Short play recommendation method based on cover data, electronic equipment and storage medium

Abstract

The invention relates to the technical field of video personalized recommendation, and provides a short play recommendation method based on cover data, electronic equipment and a storage medium, and the method comprises the steps: extracting short play cover image features, and storing the features in a database; extracting user behavior data, and calculating the score of the user for each movie; according to the image features, calculating the similarity of the short drama cover image; and according to the similarity between the short play score and the short play cover image of the user, recommending a short play to the user. According to the scheme, an automatic dynamic recommendation system is adopted, manpower and time cost is saved, the coverage range is wide, and application and popularization are facilitated; intelligent configuration is adopted, and the problem that subjectivity of manual configuration is too high is solved; the change requirements of markets and users can be dynamically met; the quality fluctuation of the recommended content is avoided, and the overall experience of the user is improved.

Description

Technical Field

[0001] This application relates to the field of software technology, and particularly to a short drama recommendation method, an electronic device, and a storage medium based on cover data. Background Art

[0002] Short dramas usually refer to TV dramas or web dramas with short lengths, and the time for each episode is generally from a few minutes to more than a dozen minutes. With the development of Internet technology and the popularity of mobile devices, short dramas are loved by more and more audiences due to their convenience and diverse content. The younger generation is more inclined to a fast-paced and high-density information intake method, and short dramas cater to this trend with their compact plots and rapid information transmission methods. The short drama industry is in a stage of rapid development and is expected to further expand the market space and improve the user experience through technological innovation and content innovation in the future. Currently, most short drama system software adopts relatively primitive recommendation methods, and some relatively good dramas are configured and recommended to users through the analysis of existing data, with a large investment in time and labor costs.

[0003] However, the existing short drama recommendation methods have the following disadvantages: high labor intensity, as the background configuration usually requires manual selection and configuration of recommended content, which will increase the workload of the operation team; slow response speed, as manual configuration cannot respond to changes in user behavior in real time, which may lead to the lag of recommended content; strong subjectivity, as subjective biases may exist when manually selecting recommended content, which will affect the objectivity and fairness of recommendations; difficult to cover all users, as it is difficult to accurately recommend according to the personalized needs of each user through manual configuration, which may lead to the generalization of recommended content; high maintenance cost, as the complexity and maintenance cost of background configuration will increase correspondingly with the increase in the number of users and the content library; insufficient response to dynamic changes, as the changes in the market and user behavior are dynamic, and it is difficult for manual configuration to adjust the recommendation strategy in a timely manner; unstable recommendation quality, as manual configuration may lead to fluctuations in the quality of recommended content, which will affect the overall user experience. Summary of the Invention

[0004] To help solve the problems of poor recommendation effect, high time and labor costs in the existing short drama recommendation system, this application provides a short drama recommendation method, an electronic device, and a storage medium based on cover data.

[0005] In a first aspect, this application provides a short drama recommendation method based on cover data for use in a software system, and the method includes:

[0006] Extract the short drama cover image features and store them in a database;

[0007] Extract user behavior data and calculate the score of the user for each short drama;

[0008] Calculate the similarity of the short drama cover image according to the image features;

[0009] Recommend short dramas for the user according to the user's short drama ratings and the similarity of the short drama cover images.

[0010] By adopting the above technical solutions, short dramas with similar covers can be recommended for users, improving the click-through rate of short dramas, the stickiness of users, and saving time and labor costs.

[0011] In a second aspect, the present application provides an electronic device, adopting the following technical solutions:

[0012] An electronic device, the electronic device includes:

[0013] At least one processor;

[0014] A memory;

[0015] At least one application program, wherein at least one application program is stored in the memory and is configured to be executed by at least one processor, and the at least one application program is configured to: execute a short drama recommendation method based on cover data.

[0016] In a third aspect, the present application provides a computer-readable storage medium, adopting the following technical solutions:

[0017] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed on a computer, the computer is made to execute any short drama recommendation method provided in the first aspect based on cover data.

[0018] In summary, the present application includes at least one of the following beneficial technical effects: an automated dynamic recommendation system, which saves labor and time costs, has a wide coverage, and is convenient for popularization and use; adopts intelligent configuration, solving the problem of too strong subjectivity in manual configuration; can dynamically change to meet the changing needs of the market and users; avoids fluctuations in the quality of recommended content and improves the overall user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a schematic flowchart of the steps of a short drama recommendation method based on cover data provided by an embodiment of the present application;

[0020] Figure 2 is a schematic diagram of modules of a short drama recommendation method based on cover data provided by an embodiment of the present application;

[0021] Figure 3 is a schematic diagram of the module principle of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions and advantages of this application more clear and understandable, the following further elaborates on this application in combination with the accompanying Figures 1-3 drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0023] The short drama recommendation method based on cover data provided by the embodiments of this application is used in a short drama recommendation system.

[0024] This embodiment takes short drama recommendation as an application example to specifically illustrate the above method. In actual implementation, the above method can also be used in other video recommendation systems, and this embodiment does not limit this.

[0025] Specifically referring to Figure 1 and Figure 2 as shown, this application proposes a short drama recommendation method based on cover data, which is used in a short drama recommendation system. The method includes:

[0026] S10: Extract short drama cover image features and store them in a database;

[0027] S20: Extract user behavior data and calculate the scores of this user for each short drama;

[0028] S30: Calculate the similarity of short drama cover images according to the image features;

[0029] S40: Recommend short dramas for this user according to the user's short drama scores and the similarity of short drama cover images.

[0030] For example, recommend the short drama corresponding to the cover with a high display degree to this user to increase its click-through rate; or select a cover with a high similarity as the cover of the short drama planned to be recommended.

[0031] The solutions of the embodiments of this invention have the following advantages: They can help users quickly find short dramas they are interested in among a vast amount of content, saving time and effort; they can increase the frequency of users watching short dramas and improve user activity; the system can extend the user's stay time on the platform by recommending content that users are interested in; the accurate recommendation of the system can improve user satisfaction and reduce user churn rate; it can become one of the core competitiveness of the platform, attracting more users and content creators; this system can enhance user activity and increase user interaction and participation on the platform; the system can also attract new users to join the platform through recommendations, expanding the user base.

[0032] Among them, in this embodiment, when the similarity between the short drama score of a user and the cover of a short drama to be recommended is lower than a preset value, we consider that the short drama cover does not meet the recommendation criteria. At this time, a still frame screenshot of the short drama is selected to replace the cover. The specific method is as follows:

[0033] Obtain multiple stills of the screenshot short drama and acquire the image features of each still;

[0034] Calculate the similarity between each still and the short drama rating given by the user;

[0035] Obtain the still with the highest similarity as the new cover of the short drama and recommend it to the user.

[0036] In the embodiments of the present invention, in order to avoid the existing cover not meeting the user's preferences, through the above method, a personalized cover that meets the user's preferences is automatically generated for the short drama to be recommended, so as to increase the click probability of the user, thereby improving the revenue and user stickiness.

[0037] Further, in step S10, the extracted short drama cover image features include:

[0038] Color feature: Calculate the first-order matrix and second-order matrix features of the image;

[0039] Shape feature: The contour information in the image.

[0040] Texture feature: Gray-level co-occurrence matrix (GLCM), and CLCM is used to describe the spatial relationship between pixel pairs in the image. GLCM is a two-dimensional matrix representing the co-occurrence frequency of the gray values of two pixels in the image at a specific direction and distance. Specifically, each element (i, j) of GLCM represents the number of times the pixel with gray value i and the pixel with gray value j appear in the same direction and distance in the image.

[0041] Include the local gray-scale difference in the image, the difference in gray values of adjacent pixels in the image, the similarity of gray values of adjacent pixels in the image, the correlation of gray values of adjacent pixels in the image, and the total energy of the gray-level co-occurrence matrix in the image;

[0042] Further, the formula for calculating the local gray-scale difference in the image is:

[0043] Contrast=∑i L =- 0 1 ∑j L =- 0 1 p(i,j).(i-j) 2 ;

[0044] It can be understood that a higher contrast indicates greater gray-scale changes and coarser texture in the image.

[0045] Among them, the formula for calculating the difference in gray values of adjacent pixels in the image is:

[0046]

[0047] It can be understood that a larger dissimilarity indicates that the gray-scale changes in the image are more complex.

[0048] Among them, the formula for calculating the similarity degree of gray-scale values of adjacent pixels in the image is:

[0049]

[0050] It can be understood that its higher homogeneity indicates that the gray-scale changes in the image are smaller and the texture is more uniform.

[0051] Among them, the formula for calculating the correlation of gray-scale values of adjacent pixels in the image is:

[0052]

[0053] It can be understood that in this parameter, a higher correlation indicates that the gray-scale changes in the image are more regular.

[0054] Among them, σ i and σ j are the standard deviations of gray-scale values i and j respectively, indicating the degree of fluctuation of gray-scale values. The role of the standard deviation is to standardize the product of deviations in the numerator so that the value range of the correlation is between [-1, 1][-1, 1].

[0055] Explanation of the meaning of the value range of the correlation: 1 indicates a perfect positive correlation, that is, there is a strong linear relationship between gray-scale values i and j, and as i increases, j also increases accordingly; -1 indicates a perfect negative correlation, that is, there is a strong linear relationship between gray-scale values i and j, but as i increases, j decreases instead; 0: indicates no linear correlation, that is, there is no obvious linear relationship between gray-scale values i and j.

[0056] In this embodiment, the formula for calculating the total energy of the gray-level co-occurrence matrix in the image is:

[0057]

[0058] In this parameter, a higher ASM indicates that the gray-scale distribution in the image is more uniform.

[0059] Among them, p(i, j) is the normalized probability in the gray-level co-occurrence matrix, indicating the probability that pixels with gray-scale value i and pixels with gray-scale value j appear in the same direction and distance; L is the number of gray levels, indicating the total number of gray-scale values in the image, which is 256 levels in this embodiment; ∣i - j∣ represents the absolute difference between gray-scale values i and j; (i - j) 2 represents the square of the difference between gray-scale values i and j; μ i and μ j are the means of gray-scale values i and j respectively; σ i and σ jThey are the standard deviations of the grayscale values i and j respectively.

[0060] In the embodiment of the present invention, further in step S30, the process of calculating the similarity of the short drama cover images includes:

[0061] Calculate the perceptual hash values of the two pictures respectively; in this embodiment, dHash is calculated, and of course, other types of perceptual hash values such as pHash, dHash or aHash can also be calculated.

[0062] It can be understood that the perceptual hash can generate the fingerprint of the image for quickly comparing the similarity of images.

[0063] Calculate the Hamming distance between the two pictures through the perceptual hash values, and judge the similarity degree of the two pictures according to the size of the Hamming distance.

[0064] Among them, specifically, the calculation process of the perceptual hash includes:

[0065] Shrink the image, convert the image into a grayscale image to obtain a grayscale value matrix G, and generate a difference value;

[0066] Among them, the process of generating the difference value includes: calculating the difference between each pixel G(x,y) and its adjacent pixel G(x + 1,y) on the right:

[0067] Connect all D(x,y) values to form a binary string.

[0068] Furthermore, in this embodiment during similarity search, a ball tree index structure (Ball Tree) is used for acceleration. It can be understood that the ball tree index structure has the following several data processing advantages: the data dimension is relatively high, it is more suitable for processing a relatively large data scale, the construction process is more robust, and it has better processing ability for large-scale data.

[0069] In another embodiment of the present invention, the method for judging the similarity of the two covers may further include color histogram, SIFT feature, CNN feature and short drama category feature.

[0070] Color histogram: In this example, the standard color histogram feature, that is, RGB color, is adopted to judge the similarity of pictures.

[0071] SIFT feature: SIFT (Scale-invariant feature transform) has good stability. For the same object photographed from different angles in the same scene, the SIFT feature can still be effectively recognized. In short drama stills, this type of data accounts for a relatively large proportion. Therefore, this embodiment adopts this feature, which plays a very good role in effectively aggregating picture data.

[0072] CNN features: Features of a deep convolutional neural network (Convoutional Neural Network). Due to its strong ability to express abstract features, the deep convolutional network has achieved a level beyond human classification in image classification tasks. Since it is a multi-layer convolutional network, different levels of abstract features can be extracted from different layers. In this solution, existing tools are used to extract multi-layer abstract features from short drama posters and stills to assist the short drama recommendation task. In the example of the present invention, an 8-layer convolutional network is adopted, and the features of the 5th and 6th layers are extracted respectively.

[0073] This embodiment can also refer to the category features of short dramas: used to identify the categories to which short drama covers and stills belong. The category system of ImageNet that can be adopted in this example is also used to predict the categories to which posters and stills belong by using the network for extracting CNN features.

[0074] In the embodiment of the present invention, in step S20, the time decay effect of user behavior is considered during calculation, and the scores of earlier behaviors gradually decrease, and the scores decay over time. And in this embodiment, the weights of different behaviors need to be adjusted according to business requirements.

[0075] Among them, the process of calculating the score of a short drama for a user includes:

[0076] Data collection and storage, data preprocessing, and score calculation.

[0077] Among them, the data collection methods include reporting of buried point data and reporting through interfaces. The buried point data records of user behavior include behavior types, progress, video information collection, etc. Interface collection is to collect user behavior data through API interfaces.

[0078] Data preprocessing is to remove invalid and duplicate data through data cleaning.

[0079] In this embodiment, the score calculation calculates a comprehensive score according to the weights of each user behavior and their corresponding time decay. Specifically, the user behavior data includes user viewing, interaction, search, and subscription behavior data of short dramas. In this embodiment, the user behaviors and their corresponding scoring methods include the content in the following table:

[0080]

[0081]

[0082] Of course, the present invention can also refer to other user behaviors for score calculation, which is not limited to the description in this embodiment.

[0083] The embodiment of the present application also provides an electronic device, as Figure 3 shown Figure 3The electronic device 700 shown includes: a processor 701 and a memory 703. Among them, the processor 701 and the memory 703 are connected, such as connected through a bus 702. Optionally, the electronic device 700 may further include a transceiver 704. It should be noted that in actual applications, the transceiver 704 is not limited to one, and the structure of the electronic device 700 does not constitute a limitation to the embodiments of the present application.

[0084] The processor 701 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logic blocks, modules, and circuits described in connection with the disclosure of the present application. The processor 701 may also be a combination that implements a computing function, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0085] The bus 702 may include a path for transmitting information between the above components. The bus 702 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 702 may be divided into an address bus, a data bus, etc. For the sake of representation, Figure 3 only a thick line is used to represent it in the figure, but it does not mean that there is only one bus or one type of bus.

[0086] The memory 703 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a disk storage medium, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0087] The memory 703 is used to store the application program code for implementing the solution of this application, and is controlled by the processor 701 for execution. The processor 701 is used to execute the application program code stored in the memory 703 to implement the content shown in the foregoing method embodiments.

[0088] Among them, the electronic device includes but is not limited to: mobile terminals such as mobile phones, laptop computers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), etc. and fixed terminals such as digital TVs, desktop computers, etc. It can also be a server, etc. Figure 3 The electronic device shown is only an example and should not impose any limitations on the functions and usage scope of the embodiments of this application.

[0089] The embodiments of this application also provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed on a computer, the computer is made to execute the short drama recommendation method based on cover data provided in the foregoing embodiments.

[0090] It should be understood that although the steps in the flowchart of the accompanying drawings are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order restriction and can be executed in other orders.

[0091] The above are only partial embodiments of this application. It should be noted that for those of ordinary skill in the art, without departing from the principle of this application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. A method for recommending short plays based on cover data, characterized in that: Used in a software system, the method comprises: Extract the features of the skit cover image and store them in the database; Extract user behavior data and calculate the user's rating for each skit; Calculate the similarity of the skit cover images based on the image features; Recommend skits to the user based on the user's skit ratings and the similarity of the skit cover images.

2. The method for recommending short plays based on cover data according to claim 1, characterized in that: The extracted features of the skit cover image include: Color features: calculate the first-order matrix and second-order matrix features of the image; Texture features: including local grayscale differences in the image, differences in grayscale values ​​of adjacent pixels in the image, similarities in grayscale values ​​of adjacent pixels in the image, correlations in grayscale values ​​of adjacent pixels in the image, and the total energy of the grayscale co-occurrence matrix in the image; Shape features: contour information in an image.

3. The method for recommending short plays based on cover data according to claim 2, characterized in that: The calculation formula of local grayscale difference in image is: The formula for calculating the difference in grayscale values ​​of adjacent pixels in an image is: The similarity calculation formula of the grayscale values ​​of adjacent pixels in an image is: The correlation calculation formula of the grayscale values ​​of adjacent pixels in the image is: The total energy calculation formula of the gray level co-occurrence matrix in the image is: ASM=∑i L =-o 1 ∑j L =-0 1 p(i,j) 2 ; Where p(i,j) is the normalized probability in the gray-level co-occurrence matrix, which indicates the probability that a pixel with gray-level value i and a pixel with gray-level value j appear in the same direction and distance; L indicates the total number of gray-level values ​​in the image; |ij| indicates the absolute difference between gray-level values ​​i and j; (ij) 2 It represents the square of the difference between grayscale values ​​i and j; μ i and μ j are the means of grayscale values ​​i and j respectively; σ i and σ j are the standard deviations of grayscale values ​​i and j, respectively.

4. The method for recommending short plays based on cover data according to claim 1, characterized in that: When the similarity between the user's drama rating and the cover of a recommended drama is lower than the preset value: Screenshot multiple stills of the short play and obtain the image features of each still; Calculate the similarity between each still and the user's skit rating; The still photo with the highest similarity is obtained as the new cover of the short play and recommended to the user.

5. The method for recommending short plays based on cover data according to claim 1, characterized in that: The process of calculating the similarity of skit cover images includes: Calculate the perceptual hash values ​​of the two images respectively; The Hamming distance between two images is calculated by perceptual hash value, and the similarity between the two images is determined by the size of the Hamming distance.

6. The method for recommending short plays based on cover data according to claim 5, characterized in that: The calculation process of perceptual hashing includes: Reduce the image, convert the image into a grayscale image, obtain the grayscale value matrix G, and generate the difference value; The process of generating the difference value includes: calculating the difference between each pixel G(x,y) and its right adjacent pixel G(x+1,y): Concatenate all D(x,y) values ​​to form a binary string; Among them, when performing similarity search, a ball tree index structure is used for acceleration.

7. The method for recommending short plays based on cover data according to claim 1, characterized in that: in, The process of calculating the rating of a user's skit includes: Data collection and storage, data preprocessing, and score calculation; The scoring calculation calculates a comprehensive score based on the weight of each user behavior and its corresponding time decay.

8. The method for recommending short plays based on cover data according to claim 1, characterized in that: The user behavior data includes the user's viewing, interaction, search, and subscription behavior data for the short drama.

9. An electronic device, characterized in that: The electronic device includes: at least one processor; Memory; At least one application, wherein the at least one application is stored in a memory and configured to be executed by at least one processor, and the at least one application is configured to: execute the short drama recommendation method based on cover data as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed in a computer, the computer is caused to execute the short drama recommendation method based on cover data as described in any one of claims 1 to 8.

Images