Content recommendation methods, author model generation methods, devices and electronic equipment

Abstract

This disclosure provides a content recommendation method, an author model generation method, apparatus, electronic device, and storage medium, relating to data processing and deep learning technologies within the field of artificial intelligence. The method includes: obtaining a first author vector of authors followed by a target user; obtaining a second author vector of authors to be recommended; determining, based on the first and second author vectors, authors similar to those followed by the target user as target authors to be recommended; and recommending content published by the target authors to the target user. By identifying authors similar to those followed by the target user, new authors that the target user may be interested in are determined, and new content published by these new authors is recommended to the target user, ensuring accurate exposure of new resources published by new authors to users and solving the problems of insufficient exposure and accuracy.

Description

Technical Field

[0001] This disclosure relates to the fields of data processing and deep learning in the field of artificial intelligence technology, and in particular to a content recommendation method, an author model generation method, an apparatus, and an electronic device. Background Technology

[0002] In related technologies, recommendation systems typically employ both explicit and implicit following methods when recommending newly published content. This requires information on explicit following and content consumption co-occurrence to identify users interested in the current author or content. However, for new authors, recommendation systems lack historical information such as explicit following and content consumption co-occurrence, resulting in inaccurate exposure of their new resources to users, leading to issues of insufficient exposure and inaccuracy. Summary of the Invention

[0003] This invention provides a content recommendation method, an author model generation method, a device, and an electronic device.

[0004] According to the first aspect, a content recommendation method is provided, comprising: obtaining a first author vector of authors followed by a target user; obtaining a second author vector of an author to be recommended; determining the author to be recommended as a target author to be recommended based on the first author vector and the second author vector; and recommending content published by the target author to the target user.

[0005] According to the second aspect, a method for generating an author model is provided, comprising: obtaining sample feature vectors of content published by sample authors in the same vertical category; obtaining similarity between sample authors in the same vertical category; and training an author model to be trained based on the sample feature vectors and the similarity between sample authors in the same vertical category to generate an author model, wherein the author model is used to generate a first author vector of authors followed by a target user and a second author vector of authors to be recommended, wherein the first author vector and the second author vector are used to determine target authors to be recommended that are similar to the authors followed by the target user among the authors to be recommended.

[0006] According to a third aspect, a content recommendation device is provided, comprising: a first acquisition module for acquiring a first author vector of an author followed by a target user; a second acquisition module for acquiring a second author vector of an author to be recommended; a determination module for determining, based on the first author vector and the second author vector, an author to be recommended that is similar to an author followed by the target user as a target author to be recommended; and a recommendation module for recommending content published by the target author to the target user.

[0007] According to the fourth aspect, an author model generation apparatus is provided, comprising: a fifth acquisition module for acquiring sample feature vectors of content published by sample authors in the same vertical category; a sixth acquisition module for acquiring similarity between sample authors in the same vertical category; and a second training module for training an author model to be trained based on the sample feature vectors and the similarity between sample authors in the same vertical category to generate an author model, wherein the author model is used to generate a first author vector of authors followed by a target user and a second author vector of authors to be recommended, wherein the first author vector and the second author vector are used to determine target authors to be recommended that are similar to the authors followed by the target user among the authors to be recommended.

[0008] According to a fifth aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the content recommendation method described in the first aspect of this disclosure, or to perform the author model generation method described in the second aspect of this disclosure.

[0009] According to a sixth aspect, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the content recommendation method according to the first aspect of this disclosure, or to perform the author model generation method according to the second aspect of this disclosure.

[0010] According to a seventh aspect, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the steps of the content recommendation method according to the first aspect of this disclosure, or the steps of the author model generation method according to the second aspect of this disclosure.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0013] Figure 1 This is a flowchart illustrating the content recommendation method according to the first embodiment of this disclosure;

[0014] Figure 2 This is a flowchart illustrating the content recommendation method according to the second embodiment of this disclosure;

[0015] Figure 3This is a flowchart illustrating the author model generation method according to the first embodiment of this disclosure;

[0016] Figure 4 This is a flowchart illustrating the author model generation method according to the second embodiment of this disclosure;

[0017] Figure 5 This is a flowchart illustrating the author model generation method according to the third embodiment of this disclosure;

[0018] Figure 6 This is a schematic diagram of a student model and a teacher model according to embodiments of this disclosure;

[0019] Figure 7 This is a block diagram of a content recommendation device according to a first embodiment of the present disclosure;

[0020] Figure 8 This is a block diagram of a content recommendation device according to a second embodiment of the present disclosure;

[0021] Figure 9 This is a block diagram of an author model generation apparatus according to a first embodiment of the present disclosure;

[0022] Figure 10 This is a block diagram of an author model generation apparatus according to a second embodiment of the present disclosure;

[0023] Figure 11 This is a block diagram of an electronic device used to implement the methods of the embodiments of this disclosure. Detailed Implementation

[0024] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0025] Artificial intelligence (AI) is a technical science that studies and develops theories, methods, technologies, and application systems to simulate, extend, and expand human intelligence. Currently, AI technology has the advantages of high automation, high accuracy, and low cost, and has been widely applied.

[0026] Data processing (DP) is the acquisition, storage, retrieval, processing, transformation, and transmission of data. The fundamental purpose of data processing is to extract and derive valuable and meaningful data from large, potentially chaotic, and difficult-to-understand datasets. Data processing is a fundamental component of systems engineering and automatic control. It permeates all areas of social production and life. The development of data processing technology and the breadth and depth of its applications have profoundly influenced the progress of human society.

[0027] Deep learning (DL) is a new research direction in the field of machine learning (ML). It learns the inherent patterns and hierarchical representations of sample data. The information gained during this learning process greatly aids in interpreting data such as text, images, and sound. Its ultimate goal is to enable machines to possess analytical and learning capabilities like humans, allowing them to recognize text, images, and sound. Specific research areas mainly include convolutional neural networks (CNNs) based on convolutional operations; autoencoder neural networks based on multi-layered neurons; and deep belief networks that are pre-trained using multi-layered autoencoder neural networks and then further optimized by incorporating discriminative information. Deep learning has achieved significant results in search technology, data mining, machine learning, machine translation, natural language processing, multimedia learning, speech recognition, recommendation and personalization technologies, and other related fields. Deep learning enables machines to mimic human activities such as sight, hearing, and thinking, solving many complex pattern recognition problems and leading to significant advancements in artificial intelligence technologies.

[0028] The following describes, with reference to the accompanying drawings, the content recommendation method, the author model generation method, the apparatus, and the electronic device of this disclosure.

[0029] Figure 1 This is a schematic flowchart of a content recommendation method according to the first embodiment of this disclosure.

[0030] like Figure 1 As shown, the content recommendation method of this disclosure embodiment may specifically include the following steps:

[0031] S101, obtain the first author vector of the authors followed by the target user.

[0032] Specifically, the execution entity of the content recommendation method in this embodiment of the disclosure may be the content recommendation device provided in this embodiment of the disclosure. This content recommendation device may be a hardware device with data processing capabilities and / or the necessary software to drive the hardware device. Optionally, the execution entity may include a workstation, server, computer, user terminal, and other devices. The user terminal may specifically include, but is not limited to, mobile phones, computers, intelligent voice interaction devices, smart home appliances, and in-vehicle terminals.

[0033] In this embodiment of the disclosure, the purpose is to accurately recommend new content published by new authors to suitable users, that is, to accurately recommend new content published by new authors that users are interested in.

[0034] Users are those who view and play multimedia content; they are the consumers of the content. Authors are those who create and publish content; they are the producers of the content. Target users are the users to whom the content is recommended, i.e., the users who are the target audience for content recommendation. The process involves obtaining a list of authors followed by the target users, which includes multiple authors followed by the target users. An author vector is a feature vector representing the author, determined based on the content the author is interested in. The author vector corresponding to the authors followed by the target users is obtained, denoted as the first author vector.

[0035] S102, obtain the second author vector of the author to be recommended.

[0036] In this embodiment of the disclosure, the author to be recommended is the newly added author, referred to as the new author. The author vector corresponding to the author to be recommended is obtained and denoted as the second author vector.

[0037] S103, based on the first author vector and the second author vector, identify the authors to be recommended that are similar to the authors followed by the target user as the target authors to be recommended.

[0038] In this embodiment of the disclosure, the target author to be recommended is the new author selected from among multiple authors to be recommended to the target user; that is, the new author whose content should be recommended to the target user. Based on the first author vector of the authors followed by the target user obtained in step S101 and the second author vector of the author to be recommended obtained in step S102, the similarity between the author to be recommended and the authors followed by the target user is calculated, and the author with the highest similarity is determined as the target author to be recommended. For example, authors with a similarity greater than a set similarity threshold, or those with the highest similarity among a set number (e.g., the top 40 or the top 10%), can be determined as the target authors to be recommended.

[0039] Among them, the Annoy algorithm can be used to identify authors similar to those followed by the target user. The Annoy algorithm is an approximate nearest neighbor search algorithm, which aims to quickly find the most similar authors among a large number of authors to be recommended.

[0040] S104 recommends content published by the target author to the target user.

[0041] In this embodiment of the disclosure, the content published by the target authors to be recommended, as determined in step S103, is recommended to the target users. This means distributing new content published by the target authors to the target users, thus achieving accurate distribution of new content published by new authors. Specifically, the content published by the target authors to be recommended is recalled, and after coarse ranking, fine ranking, and fusion, a recommendation result is generated and recommended to the target users.

[0042] In summary, the content recommendation method of this disclosure, based on the first author vector of authors followed by the target user and the second author vector of the authors to be recommended, identifies authors similar to those followed by the target user as target authors to be recommended, and recommends the content published by these target authors to the target user. This disclosure, by identifying authors similar to those followed by the target user, further identifies new authors that the target user may be interested in, and recommends the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the cold start problem for authors.

[0043] Figure 2 This is a schematic flowchart of a content recommendation method according to a second embodiment of the present disclosure.

[0044] like Figure 2 As shown, in Figure 1 Based on the illustrated embodiments, the content recommendation method of this disclosure may specifically include the following steps:

[0045] The step S101 "obtaining the first author vector of the authors followed by the target user" in the above embodiment may specifically include the following steps S201-S202.

[0046] S201, obtain the first feature vector of the content published by the authors followed by the target user.

[0047] In this embodiment, the feature vector of the content can specifically be a video understanding multimodal vector. For example, content published by the author in the past 60 days can be filtered, sorted by popularity and publication time, and 10 representative video resources can be selected to concatenate a multimodal vector to represent the author. The feature vector of the content published by the author followed by the target user is obtained and denoted as the first feature vector.

[0048] S202, Obtain the first author vector based on the first feature vector.

[0049] In this embodiment of the disclosure, a first author vector is generated to represent the authors followed by the target user, based on the multimodal first feature vector of the content published by the authors followed by the target user obtained in step S201.

[0050] The step S102 "obtaining the second author vector of the author to be recommended" in the above embodiment may specifically include the following steps S203-S204.

[0051] S203, obtain the second feature vector of the content published by the author to be recommended.

[0052] In this embodiment of the disclosure, the feature vector of the content published by the author to be recommended, i.e., the new author, is obtained and denoted as the second feature vector.

[0053] S204, Obtain the second author vector based on the second feature vector.

[0054] In this embodiment of the disclosure, a second author vector is generated to characterize the content published by the author to be recommended, i.e., the new author, based on the multimodal first feature vector of the content published by the author to be recommended, i.e., the new author, obtained in step S203.

[0055] S205, based on the first author vector and the second author vector, identify the authors to be recommended that are similar to the authors followed by the target user as the target authors to be recommended.

[0056] S206 recommends content published by the target author to the target user.

[0057] In this embodiment of the disclosure, steps S205-S206 are the same as steps S103-S104 described above, and will not be repeated here.

[0058] Based on the above embodiments, the step S202 above, "obtaining the first author vector based on the first feature vector", may specifically include the following steps: inputting the first feature vector into the author model to generate the first author vector.

[0059] In this embodiment, the first feature vector of the multimodal model is input into the trained author model, and the author model outputs the first author vector. Specifically, the author model can be a Siamese network model. The author model can be obtained using the author model generation method described in the following embodiments.

[0060] Based on the above embodiments, the step S204 above, "obtaining the second author vector based on the second feature vector", may specifically include the following steps: inputting the second feature vector into the author model to generate the second author vector.

[0061] In this embodiment of the disclosure, the second feature vector of the multimodal model is input into the trained author model, and the author model outputs the second author vector. The author model can be obtained using the author model generation method described in the following embodiments.

[0062] In summary, the content recommendation method of this disclosure, based on the first author vector of authors followed by the target user and the second author vector of the authors to be recommended, identifies authors similar to those followed by the target user as target authors to be recommended, and recommends the content published by these target authors to the target user. This disclosure, by identifying authors similar to those followed by the target user, further identifies new authors that the target user may be interested in, and recommends the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the cold start problem for authors. The method determines author vectors based on the multimodal feature vectors of the author's published content, thereby identifying similar authors. This method does not require historical interaction information and is therefore applicable to new authors as well, solving the cold start problem for new authors. Furthermore, the author vectors can be generated quickly and accurately based on an offline-trained author model, improving the recommendation speed.

[0063] Figure 3 This is a flowchart illustrating the method for generating an author model according to the first embodiment of this disclosure.

[0064] like Figure 3 As shown, the method for generating the author model in this embodiment of the disclosure may specifically include the following steps:

[0065] S301, obtain the sample feature vector of the content published by the same vertical category sample authors.

[0066] In this embodiment, the vertical category with the most recent publications for each author is considered the vertical category to which the author belongs. The sample author is the author used as a sample in the author model training. The feature vector of the content published by the sample author is the sample feature vector. The sample feature vectors of the content published by sample authors belonging to the same vertical category are obtained. The specific process for obtaining these sample feature vectors can be found in the relevant descriptions in steps S201 and S203 above, and will not be repeated here.

[0067] S302, obtain the similarity between authors of samples in the same vertical category.

[0068] In this embodiment of the disclosure, the similarity between sample authors in the same vertical category can be obtained based on the historical interaction co-occurrence information of sample authors in the same vertical category.

[0069] S303, based on the similarity between sample feature vectors and authors of samples in the same vertical category, train the author model to be trained to generate the author model.

[0070] In this embodiment of the disclosure, the sample feature vectors of the content published by the same vertical category sample authors obtained in step S301 can be used as the input of the author model to be trained, and the similarity between the same vertical category sample authors obtained in step S302 can be used as the target to train the author model to be trained, so as to generate a trained author model.

[0071] The author model is used to generate a first author vector of the authors followed by the target user and a second author vector of the authors to be recommended. The first author vector and the second author vector are used to identify target authors to be recommended that are similar to the authors followed by the target user, so that the content recommendation device can recommend the content published by the target authors to the target user. For details, please refer to the relevant description in the above content recommendation method embodiment, which will not be repeated here.

[0072] Those skilled in the art will understand that during the training of the author model, a set number of the most similar sample authors in the same vertical category, such as the top 40 sample authors, can be used as positive samples, and a set number of dissimilar sample authors in the same vertical category, such as 40, can be randomly selected as negative samples.

[0073] The author model generation method of this disclosure involves obtaining sample feature vectors of content published by authors within the same vertical category and the similarity between these authors. Based on these similarity vectors, the author model to be trained is trained to generate the author model. The author model then obtains the first author vector of authors followed by the target user and the second author vector of authors to be recommended. This allows for the identification of new authors similar to those followed by the target user, thus identifying new authors the target user might be interested in. New content published by these new authors is then recommended to the target user, ensuring accurate exposure to users and addressing issues of insufficient exposure and accuracy, while avoiding the cold start problem for authors. The method determines author vectors based on the multimodal feature vectors of the author's published content, thereby identifying similar authors. This method does not require historical interaction information and is therefore applicable to new authors as well, solving the cold start problem for new authors. Furthermore, the offline-trained author model allows for rapid and accurate generation of author vectors, improving recommendation speed.

[0074] Figure 4 This is a flowchart illustrating the method for generating an author model according to the second embodiment of this disclosure.

[0075] like Figure 4 As shown, in Figure 3 Based on the illustrated embodiments, the method for generating the author model in this disclosure may specifically include the following steps:

[0076] S401, obtain the sample feature vector of the content published by the same vertical category sample authors.

[0077] In this embodiment of the disclosure, step S401 is the same as step S301 described above, and will not be repeated here.

[0078] The step S302 "obtaining the similarity between authors of samples in the same vertical category" in the above embodiment may specifically include the following steps S402-S403.

[0079] S402, Obtain positive feedback behavior from sample users towards the content published by the sample author.

[0080] In this embodiment of the disclosure, the sample users are the users used as samples for training the author model. For example, behavioral data of multiple users on the content published by the sample author within 30 days can be obtained, and positive feedback behaviors such as single playback duration exceeding the bucket threshold or interaction (likes, comments, follows, and shares) can be filtered out.

[0081] S403 generates similarity scores between authors of samples in the same vertical category based on positive feedback behavior.

[0082] In this embodiment of the disclosure, the similarity between authors of samples in the same vertical category is generated based on the positive feedback behavior obtained in step S402.

[0083] S404: Based on the similarity between sample feature vectors and authors of samples in the same vertical category, train the author model to be trained to generate the author model.

[0084] In this embodiment of the disclosure, step S404 is the same as step S303 described above, and will not be repeated here.

[0085] Among them, such as Figure 5 As shown, step S403, "generating similarity between authors of samples in the same vertical category based on positive feedback behavior," may specifically include the following steps:

[0086] S501, based on the positive feedback behavior, obtain the sum of the number of positive feedbacks from authors of the same vertical category, the number of co-occurrences of authors of the same vertical category, and the number of positive feedbacks from sample users who co-occur with authors of the same vertical category.

[0087] In this embodiment of the disclosure, taking sample authors a and b of the same vertical category as examples, the number of positive feedbacks for sample authors a and b is I. a I b That is, the number of positive feedback behaviors of sample authors a and b respectively, and the number of co-occurrences of sample authors a and b in the same vertical category. ab That is, the number of times positive feedback behaviors of sample authors a and b co-occur, which is the sum of the number of positive feedback behaviors of sample users whose sample authors a and b co-occur in the same vertical category. That is, the sum of the number of positive feedback responses from all sample users where authors a and b co-occur, I uLet I be the number of positive feedback responses from the sample users. Based on the positive feedback behaviors of sample users a and b, I is calculated. a I b I ab and

[0088] S502, generate the similarity between authors in the same vertical category based on the sum of the number of positive feedbacks from authors in the same vertical category, the number of co-occurrences of authors in the same vertical category, and the number of positive feedbacks from users of samples that co-occur with authors in the same vertical category.

[0089] In this embodiment of the disclosure, I is calculated according to step S501. a I b I ab and Generate the similarity sim(a,b) between authors of samples from the same vertical category. This can be calculated using the following formula:

[0090]

[0091] It should be noted here that an author similarity model can be used to generate similarity scores between authors within the same vertical category. Positive feedback behaviors from sample users are input into the author similarity model, which then calculates the similarity score. a I b I ab and The similarity sim(a,b) between authors of samples in the same vertical category is calculated according to the above formula.

[0092] Among them, such as Figure 6 As shown, the author similarity model can serve as the teacher model, and the author model can serve as the student model. The teacher model's input is the positive feedback behavior of sample users' interactions, and its output is the similarity of sample authors within the same vertical category. The output of the teacher model, the similarity of sample authors within the same vertical category, is distilled into the student model as its target. The student model's input is the multimodal feature vector of sample videos containing content published by sample authors, and its output is the sample author vector. The target is the similarity of sample authors within the same vertical category.

[0093] In summary, the author model generation method of this disclosure obtains the sample feature vectors of content published by sample authors in the same vertical category and the similarity between sample authors in the same vertical category. Based on the sample feature vectors and the similarity between sample authors in the same vertical category, the author model to be trained is trained to generate an author model. The author model obtains the first author vector of authors followed by the target user and the second author vector of authors to be recommended, thereby determining authors similar to those followed by the target user, i.e., identifying new authors that the target user may be interested in, and recommending the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the author cold start problem. The author vector is determined based on the multimodal feature vectors of the author's published content, thereby identifying similar authors. This method does not require historical interaction information and is therefore applicable to new authors, solving the new author cold start problem. Furthermore, the author vector can be generated quickly and accurately based on the offline trained author model, improving the recommendation speed. Using the teacher model to obtain the similarity of sample authors in the same vertical category as the target of the student model, i.e., the author model, improves the performance of the author model, thereby improving the accuracy of recommendations.

[0094] Figure 7 This is a block diagram of a content recommendation device according to a first embodiment of the present disclosure.

[0095] like Figure 7 As shown, the content recommendation device 700 of this disclosure includes: a first acquisition module 701, a second acquisition module 702, a determination module 703, and a recommendation module 704.

[0096] The first acquisition module 701 is used to acquire the first author vector of the authors followed by the target user.

[0097] The second acquisition module 702 is used to acquire the second author vector of the author to be recommended.

[0098] The determination module 703 is used to determine the authors to be recommended as target authors based on the first author vector and the second author vector.

[0099] The recommendation module 704 is used to recommend content published by the target authors to the target users.

[0100] It should be noted that the above explanation of the content recommendation method embodiments also applies to the content recommendation device of the present disclosure embodiments, and the specific process will not be repeated here.

[0101] In summary, the content recommendation device of this disclosure, based on the first author vector of authors followed by the target user and the second author vector of the authors to be recommended, identifies authors similar to those followed by the target user as target authors to be recommended, and recommends the content published by these target authors to the target user. This disclosure, by identifying authors similar to those followed by the target user, further identifies new authors that the target user may be interested in, and recommends the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the cold start problem for authors.

[0102] Figure 8 This is a block diagram of a content recommendation device according to a second embodiment of the present disclosure.

[0103] like Figure 8 As shown, the content recommendation device 800 of this embodiment includes: a first acquisition module 801, a second acquisition module 802, a determination module 803, and a recommendation module 804.

[0104] The first acquisition module 801 has the same structure and function as the first acquisition module 701 in the previous embodiment, the second acquisition module 802 has the same structure and function as the second acquisition module 702 in the previous embodiment, the determination module 803 has the same structure and function as the determination module 703 in the previous embodiment, and the recommendation module 804 has the same structure and function as the recommendation module 704 in the previous embodiment.

[0105] Further, the first acquisition module 801 includes: a first acquisition unit 8011, used to acquire a first feature vector of content published by authors followed by the target user; and a second acquisition unit 8012, used to acquire a first author vector based on the first feature vector; the second acquisition module 802 includes: a third acquisition unit 8021, used to acquire a second feature vector of content published by authors to be recommended; and a fourth acquisition unit 8022, used to acquire a second author vector based on the second feature vector.

[0106] Furthermore, the second acquisition unit 8012 is further configured to: input the first feature vector into the author model to generate the first author vector; the fourth acquisition unit 8022 is further configured to: input the second feature vector into the author model to generate the second author vector.

[0107] Furthermore, the content recommendation device 800 of this disclosure embodiment may further include: a third acquisition module, used to acquire sample feature vectors of content published by sample authors of the same vertical category; a fourth acquisition module, used to acquire the similarity between sample authors of the same vertical category; and a first training module, used to train the author model to be trained based on the sample feature vectors and the similarity between sample authors of the same vertical category, so as to generate an author model.

[0108] Furthermore, the fourth acquisition module includes: a fifth acquisition unit, used to acquire positive feedback behavior of sample users towards the content published by sample authors; and a first generation unit, used to generate similarity between sample authors of the same vertical category based on the positive feedback behavior.

[0109] Furthermore, the first generation unit is further configured to: obtain, based on the positive feedback behavior, the sum of the number of positive feedbacks from sample authors in the same vertical category, the number of co-occurrences of sample authors in the same vertical category, and the sum of the number of positive feedbacks from sample users who co-occur with sample authors in the same vertical category; and generate the similarity between sample authors in the same vertical category based on the sum of the number of positive feedbacks from sample authors in the same vertical category, the number of co-occurrences of sample authors in the same vertical category, and the sum of the number of positive feedbacks from sample users who co-occur with sample authors in the same vertical category.

[0110] It should be noted that the above explanation of the content recommendation method embodiments also applies to the content recommendation device of the present disclosure embodiments, and the specific process will not be repeated here.

[0111] In summary, the content recommendation device of this disclosure, based on the first author vector of authors followed by the target user and the second author vector of the authors to be recommended, identifies authors similar to those followed by the target user as target authors to be recommended, and recommends the content published by these target authors to the target user. This disclosure, by identifying authors similar to those followed by the target user, further identifies new authors that the target user may be interested in, and recommends the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the cold start problem for authors. The device determines author vectors based on the multimodal feature vectors of the author's published content, and then identifies similar authors, without requiring historical interaction information, thus also applicable to new authors and solving the cold start problem for new authors. Furthermore, the author vectors can be generated quickly and accurately based on an offline-trained author model, improving the recommendation speed.

[0112] Figure 9 This is a block diagram of an author model generation apparatus according to a first embodiment of the present disclosure.

[0113] like Figure 9As shown, the author model generation device 900 of this disclosure embodiment includes: a fifth acquisition module 901, a sixth acquisition module 902, and a second training module 903.

[0114] The fifth acquisition module 901 is used to acquire the sample feature vectors of the content published by the authors of the same vertical category.

[0115] The sixth acquisition module 902 is used to obtain the similarity between authors of samples in the same vertical category.

[0116] The second training module 903 is used to train the author model to be trained based on the similarity between the sample feature vector and the authors of the same vertical category to generate an author model. The author model is used to generate the first author vector of the authors followed by the target user and the second author vector of the authors to be recommended. The first author vector and the second author vector are used to identify the target authors to be recommended that are similar to the authors followed by the target user.

[0117] It should be noted that the above explanation of the author model generation method embodiment also applies to the author model generation apparatus of the present disclosure embodiment, and the specific process will not be repeated here.

[0118] In summary, the author model generation apparatus of this disclosure obtains the sample feature vectors of content published by sample authors in the same vertical category and the similarity between sample authors in the same vertical category. Based on the sample feature vectors and the similarity between sample authors in the same vertical category, it trains the author model to be trained to generate an author model. Based on the author model, it obtains the first author vector of authors followed by the target user and the second author vector of authors to be recommended, thereby determining authors similar to those followed by the target user, i.e., identifying new authors that the target user may be interested in, and recommending the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the cold start problem for authors. The author vector is determined based on the multimodal feature vectors of the author's published content, thereby identifying similar authors. This method does not require historical interaction information and is therefore applicable to new authors as well, solving the cold start problem for new authors. Furthermore, based on an offline-trained author model, author vectors can be generated quickly and accurately, improving the recommendation speed.

[0119] Figure 10 This is a block diagram of an author model generation apparatus according to a second embodiment of the present disclosure.

[0120] like Figure 10 As shown, the author model generation device 1000 of this disclosure embodiment includes: a fifth acquisition module 1001, a sixth acquisition module 1002, and a second training module 1003.

[0121] The fifth acquisition module 1001 has the same structure and function as the fifth acquisition module 901 in the previous embodiment, the sixth acquisition module 1002 has the same structure and function as the sixth acquisition module 902 in the previous embodiment, and the second training module 1003 has the same structure and function as the second training module 903 in the previous embodiment.

[0122] Furthermore, the sixth acquisition module 1002 includes: a sixth acquisition unit 1021, used to acquire positive feedback behavior of sample users on the content published by sample authors; and a second generation unit 1021, used to generate similarity between sample authors of the same vertical category based on the positive feedback behavior.

[0123] Furthermore, the second generation unit 1021 is further configured to: obtain, based on the positive feedback behavior, the sum of the number of positive feedbacks of sample authors in the same vertical category, the number of co-occurrences of sample authors in the same vertical category, and the sum of the number of positive feedbacks of sample users who co-occur with sample authors in the same vertical category; and generate the similarity between sample authors in the same vertical category based on the sum of the number of positive feedbacks of sample authors in the same vertical category, the number of co-occurrences of sample authors in the same vertical category, and the sum of the number of positive feedbacks of sample users who co-occur with sample authors in the same vertical category.

[0124] It should be noted that the above explanation of the author model generation method embodiment also applies to the author model generation apparatus of the present disclosure embodiment, and the specific process will not be repeated here.

[0125] In summary, the author model generation apparatus of this disclosure obtains the sample feature vectors of content published by sample authors in the same vertical category and the similarity between sample authors in the same vertical category. Based on the sample feature vectors and the similarity between sample authors in the same vertical category, it trains the author model to be trained to generate an author model. Based on the author model, it obtains the first author vector of authors followed by the target user and the second author vector of authors to be recommended, thereby determining authors similar to those followed by the target user, i.e., identifying new authors that the target user may be interested in, and recommending the new content published by these new authors to the target user. This ensures that new resources published by new authors are accurately exposed to users, solving the problems of insufficient exposure and accuracy, and avoiding the author cold start problem. The author vector is determined based on the multimodal feature vectors of the author's published content, thereby identifying similar authors. This does not require historical interaction information and is therefore applicable to new authors as well, solving the new author cold start problem. Furthermore, the author vector can be generated quickly and accurately based on the offline-trained author model, improving the recommendation speed. Using the teacher model to obtain the similarity of sample authors in the same vertical category as the target of the student model, i.e., the author model, improves the performance of the author model, thereby improving the accuracy of recommendations.

[0126] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0127] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0128] Figure 11 A schematic block diagram of an example electronic device 1100 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0129] like Figure 11 As shown, the electronic device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. The RAM 1103 may also store various programs and data required for the operation of the electronic device 1100. The computing unit 1101, ROM 1102, and RAM 1103 are interconnected via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0130] Multiple components in electronic device 1100 are connected to I / O interface 1105, including: input unit 1106, such as keyboard, mouse, etc.; output unit 1107, such as various types of displays, speakers, etc.; storage unit 1108, such as disk, optical disk, etc.; and communication unit 1109, such as network card, modem, wireless transceiver, etc. Communication unit 1109 allows electronic device 1100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0131] The computing unit 1101 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1101 performs the various methods and processes described above, such as... Figures 1 to 2 The content recommendation method shown Figures 3-6 The author model generation method is illustrated. For example, in some embodiments, the content recommendation method or the author model generation method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 1100 via ROM 1102 and / or communication unit 1109. When the computer program is loaded into RAM 1103 and executed by computing unit 1101, one or more steps of the semantic parsing method described above may be performed. Alternatively, in other embodiments, computing unit 1101 may be configured to perform the content recommendation method or the author model generation method by any other suitable means (e.g., by means of firmware).

[0132] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0133] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0134] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0135] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0136] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0137] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or simply "VPS") services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0138] According to embodiments of this disclosure, this disclosure also provides a computer program product, including a computer program, wherein the computer program, when executed by a processor, implements the steps of the content recommendation method or the author model generation method shown in the above embodiments of this disclosure.

[0139] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0140] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for generating an author model, comprising: obtaining sample feature vectors of contents published by sample authors of the same vertical category; obtaining similarities between the sample authors of the same vertical category; and training an author model to be trained according to the sample feature vectors and the similarities between the sample authors of the same vertical category, to generate the author model for generating a first author vector of an author followed by a target user and a second author vector of an author to be recommended, the first author vector and the second author vector being used to determine a target author to be recommended similar to the author followed by the target user from the author to be recommended; the obtaining of the similarities between the sample authors of the same vertical category comprises: obtaining positive feedback behaviors of sample users on the contents published by the sample authors; obtaining, according to the positive feedback behaviors, a positive feedback frequency of the sample authors of the same vertical category, a co-occurrence frequency of the sample authors of the same vertical category, and a sum of positive feedback frequencies of the sample users co-occurring with the sample authors of the same vertical category; and generating the similarities between the sample authors of the same vertical category according to the positive feedback frequency of the sample authors of the same vertical category, the co-occurrence frequency of the sample authors of the same vertical category, and the sum of the positive feedback frequencies of the sample users co-occurring with the sample authors of the same vertical category.

2. A method for recommending contents, comprising: obtaining a first feature vector of contents published by an author followed by a target user; inputting the first feature vector into an author model to generate a first author vector, the author model being generated by the method of claim 1; obtaining a second feature vector of contents published by an author to be recommended; inputting the second feature vector into the author model to generate a second author vector; determining, according to the first author vector and the second author vector, the author to be recommended similar to the author followed by the target user as a target author to be recommended; and recommending, to the target user, contents published by the target author to be recommended.

3. The method for recommending contents according to claim 2, further comprising: obtaining sample feature vectors of contents published by sample authors of the same vertical category; obtaining similarities between the sample authors of the same vertical category; and training an author model to be trained according to the sample feature vectors and the similarities between the sample authors of the same vertical category, to generate the author model. the obtaining of the similarities between the sample authors of the same vertical category comprises: obtaining positive feedback behaviors of sample users on the contents published by the sample authors; and generating the similarities between the sample authors of the same vertical category according to the positive feedback behaviors.

4. The content recommendation method of claim 3, wherein, the generating of the similarities between the sample authors of the same vertical category according to the positive feedback behaviors comprises: obtaining, according to the positive feedback behaviors, a positive feedback frequency of the sample authors of the same vertical category, a co-occurrence frequency of the sample authors of the same vertical category, and a sum of positive feedback frequencies of the sample users co-occurring with the sample authors of the same vertical category; and ​ 5. The content recommendation method of claim 4, wherein, ​ ​ generate the similarity between the same vertical category sample authors according to the sum of the positive feedback times of the same vertical category sample authors, the co-occurrence times of the same vertical category sample authors, and the positive feedback times of the sample users co-occurring with the same vertical category sample authors.

6. A content recommendation apparatus, comprising: a first obtaining module configured to obtain a first author vector of an author followed by a target user; a second obtaining module configured to obtain a second author vector of a to-be-recommended author; a determining module configured to determine, according to the first author vector and the second author vector, the to-be-recommended author similar to the author followed by the target user as a target to-be-recommended author; and a recommendation module configured to recommend content published by the target to-be-recommended author to the target user. The first obtaining module comprises: a first obtaining unit configured to obtain a first feature vector of content published by the author followed by the target user; and a second obtaining unit configured to obtain the first author vector according to the first feature vector. The second obtaining module comprises: a third obtaining unit configured to obtain a second feature vector of content published by the to-be-recommended author; and a fourth obtaining unit configured to obtain the second author vector according to the second feature vector. The second obtaining unit is further configured to: input the first feature vector into an author model to generate the first author vector, the author model being generated by the method in claim 1. The fourth obtaining unit is further configured to: input the second feature vector into the author model to generate the second author vector.

7. The content recommendation apparatus of claim 6, further comprising: a third obtaining module configured to obtain a sample feature vector of content published by same vertical category sample authors; a fourth obtaining module configured to obtain a similarity between the same vertical category sample authors; and a first training module configured to train a to-be-trained author model according to the sample feature vector and the similarity between the same vertical category sample authors to generate the author model. The fourth obtaining module comprises: a fifth obtaining unit configured to obtain positive feedback behaviors of sample users on content published by the sample authors; and 8. The content recommendation device of claim 7, wherein, a first generating unit configured to generate the similarity between the same vertical category sample authors according to the positive feedback behaviors. The first generating unit is further configured to: obtain, according to the positive feedback behaviors, the sum of the positive feedback times of the same vertical category sample authors, the co-occurrence times of the same vertical category sample authors, and the positive feedback times of the sample users co-occurring with the same vertical category sample authors; and 9. The content recommendation device of claim 8, wherein, generate the similarity between the same vertical category sample authors according to the sum of the positive feedback times of the same vertical category sample authors, the co-occurrence times of the same vertical category sample authors, and the positive feedback times of the sample users co-occurring with the same vertical category sample authors.

10. An apparatus for generating an author model, comprising: a fifth obtaining module configured to obtain a sample feature vector of content published by same vertical category sample authors; a sixth obtaining module configured to obtain a similarity between the same vertical category sample authors; and a first training module configured to train a to-be-trained author model according to the sample feature vector and the similarity between the same vertical category sample authors to generate the author model. ​ ​ The second training module is configured to train the author model to be trained according to the similarity between the sample feature vectors and the same vertical category sample authors, to generate an author model, and the author model is configured to generate a first author vector of an author focused on by a target user and a second author vector of an author to be recommended, and the first author vector and the second author vector are configured to determine a target author to be recommended similar to the author focused on by the target user from the authors to be recommended. The sixth acquisition module includes: The sixth acquisition unit is configured to acquire positive feedback behaviors of sample users on content published by the sample authors. And The second generation unit is configured to generate the similarity between the same vertical category sample authors according to the positive feedback behaviors. The second generation unit is further configured to: According to the positive feedback behaviors, acquire the number of positive feedbacks of the same vertical category sample authors, the number of co-occurrences of the same vertical category sample authors, and the sum of the number of positive feedbacks of the sample users co-occurring with the same vertical category sample authors; and According to the number of positive feedbacks of the same vertical category sample authors, the number of co-occurrences of the same vertical category sample authors, and the sum of the number of positive feedbacks of the sample users co-occurring with the same vertical category sample authors, generate the similarity between the same vertical category sample authors.

11. An electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of claim 1, or perform the method of any one of claims 2-5.

12. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method according to claim 1, or perform the method according to any one of claims 2-5.

13. A computer program product comprising a computer program which, when executed by a processor, implements the steps of the method according to claim 1, or implements the steps of the method according to any one of claims 2-5.

Images