WO2024041043A1

WO2024041043A1 - Book recommendation method, computing device, and computer storage medium

Info

Publication number: WO2024041043A1
Application number: PCT/CN2023/094237
Authority: WO
Inventors: 王海璐
Original assignee: 掌阅科技股份有限公司
Priority date: 2022-08-24
Filing date: 2023-05-15
Publication date: 2024-02-29
Also published as: CN115238194A

Abstract

Disclosed in the present invention are a book recommendation method, a computing device, and a computer storage medium. The method comprises: obtaining a plurality of candidate books and book features corresponding to the candidate books; for any candidate book, inputting the book features of the candidate book into a pre-trained book recommendation model for recommendation prediction to obtain a recommendation score corresponding to the candidate book; and according to the recommendation score, screening the plurality of candidate books for a preset number of books to be recommended, and perform book recommendation.

Description

Book recommendation methods, computing devices and computer storage media

Cross-references to related applications

This application claims priority to the Chinese patent application filed with the China Patent Office on August 24, 2022, with application number 2022110195744 and titled "Book Recommendation Method, Computing Device and Computer Storage Medium", the entire content of which is incorporated herein by reference. Applying.

Technical field

The present disclosure relates to the field of computer technology, and specifically to a book recommendation method, computing device and computer storage medium.

Background technique

Generally speaking, the potential revenue of the reading platform (on the site) is directly proportional to the access traffic. Therefore, the reading platform will increase access traffic through a variety of methods. Placing books on other websites (off-site) is one of the ways to introduce users who visit outside the site to the site.

At present, most of the books placed outside the site are manually selected and recommended by operators and others. After the placement, the user retention data of the users attracted by these books on the site are counted, such as 2 (next day), 3, 7, 14, User retention data on the 30th. For example, if the user retention data brought by a certain book on the 30th is still very high, it means that the book has been successfully launched. However, placing books off-site relies on personnel experience and consumes energy. There is an urgent need for an intelligent and automated way to screen and recommend books.

Contents of the invention

In view of the above problems, the present disclosure is proposed to provide a book recommendation method, a computing device and a computer storage medium that overcome the above problems or at least partially solve the above problems.

According to one aspect of the present disclosure, a book recommendation method is provided, including:

Obtain multiple candidate books and book characteristics corresponding to each candidate book;

For any candidate book, the book features of the candidate book are input into the pre-trained book recommendation model for recommendation prediction, and the recommendation score corresponding to the candidate book is obtained. Among them, the book recommendation model is based on the historical recommended books placed outside the site. Obtained from user retention data and book feature training;

Screen a preset number of books to be recommended from multiple candidate books based on the recommendation scores for book recommendation.

According to another aspect of the present disclosure, a computing device is provided, including: a processor, a memory, a communication interface, and a communication bus. The processor, the memory, and the communication interface complete communication with each other through the communication bus;

The memory is used to store at least one executable instruction. The executable instruction causes the processor to perform the following operations:

According to another aspect of the present disclosure, a non-volatile computer-readable storage medium is provided. At least one executable instruction is stored in the non-volatile computer-readable storage medium. The executable instruction causes the processor to execute the above-mentioned book. The operations corresponding to the recommended methods.

According to the solution provided by the present disclosure, a book recommendation model trained based on user retention data and book features corresponding to historical recommended books posted outside the site is used to determine the recommendation score of each candidate book, and the books to be recommended are screened based on the recommendation score. , thus realizing automatic screening of recommended books, while improving the success rate of book recommendation, attracting more users to use reading applications to read books, improving user retention rates, and reducing manual participation, saving costs.

The above description is only an overview of the technical solutions of the present disclosure. In order to have a clearer understanding of the technical means of the present disclosure, they can be implemented according to the content of the description, and in order to make the above and other objects, features and advantages of the present disclosure more obvious and understandable. , the specific implementation modes of the present disclosure are specifically listed below.

Figure overview

Various other advantages and benefits will become apparent upon reading the detailed description of the preferred embodiments below. This will become apparent to those of ordinary skill in the art. The drawings are for the purpose of illustrating preferred embodiments only and are not to be considered limiting of the disclosure. Also throughout the drawings, the same reference characters are used to designate the same components. In the attached picture:

Figure 1 shows a schematic flow chart of a book recommendation method according to one embodiment of the present disclosure;

Figure 2 shows a schematic flow chart of a book recommendation method according to another embodiment of the present disclosure; and

FIG. 3 shows a schematic structural diagram of a computing device according to an embodiment of the present disclosure.

Preferred embodiments of the present disclosure

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a thorough understanding of the disclosure, and to fully convey the scope of the disclosure to those skilled in the art.

Figure 1 shows a schematic flowchart of a book recommendation method according to an embodiment of the present disclosure. As shown in Figure 1, the method includes the following steps:

Step S101: Obtain multiple candidate books and book characteristics corresponding to each candidate book.

Specifically, when recommending books, multiple candidate books are obtained. In addition, for each candidate book, the book characteristics corresponding to the candidate book also need to be obtained. Candidate books refer to books that are recommended candidates when recommending books, that is, the books to be recommended are selected from the candidate books. Book features are relevant features used to characterize books.

Step S102: For any candidate book, input the book characteristics of the candidate book into the pre-trained book recommendation model for recommendation prediction, and obtain the recommendation score corresponding to the candidate book. The book recommendation model is based on historical recommendations placed outside the site. It is obtained by training the user retention data corresponding to the books and the characteristics of the books.

In order to improve the recommendation success rate, automatically screen recommended books, reduce manual participation, and save costs, this embodiment is based on a pre-trained book recommendation model to perform recommendation predictions. Specifically, multiple candidate books are obtained according to step S101 and the book features corresponding to each candidate book, for any candidate book, input the book features of the candidate book into the pre-trained book recommendation The recommendation model performs recommendation prediction and obtains the recommendation score corresponding to the candidate book. The recommendation score reflects the possibility of the candidate book being recommended as a book to be recommended. The higher the recommendation score, the more likely the candidate book is to be recommended as a book to be recommended. The higher the value, the lower the recommendation score, and the lower the possibility that the candidate book will be recommended as a book to be recommended.

Among them, the book recommendation model is trained based on the user retention data and book characteristics corresponding to the historically recommended books placed outside the site. The user retention data refers to the users attracted by the historically recommended books placed outside the site. Retain data.

Step S103: Screen a preset number of books to be recommended from multiple candidate books based on the recommendation scores for book recommendation.

After calculating the recommendation score of each candidate book, a preset number of books to be recommended can be selected from multiple candidate books according to the recommendation score for book recommendation. For example, a preset number of books with the highest recommendation score can be selected from multiple candidate books. The candidate books are used as books to be recommended for book recommendation. Here, the books to be recommended are recommended outside the site, so as to be placed outside the site to achieve user drainage.

The solution provided by this disclosure uses a book recommendation model trained based on user retention data and book features corresponding to historical recommended books posted outside the site to determine the recommendation score of each candidate book, and screen the books to be recommended based on the recommendation score. This enables automatic screening of recommended books, while improving the success rate of book recommendation, attracting more users to use reading applications to read books, improving user retention rates, and reducing manual participation, saving costs.

FIG. 2 shows a schematic flowchart of a book recommendation method according to another embodiment of the present disclosure. As shown in Figure 2, the method includes the following steps:

Step S201: For any book in the book library, calculate the similarity between the book and the historically recommended book with the first tag.

Specifically, when the number of books in the book library is large, in order to improve the recommendation efficiency, a similar book recall method is used to select candidate books from the book library, where the candidate books may be posted off-site as books to be recommended.

Books that have been recommended outside the site for off-site placement in the past period are called historically recommended books. Among the historically recommended books placed outside the site, there may be successful recommendations (recommendation success here refers to the retention rate of users who diverted the historically recommended books to the site). Books with high ) may also fail to be recommended (recommended here Recommendation failure refers to books where the retention rate of users who have been redirected to the site by the historical recommended books is low), and the first tag is the tag corresponding to the successfully recommended book. For example, the first tag is set to 1, which is used to identify the successful recommendation of the book. , the specific form is not limited.

When screening candidate books from the book library, the screening is based on similarity. Therefore, it is necessary to calculate the similarity between any book in the book library and the historically recommended book with the first tag. The similarity reflects the two The similarity between the books, the higher the similarity, the more similar the two books are, and vice versa. For example, for any book in the book library, obtain the book characteristics corresponding to the book, obtain the book characteristics corresponding to the historical recommended books with the first tag, and obtain the book characteristics corresponding to any book in the book library and the historical recommended books with the first tag. The book features corresponding to the recommended books are used to calculate the similarity between two books. For example, similarity algorithms such as i2i model, swing algorithm, and graph algorithm can be used to calculate the similarity between two books.

Step S202: Determine whether the similarity is greater than or equal to the preset similarity threshold. If so, execute step S203.

After calculating the similarity between any book in the book library and the historical recommended book with the first tag according to step S201, multiple candidate books can be screened from the book library according to the similarity. Specifically, the similarity can be compared with A preset similarity threshold is used for comparison, and books with a similarity greater than the preset similarity threshold are used as candidate books. The preset similarity threshold can be set according to actual needs. In order to avoid too many candidate books, the preset similarity threshold can be The degree threshold is set slightly larger. Alternatively, the books in the book library can be sorted based on similarity, and a specified number of books can be selected as candidate books, for example, 3000 books can be selected.

Step S203: Determine the book as a candidate book, and obtain the book characteristics corresponding to the candidate book.

When it is determined that the similarity between any book in the book library and the historically recommended book with the first tag is greater than or equal to the preset similarity threshold, the book can be determined to be a candidate book, and the book corresponding to the candidate book can be obtained Features, where book features include one or more of the following features: book word count, reading rate, author, follow-up rate, book introduction, book title, and book type. It should be noted that book features are not limited to the features listed above, and may also include other features that can be used to characterize books, which will not be listed here.

Optionally, all books in the book library can be used as candidate books. At this time, the book characteristics corresponding to each book in the book library will be directly obtained. Among them, it can be determined based on off-site delivery scenarios. Whether to select candidate books from the book library based on similar book recall, or to directly use all books in the book library as candidate books. For example, in an e-commerce delivery scenario, candidate books are selected from the book library based on similar book recall.

Step S204: For any candidate book, input the book features of the candidate book into the pre-trained book recommendation model for recommendation prediction, and obtain the recommendation score corresponding to the candidate book, where the book recommendation model is based on historical recommendations placed outside the site. It is obtained by training the user retention data corresponding to the books and the characteristics of the books.

In order to improve the recommendation success rate, automatically screen recommended books, reduce manual participation, and save costs, this embodiment is based on a pre-trained book recommendation model to perform recommendation predictions. The book recommendation model is a model that can output the recommendation score of books. , the recommendation score reflects the likelihood that the candidate book will be recommended as a book to be recommended. The higher the recommendation score, the higher the likelihood that the candidate book will be recommended as a book to be recommended. The lower the recommendation score, the candidate book will be recommended as a book to be recommended. The lower the possibility that the book to be recommended will be recommended. Specifically, after screening out the candidate books and obtaining the book features corresponding to the candidate books, for any candidate book, input the book features of the candidate book into the pre-trained book recommendation The model performs recommendation prediction and obtains the recommendation score corresponding to the candidate book. The recommendation score corresponding to the candidate book is any value within [0,1].

In an optional implementation of the present disclosure, the book recommendation model can be trained by the following method:

According to the user retention data, the historically recommended books placed outside the site are divided into positive sample books and negative sample books. The first label is set for the positive sample books and the second label is set for the negative sample books;

For any historical recommended book, obtain the book characteristics corresponding to the historical recommended book;

Model training is performed based on the book characteristics of the positive sample books, the book characteristics of the negative sample books, the first label, and the second label, and a book recommendation model is obtained.

Specifically, user retention data refers to the retention data of reading users attracted by historical recommended books posted outside the site. For example, it can be the user retention rate, which reflects the retention ratio of attracted users. Usually, the higher the user retention rate, it indicates that the historically recommended books have been successfully launched outside the site, attracting and retaining more users to use the reading application to read. On the contrary, the lower the user retention rate, indicating that the historically recommended books The unsuccessful placement outside the site did not retain more users to use the reading application for reading.

When training the model, positive and negative samples are distinguished based on user retention data. The user retention data corresponding to positive sample books is high, and the user retention data corresponding to negative sample books is low. For example, the distinction can be made as follows: For any historically recommended book, count the n-day user retention data corresponding to the historically recommended book, where n is greater than or equal to 2. Taking n=2 as an example, here is specifically: count the next-day retention data of reading users corresponding to the historically recommended book. ; Taking n=3 as an example, this step is specifically: counting the 3-day retention data of reading users corresponding to historically recommended books. In the actual application process, the specific value of n can be selected according to business needs, and this disclosure does not limit this. For example, the user retention data of the next day, the user retention data of the third day...the user retention data of the 30th day; the user retention data of n days The retention data is compared with the preset retention threshold. If the n-day user retention data is greater than or equal to the preset retention threshold, the historically recommended books will be classified as positive sample books; if the n-day user retention data is less than the preset retention threshold, the historical recommended books will be classified as positive sample books. Recommended books are divided into negative sample books. The preset retention threshold can be set based on actual experience. For example, if it is set to 15%, it requires n-day user retention data to be greater than or equal to 15%, indicating that the historically recommended books are outside the site. The delivery was very successful. This is just an example and does not have any limiting effect.

After dividing the historically recommended books into positive sample books or negative sample books, a first label is set for the positive sample books, and a second label is set for the negative sample books, where the first label can be set to 1 for It indicates that a book is successfully recommended and the user retention data corresponding to the book meets the requirements. The second tag can be set to 0 to indicate that the book recommendation fails and the user retention data corresponding to the book does not meet the requirements.

Then, for any historically recommended book, obtain the book characteristics corresponding to the historical recommended book, that is, for each positive sample book and each negative sample book, obtain the book characteristics corresponding to the book, according to the book characteristics of the positive sample book, The book features, first label, and second label of the negative sample books are used for model training to obtain a book recommendation model. Among them, deep network models, FMM models, DFM models, etc. can be used, which will not be listed here. During the training process, the recommendation results of historical recommended books will be obtained. The loss between the recommendation results and the corresponding set label (the first label or the second label) is calculated to obtain the model loss function. The model parameters are updated according to the model loss function. When the output value of the model loss function is less than the preset threshold, the model training ends.

The book recommendation model obtained by this training is a model whose input is book features and whose output is recommendation score.

In this embodiment, a model training can be performed based on all historical recommended books posted outside the site. The trained book recommendation model can determine the corresponding recommendation score for any candidate book. At this time, during training, The book type will be used as one of the book features and input to the model for training; or, for each book type, a book recommendation model will be trained. The book type is the type of the book itself, such as history, literature, social science, etc. Subject category, etc. In this case, the historically recommended books corresponding to each book type that are posted outside the site are divided into positive sample books and negative sample books, and are trained separately. The number of models finally trained is the same as the number of book types. Correspondingly, this can improve the effect of external placement of books in different scenarios.

Step S205: Sort multiple candidate books according to the recommendation scores, and select a preset number of books to be recommended from the sorted multiple candidate books for book recommendation.

After calculating the recommendation score of each candidate book, a preset number of books to be recommended can be selected from multiple candidate books according to the recommendation score for book recommendation. Specifically, the multiple candidate books can be sorted according to the recommendation score, for example, Sort multiple candidate books in order of recommendation scores from high to low, and select a preset number of books to be recommended from the sorted multiple candidate books for book recommendation.

In the embodiment of the present disclosure, the selected books to be recommended can be used for posting outside the site. By posting the selected books to be recommended outside the site, traffic can be achieved and the user retention rate can be improved.

The solution provided by this disclosure uses a book recommendation model trained based on user retention data and book features corresponding to historical recommended books posted outside the site to determine the recommendation score of each candidate book, and screen the books to be recommended based on the recommendation score. This achieves automatic screening of recommended books, while improving the success rate of book recommendation, attracting more users to use reading applications to read books, improving user retention rates, and reducing manual participation and saving costs; based on the recall of similar books Selecting candidate books effectively improves the success rate of recommendation.

Embodiments of the present disclosure also provide a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores at least one executable instruction. The computer-executable instruction can execute any of the above method embodiments. How to recommend books.

FIG. 3 shows a schematic structural diagram of a computing device according to an embodiment of the present disclosure. The specific embodiments of the present disclosure do not limit the specific implementation of the computing device.

As shown in Figure 3, the computing device may include: a processor 302, a communication interface (Communications Interface) 304, memory (memory) 306, and communication bus 308.

Among them: the processor 302, the communication interface 304, and the memory 306 complete communication with each other through the communication bus 308.

The communication interface 304 is used to communicate with network elements of other devices such as clients or other servers.

The processor 302 is configured to execute the program 310. Specifically, it can execute the relevant steps in the above book recommendation method embodiment.

Specifically, program 310 may include program code including computer operating instructions.

The processor 302 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present disclosure. The one or more processors included in the computing device may be the same type of processor, such as one or more CPUs; or they may be different types of processors, such as one or more CPUs and one or more ASICs.

Memory 306 is used to store program 310. The memory 306 may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.

The program 310 may be specifically used to cause the processor 302 to execute the book recommendation method in any of the above method embodiments.

The algorithms or displays provided herein are not inherently associated with any particular computer, virtual system, or other device. Various general-purpose systems can also be used with teaching based on this. From the above description, the structure required to construct such a system is obvious. Furthermore, embodiments of the present disclosure are not directed to any particular programming language. It should be understood that the disclosure described herein may be implemented using a variety of programming languages, and that the above descriptions of specific languages are for the purpose of disclosing the best mode for implementing the disclosure.

In the instructions provided here, a number of specific details are described. However, it is understood that embodiments of the present disclosure may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this description.

Similarly, it should be understood that in the above description of exemplary embodiments of the disclosure, various features of the embodiments of the disclosure are Features are sometimes grouped together into a single embodiment, figure, or description thereof. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed disclosure requires more features than are expressly recited in each claim. Rather, as the following claims reflect, disclosed aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this disclosure.

Those skilled in the art will understand that modules in the devices in the embodiment can be adaptively changed and arranged in one or more devices different from that in the embodiment. The modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method so disclosed may be employed in any combination, except that at least some of such features and/or processes or units are mutually exclusive. All processes or units of the equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will understand that, although some embodiments herein include certain features included in other embodiments but not others, combinations of features of different embodiments are meant to be within the scope of the present disclosure. and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

Various component embodiments of the present disclosure may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will understand that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all functions of some or all components according to embodiments of the present disclosure. The present disclosure may also be implemented as an apparatus or apparatus program (eg, computer program and computer program product) for performing part or all of the methods described herein. Such a program implementing the present disclosure may be stored on a computer-readable medium, or may be in the form of one or more signals. Such signals may be downloaded from an Internet website, or provided on a carrier signal, or in any other form.

It should be noted that the above embodiments illustrate rather than limit the present disclosure. Furthermore, those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present disclosure may be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the element claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, third, etc. does not indicate any order. These words can be interpreted as names. Unless otherwise specified, the steps in the above embodiments should not be understood as limiting the order of execution.

Claims

A book recommendation method including:

Obtain multiple candidate books and book characteristics corresponding to each candidate book;

For any candidate book, the book features of the candidate book are input into the pre-trained book recommendation model for recommendation prediction, and the recommendation score corresponding to the candidate book is obtained, wherein the book recommendation model is based on the off-site placement Obtained by training on user retention data and book features corresponding to historical recommended books;

A preset number of books to be recommended are selected from the plurality of candidate books according to the recommendation scores to perform book recommendation.
The method according to claim 1, wherein the training method of the book recommendation model includes:

According to user retention data, historically recommended books placed outside the site are divided into positive sample books and negative sample books, setting a first label for the positive sample books, and setting a second label for the negative sample books;

For any historical recommended book, obtain the book characteristics corresponding to the historical recommended book;

Model training is performed based on the book characteristics of the positive sample books, the book characteristics of the negative sample books, the first label, and the second label, and a book recommendation model is obtained.
The method according to claim 2, wherein dividing historical recommended books into positive sample books and negative sample books based on user retention data further includes:

For any historically recommended book, count n-day user retention data corresponding to the historically recommended book, where n is greater than or equal to 2;

If the n-day user retention data is greater than or equal to the preset retention threshold, then the historically recommended books are classified as positive sample books;

If the n-day user retention data is less than the preset retention threshold, the historically recommended books will be classified as negative sample books.
The method according to any one of claims 1-3, wherein the user retention data includes: user retention rate.
The method according to any one of claims 1 to 4, wherein the selecting a preset number of books to be recommended from the plurality of candidate books based on the recommendation scores for book recommendation further includes:

Multiple candidate books are sorted according to the recommendation scores, and a preset number of books to be recommended are selected from the sorted multiple candidate books for book recommendation.
The method according to any one of claims 1-5, wherein said obtaining a plurality of candidate books further includes:

For any book in the book library, calculate the similarity between the book and the historically recommended book with the first tag;

Screen multiple candidate books from the book library based on the similarity.
The method according to claim 6, wherein screening a plurality of candidate books from a book library based on the similarity further includes:

If the similarity is greater than or equal to the preset similarity threshold, the book is determined to be a candidate book.
The method according to any one of claims 1 to 7, wherein the book characteristics include one or more of the following characteristics: book word count, completion rate, author, follow-up rate, book introduction, book title , book type.
A computing device, including: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;

The memory is used to store at least one executable instruction, and the executable instruction causes the processor to perform the following operations:

Obtain multiple candidate books and book characteristics corresponding to each candidate book;

For any candidate book, the book features of the candidate book are input into the pre-trained book recommendation model for recommendation prediction, and the recommendation score corresponding to the candidate book is obtained, wherein the book recommendation model is based on the off-site placement Obtained by training on user retention data and book features corresponding to historical recommended books;

A preset number of books to be recommended are selected from the plurality of candidate books according to the recommendation scores to perform book recommendation.
The computing device of claim 9, wherein the executable instructions further cause the processor to:

According to user retention data, historically recommended books placed outside the site are divided into positive sample books and negative sample books, setting a first label for the positive sample books, and setting a second label for the negative sample books;

For any historical recommended book, obtain the book characteristics corresponding to the historical recommended book;

Model training is performed based on the book characteristics of the positive sample books, the book characteristics of the negative sample books, the first label, and the second label, and a book recommendation model is obtained.
The computing device of claim 10, wherein the executable instructions further cause the processor to:

For any historically recommended book, count n-day user retention data corresponding to the historically recommended book, where n is greater than or equal to 2;

If the n-day user retention data is greater than or equal to the preset retention threshold, then the historically recommended books are classified as positive sample books;

If the n-day user retention data is less than the preset retention threshold, the historically recommended books will be classified as negative sample books.
The computing device of any one of claims 9-11, wherein the user retention data includes: user retention rate.
The computing device of any one of claims 9-12, wherein the executable instructions further cause the processor to:

Multiple candidate books are sorted according to the recommendation scores, and a preset number of books to be recommended are selected from the sorted multiple candidate books for book recommendation.
The computing device of any one of claims 9-13, wherein the executable instructions further cause the processor to:

For any book in the book library, calculate the similarity between the book and the historically recommended book with the first tag;

Screen multiple candidate books from the book library based on the similarity.
The computing device of claim 14, wherein the executable instructions further cause the processor to:

If the similarity is greater than or equal to the preset similarity threshold, the book is determined to be a candidate book.
The computing device according to any one of claims 9-15, wherein the book characteristics include one or more of the following characteristics: book word count, completion rate, author, follow-up rate, book introduction, book Name, book type.
A non-volatile computer-readable storage medium, the non-volatile computer-readable storage medium stores at least one executable instruction, the executable instruction causes the processor to execute any one of claims 1-8 The operation corresponding to the book recommendation method described in the item.
A computer program product including a computing program stored on a non-volatile computer-readable storage medium, the computer program including program instructions that, when executed by a processor, cause the The processor performs operations corresponding to the book recommendation method according to any one of claims 1-8.