Method and apparatus for recommending short video, electronic device, and storage medium

Abstract

The present application belongs to the technical field of video recommendation, and particularly relates to a method and apparatus for recommending a short video, an electronic device, and a storage medium. The method comprises: step 1, obtaining historical viewing data of a user, determining a first short video identifier information list and a second short video identifier information list, and generating a first feature vector and a second feature vector; step 2, obtaining a short video identifier information list to be expanded, and, on the basis of the short video identifier information list to be expanded, generating a third feature vector; step 3, on the basis of the first feature vector, the second feature vector, and the third feature vector, calculating a fourth feature vector; and step 4, obtaining a short video library to be recommended, extracting a short video identifier information list to be recommended, generating a fifth feature vector, calculating the similarity between the fifth feature vector and the fourth feature vector, and, on the basis of the similarity, filtering out a short video to be recommended. According to the present application, short video identifier information is expanded, thereby improving the diversity and freshness of short video recommendation.

Description

Short video recommendation method, device, electronic device and storage medium Technical Field

[0001] The present application belongs to the field of video recommendation technology, and in particular relates to a short video recommendation method, device, electronic device and storage medium. Background Art

[0002] With the increasing popularity of mobile devices and faster internet speeds, short, snappy videos have become popular among major platforms and users, leading to the rise of short video platforms. Short video platforms contain massive amounts of video data, and how to recommend interesting short videos from this vast amount of short video data has become a key technical challenge for developers.

[0003] The existing method of recommending short videos based on content generally relies only on the user's historical preferences. Therefore, the recommendation results generated will have a very high similarity with the objects that the user has interacted with in the past, making it difficult to tap into the user's potential other interests, making the recommendations lack diversity and freshness.

[0004] Therefore, it is desired to have a technical solution to overcome or at least alleviate at least one of the above-mentioned deficiencies of the prior art.

[0005] Summary of the Invention

[0006] The purpose of this application is to provide a short video recommendation method, device, electronic device and storage medium to solve the problem that the short video recommendation methods in the existing technology lack diversity and novelty.

[0007] The technical solution of this application is:

[0008] The first aspect of the present application provides a short video recommendation method, comprising:

[0009] Step 1: Obtain user historical viewing data, determine a first short video identification information list sorted by user interest level and a second short video identification information list sorted by user dislike level based on the user historical viewing data, generate a first feature vector based on the first short video identification information list, and generate a second feature vector based on the second short video identification information list;

[0010] Step 2: Obtain a list of identification information of short videos to be expanded, and generate a third feature vector based on the list of identification information of short videos to be expanded;

[0011] Step 3: Calculate a fourth eigenvector based on the first eigenvector, the second eigenvector, and the third eigenvector;

[0012] Step 4: Obtain a short video library to be recommended, extract a list of short video identification information to be recommended for each short video in the short video library, generate a fifth eigenvector based on the list of short video identification information to be recommended, calculate the similarity between the fifth eigenvector and the fourth eigenvector, and filter out the short videos to be recommended based on the similarity.

[0013] In at least one embodiment of the present application, in step 1, obtaining user historical viewing data and determining, based on the user historical viewing data, a first list of short video identification information sorted by user interest level and a second list of short video identification information sorted by user dislike level include:

[0014] S11. Obtain user historical viewing data, wherein the user historical viewing data includes a historical viewing short video library, positive feedback behavior information, and negative feedback behavior information.

[0015] The positive feedback behavior information includes likes, follows, positive comments, and long stays;

[0016] Negative feedback behavior information includes blacklisting, clicking "not interested", negative comments, and short stay;

[0017] S12. Filtering a positive feedback short video library based on the positive feedback behavior information, and extracting short video identification information of each short video in the positive feedback short video library;

[0018] Filtering a negative feedback short video library according to the negative feedback behavior information, and extracting short video identification information of each short video in the negative feedback short video library;

[0019] S13. Generate a first short video identification information list based on the appearance frequency of each short video identification information in the positive feedback short video library, wherein the first short video identification information list includes n short video identification information sorted from most to least according to the appearance frequency, and a weight value assigned to each short video identification information according to the appearance frequency;

[0020] A second short video identification information list is generated according to the frequency of occurrence of each short video identification information in the negative feedback short video library, wherein the second short video identification information list includes m short video identification information sorted from most to least according to the frequency of occurrence, and a weight value assigned to the corresponding short video identification information according to the frequency of occurrence.

[0021] In at least one embodiment of the present application, in step 2, obtaining a list of identification information of short videos to be expanded includes:

[0022] S21. Obtain a short video identification information database;

[0023] S22: Remove the short video identification information included in the first short video identification information list and the second short video identification information list from the short video identification information library;

[0024] S23. Randomly select a predetermined number of short video identification information from the short video identification information library after removing the short video identification information included in the first short video identification information list and the second short video identification information list to generate a short video identification information list to be expanded.

[0025] In at least one embodiment of the present application, in step three, calculating the fourth eigenvector based on the first eigenvector, the second eigenvector, and the third eigenvector includes:

[0026] Among them, W u is the fourth eigenvector, I r is a collection of short video identification information that the user is interested in, I nr A collection of identification information of short videos that users dislike. e is the set of short video identification information to be expanded, w j is the feature vector of the short video identification information of the jth user in the first feature vector, w k is the feature vector of the short video identification information that the kth user hates in the second feature vector, w e is the feature vector of the e-th short video identification information to be expanded in the third feature vector, a j 、b k , A, B, C, and D are the corresponding weight values ​​respectively.

[0027] In at least one embodiment of the present application,

[0028] The weight value A is 0.8, and the weight value B is 0.2;

[0029] The weight values ​​C and D are determined as follows:

[0030] Get the total duration t of x consecutive short videos z , and the total time t for users to watch the x short videos g , calculate the viewing depth: X = t g / t z

[0031] Determine the values ​​of C and D according to the viewing depth X value: D=1-C

[0032] Among them, c1, c2, c3, α, and β are constants.

[0033] In at least one embodiment of the present application, in step 4, obtaining a library of short videos to be recommended includes:

[0034] Obtain a short video library, perform a quality score on each short video in the short video library, filter out short videos that meet the requirements based on the quality score, and obtain a short video library to be recommended;

[0035] The quality of each short video in the short video library is scored, including:

[0036] Obtaining static tags of the short video, and determining the account score based on the static tags;

[0037] Obtain dynamic tags of the short video, and determine the interaction score based on the dynamic tags;

[0038] The account score and the interaction score are added together to obtain the quality score of the short video.

[0039] In at least one embodiment of the present application, in step 4, calculating the similarity between the fifth eigenvector and the fourth eigenvector includes:

[0040] Among them, W u is the fourth eigenvector, W i is the fifth eigenvector.

[0041] A second aspect of the present application provides a short video recommendation device, comprising:

[0042] A user feature extraction module is configured to obtain user historical viewing data, determine a first short video identification information list sorted by user interest level and a second short video identification information list sorted by user dislike level based on the user historical viewing data, generate a first feature vector based on the first short video identification information list, and generate a second feature vector based on the second short video identification information list;

[0043] A feature extraction module to be expanded is used to obtain a list of identification information of short videos to be expanded, and generate a third feature vector according to the list of identification information of short videos to be expanded;

[0044] a feature calculation module, configured to calculate a fourth feature vector based on the first feature vector, the second feature vector, and the third feature vector;

[0045] The short video recommendation module is used to obtain a short video library to be recommended, extract a list of short video identification information to be recommended for each short video in the short video library, generate a fifth eigenvector based on the list of short video identification information to be recommended, calculate the similarity between the fifth eigenvector and the fourth eigenvector, and filter out the short videos to be recommended based on the similarity.

[0046] A third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the short video recommendation method as described above when executing the computer program.

[0047] A fourth aspect of the present application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the short video recommendation method described above.

[0048] The invention has at least the following beneficial technical effects:

[0049] The short video recommendation method of the present application assigns weights to user features according to the user's degree of interest and dislike, which is conducive to finding content that the user is truly interested in from massive videos; by expanding the short video identification information, the diversity and freshness of the short video recommendations are increased. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] FIG1 is a flow chart of a short video recommendation method according to an embodiment of the present application;

[0051] FIG2 is a flow chart of a method for extracting user features according to an embodiment of the present application;

[0052] FIG3 is a flow chart of a method for screening high-quality short videos according to an embodiment of the present application;

[0053] FIG4 is a schematic diagram of a short video recommendation device according to an embodiment of the present application;

[0054] FIG5 is a schematic diagram of the structure of a computer device of a terminal or server suitable for implementing the embodiments of the present application.

[0055] Among them: 100-user feature extraction module; 200-feature extraction module to be expanded; 300-feature calculation module; 400-short video recommendation module; 500-computer equipment; 501-CPU; 502-ROM; 503-RAM; 504-bus; 505-I / O interface; 506-input part; 507-output part; 508-storage part; 509-communication part; 510-drive; 511-removable media. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of the implementation of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below in conjunction with the drawings in the embodiments of this application. In the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of this application, not all of the embodiments. The embodiments described below with reference to the drawings are exemplary and are intended to be used to explain this application, and should not be understood as limitations on this application. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. The embodiments of this application are described in detail below in conjunction with the drawings.

[0057] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be understood as limiting the scope of protection of this application.

[0058] The present application will be further described in detail below with reference to Figures 1 to 5 .

[0059] The first aspect of the present application provides a short video recommendation method, as shown in FIG1 , comprising the following steps:

[0060] Step 1: Obtain user historical viewing data, determine a first short video identification information list sorted by user interest level and a second short video identification information list sorted by user dislike level based on the user historical viewing data, generate a first feature vector based on the first short video identification information list, and generate a second feature vector based on the second short video identification information list;

[0061] Step 2: Obtain a list of identification information of short videos to be expanded, and generate a third feature vector based on the list of identification information of short videos to be expanded;

[0062] Step 3: Calculate a fourth eigenvector based on the first eigenvector, the second eigenvector, and the third eigenvector;

[0063] Step 4: Obtain a library of short videos to be recommended, extract a list of identification information of short videos to be recommended for each short video in the library, generate a fifth eigenvector based on the list of identification information of short videos to be recommended, calculate the similarity between the fifth eigenvector and the fourth eigenvector, and filter out the short videos to be recommended based on the similarity.

[0064] In a preferred embodiment of the present application, as shown in FIG2 , in step 1, the process of extracting user features specifically includes:

[0065] S11. Obtain user historical viewing data, which includes a library of historically viewed short videos, positive feedback behavior information, and negative feedback behavior information.

[0066] Positive feedback behavior information includes likes, follows, positive comments, long stays, etc.

[0067] Negative feedback behavior information includes blocking, clicking "not interested", negative comments, short stay, etc.

[0068] S12. Filtering a positive feedback short video library based on the positive feedback behavior information, and extracting short video identification information of each short video in the positive feedback short video library;

[0069] Filtering a negative feedback short video library based on the negative feedback behavior information, and extracting short video identification information of each short video in the negative feedback short video library;

[0070] S13. Generate a first short video identification information list based on the appearance frequency of each short video identification information in the positive feedback short video library, wherein the first short video identification information list includes n short video identification information sorted from most to least according to the appearance frequency, and a weight value assigned to each short video identification information according to the appearance frequency;

[0071] A second short video identification information list is generated according to the frequency of occurrence of each short video identification information in the negative feedback short video library, wherein the second short video identification information list includes m short video identification information sorted from most to least according to the frequency of occurrence, and a weight value assigned to the corresponding short video identification information according to the frequency of occurrence.

[0072] The short video recommendation method of this application first categorizes user feedback behavior into positive and negative feedback based on the positivity of the feedback. Positive feedback indicates that the user is inclined to be interested in the content, while negative feedback indicates that the user is inclined to dislike the content. Then, based on the different user feedback behaviors, a library of short videos with positive and negative feedback is selected from the historically viewed short video library. Short video identification information is extracted from the corresponding short video library. Extracting short video identification information includes extracting the identifiers of the text, audio, and image content in the short video. ASR technology can be used to convert the audio content in the short video into a text description. Key frames of the short video are extracted to obtain information-rich images, which are then converted into text using optical character recognition (OCR). Finally, short video identification information is extracted from the resulting text information collection. Short video identification information includes, but is not limited to, product identifiers, such as clothing, computers, and cars; location identifiers, such as Beijing, Shanghai, and Northeast China; holiday identifiers, such as Spring Festival, Valentine's Day, and Mid-Autumn Festival; industry identifiers, such as beauty, food, and travel; and animal identifiers, such as dogs, cats, and ducks. The occurrence frequency of each short video identification information in the short video library is counted, and it is sorted from high to low according to the occurrence frequency to obtain a short video identification information list with a specific number of short video identification information, and the corresponding short video identification information is assigned a weight value according to the proportion of the occurrence frequency. Finally, the bag-of-words model, N-gram model, TF-IDF model, neural network-based model, etc. can be used to convert the text information in the short video identification information list into a feature vector. It can be understood that in terms of distributed representation, deep learning has greater advantages than traditional methods, and it is preferred to use Paragraph2Vec technology to obtain the feature vector.

[0073] In a preferred embodiment of the present application, in step 2, obtaining a list of short video identification information to be expanded includes:

[0074] S21. Obtain a short video identification information database;

[0075] S22: Remove the short video identification information included in the first short video identification information list and the second short video identification information list from the short video identification information library;

[0076] S23. Randomly select a predetermined number of short video identification information from the short video identification information library after removing the short video identification information included in the first short video identification information list and the second short video identification information list to generate a short video identification information list to be expanded.

[0077] The short video recommendation method of the present application expands user characteristics by randomly selecting a number of short video identification information from a short video identification information library, thereby exploring other potential interests of the user.

[0078] In a preferred embodiment of the present application, in step 3, the fourth eigenvector is calculated based on the first eigenvector, the second eigenvector, and the third eigenvector, specifically:

[0079] Among them, W u is the fourth eigenvector, I r is a collection of short video identification information that the user is interested in, I nr A collection of identification information of short videos that users dislike. e is the set of short video identification information to be expanded, w j is the feature vector of the short video identification information of the jth user in the first feature vector, w k is the feature vector of the short video identification information that the kth user hates in the second feature vector, w e is the feature vector of the e-th short video identification information to be expanded in the third feature vector, a j 、b k , A, B, C, and D are the corresponding weight values ​​respectively.

[0080] Since positive feedback is generally more important than negative feedback, in this embodiment, the weight value A is set to 0.8 and the weight value B is set to 0.2.

[0081] Advantageously, in this embodiment, the weight values ​​C and D are determined as follows:

[0082] Get the total duration t of x consecutive short videos z , and the total time t for users to watch the x short videos g , calculate the viewing depth: X = t g / t z

[0083] Determine the values ​​of C and D according to the viewing depth X value: D=1-C

[0084] Among them, c1, c2, c3, α, and β are constants.

[0085] The short video recommendation method of the present application, on the one hand, assigns a weight value to the feature vector of each short video identification information in the first feature vector and the second feature vector, which is conducive to screening out features that the user is more interested in; on the other hand, the first feature vector and the second feature vector are assigned weight values ​​according to the importance of positive feedback; in addition, the user feature vector and the feature vector to be expanded are assigned weight values ​​according to the viewing depth. When the viewing depth X value is too small, it means that the user is not interested in the recommended content and it is not easy to extract suitable user features. Therefore, the short video recommendation is directly implemented based on the feature vector to be expanded. When the viewing depth X value is stable at a relatively large value, it means that the user is more interested in the recommended content and it is easy to extract suitable user features. At this time, the weight value C is determined to be a fixed value c3, such as 0.8, to ensure that the recommended short videos are appropriately improved in diversity and freshness based on the user's interest. Through the feature fusion method in this embodiment, the feature vector can be updated in real time according to user feedback to achieve real-time adjustment of the recommended short videos.

[0086] In a preferred embodiment of the present application, as shown in FIG3 , in step 4, obtaining a short video library to be recommended includes:

[0087] Obtain a short video library, score the quality of each short video in the short video library, filter out short videos that meet the requirements based on the quality score, and obtain a short video library to be recommended;

[0088] Among them, each short video in the short video library is scored for quality, including:

[0089] Obtain static tags for short videos and determine account scores based on the static tags;

[0090] Obtain dynamic tags of short videos and determine interaction scores based on the dynamic tags;

[0091] Add the account score and the interaction score to get the quality score of the short video.

[0092] In this embodiment, the similarity between the fifth eigenvector and the fourth eigenvector is calculated using cosine similarity, including:

[0093] Among them, W u is the fourth eigenvector, W i is the fifth eigenvector.

[0094] The short video recommendation method of this application scores short videos based on static and dynamic tags. Static tags include account information such as avatar, nickname, signature, and gender. Accounts can be scored based on the completeness of the account information to obtain an account score. Dynamic tags include interactive information such as completion rate, likes, follows, comments, and shares. Interactions can be scored based on the amount of each interaction to obtain an interaction score. Finally, the quality score of each short video is calculated. Short videos that meet the quality score requirements are screened based on the size of the quality score to obtain a library of short videos to be recommended. Finally, short videos to be recommended are screened from the collection of high-quality short videos based on feature similarity.

[0095] The short video recommendation method of this application assigns weights to user features based on their level of interest and dislike, which helps screen out content that users are truly interested in. It also expands user features through randomly selected features, which helps increase the diversity and freshness of short videos. A unique feature fusion method facilitates real-time adjustment of recommended short videos. Quality scoring yields a high-quality collection of recommended short videos, which helps improve user satisfaction. This application can achieve more accurate personalized recommendations, increase user engagement with the platform, improve the platform's benefits, and enhance the user experience.

[0096] Based on the above-mentioned short video recommendation method, the second aspect of the present application provides a short video recommendation device, including:

[0097] The user feature extraction module 100 is configured to obtain user historical viewing data, determine a first short video identification information list sorted by user interest level and a second short video identification information list sorted by user dislike level based on the user historical viewing data, generate a first feature vector based on the first short video identification information list, and generate a second feature vector based on the second short video identification information list;

[0098] The feature extraction module 200 is used to obtain a list of identification information of short videos to be expanded, and generate a third feature vector according to the list of identification information of short videos to be expanded;

[0099] A feature calculation module 300 is configured to calculate a fourth feature vector based on the first feature vector, the second feature vector, and the third feature vector;

[0100] The short video recommendation module 400 is used to obtain a short video library to be recommended, extract a list of short video identification information to be recommended for each short video in the short video library, generate a fifth eigenvector based on the list of short video identification information to be recommended, calculate the similarity between the fifth eigenvector and the fourth eigenvector, and filter out the short videos to be recommended based on the similarity.

[0101] In another aspect of the present application, a computer device is provided, including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned short video recommendation method.

[0102] 5, which shows a schematic diagram of the structure of a computer device 500 suitable for implementing the embodiments of the present application. The computer device shown in FIG5 is only an example and should not limit the functions and scope of use of the embodiments of the present application.

[0103] As shown in FIG5 , a computer device 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage unit 508 into a random access memory (RAM) 503. Various programs and data required for the operation of the device 500 are also stored in the RAM 503. The CPU 501, ROM 502, and RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0104] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, and the like; an output section 507 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 508 including a hard disk; and a communication section 509 including a network interface card such as a LAN card or a modem. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 510 as needed, so that computer programs read therefrom can be installed into the storage section 508 as needed.

[0105] In particular, according to the embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or installed from the removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-mentioned functions defined in the method of the present application are executed. It should be noted that the computer storage medium of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code embodied on a computer-readable medium may be transmitted using any suitable medium, including, but not limited to, wireless, wire, optical cable, RF, etc., or any suitable combination thereof.

[0106] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code includes one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0107] The modules or units described in the embodiments of this application may be implemented in software or hardware. The modules or units described may also be provided in a processor, and the names of these modules or units do not, in certain circumstances, limit the modules or units themselves.

[0108] As another aspect, the present application further provides a computer-readable storage medium, which may be included in the apparatus described in the above embodiment, or may exist independently and not be incorporated into the apparatus. The computer-readable storage medium carries one or more programs, and when the one or more programs are executed by the apparatus, the data is processed according to the above short video recommendation method.

[0109] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A short video recommendation method, characterized in that: include: Step 1: obtaining user historical viewing data, determining a first short video identification information list sorted by user interest level and a second short video identification information list sorted by user dislike level according to the user historical viewing data, generating a first feature vector according to the first short video identification information list, and generating a second feature vector according to the second short video identification information list; Step 2: obtaining a list of identification information of short videos to be extended, and generating a third feature vector according to the list of identification information of short videos to be extended; Step 3: Calculate a fourth eigenvector based on the first eigenvector, the second eigenvector and the third eigenvector; Step 4: obtain a short video library to be recommended, extract a list of short video identification information to be recommended for each short video in the short video library, generate a fifth feature vector based on the list of short video identification information to be recommended, calculate the similarity between the fifth feature vector and the fourth feature vector, and filter out the short videos to be recommended based on the similarity.

2. The short video recommendation method according to claim 1, characterized in that: In step 1, the user's historical viewing data is obtained, and a first short video identification information list sorted by the user's interest level and a second short video identification information list sorted by the user's dislike level are determined according to the user's historical viewing data, including: S11, obtaining user historical viewing data, wherein the user historical viewing data includes a historical viewing short video library, positive feedback behavior information, and negative feedback behavior information, wherein: The positive feedback behavior information includes likes, follows, positive comments, and long stays; The negative feedback behavior information includes blacklisting, clicking "not interested", negative comments, and short stay; S12, filtering out a positive feedback short video library according to the positive feedback behavior information, and extracting short video identification information of each short video in the positive feedback short video library; Filtering a negative feedback short video library according to the negative feedback behavior information, and extracting short video identification information of each short video in the negative feedback short video library; S13, generating a first short video identification information list according to the appearance frequency of each short video identification information in the positive feedback short video library, wherein the first short video identification information list includes The n short video identification information is sorted from most to least according to the frequency of occurrence, and the weight value assigned to the corresponding short video identification information according to the frequency of occurrence; A second short video identification information list is generated according to the appearance frequency of each short video identification information in the negative feedback short video library, wherein the second short video identification information list includes m short video identification information sorted from most to least according to the appearance frequency, and a weight value assigned to the corresponding short video identification information according to the appearance frequency.

3. The short video recommendation method according to claim 2, characterized in that: In step 2, obtaining a list of short video identification information to be expanded includes: S21, obtaining a short video identification information database; S22, removing the short video identification information included in the first short video identification information list and the second short video identification information list from the short video identification information library; S23. Randomly select a predetermined number of short video identification information from the short video identification information library after removing the short video identification information included in the first short video identification information list and the second short video identification information list to generate a short video identification information list to be expanded.

4. The short video recommendation method according to claim 3, characterized in that: In step three, calculating a fourth eigenvector according to the first eigenvector, the second eigenvector and the third eigenvector includes: Among them, W u is the fourth eigenvector, I r is a collection of short video identification information that the user is interested in. nr is a collection of identification information of short videos that users hate, e is the set of short video identification information to be expanded, w j is the feature vector of the short video identification information of the jth user in the first feature vector, w k is the feature vector of the short video identification information that the kth user hates in the second feature vector, w e is the feature vector of the e-th short video identification information to be expanded in the third feature vector, a j , b k , A, B, C, and D are the corresponding weight values ​​respectively.

5. The short video recommendation method according to claim 4, characterized in that: The weight value A is 0.8, and the weight value B is 0.2; The weight values ​​C and D are determined as follows: Get the total duration t of x consecutive short videos z , and the total time the user watched the x short videos Long T g , calculate the viewing depth: X=t g / t z Determine the values ​​of C and D according to the viewing depth X value: D=1-C Among them, c1, c2, c3, α, and β are constants.

6. The short video recommendation method according to claim 5, characterized in that: In step 4, obtaining a short video library to be recommended includes: Obtain a short video library, score the quality of each short video in the short video library, filter out short videos that meet the requirements according to the quality scores, and obtain a short video library to be recommended; Wherein, each short video in the short video library is scored for quality, including: Obtaining static tags of short videos, and determining account scores according to the static tags; Obtaining dynamic tags of the short video, and determining an interaction score according to the dynamic tags; The account score and the interaction score are added together to obtain a quality score of the short video.

7. The short video recommendation method according to claim 6, characterized in that: In step 4, calculating the similarity between the fifth eigenvector and the fourth eigenvector includes: Among them, W u is the fourth eigenvector, W i is the fifth eigenvector.

8. A short video recommendation device, characterized in that: include: A user feature extraction module is used to obtain user historical viewing data, determine a first short video identification information list sorted by user interest level and a second short video identification information list sorted by user dislike level according to the user historical viewing data, generate a first feature vector according to the first short video identification information list, and generate a second feature vector according to the second short video identification information list; A feature extraction module to be extended, used for obtaining a list of identification information of short videos to be extended, and generating a third feature vector according to the list of identification information of short videos to be extended; A feature calculation module is used to calculate the first feature vector and the second feature vector according to the first feature vector and the second feature vector. and calculating a fourth eigenvector from the third eigenvector; The short video recommendation module is used to obtain a short video library to be recommended, extract a list of short video identification information to be recommended for each short video in the short video library, generate a fifth feature vector according to the list of short video identification information to be recommended, calculate the similarity between the fifth feature vector and the fourth feature vector, and filter out the short videos to be recommended according to the similarity.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: When the processor executes the computer program, the short video recommendation method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it is capable of implementing the short video recommendation method as described in any one of claims 1 to 7.

Images