Model determination method, model application method and related device
By obtaining sample information collection and adjusting model parameters, predicting the video display speed twice, solving the problem of not considering the intention of the video viewing object in the prior art, improving the information reception experience, and ensuring that the video speed twice meets the needs of the viewing object.
Patent Information
- Application Number
- CN202310667323.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-06
- Publication Date
- 2024-12-06
AI Technical Summary
The prior art does not consider the actual selection intention of the video viewing object when adjusting the video playback speed, which makes it difficult for the video viewing object to meet the needs of the viewing object and affects the video viewing experience.
By obtaining a collection of sample information, including the sample speed corresponding to the respective sample information, the initial speed prediction model is used to determine the pending speed, and the model parameters are adjusted according to the difference between the target sample speed and the pending speed, and the speed prediction model is obtained to predict the information Display the speed twice to ensure that the speed twice meets the actual needs of the information receiving object.
The information reception experience is improved. By analyzing the information content and the display needs of the receiving object, and adjusting the model parameters, the predicted information display speed is more in line with the actual needs of the information reception object, and the video viewing experience is improved.
Smart Images

Figure CN119094835A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing, and in particular to a model determination method, a model application method and related devices. Background Art
[0002] In the multimedia era, video has become one of the main forms of information dissemination, and watching video is an important way for people to receive information in their daily lives.
[0003] In the related art, in order to provide a more comfortable video viewing experience for the video viewers, the video content can be analyzed to adjust the video playback speed. For example, when the action of a character in a video clip is fast, the video playback speed will be automatically slowed down so that the video viewers can see the action of the character in the video clearly; when there are too many lines in a video clip, the video playback speed will be automatically slowed down so that the video viewers can hear the lines clearly.
[0004] However, the video speed determination method in the related art is only based on the video content itself, and does not take into account the actual selection intention of the video viewers for the video speed. Therefore, the determined video speed is difficult to meet the actual needs of the video viewers, and the effect of improving the video viewing experience is poor. Summary of the invention
[0005] In order to solve the above technical problems, the present application provides a model determination method, and the model determined by this method can determine the information display speed that is more in line with the actual needs of the object, thereby improving the information receiving experience.
[0006] The embodiments of the present application disclose the following technical solutions:
[0007] In a first aspect, an embodiment of the present application discloses a model determination method, the method comprising:
[0008] Acquire a sample information set, wherein the sample information set includes a plurality of sample information, and the plurality of sample information has respectively corresponding sample speeds;
[0009] The plurality of sample information are respectively used as target sample information, and a to-be-determined speed is determined according to the target sample information through an initial speed prediction model;
[0010] The model parameters corresponding to the initial speed prediction model are adjusted according to the difference between the target sample speed and the to-be-determined speed to obtain a speed prediction model, wherein the target sample speed is the information display speed when the target sample information is displayed to the information receiving object, and the target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
[0011] In a second aspect, an embodiment of the present application discloses a model application method, the method comprising:
[0012] Get the information to be displayed;
[0013] Determine, by a speed prediction model, a first information display speed corresponding to the information to be displayed according to the information to be displayed, wherein the speed prediction model is obtained by training based on a plurality of sample information, the plurality of sample information have respectively corresponding sample speeds, the target sample speed is the information display speed when displaying the target sample information to the information receiving object, the target sample speed is the sample speed corresponding to the target sample information, and the target sample information is any one of the plurality of sample information;
[0014] The information to be displayed is displayed to the object to be displayed according to the first information display speed.
[0015] In a possible implementation, the method further includes:
[0016] Acquire a speed selection operation of the to-be-displayed object for the to-be-displayed information, wherein the speed selection operation corresponds to a second information display speed;
[0017] The information display speed corresponding to the information to be displayed is updated from the first information display speed to the second information display speed.
[0018] In a third aspect, an embodiment of the present application discloses a model determination device, the device comprising a first acquisition unit, a first determination unit and an adjustment unit:
[0019] The first acquisition unit is used to acquire a sample information set, wherein the sample information set includes a plurality of sample information, and the plurality of sample information has respectively corresponding sample speeds;
[0020] The first determining unit is configured to use the plurality of sample information as target sample information respectively, and determine the to-be-determined speed according to the target sample information through an initial speed prediction model;
[0021] The adjustment unit is used to adjust the model parameters corresponding to the initial speed prediction model according to the difference between the target sample speed and the to-be-determined speed to obtain the speed prediction model. The target sample speed is the information display speed when the target sample information is displayed to the information receiving object. The target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
[0022] In a possible implementation manner, the first determining unit is specifically configured to:
[0023] Generate a content feature corresponding to the target sample information, wherein the content feature is used to characterize the information content corresponding to the target sample information;
[0024] The undetermined speed is determined according to the content characteristics.
[0025] In one possible implementation, the content features include any one or more combinations of object features, scene features, and information volume features, wherein the object features are used to identify the objects included in the target sample information, the scene features are used to identify the scenes corresponding to the target sample information, and the information volume features are used to characterize the amount of information displayed by the target sample information.
[0026] In a possible implementation manner, the target sample information is video information, the target sample information includes multiple video frames, the content feature includes the object feature, the scene feature and the information volume feature, and the first determining unit is specifically configured to:
[0027] By performing object recognition on the multiple video frames, an object feature corresponding to the target sample information is generated, wherein the object feature includes the number of video frames corresponding to the multiple objects included in the target sample information, the number of video frames corresponding to the target object is the number of video frames including the target object in the multiple video frames, and the target object is any one of the multiple objects;
[0028] By performing scene recognition on the target sample information, a scene feature corresponding to the target sample information is generated, wherein the scene feature includes probability information corresponding to a plurality of scenes respectively, and the probability information corresponding to the target scene is used to identify the probability that the target sample information corresponds to the target scene, and the target scene is any one of the plurality of scenes;
[0029] By performing text extraction on the target sample information, an information volume feature corresponding to the target sample information is generated, and the information volume feature includes at least one of the following: the total number of texts corresponding to the target sample information, and the average number of texts corresponding to the target sample information in a preset time unit.
[0030] In one possible implementation, the information receiving object corresponds to a target object type, the first information display speed is used to display the information to be displayed to an object corresponding to the target object type, the target object type is determined based on object information corresponding to the information receiving object, the object information corresponding to the information receiving object is used to characterize the acceptance of the information receiving object for information constituent elements, and the information constituent elements are used to constitute the information content corresponding to the information.
[0031] In a possible implementation, the device further includes a first display unit, a second acquisition unit, and a first update unit:
[0032] The first display unit is used to display the information to be displayed to the target information receiving object according to the first information display speed;
[0033] The second acquisition unit is used for acquiring a changed second information display speed in response to the target information receiving object changing the information display speed corresponding to the information to be displayed;
[0034] The first updating unit is used to display the speed according to the second information and update the model parameters corresponding to the speed prediction model.
[0035] In a possible implementation manner, the first updating unit is specifically configured to:
[0036] Determine the target undetermined speed according to the information to be displayed through a target speed prediction model, wherein the initial model parameters of the target speed prediction model are consistent with the model parameters of the speed prediction model adjusted by the multiple sample information, and the target speed prediction model and the speed prediction model have the same model architecture;
[0037] According to the difference between the second information display speed and the target undetermined speed, the model parameters corresponding to the target speed prediction model are adjusted to obtain the second model parameters,
[0038] Obtaining a first model parameter, where the first model parameter is a model parameter corresponding to the speed prediction model before the speed prediction model is updated based on the second information display speed;
[0039] Determine a first parameter to be updated according to the first model parameter and the first parameter, and determine a second parameter to be updated according to the second model parameter and the second parameter;
[0040] The model parameters corresponding to the speed prediction model after the model parameters are updated are determined according to the first parameters to be updated and the second parameters to be updated.
[0041] In a possible implementation manner, the plurality of sample information and the information to be displayed correspond to target information content types.
[0042] In a possible implementation, the target information is composed of N sub-information, the multiple sample information is M sub-information that has been displayed to the information receiving object, the information to be displayed is any sub-information among the N sub-information that has not been displayed to the information receiving object, the first information display speed is used to display the information to be displayed to the information receiving object, N is a positive integer greater than 2, and M is a positive integer greater than 1 and less than N.
[0043] In a possible implementation, the device further includes a segmentation unit:
[0044] The segmentation unit is used to segment the target information through scene analysis to obtain the N sub-information, and the scenes corresponding to two adjacent sub-information in the target information are different.
[0045] In a possible implementation, the model parameters include sub-model parameters corresponding to multiple speeds respectively, and the target sub-model parameters corresponding to the target speed are used to determine the probability that the information to be displayed corresponds to the target speed, and the target speed is any one of the multiple speeds.
[0046] In a fourth aspect, an embodiment of the present application discloses a model application device, the device comprising a third acquisition unit, a second determination unit, and a second display unit:
[0047] The third acquisition unit is used to acquire information to be displayed;
[0048] The second determination unit is used to determine a first information display speed corresponding to the information to be displayed according to the information to be displayed by using a speed prediction model, wherein the speed prediction model is obtained by training based on a plurality of sample information, the plurality of sample information have respectively corresponding sample speeds, the target sample speed is the information display speed when the target sample information is displayed to the information receiving object, the target sample speed is the sample speed corresponding to the target sample information, and the target sample information is any one of the plurality of sample information;
[0049] The second display unit is used to display the information to be displayed to the object to be displayed according to the first information display speed.
[0050] In a possible implementation manner, the information receiving object corresponds to a target object type, and the device further includes a fourth acquiring unit and a third determining unit:
[0051] The fourth acquisition unit is used to acquire object information corresponding to the object to be displayed, wherein the object information is used to represent the acceptance of the information constituent elements to be displayed, and the information constituent elements are used to constitute the information content corresponding to the information;
[0052] The third determining unit is used to determine the object type corresponding to the object to be displayed according to the object information;
[0053] The second display unit is specifically used for:
[0054] In response to the object type being the target object type, the information to be displayed is displayed to the object to be displayed according to the first information display speed.
[0055] In a possible implementation manner, the object information includes a to-be-determined object type, and the fourth acquiring unit is specifically configured to:
[0056] In response to obtaining a type selection operation of the object to be displayed for the to-be-determined object type, determining to obtain object information corresponding to the object to be displayed;
[0057] The third determining unit is specifically configured to:
[0058] The to-be-determined object type is determined as the object type corresponding to the to-be-displayed object.
[0059] In a possible implementation manner, the device further includes a fifth acquiring unit and a second updating unit:
[0060] The fifth acquisition unit is used to acquire a speed selection operation of the to-be-displayed object for the to-be-displayed information, wherein the speed selection operation corresponds to a second information display speed;
[0061] The second updating unit is used to update the information display speed corresponding to the information to be displayed from the first information display speed to the second information display speed.
[0062] In a possible implementation manner, the plurality of sample information corresponds to a target information content type, and the apparatus further includes a fourth determining unit:
[0063] The fourth determining unit is used to determine the to-be-determined information content type corresponding to the to-be-displayed information;
[0064] The second determining unit is specifically configured to:
[0065] In response to the to-be-determined information content type being consistent with the target information content type, a first information display speed corresponding to the to-be-displayed information is determined according to the to-be-displayed information through a speed prediction model.
[0066] In a fifth aspect, an embodiment of the present application discloses a computer device, wherein the computer device includes a processor and a memory:
[0067] The memory is used to store a computer program and transmit the computer program to the processor;
[0068] The processor is used to execute the model determination method described in any one of the first aspects, or execute the model application method described in any one of the second aspects according to the instructions in the computer program;
[0069] In a sixth aspect, an embodiment of the present application discloses a computer-readable storage medium, wherein the computer-readable storage medium is used to store a computer program, wherein the computer program is used to execute the model determination method described in any one of the first aspects, or execute the model application method described in any one of the second aspects;
[0070] In the seventh aspect, an embodiment of the present application discloses a computer program product including a computer program, which, when running on a computer device, enables the computer device to execute the model determination method described in any one of the first aspects, or execute the model application method described in any one of the second aspects.
[0071] It can be seen from the above technical solution that in order to make the information display speed of the information received by the information receiving object more in line with the actual needs of the information receiving object, multiple sample information with corresponding sample speeds will be obtained during model training. The sample speed corresponding to the sample information is the information display speed when the sample information is displayed to the information receiving object. Therefore, the sample speed can reflect the actual needs of the information receiving object for the information display speed of the sample information. Therefore, based on the difference between the sample speed and the pending speed predicted by the initial speed prediction model based on the sample information, it can reflect the degree of fit of the initial speed prediction model to the actual needs of the information receiving object when the speed prediction is based on the sample information. Therefore, adjusting the model parameters based on the difference can enable the initial speed prediction model to learn how to make the predicted information display speed more in line with the actual needs of the information receiving object, and obtain a speed prediction model. The speed prediction model can be used to predict the information display speed corresponding to the information to be displayed. It can be seen from this that the speed prediction model in the present application can, on the one hand, analyze the information content of the information to be displayed itself, so that the determined information display speed fits the information content of the information to be displayed; on the other hand, it adjusts the model parameters by referring to the information display speed actually adopted by the information receiving object, so that the predicted information display speed can better meet the actual needs of the information receiving object, thereby improving the information receiving experience of the information receiving object. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0073] Figure 1 A schematic diagram of a model determination method in an actual application scenario provided by an embodiment of the present application;
[0074] Figure 2 A flow chart of a model determination method provided in an embodiment of the present application;
[0075] Figure 3 A schematic diagram of a model determination method provided in an embodiment of the present application;
[0076] Figure 4 A flow chart of a model application method provided in an embodiment of the present application;
[0077] Figure 5 A schematic diagram of a model determination method in an actual application scenario provided by an embodiment of the present application;
[0078] Figure 6 A structural block diagram of a model determination device provided in an embodiment of the present application;
[0079] Figure 7 A structural block diagram of a model application device provided in an embodiment of the present application;
[0080] Figure 8 A structural diagram of a terminal provided in an embodiment of the present application;
[0081] Fig. 9 A structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION
[0082] The embodiments of the present application are described below in conjunction with the accompanying drawings.
[0083] In the related art, the analysis of the speed of information display only combines the content of the information itself, and does not refer to the actual speed requirements of the information recipient when receiving the information. For example, SpeedNet is a deep learning algorithm for video speed estimation. Its main idea is to use convolutional neural networks to model the motion information in video sequences to predict the speed of the video, and then increase the speed for slow-moving segments and reduce the speed for fast-moving segments, so as to bring a more balanced video motion speed to the recipient of the video information, for example, it can keep the characters in the video at a more balanced speed.
[0084] However, in actual scenarios, the information receiving object's demand for the speed of displaying information is not simply based on content factors such as motion speed balance. For example, some information receiving objects may prefer video content with faster character movement speeds. In this case, if the video speed is adjusted only based on the video content itself, the video display speed will not meet the actual needs of the information receiving object.
[0085] In order to solve the above technical problems, the present application provides a model determination method, which can adjust the model parameters based on the two dimensions of the information itself and the actual needs of the information receiving object for the information display speed. Through the difference between the pending speed determined by the model based on sample information and the sample speed, the model can learn how to predict the information display speed that fits the information receiving needs of the information receiving object, thereby improving the accuracy and rationality of the information display speed prediction and improving the information receiving experience of the information receiving object.
[0086] It is understandable that the method can be applied to a computer device, which can be, for example, a terminal device or a server. The method can be executed independently by a terminal device or a server, or can be applied to a network scenario where a terminal device and a server communicate, and is executed by the cooperation of the terminal device and the server. Among them, the terminal device can be a mobile phone, a tablet computer, a laptop computer, a desktop computer, a vehicle-mounted device, a smart TV, and other devices. The server can be understood as an application server or a Web server. In actual deployment, the server can be an independent server, a cluster server, or a cloud server, etc.
[0087] In order to facilitate understanding of the technical solution provided in the embodiment of the present application, a model determination method provided in the embodiment of the present application will be introduced below in combination with an actual application scenario.
[0088] See also Figure 1 , Figure 1 A schematic diagram of a model determination method in an actual application scenario provided in an embodiment of the present application. In the actual application scenario, a computer device is taken as server 101 as an example for explanation, and the sample information is sample video information.
[0089] like Figure 1 As shown, the server 101 can obtain a sample information set, which includes N sample video information, namely, sample video information 1, sample video information 2, ..., sample video information N, each of which has a corresponding sample speed, and the sample speed is the information display speed when the sample video information is displayed to the information receiving object. For example, the sample speed corresponding to the sample video information 1 is 0.5 times the speed, that is, the information display speed (i.e., the video playback speed) used by the information receiving object when watching the sample video information 1 is 0.5 times the speed, that is, the sample video information 1 is slowed down by one time.
[0090] In order to enable the model to predict the information display speed in combination with the information content, the server 101 can use each sample video information as a model input for model training. For example, the server 101 can use the initial speed prediction model to determine the undetermined speed according to the sample video information 1, and the undetermined speed is the information display speed determined by the model for the sample video information 1. Figure 1As shown, the pending speed determined by the model is 1.0 speed, while the sample speed 1 corresponding to the sample video information 1 is 0.5 speed, which means that the information display speed determined by the model based on the sample video information 1 is different from the information display speed actually required by the information receiving object. In order to enable the model to learn how to meet the actual information receiving needs of the information receiving object, the server 101 can adjust the model parameters corresponding to the initial speed prediction model according to the difference between the pending speed and the sample speed 1, so that the pending speed determined by the model based on the sample video information 1 gradually approaches the sample speed 1. Through this method, combined with N sample video information, a speed prediction model can be trained, and the speed prediction model has a relatively accurate speed prediction capability.
[0091] In actual applications, when there is video information to be displayed that needs to be displayed to the information receiving object, the server 101 can use the speed prediction model to determine the first information display speed corresponding to the video information to be displayed based on the video information to be displayed, and display the video information to be displayed to the information receiving object according to the first information display speed. For example, in this actual application scenario, the server 101 can display the video information to be displayed to the information receiving object at a predicted speed of 1.0, so that the information display speed of the video information to be displayed can be more in line with the actual speed requirements of the information receiving object, thereby bringing the information receiving object a better information receiving experience.
[0092] Next, the technical solution provided by this application will be generally introduced in conjunction with the accompanying drawings.
[0093] See also Figure 2 , Figure 2 A flowchart of a model determination method provided in an embodiment of the present application, the method is executed by a computer device. The method includes:
[0094] S201: Obtain a sample information set.
[0095] The sample information set includes multiple sample information, and the multiple sample information has corresponding sample speeds. The sample speed is the information display speed corresponding to when the sample information is displayed to the information receiving object, that is, the sample speed is the information display speed received by the information receiving object when the sample information is displayed to the information receiving object. Therefore, the sample speed can reflect the actual needs of the information receiving object for the information display speed. Among them, the information receiving object can be any object that can receive sample information, such as the audience of video information, the listener of audio information, etc.; the sample information can include multiple types of information that support changes in information display speed, such as video information, audio information, etc. The information display speed is the speed when the information is displayed to the information receiving object, such as the video playback speed, the audio playback speed, etc.
[0096] S202: taking the plurality of sample information as target sample information respectively, and determining the to-be-determined speed according to the target sample information through an initial speed prediction model.
[0097] In order to make the information display speed predicted by the model match the content displayed by the information, the computer device can use the sample information itself as the input of the initial speed prediction model, so that the initial speed model performs speed prediction based on the sample information. The computer device can use multiple sample information as target sample information respectively, input the target sample information into the initial speed prediction model, and obtain the undetermined speed output by the model, which is the information display speed predicted by the initial speed prediction model based on the target sample information itself.
[0098] S203: adjusting the model parameters corresponding to the initial speed prediction model according to the difference between the target sample speed and the to-be-determined speed, to obtain the speed prediction model.
[0099] Among them, the target sample speed is the sample speed corresponding to the target sample information, and the target sample speed is the information display speed when the target sample information is displayed to the information receiving object. The target sample speed can reflect the actual demand of the information receiving object for the information display speed of the target sample information. Therefore, through the difference between the target sample speed and the pending speed, it can reflect the degree of fit of the speed predicted by the initial speed prediction model to the actual demand of the information receiving object, and then the computer device can adjust the model parameters corresponding to the initial speed prediction model based on the difference, so that the speed determined by the initial speed prediction model based on the sample information gradually approaches the sample speed corresponding to the sample information. In this way, the initial speed prediction model can learn how to determine the information display speed that meets the actual information display speed demand of the information receiving object, and obtain the speed prediction model, which is the initial speed prediction model after parameter adjustment.
[0100] The speed prediction model can be used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed, and the information to be displayed can be any information that can be displayed to the information receiving object. It should be emphasized that the information receiving object corresponding to the information to be displayed and the information receiving object corresponding to the sample information can be the same information receiving object or different information receiving objects, which is not limited here.
[0101] It can be seen from the above technical solution that in order to make the information display speed of the information received by the information receiving object more in line with the actual needs of the information receiving object, multiple sample information with corresponding sample speeds will be obtained during model training. The sample speed corresponding to the sample information is the information display speed when the sample information is displayed to the information receiving object. Therefore, the sample speed can reflect the actual needs of the information receiving object for the information display speed of the sample information. Therefore, based on the difference between the sample speed and the pending speed predicted by the initial speed prediction model based on the sample information, it can reflect the degree of fit of the initial speed prediction model to the actual needs of the information receiving object when the speed prediction is based on the sample information. Therefore, adjusting the model parameters based on the difference can enable the initial speed prediction model to learn how to make the predicted information display speed more in line with the actual needs of the information receiving object, and obtain a speed prediction model. The speed prediction model can be used to predict the information display speed corresponding to the information to be displayed. It can be seen from this that the speed prediction model in the present application can, on the one hand, analyze the information content of the information to be displayed itself, so that the determined information display speed fits the information content of the information to be displayed; on the other hand, it adjusts the model parameters by referring to the information display speed actually adopted by the information receiving object, so that the predicted information display speed can better meet the actual needs of the information receiving object, thereby improving the information receiving experience of the information receiving object.
[0102] It is understandable that the content displayed through the information is the main factor affecting the information receiving object's demand for information receiving speed. For example, content factors such as the amount of information per unit time, the objects appearing in the information or the corresponding scenes in the information will affect the information receiving experience of the information receiving object. Therefore, the information receiving object will adjust the information display speed based on these content factors.
[0103] Based on this, in a possible implementation, when determining the to-be-determined speed according to the target sample information, the computer device may execute steps S2021-S2022 (not shown in the figure), and steps S2021-S2022 are a possible implementation of step S202:
[0104] S2021: Generate content features corresponding to target sample information.
[0105] The computer device can generate content features corresponding to the target sample information based on information analysis of the target sample information. The content features are used to characterize the information content corresponding to the target sample information, for example, the amount of information contained in the target sample information, the objects included or the corresponding scenes, etc.
[0106] S2022: Determine a pending speed multiple according to content characteristics.
[0107] By determining the pending speed through content features, the information display speed predicted by the model can be more closely aligned with the information display speed of the information content, thereby making the speed determination method more consistent with the adjustment logic of the information receiving object for adjusting the information display speed, and obtaining a more accurate information display speed.
[0108] Among them, content features may include multiple types. Specifically, in a possible implementation, content features include any one or more combinations of object features, scene features and information volume features. Object features are used to identify objects included in the target sample information, such as people, animals, objects, etc. that appear in video information or audio information. Since the information receiving object may have different intention levels for different objects, the information display speed may be slowed down for objects with high intention levels to be received carefully, and the information display speed may be accelerated for objects with low intention levels to be received as quickly as possible. Therefore, the objects included in the information will affect the information receiving object's demand for the information display speed to a certain extent.
[0109] The scene feature is used to identify the scene corresponding to the target sample information. Similarly, since the information recipient may have different intentions for different scenes, the scene corresponding to the information will also affect the information recipient's demand for the information display speed to a certain extent.
[0110] The information volume feature is used to characterize the amount of information displayed by the target sample information. For example, for video information or audio information, the information volume feature can be the number of words displayed per minute. The amount of information is also one of the important factors that affect the information receiving object's demand for information display speed. For example, when the amount of information is large, the information receiving object will choose to reduce the information display speed to fully receive the information; when the amount of information is small, the information receiving object will choose to increase the information display speed to improve the efficiency of information reception.
[0111] Next, we will take video information as an example to explain in detail how to extract the above-mentioned multiple content features. In this embodiment, the target sample information is video information, the target sample information includes multiple video frames, the content features include object features, scene features and information volume features, and the computer device can execute steps S20211-S20213 (not shown in the figure), and steps S20211-S20213 are a possible implementation of step S2021:
[0112] S20211: Generate object features corresponding to target sample information by performing object recognition on multiple video frames.
[0113] The object feature includes the number of video frames corresponding to the multiple objects included in the target sample information, and the number of video frames corresponding to the target object is the number of video frames including the target object in the multiple video frames, and the target object is any one of the multiple objects. The computer device can determine the objects included in each video frame through object recognition, and then count the number of video frames where each object appears to obtain the object feature. The object feature can be as follows: Figure 3 The object feature matrix shown in the figure, the numbers in the matrix represent the number of video frames in which an object appears. There are many ways to identify an object. For example, in the embodiment of the present application, the face area appearing in the video frame can be detected by first using a multi-task cascaded convolutional network (MTCNN for short), and then the face recognition algorithm ArcFace is used to identify the object in each face area.
[0114] S20212: Generate scene features corresponding to the target sample information by performing scene recognition on the target sample information.
[0115] It is understandable that when identifying a scene, the output is usually the probability of each scene corresponding to the information, and then the information is determined based on the probability to which scene it corresponds. Therefore, in this application, the probability can be directly used as a scene feature, which includes probability information corresponding to multiple scenes, and the probability information corresponding to the target scene is used to identify the probability of the target sample information corresponding to the target scene, and the target scene is any one of the multiple scenes. The scene recognition method can also include multiple methods, and in this application, a video processing model (Video SwinTransformer) based on the visual attention mechanism can be used to implement it.
[0116] S20213: Generate information volume features corresponding to the target sample information by performing text extraction on the target sample information.
[0117] Among them, the way in which the computer device extracts text may be different for different information components of the target sample information. For example, for text located on an image in video information (such as subtitles, etc.), text extraction can be performed through image text recognition. For text displayed through audio information in video information, text extraction can be performed through speech recognition.
[0118] The information volume features may include multiple types, for example, at least one of the following: the total number of texts corresponding to the target sample information, the average number of texts corresponding to the target sample information in a preset time unit, such as the number of texts corresponding to each minute, etc.
[0119] It is understandable that different information recipients may have different preferences for information display speeds. For example, some information recipients prefer people-like objects, so they would like to receive information about people-like objects at a lower information display speed, while some information recipients prefer non-people-like objects, so they would like to receive information about non-people-like objects at a lower information display speed. It can be seen that different information recipients have different acceptance of the elements that constitute the information content, and different information display speeds will be selected for the same information.
[0120] Based on this, in a possible implementation, in order to further improve the degree of matching of the information display speed with the actual needs of the information receiving object, the computer device can perform targeted speed prediction for objects with different acceptance of information content constituent elements. For example, the information receiving object corresponding to the above sample information corresponds to a target object type, and the target object type is determined based on the object information corresponding to the information receiving object. The object information corresponding to the information receiving object is used to characterize the acceptance of the information receiving object to the information constituent elements, such as the language environment of the object, the historical browsing information of the object, etc. The information constituent elements are used to constitute the information content corresponding to the information, such as the object in the information content, the scene corresponding to the information content, the text contained in the information, etc. The target object type can be any one of multiple object types. Since the object type is determined based on the object information, the acceptance of multiple objects corresponding to the same object type on the information constituent elements is relatively similar, so that when multiple objects corresponding to the same object type receive the same information, the demand for information display speed is also relatively consistent.
[0121] Based on this, since the speed prediction model is trained based on the sample information, the speed prediction model can be used to make a more reasonable speed prediction for the object of the target object type. The first information display speed determined by the speed prediction model based on the information to be displayed can be specifically used to display the information to be displayed to the object of the corresponding target object type, thereby bringing the information receiving object an information display experience that is more in line with the information display speed requirements.
[0122] Of course, the model can be determined not only for objects of a certain object type, but also for each object itself. For example, the sample speed corresponding to the sample information obtained by the computer device is the information display speed when it is displayed to a certain object, and the information to be displayed is the information that needs to be displayed to the object.
[0123] In addition to making targeted predictions based on the acceptance of the information elements of the information receiving object, in another possible implementation, the computer device can also make speed predictions for information of different information content types. It is understandable that the information receiving object may prefer different information display speeds when receiving different types of information. For example, since video information has both pictures and sounds, and audio information only has sounds, the information receiving object usually prefers a slower information display speed for video information than for audio information. Alternatively, the information receiving object may have different requirements for the information display speed for video information of two different content types, documentaries and TV series. The information display speed requirement for documentaries is usually lower than that for TV series.
[0124] Based on this, the multiple sample information and information to be displayed obtained by the computer device can all correspond to the target information content type, thereby ensuring that the information used to determine the speed during application is the same as the sample information type used during training, resulting in more accurate information display speed prediction.
[0125] It is understandable that although the speed prediction model can predict the information display speed that is more in line with the actual needs of the information recipient, in some cases, the information recipient may choose other information display speeds, which means that the information display speed predicted by the speed prediction model may be inaccurate. Based on this, in a possible implementation method, the computer device can update the parameters of the speed prediction model based on the feedback of the information recipient on the predicted information display speed during the model application process.
[0126] The computer device may display the information to be displayed to the target information receiving object according to the first information display speed, and the target information receiving object may be any information receiving object.
[0127] In response to the target information receiving object changing the information display speed corresponding to the information to be displayed, it means that the first information display speed cannot meet the information display speed requirement of the target information receiving object for the information to be displayed. At this time, the computer device can obtain the changed second information display speed. Since the second information display speed is obtained by the change of the target information receiving object, the second information display speed can more accurately reflect the information display speed requirement of the target information receiving object for the information to be displayed. Therefore, the computer device can update the model parameters corresponding to the speed prediction model according to the second information display speed, so as to timely correct the speed prediction defects of the speed prediction model in some cases, and continuously improve the speed prediction model's prediction ability for the information display speed, thereby bringing more accurate information display speed prediction.
[0128] Among them, in some scenarios, the change of information display speed by the information receiving object may be caused by some accidental factors. For example, the information receiving object may accidentally touch the control for changing the information display speed. In this case, the second information display speed may be significantly different from the information display speed of the information to be displayed by the information receiving object under normal circumstances. Therefore, if the model parameters are updated only based on the second information display speed, the model may over-learn the second information display speed, resulting in overfitting.
[0129] Based on this, in a possible implementation, when updating parameters, the computer device can combine historical model parameters to form new model parameters, reducing the overall impact of a single update on the model parameters. When the speed prediction model is obtained through the above-mentioned multiple sample information training, the computer device can copy a target speed prediction model that is the same as the speed prediction model, that is, the initial model parameters of the target speed prediction model are consistent with the model parameters obtained by adjusting the speed prediction model through multiple sample information, and the target speed prediction model has the same model architecture as the speed prediction model. The model parameters of the target speed prediction model are only affected by the actual information display speed that is inconsistent with the predicted information display speed during the application of the speed prediction model.
[0130] For example, when updating the model parameters corresponding to the speed prediction model according to the display speed of the second information, the computer device can first determine the target speed according to the information to be displayed through the target speed prediction model, and then adjust the model parameters corresponding to the target speed prediction model according to the difference between the second information display speed and the first information display speed to obtain the second model parameters. It can be seen that the model parameter update of the target speed prediction model is only affected by the new training samples. The target speed can reflect the speed prediction model's ability to predict the speed of the information to be displayed under the continuous update of new training samples.
[0131] The computer device may also obtain a first model parameter, which is a model parameter corresponding to the speed prediction model before the model parameter of the speed prediction model is updated based on the speed displayed by the second information. When updating the model parameter corresponding to the speed prediction model, the computer device may determine a first parameter to be updated based on the first model parameter and the first parameter, and determine a second parameter to be updated based on the second model parameter and the second parameter, the first parameter being used to identify the degree of influence of the first model parameter on the updated model parameter, and the second parameter being used to identify the degree of influence of the second model parameter on the updated model parameter.
[0132] The computer device can determine the corresponding model parameters after the model parameters of the speed prediction model are updated according to the first parameter to be updated and the second parameter to be updated, so that the updated model parameters are not only determined by the information display speed obtained during a single model parameter update, but are also constrained by the model parameters before the update. That is, through the first model parameters, the computer device can continuously refer to the historical model parameters, and through the second model parameters, the computer device can learn the parameter improvements brought about by multiple new training samples that appear in the application process, thereby avoiding the overfitting problem of a single model parameter adjustment, while ensuring the accuracy of model prediction, and further improving the stability of model parameter adjustment.
[0133] For example, the update of model parameters can be expressed as follows:
[0134] Θ ′ ←τΘ ′ +(1-τ)Θ
[0135] Among them, Θ on the left side of the arrow ′ is the updated model parameter corresponding to the double-speed prediction model, and Θ on the right side of the arrow ′ is the first model parameter mentioned above, Θ is the second model parameter mentioned above, τ∈[0,1] is a pre-specified hyperparameter as the first parameter, and (1-τ) is the second parameter.
[0136] In a possible implementation, the sample information and the information to be displayed may be different components of the same information, that is, the model determination process may be performed during the process of the information receiving object receiving information. For example, the target information may be composed of N sub-information, and the target information may be any information that can be divided into multiple sub-information, the multiple sample information may be M sub-information that has been displayed to the information receiving object, and the information to be displayed may be any sub-information among the N sub-information that has not been displayed to the information receiving object, and the first information display speed is used to display the information to be displayed to the information receiving object, N is a positive integer greater than 2, and M is a positive integer greater than 1 and less than N.
[0137] That is, in the process of displaying the target information to the information receiving object, if M sub-information has been displayed to the information receiving object, the computer device can obtain the M sub-information and the corresponding information display speed of the M sub-information when the information is displayed to perform model determination, obtain the speed prediction model, and then use the speed prediction model to predict the speed of the subsequent sub-information that is not displayed. In this way, the computer device can carry out targeted model training based on part of the content of the target information, so that the speed prediction model can carry out targeted learning of the content characteristics of the target information and the information receiving object's display speed requirements for the target information, so that the model can make more accurate speed determination for the remaining information to be displayed in the target information.
[0138] There are many ways to segment the target information. For example, in one possible implementation, since the scene is one of the important elements of information, the scene can play a key role in the information receiving object's selection of the information display speed. Based on this, the computer device can segment the target information based on the scene to obtain N sub-information.
[0139] Before obtaining the sample information set, the computer device can segment the target information through scene analysis to obtain N sub-information, and the scenes corresponding to two adjacent sub-information in the target information are different. It should be emphasized that here, it is only necessary to determine that the scenes corresponding to the two adjacent sub-information are different to segment the two sub-information, without specifically identifying the scene corresponding to each sub-information, which is different from the above-mentioned scene recognition.
[0140] In addition, the model parameters may also include multiple types. In a possible implementation, the model parameters may include sub-model parameters corresponding to multiple speeds, and the target sub-model parameters corresponding to the target speed are used to determine the probability that the information to be displayed corresponds to the target speed, and the target speed is any one of the multiple speeds. Thus, by continuously adjusting the model parameters, the model can make relatively accurate predictions for each speed.
[0141] Based on the model determination method provided in the above embodiment, the present application also provides a model application method, which can be applied to the speed prediction model described in any one of the above embodiments. Figure 4 , Figure 4 A flowchart of a model application method provided in an embodiment of the present application, the method can be executed by a computer device, and the computer device can be a terminal device or a server. It can be understood that the computer device can be a computer device for performing the above-mentioned model determination, or can be a different computer device.
[0142] The method includes:
[0143] S401: Acquire information to be displayed.
[0144] The information to be displayed may be any information that can be displayed.
[0145] S402: Determine a first information display speed corresponding to the information to be displayed according to the information to be displayed by using a speed prediction model.
[0146] The speed prediction model is obtained by training based on multiple sample information, and the multiple sample information has corresponding sample speeds respectively. The target sample speed is the information display speed when displaying the target sample information to the information receiving object. The target sample speed is the sample speed corresponding to the target sample information, and the target sample information is any one of the multiple sample information. Therefore, the speed prediction model can more accurately predict the information display speed requirements of the display object for the information to be displayed based on the information to be displayed itself, and obtain a more reasonable first information display speed.
[0147] S403: Displaying the information to be displayed to the object to be displayed according to the first information display speed.
[0148] In one possible implementation, the information receiving object used for model training corresponds to the target object type, so the speed prediction model can specifically learn the information receiving object's requirements for information display speed of the target object type, thereby being able to accurately predict the information display speed requirements of the information receiving object corresponding to the target object type.
[0149] Based on this, when displaying the information to be displayed, the computer device can first obtain the object information corresponding to the object to be displayed, and the object information is used to characterize the acceptance of the information constituent elements to be displayed, and the information constituent elements are used to constitute the information content corresponding to the information. The object information can be used to determine the object type of the object to be displayed, that is, the object type is distinguished based on the object's acceptance of the information constituent elements. The information receiving objects corresponding to the same object type are relatively similar in the acceptance of the information constituent elements, and thus the requirements for the information display speed are also relatively close.
[0150] The computer device can determine the object type corresponding to the object to be displayed based on the object information, and determine whether the object type is the target object type. In response to the object type being the target object type, it means that the speed prediction model can more accurately analyze the actual demand of the object to be displayed for the information display speed, so the determined first information display speed can more accurately meet the demand of the object to be displayed for the information display speed of the information to be displayed, and the computer device can display the information to be displayed to the object to be displayed according to the first information display speed, so as to provide the object to be displayed with an information display experience that better meets the demand.
[0151] The object information may be information that can reflect the acceptance of the object to be displayed in terms of information elements, such as historical information reception information of the object to be displayed, or may be an object type directly selected by the object to be displayed. That is, in a possible implementation, the computer device may provide the object to be displayed with the opportunity to select its corresponding object type and generate object information corresponding to the object to be displayed, so that the object information can reflect the acceptance of the object to be displayed in terms of image elements.
[0152] In this implementation, the object information may include a pending object type, which is the object type selected by the object to be displayed. When obtaining the object information corresponding to the object to be displayed, the computer device may obtain the type selection operation of the object to be displayed for the pending object type, and the type selection operation is used to select the pending object type as the object type corresponding to the object to be displayed. For example, the computer device may display an object type selection interface to the object to be displayed, and the object type selection interface may include selection controls corresponding to multiple object types, wherein the selection control corresponding to the pending object type is used to select the pending object type, and the multiple object types may include, for example, "like character objects", "like animation objects", "like mechanical objects" and other types that can reflect the acceptance of information constituent elements. If the object to be displayed wants to use the pending object type as its corresponding object type, it can trigger the selection control corresponding to the pending object type as a type selection operation for the pending object type.
[0153] In response to obtaining the type selection operation of the object to be displayed for the undetermined object type, the computer device can determine to obtain the object information corresponding to the object to be displayed, and can directly determine the undetermined object type as the object type corresponding to the object to be displayed. Through this object type determination method, a more flexible type selection method can be provided to the object to be displayed, so that the object to be displayed can more accurately express its acceptance of the information constituent elements, thereby providing a more accurate multiple speed prediction function for the object to be displayed.
[0154] In a possible implementation, in addition to directly displaying the information to be displayed to the object to be displayed based on the first information display speed determined by the model, in order to provide the object to be displayed with a more free information display experience, the computer device can also provide the object to be displayed with the function of selecting the information display speed by itself.
[0155] The computer device may obtain a speed selection operation of the to-be-displayed object for the to-be-displayed information, the speed selection operation corresponding to the second information display speed, and the speed selection operation is used to adjust the information display speed of the to-be-displayed information to the second information display speed.
[0156] After obtaining the speed selection operation, the computer device can determine that the second information display speed is the information display speed actually required by the object to be displayed, and thus the information display speed corresponding to the information to be displayed can be updated from the first information display speed to the second information display speed. In this way, when the speed prediction model does not fully meet the information display speed requirements of the object to be displayed, the object to be displayed can express its own information display speed requirements to meet them, further improving the probability of meeting the information display speed requirements of the object to be displayed and improving the information display experience of the object to be displayed.
[0157] In addition, the computer device can also use the above-mentioned model parameter updating method to continuously update the speed training model based on the changed second information display speed when the information display speed determined by the model of the object to be displayed is changed, so that the model can further learn how to accurately meet the information display speed requirements of the object to be displayed.
[0158] As mentioned above, in addition to the object type, the computer device can also train the model specifically for the information content type, so as to obtain a speed prediction model that can accurately predict the speed for each information content type. For example, in one possible implementation, the above-mentioned multiple sample information corresponds to the target information content type, so the speed prediction model can be used to determine the information display speed more accurately for the information of the target information content type.
[0159] Before determining the information display speed corresponding to the information to be displayed through the speed prediction model, the computer device may first determine the pending information content type corresponding to the information to be displayed. In response to the pending information content type being consistent with the target information content type, it means that the speed prediction model has a relatively accurate speed prediction capability for the information to be displayed. Based on this, the computer device may determine the first information display speed corresponding to the information to be displayed according to the information to be displayed through the speed prediction model.
[0160] In order to facilitate understanding of the technical solution provided in the embodiment of the present application, a model determination method provided in the embodiment of the present application will be introduced below in combination with an actual application scenario.
[0161] See also Figure 5 , Figure 5 A schematic diagram of a model determination method in an actual application scenario provided for an embodiment of the present application. In this actual application scenario, the computer device may be a terminal device or a server with model determination capability, or a community consisting of a terminal device and a server.
[0162] like Figure 5As shown, the sample information is sample video information, and the computer device can perform video storyboard segmentation on the sample video information based on the scene to obtain multiple video clips. The present application can use the video shot boundary detection (TransNetV2) algorithm to segment video clips. TransNetV2 is an algorithm for video temporal segmentation, which can divide a video sequence into multiple action clips, each action clip corresponds to a video scene. It can be understood that when the action in the video changes suddenly and drastically, it is highly likely that the storyboards before and after the video will be different, that is, the corresponding video scenes will be different. The algorithm can classify and locate each clip. The input of the TransNetV2 algorithm can be a video of any length, and the algorithm includes the following steps:
[0163] 1. Feature extraction: Use convolutional neural networks such as deep residual network (ResNet) to extract features from the video and obtain feature sequences, which are used to represent the corresponding actions in the video;
[0164] 2. Temporal segmentation: Use the temporal segmentation module to divide the feature sequence into multiple segments, each of which contains a series of continuous feature vectors, representing an action in the video;
[0165] 3. Clip classification: Classify each clip using a classification module and output an action label that can identify the corresponding scene information;
[0166] 4. Fragment positioning: locate each fragment, use the positioning module to locate each fragment, and output the start and end timestamps of each fragment;
[0167] 5. Result output: The classification and positioning results are output as the final video segmentation result, that is, multiple video clips are obtained.
[0168] Then, the computer device can extract object features, scene features, and information features from each video clip. In this application, the object feature is a video character object coding vector, which is used to identify the frequency of each character appearing in the video clip, for example, Figure 3 The scene feature is a video scene classification coding vector, which is used to identify the scene type corresponding to each video judgment. The information feature can be obtained by identifying the number of character lines in the video clip, for example, it can include the total number of lines corresponding to the video clip, the average number of lines per minute, and the length of the video clip.
[0169] Among them, object features can be obtained in the following ways:
[0170] For each video clip, the computer device can extract frames at an interval of three frames per second. For each video frame, the multi-task cascaded convolutional network (MTCNN) is first used to detect the face area appearing in the video frame, and then the face recognition algorithm ArcFace is used to identify the person in each face area.
[0171] MTCNN is a deep learning algorithm for face detection and facial key point detection. The MTCNN algorithm uses a cascade approach and divides face detection and facial key point detection into three stages: the Proposal Network (P-Net) is used to generate candidate face frames; the Refine Network (R-Net) is used to screen candidate face frames and remove unqualified frames; the Output Network (O-Net) is used to detect facial key points, and finally the regression model and classification model are used to predict the position and confidence of the face frame.
[0172] The ArcFace algorithm is a deep learning algorithm for face recognition. The core of the ArcFace algorithm is a new loss function that adds a cosine similarity constraint to the traditional Softmax classifier, making faces of the same category closer in the feature space and the distance between faces of different categories more obvious. This can improve the accuracy and robustness of face recognition.
[0173] The computer device can obtain the character recognition results for each frame of the video frame image obtained by sampling, and count the frequency of each character appearing, for example, character A appears in 3 frames, character B does not appear, character C appears in 2 frames, etc., let K p Indicates the total number of characters appearing in the target video information, and records the frequency of characters appearing in K p In the video character object encoding vector of dimension Figure 3 The vector shown.
[0174] The extraction of scene features can be achieved in the following ways:
[0175] The computer device can classify each video clip and determine the scene corresponding to each video clip, such as outdoor, indoor, urban, rural, beach, forest, park, etc. This application can use the video processing model Video Swin Transformer to classify video scenes. Video Swin Transformer is a video processing model based on the visual attention mechanism. The specific classification process is as follows:
[0176] 1. Preprocessing: The computer device can divide the video clip into several slices (clip), each clip is composed of several consecutive frames of images, and then the image of each clip is resized (resized) and normalized.
[0177] 2. Spatial feature extraction: The computer device can use Swin Transformer to process the spatial features of each clip. First, the image of each clip is divided into several patches, and features are extracted for each patch; then, these patches are arranged into a matrix similar to an image and input into Swin Transformer for processing. In this way, the spatial feature representation of each clip can be obtained.
[0178] 3. Temporal feature extraction: The spatial feature representation of each clip is arranged in time order to form a matrix similar to a sequence, and then input into another Swin Transformer for processing. In this way, the temporal feature representation of each clip can be obtained.
[0179] 4. Classification: The spatial feature representation and temporal feature representation of each clip are concatenated together and input into a fully connected layer for classification to obtain the scene corresponding to each video clip.
[0180] For the segmented video clips, since the probability of the video clip corresponding to each scene is first obtained during video classification, and then the scene corresponding to the video clip is determined based on the probability, the computer device can extract the probability of the video clip corresponding to each scene to generate a video scene classification coding vector as a scene feature. That is, if the scene classification module has a total of Ks types of scenes, a Ks-dimensional vector will be obtained, and the sum of the dimensions of the vector is equal to 1 (that is, the sum of the probabilities of each scene is 1).
[0181] Through the above video task object coding vector, video scene classification coding vector, total number of lines, average number of lines per minute and video clip length, we can construct a Kp+Ks+3-dimensional feature vector x, which corresponds to the video clip one by one. On the basis of the feature vector x, we can add a one-dimensional constant 1 to form an augmented feature vector as input to the regression model.
[0182] The regression model can adopt a normalized regression model (Softmax Regression). Softmax regression, also known as multi-class logarithmic probability regression (Logistic Regression), is a classification method suitable for multi-classification problems. The main idea of Softmax regression is to map a vector to another vector output through a linear transformation and a Softmax activation function, so that the value of each element of the output vector is between 0 and 1, and the sum of all elements is 1. The output vector is a multidimensional vector, and the kth dimension of the output vector represents the probability that the input vector belongs to the kth class. In this application, each class represents a speed, and the kth dimension of the output vector represents the probability of the corresponding kth class speed.
[0183] The formula for determining the output of this regression model can be shown as follows:
[0184]
[0185] Wherein, C represents the number of categories of information display speed. In this application, C=4 categories are used, corresponding to 1x speed, 1.5x speed, 2x speed, and 2.5x speed respectively; θ c represents the learnable parameter of the cth category in the model, which is used to determine the probability of the video clip corresponding to the cth category information display speed, and p(y=k) identifies the probability of the video clip corresponding to the kth category information display speed. The complete output vector is:
[0186]
[0187] Let Θ be as shown in the following formula:
[0188]
[0189] Θ is the set of learnable parameters corresponding to the display speed of C type information, and R is a set of real numbers.
[0190] Softmax regression can be written as the following matrix multiplication form:
[0191]
[0192]
[0193] Based on the information display speed predicted by the model and the information display speed used by the information recipient when watching the video clip, a loss function can be generated. In this practical application scenario, a cross entropy loss function £ can be used, as shown in the following formula:
[0194] Where λ is a pre-set hyperparameter used to control the degree of regularization. This application may also use other loss functions, which are not limited here.
[0195] The computer device can perform stochastic gradient descent on the loss function to improve the accuracy of adjusting the model parameters based on the loss function. The derivative of the loss function with respect to the parameters is:
[0196]
[0197] in The kth element of is 1, and the rest are 0. Stochastic gradient descent is used in the module, that is, a sample pair is used to estimate the gradient of the loss function to the parameter, as shown in the following formula:
[0198]
[0199] By adjusting the parameters in the above model training method, a multiple speed prediction model can be obtained. In the present application, the computer device can construct two models f and g, the structures of the two models are the same, and the initial parameters are model parameters obtained by training with sample video information. Let the parameters of model f be Θ, the parameters of model g be Θ', and the parameters of g are the exponential moving average (EMA) of the parameters of f, as shown in the following formula:
[0200] Θ ′ ←τΘ ′ +(1-τ)Θ
[0201] Where τ∈[0,1] is a pre-specified hyperparameter. Every time a new training sample (x, y) appears (where x is the display information and y is the actual information display speed at which the display information is displayed to the information receiving object), model f will update the model parameters based on the new training sample, and the parameters of model g are updated based on the parameters of model g before the update and the parameters of model f after the update based on the new training sample. EMA is a smoothing method that can perform weighted averaging on historical data, so that the impact of new data on model parameters gradually decreases, while the impact of old data on model parameters gradually increases. Updating the model parameters by EMA is to reduce the interference caused by the accidental behavior of the information receiving object on the model parameter update, thereby making the model training more stable and reducing the risk of overfitting.
[0202] Based on the model determination method provided in the above embodiment, the present application also provides a model determination device, see Figure 6 , Figure 6 This is a structural block diagram of a model determination device 600 provided in an embodiment of the present application, wherein the device includes a first acquisition unit 601, a first determination unit 602, and an adjustment unit 603:
[0203] The first acquisition unit 601 is used to acquire a sample information set, where the sample information set includes a plurality of sample information, and the plurality of sample information has a corresponding sample speed;
[0204] The first determining unit 602 is configured to use the plurality of sample information as target sample information respectively, and determine a to-be-determined speed according to the target sample information through an initial speed prediction model;
[0205] The adjustment unit 603 is used to adjust the model parameters corresponding to the initial speed prediction model according to the difference between the target sample speed and the to-be-determined speed to obtain the speed prediction model. The target sample speed is the information display speed when the target sample information is displayed to the information receiving object. The target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
[0206] In a possible implementation manner, the first determining unit 602 is specifically configured to:
[0207] Generate a content feature corresponding to the target sample information, wherein the content feature is used to characterize the information content corresponding to the target sample information;
[0208] The undetermined speed is determined according to the content characteristics.
[0209] In one possible implementation, the content features include any one or more combinations of object features, scene features, and information volume features, wherein the object features are used to identify the objects included in the target sample information, the scene features are used to identify the scenes corresponding to the target sample information, and the information volume features are used to characterize the amount of information displayed by the target sample information.
[0210] In a possible implementation, the target sample information is video information, the target sample information includes multiple video frames, the content feature includes the object feature, the scene feature and the information volume feature, and the first determining unit 602 is specifically configured to:
[0211] By performing object recognition on the multiple video frames, an object feature corresponding to the target sample information is generated, wherein the object feature includes the number of video frames corresponding to the multiple objects included in the target sample information, the number of video frames corresponding to the target object is the number of video frames including the target object in the multiple video frames, and the target object is any one of the multiple objects;
[0212] By performing scene recognition on the target sample information, a scene feature corresponding to the target sample information is generated, wherein the scene feature includes probability information corresponding to a plurality of scenes respectively, and the probability information corresponding to the target scene is used to identify the probability that the target sample information corresponds to the target scene, and the target scene is any one of the plurality of scenes;
[0213] By performing text extraction on the target sample information, an information volume feature corresponding to the target sample information is generated, and the information volume feature includes at least one of the following: the total number of texts corresponding to the target sample information, and the average number of texts corresponding to the target sample information in a preset time unit.
[0214] In one possible implementation, the information receiving object corresponds to a target object type, the first information display speed is used to display the information to be displayed to an object corresponding to the target object type, the target object type is determined based on object information corresponding to the information receiving object, the object information corresponding to the information receiving object is used to characterize the acceptance of the information receiving object for information constituent elements, and the information constituent elements are used to constitute the information content corresponding to the information.
[0215] In a possible implementation, the device further includes a first display unit, a second acquisition unit, and a first update unit:
[0216] The first display unit is used to display the information to be displayed to the target information receiving object according to the first information display speed;
[0217] The second acquisition unit is used for acquiring a changed second information display speed in response to the target information receiving object changing the information display speed corresponding to the information to be displayed;
[0218] The first updating unit is used to display the speed according to the second information and update the model parameters corresponding to the speed prediction model.
[0219] In a possible implementation manner, the first updating unit is specifically configured to:
[0220] Determine the target undetermined speed according to the information to be displayed through a target speed prediction model, wherein the initial model parameters of the target speed prediction model are consistent with the model parameters of the speed prediction model adjusted by the multiple sample information, and the target speed prediction model and the speed prediction model have the same model architecture;
[0221] According to the difference between the second information display speed and the target undetermined speed, the model parameters corresponding to the target speed prediction model are adjusted to obtain the second model parameters,
[0222] Obtaining a first model parameter, where the first model parameter is a model parameter corresponding to the speed prediction model before the speed prediction model is updated based on the second information display speed;
[0223] Determine a first parameter to be updated according to the first model parameter and the first parameter, and determine a second parameter to be updated according to the second model parameter and the second parameter;
[0224] The model parameters corresponding to the speed prediction model after the model parameters are updated are determined according to the first parameters to be updated and the second parameters to be updated.
[0225] In a possible implementation manner, the plurality of sample information and the information to be displayed correspond to target information content types.
[0226] In a possible implementation, the target information is composed of N sub-information, the multiple sample information is M sub-information that has been displayed to the information receiving object, the information to be displayed is any sub-information among the N sub-information that has not been displayed to the information receiving object, the first information display speed is used to display the information to be displayed to the information receiving object, N is a positive integer greater than 2, and M is a positive integer greater than 1 and less than N.
[0227] In a possible implementation, the device further includes a segmentation unit:
[0228] The segmentation unit is used to segment the target information through scene analysis to obtain the N sub-information, and the scenes corresponding to two adjacent sub-information in the target information are different.
[0229] In a possible implementation, the model parameters include sub-model parameters corresponding to multiple speeds respectively, and the target sub-model parameters corresponding to the target speed are used to determine the probability that the information to be displayed corresponds to the target speed, and the target speed is any one of the multiple speeds.
[0230] Based on the model application method provided in the above embodiment, the present application also provides a model application device, see Figure 7 , Figure 7 A structural block diagram of a model application device 700 provided in an embodiment of the present application, the device includes a third acquisition unit 701, a second determination unit 702 and a second display unit 703:
[0231] The third acquisition unit 701 is used to acquire information to be displayed;
[0232] The second determination unit 702 is used to determine a first information display speed corresponding to the information to be displayed according to the information to be displayed by using a speed prediction model, wherein the speed prediction model is obtained by training based on a plurality of sample information, the plurality of sample information have respectively corresponding sample speeds, the target sample speed is the information display speed when the target sample information is displayed to the information receiving object, the target sample speed is the sample speed corresponding to the target sample information, and the target sample information is any one of the plurality of sample information;
[0233] The second display unit 703 is used to display the information to be displayed to the display object according to the first information display speed.
[0234] In a possible implementation manner, the information receiving object corresponds to a target object type, and the device further includes a fourth acquiring unit and a third determining unit:
[0235] The fourth acquisition unit is used to acquire object information corresponding to the object to be displayed, wherein the object information is used to represent the acceptance of the information constituent elements to be displayed, and the information constituent elements are used to constitute the information content corresponding to the information;
[0236] The third determining unit is used to determine the object type corresponding to the object to be displayed according to the object information;
[0237] The second display unit 703 is specifically used for:
[0238] In response to the object type being the target object type, the information to be displayed is displayed to the object to be displayed according to the first information display speed.
[0239] In a possible implementation manner, the object information includes a to-be-determined object type, and the fourth acquiring unit is specifically configured to:
[0240] In response to obtaining a type selection operation of the object to be displayed for the to-be-determined object type, determining to obtain object information corresponding to the object to be displayed;
[0241] The third determining unit is specifically configured to:
[0242] The to-be-determined object type is determined as the object type corresponding to the to-be-displayed object.
[0243] In a possible implementation manner, the device further includes a fifth acquiring unit and a second updating unit:
[0244] The fifth acquisition unit is used to acquire a speed selection operation of the to-be-displayed object for the to-be-displayed information, wherein the speed selection operation corresponds to a second information display speed;
[0245] The second updating unit is used to update the information display speed corresponding to the information to be displayed from the first information display speed to the second information display speed.
[0246] In a possible implementation manner, the plurality of sample information corresponds to a target information content type, and the apparatus further includes a fourth determining unit:
[0247] The fourth determining unit is used to determine the to-be-determined information content type corresponding to the to-be-displayed information;
[0248] The second determining unit 702 is specifically configured to:
[0249] In response to the to-be-determined information content type being consistent with the target information content type, a first information display speed corresponding to the to-be-displayed information is determined according to the to-be-displayed information through a speed prediction model.
[0250] The present application also provides a computer device, see Figure 8 As shown, the computer device may be a terminal device, and a mobile phone is taken as an example:
[0251] Figure 8 FIG. 1 is a block diagram showing a partial structure of a mobile phone related to a terminal device provided in an embodiment of the present application. Figure 8 The mobile phone includes: a radio frequency (RF) circuit 710, a memory 720, an input unit 730, a display unit 740, a sensor 750, an audio circuit 760, a wireless fidelity (WiFi) module 770, a processor 780, and a power supply 790. Those skilled in the art will understand that Figure 8 The mobile phone structure shown in the figure does not constitute a limitation on the mobile phone, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.
[0252] Combine the following Figure 8 A detailed introduction to the various components of the mobile phone:
[0253] The RF circuit 710 can be used for receiving and sending signals during the process of sending and receiving information or making calls. In particular, after receiving the downlink information of the base station, it is sent to the processor 780 for processing; in addition, the designed uplink data is sent to the base station. Usually, the RF circuit 710 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, referred to as LNA), a duplexer, etc. In addition, the RF circuit 710 can also communicate with the network and other devices through wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to the Global System of Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.
[0254] The memory 720 can be used to store software programs and modules. The processor 780 executes various functional applications and data processing of the mobile phone by running the software programs and modules stored in the memory 720. The memory 720 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 720 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0255] The input unit 730 can be used to receive input digital or character information, and to generate key signal input related to the user settings and function control of the mobile phone. Specifically, the input unit 730 may include a touch panel 731 and other input devices 732. The touch panel 731, also known as a touch screen, can collect the user's touch operation on or near it (such as the user's operation on the touch panel 731 or near the touch panel 731 using any suitable object or accessory such as a finger, stylus, etc.), and drive the corresponding connection device according to a pre-set program. Optionally, the touch panel 731 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into contact coordinates, and then sends it to the processor 780, and can receive and execute commands sent by the processor 780. In addition, the touch panel 731 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic waves. In addition to the touch panel 731, the input unit 730 may further include other input devices 732. Specifically, the other input devices 732 may include but are not limited to one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, and the like.
[0256] The display unit 740 can be used to display information input by the user or information provided to the user and various menus of the mobile phone. The display unit 740 may include a display panel 741. Optionally, the display panel 741 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an organic light-emitting diode (Organic Light-Emitting Diode, OLED), etc. Further, the touch panel 731 may cover the display panel 741. When the touch panel 731 detects a touch operation on or near it, it is transmitted to the processor 780 to determine the type of touch event. Subsequently, the processor 780 provides a corresponding visual output on the display panel 741 according to the type of touch event. Although in Figure 8 In the embodiment, the touch panel 731 and the display panel 741 are used as two independent components to realize the input and output functions of the mobile phone, but in some embodiments, the touch panel 731 and the display panel 741 can be integrated to realize the input and output functions of the mobile phone.
[0257] The mobile phone may also include at least one sensor 750, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 741 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 741 and / or the backlight when the mobile phone is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that identify the posture of the mobile phone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that can be configured in the mobile phone, such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., they will not be repeated here.
[0258] The audio circuit 760, the speaker 761, and the microphone 762 can provide an audio interface between the user and the mobile phone. The audio circuit 760 can transmit the received audio data to the speaker 761 after converting the received audio data into an electrical signal, which is converted into a sound signal for output; on the other hand, the microphone 762 converts the collected sound signal into an electrical signal, which is received by the audio circuit 760 and converted into audio data, and then the audio data is output to the processor 780 for processing, and then sent to another mobile phone through the RF circuit 710, or the audio data is output to the memory 720 for further processing.
[0259] WiFi is a short-range wireless transmission technology. The mobile phone can help users send and receive emails, browse web pages and access streaming media through the WiFi module 770. It provides users with wireless broadband Internet access. Figure 8 A WiFi module 770 is shown, but it is understandable that it is not an essential component of the mobile phone and can be omitted as needed without changing the essence of the invention.
[0260] The processor 780 is the control center of the mobile phone. It uses various interfaces and lines to connect various parts of the entire mobile phone. By running or executing software programs and / or modules stored in the memory 720, and calling data stored in the memory 720, it executes various functions of the mobile phone and processes data, thereby performing overall detection of the mobile phone. Optionally, the processor 780 may include one or more processing units; preferably, the processor 780 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 780.
[0261] The mobile phone also includes a power supply 790 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 780 through a power management system, so that the power management system can manage charging, discharging, power consumption and other functions.
[0262] Although not shown, the mobile phone may also include a camera, a Bluetooth module, etc., which will not be described in detail here.
[0263] In this embodiment, the processor 780 included in the terminal device also has the following functions:
[0264] Acquire a sample information set, wherein the sample information set includes a plurality of sample information, and the plurality of sample information has respectively corresponding sample speeds;
[0265] The plurality of sample information are respectively used as target sample information, and a to-be-determined speed is determined according to the target sample information through an initial speed prediction model;
[0266] The model parameters corresponding to the initial speed prediction model are adjusted according to the difference between the target sample speed and the to-be-determined speed to obtain a speed prediction model, wherein the target sample speed is the information display speed when the target sample information is displayed to the information receiving object, and the target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
[0267] Or, with the following functionality:
[0268] The first acquisition unit is used to acquire a sample information set, wherein the sample information set includes a plurality of sample information, and the plurality of sample information has respectively corresponding sample speeds;
[0269] The first determining unit is configured to use the plurality of sample information as target sample information respectively, and determine the to-be-determined speed according to the target sample information through an initial speed prediction model;
[0270] The adjustment unit is used to adjust the model parameters corresponding to the initial speed prediction model according to the difference between the target sample speed and the to-be-determined speed to obtain the speed prediction model. The target sample speed is the information display speed when the target sample information is displayed to the information receiving object. The target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
[0271] The present application also provides a server. Fig. 9 As shown, Fig. 9The structural diagram of the server 800 provided in the embodiment of the present application, the server 800 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 822 (for example, one or more processors) and a memory 832, and one or more storage media 830 (for example, one or more mass storage devices) storing application programs 842 or data 844. Among them, the memory 832 and the storage medium 830 can be temporary storage or permanent storage. The program stored in the storage medium 830 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 822 can be configured to communicate with the storage medium 830 and execute a series of instruction operations in the storage medium 830 on the server 800.
[0272] The server 800 may also include one or more power supplies 826, one or more wired or wireless network interfaces 850, one or more input and output interfaces 858, and / or one or more operating systems 841, such as Windows Server 2000. TM , Mac OS X TM , Unix TM ,Linux TM , FreeBSD TM etc.
[0273] The steps performed by the server in the above embodiment can be based on Fig. 9 The server structure shown.
[0274] An embodiment of the present application also provides a computer-readable storage medium for storing a computer program, which is used to execute any implementation of the model determination method or model application method described in the aforementioned embodiments.
[0275] An embodiment of the present application also provides a computer program product including a computer program, which, when executed on a computer device, enables the computer device to execute the model determination method or model application method described in any one of the above embodiments.
[0276] It can be understood that in the specific implementation of the present application, related data such as object information (such as the sample speed when the object receives sample information) is involved. When the above embodiments of the present application are applied to specific products or technologies, it is necessary to obtain the object's permission or consent, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.
[0277] A person of ordinary skill in the art can understand that all or part of the steps of implementing the above-mentioned method embodiment can be completed by hardware related to program instructions, and the above-mentioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above-mentioned method embodiment; and the above-mentioned storage medium can be at least one of the following media: read-only memory (English: read-only memory, abbreviated: ROM), RAM, magnetic disk or optical disk, etc. Various media that can store program codes.
[0278] It should be noted that each embodiment in this specification is described in a progressive manner, and the same and similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device and system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments. The device and system embodiments described above are merely schematic, in which the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.
[0279] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
Claims
1. A model determination method, characterized in that: The method comprises: Acquire a sample information set, wherein the sample information set includes a plurality of sample information, and the plurality of sample information has respectively corresponding sample speeds; The plurality of sample information are respectively used as target sample information, and a to-be-determined speed is determined according to the target sample information through an initial speed prediction model; The model parameters corresponding to the initial speed prediction model are adjusted according to the difference between the target sample speed and the to-be-determined speed to obtain a speed prediction model, wherein the target sample speed is the information display speed when the target sample information is displayed to the information receiving object, and the target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
2. The method according to claim 1, characterized in that The determining of the to-be-determined speed according to the target sample information includes: Generate a content feature corresponding to the target sample information, wherein the content feature is used to characterize the information content corresponding to the target sample information; The undetermined speed is determined according to the content characteristics.
3. The method according to claim 2, characterized in that The content features include any one or more combinations of object features, scene features and information volume features. The object features are used to identify the objects included in the target sample information. The scene features are used to identify the scenes corresponding to the target sample information. The information volume features are used to characterize the amount of information displayed by the target sample information.
4. The method according to claim 3, characterized in that The target sample information is video information, the target sample information includes a plurality of video frames, the content feature includes the object feature, the scene feature and the information volume feature, and the generating of the content feature corresponding to the target sample information includes: By performing object recognition on the multiple video frames, an object feature corresponding to the target sample information is generated, wherein the object feature includes the number of video frames corresponding to the multiple objects included in the target sample information, the number of video frames corresponding to the target object is the number of video frames including the target object in the multiple video frames, and the target object is any one of the multiple objects; By performing scene recognition on the target sample information, a scene feature corresponding to the target sample information is generated, wherein the scene feature includes probability information corresponding to a plurality of scenes respectively, and the probability information corresponding to the target scene is used to identify the probability that the target sample information corresponds to the target scene, and the target scene is any one of the plurality of scenes; By performing text extraction on the target sample information, an information volume feature corresponding to the target sample information is generated, and the information volume feature includes at least one of the following: the total number of texts corresponding to the target sample information, and the average number of texts corresponding to the target sample information in a preset time unit.
5. The method according to claim 1, characterized in that The information receiving object corresponds to a target object type, the first information display speed is used to display the information to be displayed to an object corresponding to the target object type, the target object type is determined based on object information corresponding to the information receiving object, the object information corresponding to the information receiving object is used to characterize the acceptance of the information receiving object for information constituent elements, and the information constituent elements are used to constitute the information content corresponding to the information.
6. The method according to claim 1, characterized in that The method further comprises: Displaying the information to be displayed to a target information receiving object according to the first information display speed; In response to the target information receiving object changing the information display speed corresponding to the information to be displayed, obtaining a changed second information display speed; The speed is displayed according to the second information, and the model parameters corresponding to the speed prediction model are updated.
7. The method according to claim 6, characterized in that The displaying speed according to the second information and updating the model parameters corresponding to the speed prediction model include: Determine the target undetermined speed according to the information to be displayed through a target speed prediction model, wherein the initial model parameters of the target speed prediction model are consistent with the model parameters of the speed prediction model adjusted by the multiple sample information, and the target speed prediction model and the speed prediction model have the same model architecture; According to the difference between the second information display speed and the target undetermined speed, the model parameters corresponding to the target speed prediction model are adjusted to obtain the second model parameters, Obtaining a first model parameter, where the first model parameter is a model parameter corresponding to the speed prediction model before the speed prediction model is updated based on the second information display speed; Determine a first parameter to be updated according to the first model parameter and the first parameter, and determine a second parameter to be updated according to the second model parameter and the second parameter; The model parameters corresponding to the speed prediction model after the model parameters are updated are determined according to the first parameters to be updated and the second parameters to be updated.
8. The method according to claim 1, characterized in that The plurality of sample information and the information to be displayed correspond to target information content types.
9. The method according to claim 1, characterized in that: The target information is composed of N sub-information, the multiple sample information is M sub-information that has been displayed to the information receiving object, the information to be displayed is any sub-information among the N sub-information that has not been displayed to the information receiving object, the first information display speed is used to display the information to be displayed to the information receiving object, N is a positive integer greater than 2, and M is a positive integer greater than 1 and less than N.
10. The method according to claim 9, characterized in that Before obtaining the sample information set, the method further includes: The target information is segmented through scene analysis to obtain the N sub-information, and two adjacent sub-information in the target information correspond to different scenes.
11. The method according to claim 1, characterized in that: The model parameters include sub-model parameters corresponding to a plurality of speeds, and the target sub-model parameters corresponding to the target speed are used to determine the probability that the information to be displayed corresponds to the target speed, and the target speed is any one of the plurality of speeds.
12. A model application method, characterized in that: The method comprises: Get the information to be displayed; Determine, by a speed prediction model, a first information display speed corresponding to the information to be displayed according to the information to be displayed, wherein the speed prediction model is obtained by training based on a plurality of sample information, the plurality of sample information have respectively corresponding sample speeds, the target sample speed is the information display speed when displaying the target sample information to the information receiving object, the target sample speed is the sample speed corresponding to the target sample information, and the target sample information is any one of the plurality of sample information; The information to be displayed is displayed to the object to be displayed according to the first information display speed.
13. The method according to claim 12, characterized in that The information receiving object corresponds to a target object type, and the method further includes: Obtaining object information corresponding to the object to be displayed, wherein the object information is used to represent the acceptance of the information constituent elements to be displayed, and the information constituent elements are used to constitute the information content corresponding to the information; Determining, according to the object information, an object type corresponding to the object to be displayed; The step of displaying the information to be displayed to the object to be displayed according to the first information display speed includes: In response to the object type being the target object type, the information to be displayed is displayed to the object to be displayed according to the first information display speed.
14. The method according to claim 13, characterized in that The object information includes a to-be-determined object type, and the obtaining of the object information corresponding to the to-be-displayed object includes: In response to obtaining a type selection operation of the object to be displayed for the to-be-determined object type, determining to obtain object information corresponding to the object to be displayed; The determining, according to the object information, the object type corresponding to the object to be displayed includes: The to-be-determined object type is determined as the object type corresponding to the to-be-displayed object.
15. The method according to claim 11, characterized in that The plurality of sample information correspond to target information content types, and the method further includes: Determine the type of pending information content corresponding to the information to be displayed; The determining, according to the information to be displayed, a first information display speed corresponding to the information to be displayed by using a speed prediction model includes: In response to the to-be-determined information content type being consistent with the target information content type, a first information display speed corresponding to the to-be-displayed information is determined according to the to-be-displayed information through a speed prediction model.
16. A model determination device, characterized in that: The device comprises a first acquisition unit, a first determination unit and an adjustment unit: The first acquisition unit is used to acquire a sample information set, wherein the sample information set includes a plurality of sample information, and the plurality of sample information has respectively corresponding sample speeds; The first determining unit is configured to use the plurality of sample information as target sample information respectively, and determine the to-be-determined speed according to the target sample information through an initial speed prediction model; The adjustment unit is used to adjust the model parameters corresponding to the initial speed prediction model according to the difference between the target sample speed and the to-be-determined speed to obtain the speed prediction model. The target sample speed is the information display speed when the target sample information is displayed to the information receiving object. The target sample speed is the sample speed corresponding to the target sample information. The speed prediction model is used to predict the first information display speed corresponding to the information to be displayed, and the first information display speed is used to display the information to be displayed.
17. A model application device, characterized in that: The device comprises a third acquisition unit, a second determination unit and a second display unit: The third acquisition unit is used to acquire information to be displayed; The second determination unit is used to determine a first information display speed corresponding to the information to be displayed according to the information to be displayed by using a speed prediction model, wherein the speed prediction model is obtained by training based on a plurality of sample information, the plurality of sample information have respectively corresponding sample speeds, the target sample speed is the information display speed when the target sample information is displayed to the information receiving object, the target sample speed is the sample speed corresponding to the target sample information, and the target sample information is any one of the plurality of sample information; The second display unit is used to display the information to be displayed to the object to be displayed according to the first information display speed.
18. A computer device, characterized in that: The computer device comprises a processor and a memory: The memory is used to store a computer program and transmit the computer program to the processor; The processor is used to execute the model determination method described in any one of claims 1-11, or execute the model application method described in any one of claims 12-15 according to the instructions in the computer program.
19. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the model determination method described in any one of claims 1-11, or execute the model application method described in any one of claims 12-15.
20. A computer program product comprising a computer program, which, when executed on a computer device, enables the computer device to execute the model determination method described in any one of claims 1 to 11, or the model application method described in any one of claims 12 to 15.
