing probability label of the sample video, α represents the hyperparameter, with a default value of -0.02, y represents the viewing duration of the sample video, noise_bias represents the noise bias expectation, and noise_bias = min(mean1, mean2), that is, the noise bias expectation can be the smaller value of the two expectations of the output expectation of the expectation module. The duration_bias represents the duration bias expectation, and duration_bias = max(mean1, mean2), that is, the duration bias expectation can be the larger value of the two expectations of the output expectation of the expectation module. The value of the interest debiasing probability label is between 0 and 1.

[0050] Step 202: Based on the sample video, the associated information of the sample video, and the interest debiasing probability label of the sample video, train the basic neural network to obtain an interest probability model.

[0051] The associated information of the sample video can include user class information and video class information of the sample video. The user class information can include user-related information such as user historical video behavior data, user location, and user gender obtained after obtaining user authorization. The video class information can include video-related information such as video identifiers and video classifications. The video classification can include, for example, film and television, food, sports, etc.

[0052] After determining the interest debiasing probability label of the sample video, the video recommendation device can use the sample video and the associated information of the sample video as input data, and the interest debiasing probability label of the sample video as output data to train the basic neural network. Taking the cross-entropy loss function as an example, the loss function of this basic neural network can be expressed as the following formula: loss = -(y * log(p) + (1 - y) * log(1 - p)), where y represents the viewing duration of the sample video, and p represents the output data of the basic neural network. Until the loss function meets the conditions, the basic neural network at this time is determined as the interest probability model.

[0053] In the above solution, by designing a mixed distribution prediction model to learn the two expectations of the mixed Gaussian distribution of the sample video in terms of viewing duration, that is, the above-mentioned noise bias expectation and duration bias expectation, and since this mixed distribution prediction model is continuously trained and updated based on streaming data, the dynamic update of the parameters of the predicted mixed Gaussian distribution is realized, ensuring sustainability and self-adaptability, and being able to adapt to different distribution changes, greatly improving the accuracy of determining the interest debiasing probability of the video.

[0054] Step 103: Determine the recommended video based on the interest debiasing probability of the target video.

[0055] Among them, the recommended videos can be videos screened from multiple videos according to the interest bias elimination probability of the videos and recommended to the user. The number of recommended videos can be one or more, which is specifically set according to the actual situation.

[0056] When the video recommendation device determines the interest bias elimination probability of the target video, it can extract the target videos with the interest bias elimination probability greater than the preset probability and determine them as recommended videos. The preset probability can be set according to the actual situation and is not specifically limited. After that, the recommended videos can be pushed to the user for viewing.

[0057] The video recommendation solution provided by the embodiments of the present disclosure obtains the target video and the associated information of the target video; inputs the target video and the associated information of the target video into the interest probability model to obtain the interest bias elimination probability of the user for the target video. Among them, the interest probability model is trained based on the sample video and its corresponding interest bias elimination probability label. The interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probability of the user for the target video and the sample video to eliminate noise bias and duration bias; determine the recommended video based on the interest bias elimination probability of the target video. By adopting the above technical solution, the interest bias elimination probability label that eliminates noise bias and duration bias obtained by model learning of the sample video can be used to train the interest probability model. The interest probability model can be used to determine the interest bias elimination probability of the target video and then determine the recommended video. Since the interest bias elimination probability label considering the distribution of noise bias and duration bias of the sample video is obtained through model learning, and the model is continuously updated dynamically to adapt to different distribution transformations. Compared with the static implementation in the related art, it can not only eliminate noise bias and duration bias, but also greatly improve the accuracy of determining the interest bias elimination probability of the video, thereby improving the video recommendation effect.

[0058] Next, the video recommendation method of the embodiments of the present disclosure will be further described through specific examples. Exemplarily, Figure 5 is a schematic diagram of the training process of the interest probability model provided by some embodiments of the present disclosure. As Figure 5 shown, the training process of the interest probability model can include: inputting the viewing duration of the sample video into the mixed distribution prediction model; the mixed distribution prediction model outputs the noise bias expectation and duration bias expectation of the sample video; generating the interest bias elimination probability label of the sample video; training to obtain the interest probability model. The specific implementation of each step can be referred to the above embodiments and will not be elaborated here one by one.

[0059] For the structure of the mixed distribution prediction model, refer to Figure 4, the mixture distribution prediction model may include a probability module, an expectation module, and a variance module. All three modules can be dense connection networks (Dense network) of [32, 2]; the activation function used after the probability module takes the normalized activation function as an example; the activation functions used after the expectation module and the variance module take the nnelu activation function as an example, and the nnelu activation function is a variant of the elu activation function; the loss function of the mixture distribution prediction model takes the loss function that defines the error based on negative log-likelihood as an example and can be expressed as Probability density function where prob represents the probability of belonging to a certain Gaussian distribution, there are two Gaussian distributions in total, mean represents the expectation of belonging to a certain Gaussian distribution, var represents the variance of belonging to a certain Gaussian distribution, and g represents the viewing duration of the video.

[0060] Exemplarily, Figure 6 is a schematic diagram of the prediction process of the interest probability model provided by some embodiments of the present disclosure. As Figure 6 shown, the prediction process may include: inputting the target video and the associated information of the target video into the interest probability model; outputting the interest debiasing probability of the target video for determining the recommended video.

[0061] This solution provides an adaptive interest debiasing probability prediction method for dynamic sustainable learning, which can more accurately predict the interest debiasing probability of the video and solve the influence of duration bias and noise viewing.

[0062] Figure 7 is a schematic structural diagram of a video recommendation device provided by an embodiment of the present disclosure. The device can be implemented by software and / or hardware and is generally integrated in an electronic device. As Figure 7 shown, the device includes:

[0063] An acquisition module 701, configured to acquire the target video and the associated information of the target video;

[0064] A probability module 702, configured to input the target video and the associated information of the target video into the interest probability model to obtain the interest debiasing probability of the user for the target video, where the interest probability model is trained based on the sample video and its corresponding interest debiasing probability label, and the interest debiasing probability and the interest debiasing probability label respectively represent the interest probabilities of the user for the target video and the sample video to eliminate noise bias and duration bias;

[0065] A determination module 703, configured to determine the recommended video based on the interest debiasing probability of the target video.

[0066] Optionally, the device further includes a model training module, and the model training module includes:

[0067] A first unit for determining an interest bias elimination probability label of a sample video based on a mixed distribution prediction model;

[0068] A second unit for training a basic neural network based on the sample video, the associated information of the sample video, and the interest bias elimination probability label of the sample video to obtain an interest probability model.

[0069] Optionally, the first unit is configured to:

[0070] Use the viewing duration of the sample video as input data, and train it through a loss function to obtain the noise bias expectation and duration bias expectation determined by the mixed distribution prediction model for the sample video;

[0071] Generate an interest bias elimination probability label for the sample video based on the viewing duration, noise bias expectation, and duration bias expectation of the sample video.

[0072] Optionally, the first unit is specifically configured to:

[0073] Input the viewing duration, noise bias expectation, duration bias expectation, and hyperparameters of the sample video into a preset formula for calculation to determine the interest bias elimination probability label of the sample video, where the value of the interest bias elimination probability label is between 0 and 1.

[0074] Optionally, the mixed distribution prediction model includes a probability module, an expectation module, and a variance module;

[0075] The loss function in the mixed distribution prediction model is related to the viewing duration of the sample video, the output probability of the probability module, the noise bias expectation and duration bias expectation output by the expectation module, and the output variance of the variance module.

[0076] Optionally, the noise bias expectation represents the expected proportion of the number of viewing durations affected by noise bias in the state where the user is not interested, and the duration bias expectation represents the expected proportion of the number of viewing durations affected by duration bias in the state where the user is interested.

[0077] Optionally, the associated information of the target video includes user class information and video class information.

[0078] The video recommendation device provided by the embodiments of the present disclosure can execute the video recommendation method provided by any embodiment of the present disclosure, and has corresponding functional modules and beneficial effects for executing the method.

[0079] The embodiments of the present disclosure also provide a computer program product, including computer programs / instructions, and when the computer programs / instructions are executed by a processor, the video recommendation method provided by any embodiment of the present disclosure is implemented.

[0080] Figure 8 The structural schematic diagram of the electronic device provided for some embodiments of the present disclosure. Specifically refer to the following Figure 8 , which shows the structural schematic diagram of the electronic device 800 suitable for implementing the embodiments of the present disclosure. The electronic device 800 in the embodiments of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The electronic device shown is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present disclosure.

[0081] As Figure 8 shown, the electronic device 800 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 801, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 802 or the program loaded from the storage device 808 into the random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 are also stored. The processing device 801, the ROM 802, and the RAM 803 are connected to each other through a bus 804. The input / output (I / O) interface 805 is also connected to the bus 804.

[0082] Generally, the following devices may be connected to the I / O interface 805: an input device 806 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 807 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 808 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 809. The communication device 809 can allow the electronic device 800 to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 8 the electronic device 800 with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. Instead, more or fewer devices may be implemented or had.

[0083] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product that includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device 809, or installed from a storage device 808, or installed from a ROM 802. When the computer program is executed by a processing device 801, the above-described functions defined in the video recommendation method of the embodiments of the present disclosure are performed.

[0084] It should be noted that the above computer-readable medium in the present disclosure can be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. And in the present disclosure, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, and the computer-readable signal medium can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0085] In some embodiments, the client and the server can communicate using any currently known or future-developed network protocol such as HTTP (HyperText Transfer Protocol), and can be interconnected with digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include local area networks ("LANs"), wide area networks ("WANs"), the Internet (e.g., the Internet), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0086] The above computer-readable medium can be included in the above electronic device; or can exist separately without being assembled into the electronic device.

[0087] The above computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device is caused to: obtain a target video and associated information of the target video; input the target video and the associated information of the target video into an interest probability model to obtain an interest bias elimination probability of the user for the target video, where the interest probability model is trained based on sample videos and their corresponding interest bias elimination probability labels, and the interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probability of the user for the target video and the sample video to eliminate noise bias and duration bias; determine a recommended video based on the interest bias elimination probability of the target video.

[0088] Computer program code for performing the operations of the present disclosure can be written in one or more programming languages or combinations thereof. The above programming languages include, but are not limited to, object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., by using an Internet service provider to connect through the Internet).

[0089] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in a different order than noted in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or by a combination of dedicated hardware and computer instructions.

[0090] The units involved in the embodiments described in the present disclosure can be implemented in software or in hardware. In some cases, the name of the unit does not constitute a limitation on the unit itself.

[0091] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. By way of example, and without limitation, the types of hardware logic components that may be used include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0092] In the context of the present disclosure, a machine-readable medium may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0093] It should be understood that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, usage scopes, usage scenarios, etc. of the information involved in the present disclosure should be informed to users and the authorization of users should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0094] The above description is only a preferred embodiment of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosure concept. For example, the technical solutions formed by mutually replacing the above features with the technical features (but not limited to) having similar functions disclosed in the present disclosure.

[0095] In addition, although the operations are depicted in a particular order, this should not be construed as requiring that the operations be performed in the particular order shown or in sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features described in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, the various features described in the context of a single embodiment may also be implemented separately or in any suitable sub-combination in multiple embodiments.

[0096] Although the subject matter has been described in language specific to structural features and / or methodological logical acts, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. On the contrary, the specific features and acts described above are merely example forms for implementing the claims.

Claims

1. A video recommendation method, characterized in that, Including: Obtain a target video and associated information of the target video; Input the target video and the associated information of the target video into an interest probability model to obtain the interest bias elimination probability of the user for the target video, where the interest probability model is trained based on a sample video and its corresponding interest bias elimination probability label, and the interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probabilities of the user for the target video and the sample video to eliminate noise bias and duration bias; Determine a recommended video based on the interest bias elimination probability of the target video; Wherein, the method further includes: Determine the interest bias elimination probability label of the sample video based on a mixture distribution prediction model; the mixture distribution prediction model is a model trained based on a mixture density network, and the mixture density network is used to estimate the parameters of the potential mixture Gaussian distribution of the video; the mixture distribution prediction model includes a probability module, an expectation module, and a variance module, and the loss function in the mixture distribution prediction model is related to the viewing duration of the sample video, the output probability of the probability module, the noise bias expectation and duration bias expectation output by the expectation module, and the output variance of the variance module; Train a basic neural network based on the sample video, the associated information of the sample video, and the interest bias elimination probability label of the sample video to obtain an interest probability model.

2. The method according to claim 1, characterized in that, The determining the interest bias elimination probability label of the sample video based on the mixture distribution prediction model includes: Use the viewing duration of the sample video as input data and train it through a loss function to obtain the noise bias expectation and duration bias expectation determined by the mixture distribution prediction model for the sample video; Generate the interest bias elimination probability label of the sample video based on the viewing duration, noise bias expectation, and duration bias expectation of the sample video.

3. The method according to claim 2, characterized in that, Generating the interest bias elimination probability label of the sample video based on the viewing duration, noise bias expectation, and duration bias expectation of the sample video includes: Calculate the viewing duration, noise bias expectation, duration bias expectation, and hyperparameters of the sample video to determine the interest bias elimination probability label of the sample video, where the value of the interest bias elimination probability label is between 0 and 1.

4. The method according to claim 2, characterized in that The noise bias expectation represents the expected proportion of the viewing duration affected by the noise bias in the state where the user is not interested, and the duration bias expectation represents the expected proportion of the viewing duration affected by the duration bias in the state where the user is interested.

5. The method according to claim 1, wherein The associated information of the target video includes user class information and video class information.

6. A video recommendation device, characterized in that, Including: An acquisition module for acquiring a target video and the associated information of the target video; A probability module for inputting the target video and the associated information of the target video into an interest probability model to obtain the interest bias elimination probability of the user for the target video, where the interest probability model is trained based on a sample video and its corresponding interest bias elimination probability label, and the interest bias elimination probability and the interest bias elimination probability label respectively represent the interest probabilities of the user for the target video and the sample video to eliminate noise bias and duration bias; A determination module, configured to determine a recommended video based on the interest bias elimination probability of the target video; Wherein, the apparatus further includes a model training module, and the model training module includes: A first unit, configured to determine an interest bias elimination probability label of a sample video based on a mixture distribution prediction model; the mixture distribution prediction model is a model trained based on a mixture density network, and the mixture density network is used to estimate parameters of a potential mixture Gaussian distribution of a video; the mixture distribution prediction model includes a probability module, an expectation module, and a variance module, and a loss function in the mixture distribution prediction model is related to the viewing duration of the sample video, the output probability of the probability module, the expected noise deviation and duration deviation expected by the expectation module, and the output variance of the variance module; A second unit, configured to train a basic neural network based on the sample video, the associated information of the sample video, and the interest bias elimination probability label of the sample video to obtain an interest probability model.

7. An electronic device, characterized in that, The electronic device includes: A processor; A memory for storing executable instructions of the processor; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the video recommendation method according to any one of claims 1-5 above.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and the computer program is used to execute the video recommendation method according to any one of claims 1-5 above.

Images