WO2022199494A1 - User interest-based content recommendation method, and terminal device - Google Patents

Abstract

A user interest-based content recommendation method, and a terminal device, which are used for improving the accuracy and efficiency of user interest-based content recommendation on the premise that user privacy data is protected. The terminal device collects behavior data of multiple user operations input by a user who uses a target application one or more times in a set duration; the collected behavior data of multiple user operations is desensitized; the desensitized behavior data of multiple user operations is sent to a server, such that the server analyses the desensitized behavior data of multiple user operations to obtain a topic interest table of the target application used by the user. The terminal device receives a topic interest table, and displays a first recommendation interface when the user starts the target application, wherein the first recommendation interface comprises at least one recommended content item which is determined according to the topic interest table.

Description

A content recommendation method and terminal device based on user interests

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the priority of the Chinese patent application filed on March 23, 2021 with the application number 202110307500.X and the application name "A Content Recommendation Method and Terminal Device Based on User Interests", the entire contents of which are Incorporated herein by reference.

technical field

The present application relates to the technical field of data processing, and in particular, to a content recommendation method and terminal device based on user interests.

Background technique

With the development of Internet technology, in order to improve the user experience, more and more applications or systems pay more attention to the content recommendation technology of thousands of people. Content recommendation systems based on user interests often need to be built on massive user behavior data, which brings with it the protection of users' private data. Among them, the user's private data such as the user's personal information, the user's personalized interests, and the like.

In the prior art, there is a technical solution for protecting user's private data by using a secure encryption technology (eg, homomorphic encryption technology). Although the security encryption technology is adopted for the user's private data when the terminal device and the server side transmit and store the user's operation behavior data, there is still a security risk of privacy leakage during the transmission of the user's operation behavior data. In the prior art, there is also the use of distributed federated learning technology to train recommendation models. This technology can protect user privacy by keeping user operation behavior data on the terminal device side, but requires the terminal device side to have the ability to train the model. , so the performance requirements on the terminal device side are relatively high, and there is a disadvantage of low recommendation efficiency.

Therefore, how to improve the accuracy and efficiency of content recommendation based on user interests on the basis of protecting user privacy data is still a great challenge.

SUMMARY OF THE INVENTION

Embodiments of the present application provide a method and terminal device for content recommendation based on user interests, so as to improve the accuracy and efficiency of content recommendation based on user interests on the basis of protecting user privacy data.

In a first aspect, an embodiment of the present application provides a method for recommending content based on user interests, and the method can be applied to a terminal device. The method includes: collecting a plurality of user operation behavior data input by a user when the user uses a target application program one or more times in a set duration; performing desensitization processing on the collected user operation behavior data, and the desensitization The processing is to filter out the private data related to the user in the user operation behavior data; send the desensitized multiple user operation behavior data to the server, so that the server can desensitize the multiple user operation behaviors after the desensitization process. The data is analyzed to obtain the subject interest table of the user using the target application.

With the method provided by the embodiment of this application, the terminal device will collect the user operation behavior data in the target application program, but the user privacy data will be desensitized on the terminal device before being uploaded to the server, so the server will not obtain the data. To user privacy data, it can better protect user privacy and improve user experience. Wherein, the target application may be any application suitable for recommending content based on user interests provided in the embodiment of the present application, such as a browser and the like.

In a possible design, the terminal device collects multiple user operation behavior data input by the user when the user uses the target application one or more times in the set duration, which may be implemented as: detecting that the user starts the target application When the program is instructed, start the target application; after the target application is started, collect at least one operation data performed by the user on the target application; detect that the user exits the target application When instructed, close the target application; store at least one operation data collected during the process of starting to closing the target application as a set of user operation behavior data.

In this design, the user operation behavior data when the user uses the target application is collected through the terminal device, and the user operation behavior data is desensitized and then uploaded to the server. Therefore, it is convenient for the server to perform big data analysis based on the desensitized user operation behavior data, so as to obtain the user interest of the user in the target application program.

In a possible design, the terminal device randomly replaces the user operation behavior data under the same interest topic based on the differential privacy algorithm for one or more user operation behavior data in the plurality of user operation behavior data, the The topic of interest is determined according to the topic interest table; the user information contained in the plurality of user operation behavior data is stripped.

In this design, by randomly replacing the user operation behavior data based on the differential privacy algorithm, the terminal device can achieve the effect of masking the real user operation behavior data, thereby achieving the purpose of protecting user privacy. In addition, before uploading the user operation behavior data to the server, the terminal device can also strip the user information so that the server cannot collect the user's private data, thereby ensuring the privacy and security of the user data.

In a possible design, before the random replacement of user operation behavior data under the same topic of interest based on the differential privacy algorithm, the terminal device may also determine the sequence length of each user operation behavior data; according to the preset value The user operation behavior data is truncated and compensated to obtain user operation behavior data with a specified sequence length.

In this design, if the user operation behavior data contains too little operation data, that is, the sequence length is short, more accurate user interests cannot be analyzed from the user operation behavior data. However, the user operation behavior data contains too much operation data, that is, the sequence length is long, which will lead to the problem of excessive calculation. Therefore, by sampling the user operation behavior data according to the preset value to obtain the user operation behavior data with a relatively uniform sequence length, the processing efficiency of desensitizing the user operation behavior data can be improved, and the server's ability to detect the user operation behavior can be improved. Efficiency and accuracy of statistical analysis of data.

Optionally, the terminal device performs truncation and compensation processing on the user operation behavior data according to a preset value, and obtains user operation behavior data with a specified sequence length. The preset value is to supplement the user operation behavior data with pre-defined user operation behavior data of a target length to obtain user operation behavior data of a specified sequence length; or, if the sequence length of the user operation behavior data is greater than the A preset value, truncating the target length of the user operation behavior data to obtain the user operation behavior data of the specified sequence length; wherein, the target length is the difference between the sequence length of the user operation behavior data and a preset value the absolute value of .

In this design, a specific scenario of sampling according to the preset value is given. By judging the sequence length of the user operation behavior data, if the sequence length is less than the preset value of the user operation behavior data, the predefined default user can be used. The operation behavior data is supplemented, and the user operation behavior data whose sequence length is greater than the preset value can be randomly truncated. Therefore, after sampling the user operation behavior data according to the preset value, the user operation behavior data with a relatively uniform sequence length can be obtained to facilitate desensitization processing.

In a second aspect, an embodiment of the present application provides a method for recommending content based on user interests, and the method can be applied to a server. The method includes: receiving a plurality of desensitized user operation behavior data sent by one or more terminal devices; the desensitized user operation behavior data is collected by the one or more terminal devices for a set duration A plurality of user operation behavior data entered by the user when the user uses the target application program one or more times, and obtained by desensitizing the collected user operation behavior data; the desensitization process is to desensitize the Filter out the privacy data related to the user in the user operation behavior data; analyze the multiple user operation behavior data after the desensitization processing to obtain the subject interest table of the user using the target application; send the subject interest table to the one or more terminal devices.

In this method, because the server has better computing power than the terminal device, and the server can integrate the user operation behavior data sent by a plurality of terminal devices to analyze the group user operation behavior data, it is possible to obtain a more timely result. It can also be understood as a topic of interest that attracts the attention of most users at the moment, and a topic interest table is generated. And, the server can send the topic interest table to the terminal device, so that the terminal device can perform real-time recommendation in combination with the topic interest table, so as to improve user experience.

In a possible design, the server analyzes a plurality of user operation behavior data after the desensitization processing, and obtains the subject interest table of the user using the target application. The obtained multiple user operation behavior data is input into a pre-built topic interest model, so as to perform unsupervised learning on the desensitized multiple user operation behavior data; the topic interest table output by the pre-built topic interest model is obtained. .

In this design, through the pre-built topic interest model, a topic interest table can be obtained according to a large amount of user operation behavior data sent by multiple terminal devices, and the obtained topic interest table can better reflect the current interest that is more concerned by most users. subject of interest.

In a third aspect, an embodiment of the present application provides a method for recommending content based on user interests, and the method can be applied to a terminal device. The method includes: receiving a topic interest table sent by a server, where the topic interest table is obtained by analyzing a plurality of user operation behavior data after desensitization processing by the server; the user operation behavior data after desensitization processing is: One or more terminal devices collect multiple user operation behavior data input by the user when the user uses the target application program one or more times in the set duration, and desensitize the collected multiple user operation behavior data The desensitization process is to filter out the private data involving the user in the user operation behavior data; when detecting the user's instruction to start the target application, start the target application and A first recommendation interface is displayed, and the first recommendation interface includes at least one recommended content; the at least one recommended content is determined according to the topic interest table.

In this method, the terminal device can perform real-time recommendation in combination with the topic interest table sent by the server. In this scenario, if the user belongs to a new user, or the terminal device does not store the user's historical user interests, the terminal device can make recommendations according to the topic interest table, and by comparing the current interest topics that most users pay attention to The recommendation of the corresponding content can avoid cold-start recommendation in the target application, that is, the target application recommends content corresponding to some unpopular interest topics and cannot even be recommended, resulting in failure to arouse the user's browsing interest.

In a possible design, the starting the target application and displaying the first recommendation interface may be implemented as: starting the target application; displaying the first recommendation interface after starting the target application; One or more interest topics included in the topic interest table are taken as user interests, and at least one recommended content is obtained according to the user interest, and the obtained at least one recommended content is displayed in the first recommendation interface; wherein, Each of the interest topics has an associated weight value, and the greater the associated weight value of the interest topic, the higher the proportion of the recommended content including the related content of the interest topic.

In this design, by performing content recommendation according to the weight value associated with the interest topic contained in the topic interest table, it can achieve a higher proportion of the relevant recommended content corresponding to the more popular interest topic in the recommendation interface, thereby improving the user experience. Interested in the possibility to enhance the user experience.

In a possible design, after the target application is started and the first recommendation interface is displayed, it is implemented as: receiving and collecting one or more user operation behavior data input by the user when the user uses the target application; When detecting the user's instruction to refresh the first recommendation interface, a second recommendation interface is displayed; the recommended content included in the second recommendation interface is based on the one or more user operation behavior data and the topic interest table determined.

In this design, when the application is implemented, after the target application acquires the user's real-time operation data, the user's real-time operation data can be analyzed to obtain the interest topics that the user pays more attention to, so as to facilitate the timely analysis of the user's interest topics. interests are adjusted, so that a recommendation interface more matching the user's interests can be displayed in time.

In a possible design, the displaying of the second recommendation interface may be implemented as: determining one or more corresponding interest topics according to the one or more user operation behavior data, and assigning an association to each of the interest topics take one or more interest topics corresponding to the user operation behavior data and one or more of the interest topics included in the topic interest table as the user interest, and obtain at least one recommendation according to the user interest content, and display at least one item of recommended content obtained in the second recommendation interface; wherein, each of the interest topics included in the topic interest table has an associated weight value, and the greater the weight value associated with the interest topic , the higher the proportion of content related to the topic of interest included in the recommended content.

In this design, by combining the user's real-time user operation behavior data and the topic interest table generated by the server, on the basis of the current popular interest topics, the user's personal interest can also be taken into account, so that the recommended content that is more in line with the user's interest can be obtained. , in order to improve the user experience.

In a possible design, the acquiring at least one piece of the recommended content according to the user's interest may be implemented as: searching for the recommended content corresponding to the user's interest from locally cached content; The content providing server of the recommended content corresponding to the user's interest acquires the recommended content corresponding to the user's interest.

In this design, after the terminal device determines the user's interest, it can obtain the recommended content related to the user's interest in various possible ways, such as hot articles, hot news, etc., to improve the diversity of the recommended content.

In a fourth aspect, an embodiment of the present application further provides a terminal device, including: one or more processors; one or more memories; the one or more memories for storing one or more computer programs and data information; wherein the one or more computer programs include instructions; when the instructions are executed by the one or more processors, the terminal device is caused to perform the method according to any one of the above first aspects, Or perform the method according to any one of the above third aspects.

In a fifth aspect, an embodiment of the present application further provides a server, including: one or more processors; one or more memories; the one or more memories for storing one or more computer programs and data information ; wherein the one or more computer programs comprise instructions; when executed by the one or more processors, the instructions cause the server to perform the method of any one of the second aspects above.

In a sixth aspect, an embodiment of the present application further provides a communication system, including: a terminal device and a server; the terminal device can perform the steps of the terminal device in the method provided in the first aspect above, or perform the steps in the third aspect above. The steps of the terminal device in the provided method; the server may execute the steps of the server in the method provided in the second aspect above.

In a seventh aspect, an embodiment of the present application provides a computer-readable storage medium, where the computer-readable medium stores a computer program (also referred to as code, or instruction) when it runs on a computer, so that the computer executes the above-mentioned first The method in any possible implementation of the aspect, or the method in any of the possible implementations of the second aspect above, or the method in any possible implementation of the third aspect above.

In an eighth aspect, an embodiment of the present application provides a computer program product. The computer program product includes: a computer program (also referred to as code, or an instruction), which, when the computer program is executed, causes the computer to execute any of the above-mentioned first aspects. The method in one possible implementation manner, or the method in any one possible implementation manner of the foregoing second aspect, or the method in any one possible implementation manner in the foregoing third aspect.

In a ninth aspect, an embodiment of the present application further provides a graphical user interface on a terminal device, where the terminal device has a display screen, one or more memories, and one or more processors, where the one or more processors are used for executing one or more computer programs stored in the one or more memories, the graphical user interface includes a graphical user interface displayed when the terminal device executes any possible implementation manner of the first aspect of the embodiments of the present application, Or a graphical user interface displayed when any possible implementation manner of the third aspect of the embodiments of the present application is executed.

For the beneficial effects of any one of the fourth aspect to the ninth aspect, please refer to the beneficial effects of various possible designs in the first aspect to the third aspect, which will not be repeated here.

Description of drawings

1 is an application scenario diagram of a method for recommending content based on user interests provided by an embodiment of the present application;

2a is a schematic diagram of a hardware architecture of a terminal device provided by an embodiment of the application;

FIG. 2b is a block diagram of a software structure of a terminal device provided by an embodiment of the application;

3 is a schematic structural diagram of a content recommendation method based on user interests provided by an embodiment of the present application;

4 is one of schematic diagrams of user interfaces of a method for recommending content based on user interests provided by an embodiment of the present application;

FIG. 5 is a schematic diagram of a radar chart of a topic of user interest provided by an embodiment of the present application;

FIG. 6 is a schematic flowchart of a method for recommending content based on user interests according to an embodiment of the present application;

FIG. 7 is a second schematic diagram of a user interface of a method for recommending content based on user interests according to an embodiment of the present application;

FIG. 8a is a third schematic diagram of a user interface of a method for recommending content based on user interests provided by an embodiment of the present application;

FIG. 8b is a schematic diagram of a user interface of a content recommendation method based on user interests provided by an embodiment of the present application

Four;

FIG. 8c is a fifth schematic diagram of a user interface of a method for recommending content based on user interests according to an embodiment of the present application;

FIG. 9 is a schematic structural diagram of a terminal device or a server according to an embodiment of the present application.

Detailed ways

With the rapid development of society, terminal devices such as mobile phones are becoming more and more popular. Terminal devices not only have communication functions, but also have powerful processing capabilities, storage capabilities, and camera functions. The terminal device executes the corresponding application program through the operating system (for example, the Android operating system), and the user can use the terminal device to make calls, send short messages, browse web pages, take pictures, play games, watch videos, and so on. Among them, in some application programs (applications, APPs) with data of different interest fields, the terminal device can recommend content according to the user's interests. When the APP browses small videos, or when shopping on the shopping APP, the terminal device can recommend the content that the user may be interested in according to the user's operation behavior such as the user's search words, search history, and current browsing content. Among them, the user interest can be a custom theme interest classification in the application, or a commonly used theme interest classification. For example, the theme interest classification in the browser can include sports, finance, current affairs, etc., and the theme interest classification in the shopping APP. There can be clothing, daily necessities, food, etc.

In combination with the description in the background art, a content recommendation system based on user interests needs to collect a large amount of user operation behavior data to complete content recommendation. However, at present, on the basis of protecting user privacy, the recommendation accuracy of a content recommendation system based on user interests is improved. And there is no good solution for recommending efficiency yet.

In view of this, the present application provides a content recommendation method based on user interests, by collecting user operation behavior data on the terminal device side, and then uploading the user operation behavior data after stripping the user's sensitive private data to the server side. The server side builds a topic interest table according to a large amount of user operation behavior data uploaded by multiple terminal devices, and returns the topic interest table to the terminal device. The terminal device may determine the user's interest in combination with the topic interest table delivered by the server side, the user's real-time operation behavior data on the terminal device, historical user interests, and other factors. Finally, according to the determined user interest, the terminal device may request the server side for the relevant content of the user's interest, or may also obtain the relevant content of the user's interest from the local cache of the terminal device, so as to realize real-time Content recommendation.

The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

FIG. 1 is an application scenario diagram of a content recommendation method based on user interests provided by an embodiment of the present application. The application scenario may include a terminal device 110, a server 120, and a database 130. An application program may be installed in the terminal device 110. The server 120 may be a background server that communicates with the terminal device, or may be a separate server for mining potential objects. server. The application may be a web version application or an application pre-installed in the terminal device 110. The application in this application may be, for example, a browser application, a small video application, a shopping application, etc. Any type of application that makes content recommendations. Both the terminal device 110 and the server 120 can access the database 130 , and store the access logs generated during the user's access in the database 130 . The database 130 may be set on the server 120, or may be set relatively independently from the server 120. For example, the database 130 may be implemented by a server cluster, a cloud server, or a distributed storage server. It should be noted that this application does not limit the number and types of terminal devices 110, servers 120 and databases 130 included in the application scenario, for example, there may be multiple terminal devices 110 shown in FIG. 1 .

For example, when the current user generates user operation behavior data by starting an application program in the terminal device 110 and performing user operations on the application program, the user operation behavior data may be, for example, the user's operations such as searching and browsing in the browser. data, the terminal device 110 can perform content recommendation according to the user operation behavior data and the topic interest table issued by the server 120 . The terminal device 110 can desensitize the user operation behavior data and upload it to the server 120 after closing the application program, and then the server 120 can generate or update the topic interest table according to the massive user operation behavior data, and then return it to the server 120. Terminal device 110. In addition, the server 120 may also store the generated topic interest table or the received user operation behavior data in the database 130 .

It can be understood that the terminal device 110 in this embodiment of the present application may be, for example, a mobile phone, a tablet computer, a wearable device (for example, a watch, a wristband, a helmet, a headset, etc.), a vehicle-mounted device, an augmented reality (AR)/ Virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), smart home devices (e.g., smart TVs, smart speakers, smart cameras, etc.), etc. It can be understood that, the embodiment of the present application does not impose any limitation on the specific type of the terminal device 110 .

The terminal device 110 to which the embodiments of this application can be applied, and exemplary embodiments include but are not limited to carrying

Or portable terminal devices with other operating systems. The above-mentioned portable terminal device may also be other portable terminal devices, such as a laptop computer (Laptop) or the like having a touch-sensitive surface (eg, a touch panel).

Figure 2a shows a schematic diagram of the hardware structure of a possible terminal device. The terminal device 110 includes: a radio frequency (RF) circuit 210, a power supply 220, a processor 230, a memory 240, an input unit 250, a display unit 260, an audio circuit 270, a communication interface 280, and a wireless fidelity ( components such as wireless fidelity, Wi Fi) module 290. Those skilled in the art can understand that the hardware structure of the terminal device shown in FIG. 2a does not constitute a limitation on the terminal device, and the terminal device provided in this embodiment of the present application may include more or less components than those shown in the figure, and may be combined Two or more components, or may have different component configurations. The various components shown in Figure 2a may be implemented in hardware, software, or a combination of hardware and software, including one or more signal processing and/or application specific integrated circuits.

Each component of the terminal device 110 will be specifically introduced below with reference to FIG. 2a:

The RF circuit 210 can be used for data reception and transmission during communication or conversation. In particular, after receiving the downlink data of the base station, the RF circuit 210 sends it to the processor 230 for processing; in addition, it sends the uplink data to be sent to the base station. Typically, the RF circuit 210 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, and the like.

In addition, the RF circuit 210 may also communicate with networks and other devices via wireless communication. The wireless communication can use any communication standard or protocol, including but not limited to global system of mobile communication (GSM), general packet radio service (GPRS), code division multiple access (code division multiple access) division multiple access, CDMA), wideband code division multiple access (WCDMA), long term evolution (long term evolution, LTE), email, short message service (short messaging service, SMS), etc.

The Wi-Fi technology belongs to the short-distance wireless transmission technology, and the terminal device 110 can be connected to an access point (access point, AP) through the Wi-Fi module 290, thereby realizing the access of the data network. The WiFi module 290 can be used for data reception and transmission during the communication process.

The terminal device 110 can be physically connected with other devices through the communication interface 280 . Optionally, the communication interface 280 is connected with the communication interface of the other device through a cable to realize data transmission between the terminal device 110 and the other device.

In this embodiment of the present application, the terminal device 110 can implement communication services and interact with the server side, so the terminal device 110 needs to have a data transmission function, that is, the terminal device 110 needs to include a communication module. Although FIG. 2 a shows communication modules such as the RF circuit 210 , the WiFi module 290 , and the communication interface 280 , it can be understood that the terminal device 110 has at least one or other of the above components. A communication module (such as a Bluetooth module) used to implement communication for data transmission.

For example, when the terminal device 110 is a mobile phone, the terminal device 110 may include the RF circuit 210, and may also include the WiFi module 290; when the terminal device 110 is a computer, the terminal device 110 The communication interface 280 may be included, and the WiFi module 290 may also be included; when the terminal device 110 is a tablet computer, the terminal device 110 may include the WiFi module.

The memory 240 may be used to store software programs and modules. The processor 230 executes various functional applications and data processing of the terminal device 110 by running software programs and modules stored in the memory 240 . Optionally, the memory 240 may mainly include a program storage area and a data storage area. The storage program area can store the operating system (mainly including the corresponding software programs or modules of the kernel layer, the system layer, the application program framework layer, and the application program layer). The application layer may include various applications, and among the applications that can be recommended, content recommendation based on user interests can be implemented by using the method provided by the embodiments of the present application.

In addition, the memory 240 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

The input unit 250 can be used to receive editing operations of various types of data objects such as numbers or character information input by the user, and generate key signal input related to user settings and function control of the terminal device 110 . Optionally, the input unit 250 may include a touch panel 251 and other input devices 252 .

Wherein, the touch panel 251, also called a touch screen, can collect the user's touch operations on or near it (for example, the user uses any suitable objects or accessories such as fingers, stylus, etc. on the touch panel 251 or on the touch panel 251). operation near the touch panel 251 ), and drive the corresponding connection device according to a preset program.

Optionally, the other input devices 252 may include, but are not limited to, one or more of physical keyboards, function keys (such as volume control keys, switch keys, etc.), trackballs, mice, joysticks, and the like.

The display unit 260 may be used to display information input by the user or information provided to the user and various menus of the terminal device 110 . The display unit 260 is the display system of the terminal device 110, and is used for presenting an interface and realizing human-computer interaction. The display unit 260 may include a display panel 261 . Optionally, the display panel 261 may be configured in the form of a liquid crystal display (liquid crystal display, LCD), an organic light-emitting diode (organic light-emitting diode, OLED) or the like. In this embodiment of the present application, for example, the display unit 260 may display a visual page corresponding to the user's operation on the terminal device. For example, after the user enters a search term, the display unit 260 displays the information flow, web page, etc.

The processor 230 is the control center of the terminal device 110, uses various interfaces and lines to connect various components, runs or executes the software programs and/or modules stored in the memory 240, and invokes the software programs and/or modules stored in the The data in the memory 240 executes various functions of the terminal device 110 and processes data, thereby realizing various services based on the terminal device. In the embodiment of the present application, the processor 230 is configured to implement the method provided by the embodiment of the present application, so as to perform more accurate content recommendation for the user.

The terminal device 110 also includes a power source 220 (such as a battery) for powering the various components. Optionally, the power supply 220 may be logically connected to the processor 230 through a power management system, so as to implement functions such as managing charging, discharging, and power consumption through the power management system.

As shown in FIG. 2 a , the terminal device 110 further includes an audio circuit 270 , a microphone 271 and a speaker 272 , which can provide an audio interface between the user and the terminal device 110 . The audio circuit 270 can be used to convert the audio data into a signal that can be recognized by the speaker 272, and transmit the signal to the speaker 272, and the speaker 272 converts it into a sound signal and outputs it. The microphone 271 is used to collect external sound signals (such as voices of people speaking or other sounds, etc.), convert the collected external sound signals into signals that can be recognized by the audio circuit 270 , and send them to the audio circuit 270 . The audio circuit 270 can also be used to convert the signal sent by the microphone 271 into audio data, and then output the audio data to the RF circuit 220 for transmission to, for example, another terminal, or output the audio data to the memory 240 for subsequent further processing.

Although not shown, the terminal device 110 may further include at least one type of sensor, camera, etc., which will not be repeated here.

The operating system (operating system, OS) involved in the embodiments of the present application is the most basic system software running on the terminal device 110 . Taking a smartphone as an example, the operating system may be an Android system or an IOS system. The following embodiments take the android system as an example for introduction. Those skilled in the art can understand that in other operating systems, a similar method can also be used for implementation.

The software system of the terminal device 110 may adopt a layered architecture, an event-driven architecture, a microkernel architecture, a microservice architecture, or a cloud architecture. The embodiments of the present application take an android system using a layered architecture as an example to illustrate the software structure of the terminal device 110 as an example.

FIG. 2b shows a software structural block diagram of an android system provided by an embodiment of the present application. The layered architecture divides the software into several layers, and each layer has a clear role and division of labor. Layers communicate with each other through software interfaces. In some embodiments, the android system is divided into five layers, from top to bottom, the application layer, the application framework (framework) layer, the Android runtime (android runtime) and system library, the hardware abstraction layer, and the kernel layer. .

The application layer is the top layer of the operating system and can include a series of application packages. As shown in Figure 2b, the application layer may include native applications of the operating system and third-party applications, wherein the native applications of the operating system may include user interface (UI), browser, camera, settings, mobile phone Butler, music, text messages, calls, etc., third-party applications can include maps, shopping APPs, small video APPs, etc. The applications mentioned below may be native applications of the operating system installed on the terminal device 110 when it leaves the factory, or may be third-party applications downloaded from the network or acquired from other terminal devices 110 by the user during the use of the terminal device 110 .

In some embodiments of the present application, the application layer may be used to implement the presentation of an editing interface, and the above-mentioned editing interface may be used by a user to perform operations. For example, the user may perform user operations such as inputting a search term on the editing interface correspondingly presented by the browser.

In a possible implementation, the application can be developed using the java language, and it can be done by calling the application programming interface (API) provided by the application framework layer. The bottom layers of the system (such as hardware abstraction layer, kernel layer, etc.) interact to develop their own applications. The application framework layer is mainly a series of services and management systems of the operating system.

The application framework layer provides application programming interfaces and programming frameworks for applications in the application layer. The application framework layer includes some predefined functions. As shown in Figure 2b, the application framework layer may include a window manager, content provider, view system, telephony manager, resource manager, notification manager, etc.

A window manager is used to manage window programs. The window manager can get the size of the display screen, determine whether there is a status bar, lock the screen, take screenshots, etc. Content providers are used to store and retrieve data and make these data accessible to applications. The data may include video, images, audio, calls made and received, browsing history and bookmarks, phone book, etc. The view system includes visual controls, such as text controls that display text, picture controls that display pictures, and so on. View systems can be used to build applications. A display interface can consist of one or more views. The telephony manager is used to provide communication functions of the terminal device 110, such as management of call status display (including connecting, hanging up, etc.). The resource manager provides various resources for the application, such as localization strings, icons, pictures, layout files, video files, etc.

In some embodiments of the present application, the application framework layer is mainly responsible for invoking a service interface that communicates with the hardware abstraction layer, so as to transmit the operation request of the user to the hardware abstraction layer, and the operation request may include that the user opens a certain An operation request corresponding to an APP, or an operation request corresponding to a search term entered by a user in an APP, etc. may be included. The hardware abstraction layer generates the corresponding content recommendation service according to the operation request passed by the application layer.

Exemplarily, the content recommendation service may include a data collection module, a data calibration module, a real-time recommendation module, a privacy protection module, and the like for implementing the method provided by the present application. Wherein, the data collection module is used to collect the user operation behavior of the user on the client terminal on the terminal device, so as to obtain the user operation behavior data. The data calibration module is used for preprocessing the user operation behavior data collected by the data acquisition module to obtain user operation behavior data with a relatively uniform sequence length. The privacy protection module is used for desensitizing the collected user operation behavior data, stripping or replacing the user privacy data involved in the user operation behavior data, etc., so as to obtain user operation behavior data that does not reveal the user's privacy. The desensitized user operation behavior data is transmitted to the server side, and the desensitized user operation behavior data is used to construct a topic interest model and generate a topic interest table. The real-time recommendation module is used to perform real-time content recommendation according to the determined user interests.

Android runtime (android runtime) includes core libraries and virtual machines. The android runtime is responsible for the scheduling and management of the Android system. The core library of the Android system consists of two parts: one is the function functions that the java language needs to call, and the other is the core library of Android.

The application layer and the application framework layer run in virtual machines. Taking java as an example, the virtual machine executes the java files of the application layer and the application framework layer as binary files. The virtual machine is used to perform functions such as object lifecycle management, stack management, thread management, safety and exception management, and garbage collection.

A system library can include multiple functional modules. For example: surface manager (surface manager), media library (media library), three-dimensional graphics processing library (eg: OpenGL ES), two-dimensional (2D) graphics engine (eg: SGL) and so on.

The Surface Manager is used to manage the display subsystem and provides the fusion of 2D and 3D layers for multiple applications. The media library supports playback and recording of a variety of commonly used audio and video formats, as well as still image files. The media library can support a variety of audio and video encoding formats, such as: MPEG4, H.264, MP3, AAC, AMR, JPG, PNG, etc. The 3D graphics processing library is used to implement 3D graphics drawing, image rendering, compositing, and layer processing. 2D graphics engine is a drawing engine for 2D drawing.

The hardware abstraction layer (HAL) is the support of the application framework layer and an important link between the application framework layer and the kernel layer. It can provide services for developers through the application framework layer.

Exemplarily, the function of the content recommendation service in the embodiment of the present application may be implemented by configuring a first process in the hardware abstraction layer, and the first process may be a sub-process independently constructed in the hardware abstraction layer. The first process may include modules such as a content recommendation service configuration interface, a content recommendation service controller, and the like. The content recommendation service configuration interface is a service interface that communicates with the application framework layer. The content recommendation service controller is used to monitor the content recommendation service configuration interface, for example, to control whether the content recommendation service needs to be authenticated, etc., and is also responsible for monitoring whether the data input in the terminal device 110 needs to be cached or updated. When the input data needs to be cached or updated, the application framework layer can be notified to cache or update the corresponding data to ensure that the latest data is displayed on the display interface. The hardware abstraction layer may further include a daemon process, and the daemon process may be used to cache data in the first process, and the daemon process may also be a subprocess constructed separately in the hardware abstraction layer.

The kernel layer can be the Linux kernel (Linux kernel) layer, which is an abstraction layer between hardware and software. The kernel layer has many drivers related to the terminal device 110, including at least display drivers; Linux-based frame buffer drivers; keyboard drivers and mouse drivers as input devices; Flash drivers based on memory technology devices; audio drivers; Bluetooth drivers, etc., This embodiment of the present application does not impose any limitation on this. The Linux kernel layer is used to provide the core system services of the operating system, such as security, memory management, process management, network protocol stack and driver model, all based on the Linux kernel.

Usually, the terminal device 110 can run multiple applications at the same time. Simple, one application can correspond to one process, and more complex, one application can correspond to multiple processes. Each process has a process number (process ID).

In combination with the introduction of the hardware structure of the terminal device 110 in FIG. 2a and the introduction of the software framework of the terminal device 110 in FIG. 2b, the following is an example to illustrate that the terminal device 110 performs the implementation of the present application for the scenario of content recommendation based on user interests. The working principle of the software and hardware of the content recommendation method based on user interest proposed in the example.

It should be understood that, in the embodiments of the present application, "at least one" refers to one or more, and "a plurality" refers to two or more. "And/or", which describes the association relationship of the associated objects, means that there can be three kinds of relationships, for example, A and/or B, it can mean that A exists alone, A and B exist at the same time, and B exists alone, where A, B can be singular or plural. The character "/" generally indicates that the associated objects are an "or" relationship. "At least one (item) of the following" or its similar expression refers to any combination of these items, including any combination of single item (item) or plural item (item). For example, at least one (a) of a, b or c may represent: a, b, c, a and b, a and c, b and c, or a, b and c, where a, b, c Can be single or multiple.

The multiple involved in the embodiments of the present application refers to greater than or equal to two.

In addition, it should be understood that in the description of this application, words such as "first" and "second" are only used for the purpose of distinguishing the description, and should not be understood as indicating or implying relative importance, nor should it be understood as indicating or implied order.

In addition, in the embodiments of the present application, "terminal device", "device", "mobile phone", etc. may be used interchangeably, that is, various devices that can be used to implement the embodiments of the present application; "Application" can also be mixed, both refer to programs or clients that have certain service provision capabilities, that is to say, applications and clients can also be mixed, such as browser clients and game clients can also be called browser applications or game applications, etc.

It should be understood that the hardware structure of the terminal device may be as shown in FIG. 2a, and the software architecture may be as shown in FIG. 2b, wherein, the software programs and/or modules corresponding to the software architecture in the terminal device may be stored in the memory 240, and the processor 230 The software programs and applications stored in the memory 240 may be executed to execute the flow of the method for recommending content based on user interests provided by the embodiments of the present application.

In order to facilitate the understanding of the content recommendation method based on user interests provided by the present application, the steps involved in adopting the method provided by the present application will be introduced below with reference to the schematic structural diagram shown in FIG. 3 . It should be noted that, the division of modules in the embodiments of the present application is schematic, and is only a logical function division, and there may be other division manners in actual implementation. In addition, each functional module in each embodiment of the present application may be integrated in one processor, or may exist independently physically, or two or more modules may be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software programs.

As shown in FIG. 3 , when the application is implemented, according to the division of logical functions, the terminal device side may include a data collection module 301, a data calibration module 302, a privacy protection module 303 and a real-time recommendation module 306; the server side may include a data statistics module 304 and a real-time recommendation module 306. Data analysis module 305 .

Among them, the data collection module 301 on the terminal device side is used to collect user operation behavior data, and the collected user operation behavior data can not only continue to be preprocessed by the data calibration module 302, but also can be sent to the real-time recommendation module 306 for user interest recommended content. The privacy protection module 303 on the terminal device side is configured to further perform desensitization processing on the user operation behavior data preprocessed by the data calibration module 302, and then send it to the data statistics module 304 on the server side.

The data statistics module 304 on the server side is configured to send data to one or more terminal devices (only one terminal device is shown as an example in FIG. 3 , if there are multiple terminal devices, the processing procedures of other terminal devices are similar, and will not be repeated here) The desensitized user operation behavior data is statistically summarized, and then sent to the data analysis module 305, and the data analysis module 305 performs training according to a large number of aggregated user operation behavior data to generate a topic interest table. The data analysis module 305 on the server side can return the generated topic interest table to the terminal device. On the one hand, the data analysis module 305 can not only send the generated topic interest table to the data calibration module 302, so that the data calibration module 302 can use it as a reference when preprocessing the user operation behavior data, on the other hand, the data analysis module 305 can also The generated topic interest table is sent to the real-time recommendation module 306, so that the real-time recommendation module 306 can perform real-time content recommendation based on user interests in combination with the topic interest table. The data statistics module 304 on the server side can also be integrated with the data analysis module 305 into one module.

Based on the above-mentioned introduction to the schematic structural diagram of FIG. 3 , the method provided by the present application can be mainly divided into several stages, which are introduced in detail below:

In stage 1, the data collection module 301 on the terminal device collects user operation behavior data.

In order to ensure that the content recommended on the terminal device can more accurately track the user's preferences and habits, the implementation of the content recommendation method based on the user's interests often needs to be based on massive user operation behavior data. A large amount of user operation behavior data is generally generated by the user in the process of using the terminal device. Usually, for any application included in the terminal device, the user operation from the time when the user starts the application to when the user closes the application can be regarded as a complete set of user operation behaviors, such as through a session ( Refers to the time interval between a terminal user and a server that provides application services, usually refers to the time elapsed between the time the user registers and enters the server that provides the service to the time that he logs out of the server) object to record the operation behavior of this group of users Contains operational data. Wherein, a session object may contain one or more operation data, and the number and type of operation data are not limited in this application.

As an example, assuming that user A performs some search operations through a browser application, the time from user A opening the browser to exiting the browser may be recorded as a session. Specifically, in combination with the content shown in FIG. 4 , user A opens the browser, which may be recorded as the beginning of a session, such as interface 1 in FIG. 4 . When user A is using the browser, the terminal device can record all operations of user A in this session through the data collection module 301, as shown in interface 2 in FIG. 4, which shows the home page interface of the browser. In the home page interface, you can search for keywords through the search box, or browse the information flow (the information flow is such as the entries displayed on the home page interface of the browser, and user A can also browse more information flows by sliding up and down, Figure 4 This operation is not shown in the page), or browse the webpage by clicking the information flow on the homepage interface of the browser (for example, user A clicks the information flow of "Basketball Association - Homepage" to enter the details interface to browse more about basketball Consult the web page, this operation is not shown in Figure 4). Finally, if the terminal device detects the user operation behavior of user A exiting the browser, it is recorded as the end of this session; for example, the terminal device detects that the current display interface is changed to the main interface of the mobile phone (interface 3 in Figure 4), Or the terminal device detects that the current display interface is switched to the display interface of another application program, etc., which indicates that the current display interface of the mobile phone no longer stays on the browser.

There are many possible application scenarios based on the user's operation behavior, and the user operation behavior data collected by the data collection module 301 is embodied in the characteristics of multi-domain behavior. Among them, the multi-domain behavior refers to the operation performed by the user on different display pages with different protocols, and/or domain names, and/or ports. For example, if any two display pages included in the user operation behavior data use the same protocol, domain name and port, etc., the two display pages belong to the same domain; on the contrary, if any two display pages included in the user operation behavior data use different protocols, the two display pages belong to different domains, that is, User operation behavior data is represented as multi-domain behavior. For example, when a user conducts searches, information flow browsing, web browsing and other behaviors in the browser, the webpages corresponding to different information flows generally belong to different domains, so the user operation behavior data of the user in the browser generally has the characteristics of multi-domain behavior. , that is, the characteristics of cross-domain behavior.

Through the collection of user operation behavior data with multi-domain behavior characteristics, after statistical analysis of massive user operation behavior data, the interest topic tables in multiple domains can be synchronized, and the interests of user operation behaviors in multiple domains can be synthesized. user interests after the features, so that more accurate content recommendations can be made.

In stage 2, the data calibration module 302 on the terminal device preprocesses the user operation behavior data collected by the data collection module 301 .

Exemplarily, during each session, due to the different operations performed by the user each time, the sequence length of the user operation behavior data during each session may be inconsistent, wherein the sequence length of the user operation behavior data is determined according to the number of operations performed by the user. For example, if the user performs a few user operation behaviors in the browser after opening the browser, the sequence length of the user operation behavior data collected during this session is shorter; if the user opens the browser and searches in the browser , information flow browsing, web browsing and other user operation behaviors, the sequence length of the user operation behavior data collected during this session is longer. Table 1 is an example of user operation behavior data collected during a session, as follows:

Table 1

In the embodiment of the present application, a row of data in the user operation behavior data in Table 1 above represents a set of sequences, and Table 1 contains 4 rows of actual user operation behavior data. Therefore, the sequence length in Table 1 can be considered to be 4, and the subsequent Similar tables in the embodiments have the same definitions, and in the specific introduction process, repeated points will not be repeated. The information types of the user operation behavior data shown in Table 1 may include: user ID, key identifiers used to reflect user operation behaviors, user operation behavior types, interest topics corresponding to the key identifiers, and additional summary information (optional) ) and other information. Besides, the user operation behavior data may also include more or less different information types than those in Table 1 above, for example, application program identifiers, etc., which are not limited in this application. In addition, the storage form of the user operation behavior data may be in a tabular form or in other forms, which is also not limited in this application.

In the above example, the key identifiers in Table 1 can be obtained according to user operation behavior, and the key identifiers can be search words, information flow keywords, web page addresses, and the like. For example, if the user operation behavior corresponds to a search operation and the search term is "NBA", the key identifier may be "NBA". Alternatively, if the user's operation behavior corresponds to information flow browsing, and the information flow keyword is "basketball" (the second and third information flows as shown in 2 in Figure 4), the key identifier may be "basketball". Or, if the user's operation behavior corresponds to web browsing, and the keyword contained in the web page is "real economy", the key identifier may be the web page keyword "real economy", or the URL of the currently browsed web page, etc.

In a possible preprocessing scenario, the interest topics in Table 1 may not be obtained from the user's operation behavior (for example, when the user performs a search word operation, the terminal device can obtain the key identifier from the user's operation behavior, but cannot be determined. interest topic), the terminal device can determine it according to the key identifier in the user's operation behavior and the topic interest table obtained from the server side. Among them, the topic interest table is used to indicate the mapping relationship between key identifiers and interest topics, and the server side performs statistics based on the user operation behavior data received from one or more terminal devices and processed by the data calibration module 302 and the privacy protection module 303. After analysis, the specific generation method will be described in detail in subsequent embodiments, and will not be described in detail here. Further, the topic interest table on the terminal device side may be acquired from the server side periodically and stored on the terminal device. The acquisition method may be actively requested by the terminal device, or periodically issued by the server, or automatically issued by the server after detecting that the topic interest table has been updated. In this application, the terminal device acquires the topic interest from the server side. The implementation of the table is not limited.

Specifically, after the terminal device obtains the key identifier from the user operation behavior, it can further determine the interest topic corresponding to at least one key identifier included in the user operation behavior according to the topic interest table. For example, if the key identifier obtained by the terminal device is "NBA", and the interest topic table contains the mapping relationship between "NBA" and the interest topic "Sports", the terminal device can determine the key identifier "Sports" by querying the interest topic table. The interest topic corresponding to NBA" is "sports", and the interest topic and key identifiers are jointly stored as user operation behavior data, such as the data content shown in the second row in Table 1 above. In another possible preprocessing scenario, in order to ensure the accuracy and processing efficiency of modeling based on a large number of user operation behavior data collected during the session, when the application is implemented, the user operation behavior data collected during the session is pre- Set the value for sampling, and then realize that the user operation behavior data collected during each session has a relatively fixed sequence length. Specifically, the user operation behavior data with a shorter sequence length is supplemented by the default user operation behavior, and the user operation behavior data with a longer sequence length is randomly truncated and sampled. In this way, user operation behavior data with a relatively uniform behavior sequence length can be obtained, which can avoid the short sequence length of the user operation behavior data, or it can be understood that the sample data is too small to accurately analyze the user's interests, and can avoid the user's interest due to the short sequence length of the user operation behavior data. The sequence length of the operation behavior data is long, which can also be understood as the problem that the sample data is too much, and the analysis is redundant, which leads to the problem of low processing efficiency.

As an example, the following describes the processing performed by the data calibration module 302 in conjunction with Table 2 (including Table 2a, Table 2b) and Table 3 (including Table 3a, Table 3b), as follows:

Table 2a

key identification	topic of interest
NBA	physical education
real economy	Finance

Wherein, Table 2a is an example of user operation behavior data collected during any session, and it can be obtained that the sequence length of the user operation behavior data is short.

Table 2b

key identification	topic of interest
NBA	physical education
real economy	Finance
"0"	"0"

According to the above Table 2a and Table 2b, it can be obtained, assuming that the data calibration module 302 performs sampling according to the preset value of the sequence length of the user operation behavior data is 3, when the sequence length of the user operation behavior data collected by the data acquisition module 301 is short ( It can also be understood that when it is less than the preset value, such as the sequence length in Table 2a is 2 less than the preset value 3), the pre-defined default user operation behavior data of the target length can be supplemented, thereby obtaining a user operation with a sequence length of 3. behavioral data. The target length is the absolute value of the difference between the sequence length of the user operation behavior data and the preset value. For example, if the sequence length in Table 2a is 2 and the preset value is 3, the target length is |2-3| =1, then add one sequence of user operation behavior data in Table 2b. It should be noted that, in the above Table 2b, "0" is used to represent the default key identifier, and "0" is used to represent the default interest topic corresponding to the mapping of the default key identifier. Among them, the default key identifier and the default interest topic can be set in advance, or determined according to certain rules (for example, according to the user operation behavior of the current hot spot), etc., for example, the default key identifier can be "new crown", and the default interest topic is "current affairs" .

Table 3a

key identification	topic of interest
NBA	physical education
real economy	Finance
football	physical education
https://china.nba.com/	physical education

new crown

current affairs

Wherein, Table 3a is an example of user operation behavior data collected during any session, and it can be obtained that the sequence length of the user operation behavior data is relatively long.

Table 3b

key identification	topic of interest
NBA	physical education
football	physical education
new crown	current affairs

According to the above Table 3a and Table 3b, it can be obtained, assuming that the data calibration module 302 performs sampling according to the preset value of the sequence length of the user operation behavior data is 3, when the sequence length of the user operation behavior data collected by the data acquisition module 301 is long ( It can also be understood that when it is greater than the preset value, for example, the sequence length in Table 3a is 5 greater than the preset value 3), the target length can be randomly truncated to obtain user operation behavior data with sequence length 3. The target length is the absolute value of the difference between the sequence length of the user operation behavior data and the preset value. For example, if the sequence length in Table 3a is 5 and the preset value is 3, the target length is |5-3| = 2, the user operation behavior data of the two sequences are truncated in Table 3b. Or, optionally, when the sequence length of the user operation behavior data is greater than the preset value of the sampling, in addition to the implementation method of randomly truncating the target length, the implementation method of weighted sampling can also be performed according to the behavior type of the user operation behavior, and then according to the implementation method of randomly truncating the target length. The preset value selects several groups of sequences included in the user operation behavior data with larger weights. For example, after the terminal device obtains the weights of the 5 groups of sequences contained in Table 3a according to the type of user operation behavior, selects 3 groups of larger weights. Sequences, such as the 3 sets of sequences shown in Table 3b. Alternatively, other implementation manners of sampling to obtain user operation behavior data whose sequence length is a preset value may also be adopted during implementation, which is not limited in this application.

During the implementation of this application, through the processing of stage 2, the user operation behavior data collected in stage 1 can be preprocessed into a data structure with a uniform sequence length, so that various user operation behavior data generated in various scenarios can be processed. After the preprocessing, it is convenient to continue the desensitization processing through the privacy protection module 303, and it is convenient to realize the modeling of the content recommendation system based on the user's interest. The modeling for realizing the content recommendation system based on the user's interests will be introduced in detail in the following embodiments, and will not be repeated here.

In stage 3, the privacy protection module 303 on the terminal device performs desensitization processing on the user operation behavior data preprocessed by the data calibration module 302 to obtain desensitized user operation behavior data.

When this application is implemented, the user operation behavior data may be desensitized by one or a combination of the following methods, or the user operation behavior data may also be desensitized in other possible ways, which are not limited in this application. Exemplary, including:

Mode 1: The terminal device uses a differential privacy algorithm to perform random replacement processing on the user operation behavior data under the same interest theme. Specifically, for each user operation behavior included in the corresponding user operation behavior data during each session, there is a certain probability that it remains unchanged, and there is a certain probability that it is randomly replaced. Optionally, the terminal device is implemented to keep the topic interest unchanged, by randomly searching the topic interest table, selecting a key identifier under the topic interest, and replacing the original key identifier. Or, regenerate additional summary information and the like in the user operation behavior data. Following the example of user operation behavior data in Table 1, it is assumed that the privacy processing performed by the terminal device is to keep the subject interest "sports" in the second row in Table 1 unchanged, and the key identifier in the second row is replaced by "NBA" with "Basketball", and the additional summary information in row 5 of Table 1 is changed from "Vaccine, Pneumonia, Asymptomatic" to "Wuhan City", the user operation behavior data after privacy processing can be shown in Table 4 below:

Table 4

User ID	key identification	User action type	topic of interest	Additional summary information (not required)
User A	basketball	search	physical education
User A	https://china.nba.com/	Web browsing	physical education	NBA
User A	real economy	Information flow browsing	Finance
User A	new crown	Web browsing	current affairs	Wuhan

Manner 2: The terminal device strips the user privacy data in the user operation behavior data. Optionally, if the user operation behavior data collected by the terminal device includes user privacy data such as user ID, user operation behavior type, etc., these privacy data may be stripped. Following the example of user operation behavior data in Table 1, Table 5 is an example of user operation behavior data after stripping user-related information, as follows:

table 5

key identification	topic of interest	Additional summary information (not required)
NBA	physical education
https://china.nba.com/	physical education	NBA
real economy	Finance
new crown	current affairs	Vaccine, pneumonia, asymptomatic

In the above example, if the terminal device performs desensitization processing on the user operation behavior data through Mode 1 and Mode 2, the processed user operation behavior data can be processed in two ways. Exemplarily, Table 6 is an example of the desensitized user operation behavior data after the user operation behavior data is desensitized by way 1 and way 2, as follows:

Table 6

key identification	topic of interest	Additional summary information (not required)
basketball	physical education
https://china.nba.com/	physical education	NBA
real economy	Finance
new crown	current affairs	Wuhan

According to the above Table 6, it can be known that after the preprocessed user operation behavior data is further desensitized in stage 3, the obtained desensitized user operation behavior data can better protect user privacy, mainly for the purpose of Highlight the user interests of the current user operation behaviors, so as to obtain the current hot topic interests and the hot key identifiers under each topic interest, so as to facilitate the server side to generate the topic interest table or update the topic interest table, so as to improve the performance of the topic interest table based on user interests. Timeliness and accuracy of content recommendations.

In stage 4, the privacy protection module 303 on the terminal device uploads the processed and desensitized user operation behavior data to the data statistics module 304 on the server side.

The server side can be connected to one or more terminal devices, so the server side can obtain multiple sets of desensitized user operation behavior data uploaded by the privacy protection module 303 in one or more terminal devices.

In stage 5, the data statistics module 304 in the server performs summary statistics on the received desensitized user operation behavior data uploaded by the privacy protection module 303 in one or more terminal devices. And, the data analysis module 305 in the server analyzes the desensitized user operation behavior data collected and counted by the data statistics module 304 to generate a topic interest table.

It should be noted that, when the application is implemented, considering the computing capability of the terminal device side, if the terminal device is trained to generate a topic interest table, the terminal device needs to have high performance requirements. However, this implementation has high costs and cannot be used. It is a good defect to improve the recommendation efficiency. Therefore, the topic interest table obtained by training according to the collected massive user operation behavior data can generally be implemented on the server side. In addition, in order to protect user privacy data, before uploading the user operation behavior data to the server, the terminal device performs sensitive data desensitization processing on the user operation behavior data. The desensitization processing introduced; then, the terminal device sends the desensitized user operation behavior data obtained after the processing to the server to perform the operation of generating the topic interest table.

Exemplarily, based on the mapping relationship between a plurality of key identifiers and topics of interest contained in each user operation behavior data, the data statistics module 304 may perform statistical analysis on the received large amount of desensitized user operation behavior data, wherein the statistical analysis is performed. The analysis may include one or a combination of the corresponding relationship between the desensitized user operation behavior and the key identifier, the corresponding relationship between the desensitized user operation behavior data and the interest topic, and the corresponding relationship between the interest topic and the key identifier.

1) By analyzing the correspondence between the desensitized user operation behavior and the key identifiers, the popular key identifiers currently searched or browsed by the current user can be obtained, so that the terminal device can perform content recommendation according to the popular key identifiers, that is, mainly obtain the information related to the popular key identifiers. Contents such as information flow and web pages related to the key identification are recommended, so that the content recommended by the terminal device is the content that the user may be more interested in. For example, if the statistical analysis result indicates that the key identifier "NBA" is included the most times in a large amount of desensitized user operation behavior data, the terminal device determines that "NBA" is a popular key identifier currently searched or browsed by the user. , in this scenario, the terminal device can obtain some content related to "NBA" for recommendation.

2) By analyzing the correspondence between the desensitized user operation behavior data and the topic of interest, the popular topics of interest currently searched or browsed by the user can be obtained, so that the terminal device can perform content recommendation according to the topic of interest. For example, the statistical analysis result indicates that among a large amount of desensitized user operation behavior data, the interest topic "sports" is included the most times, and the terminal device determines that "sports" is a popular interest topic currently searched or browsed by the user. , in this scenario, the terminal device can obtain some related content under the "sports" interest topic for recommendation.

3) By analyzing the correspondence between interest topics and key identifiers, the popular key identifiers under each interest topic can be obtained, so that the terminal device can further perform content based on popular key identifiers in the scenario of content recommendation based on popular interest topics. recommend. For example, the statistical analysis result indicates that in a large amount of user operation behavior data after desensitization processing, it is determined that the key identifiers included under the interest topic "sports" include "NBA", "soccer", etc. If the number of times is the most, the terminal device determines that the popular key identifier under the "sports" interest topic is "NBA". You can get more content related to "NBA" for recommendation.

Taking the data statistics module 304 statistics on the corresponding relationship between interest topics and key identifiers as an example, the specific implementation is that the data statistics module 304 receives a large number of desensitization processing uploaded by the privacy protection module 303 in one or more terminal devices. After the user operation behavior data is obtained, the mapping relationship between multiple key identifiers and topic interests is determined according to multiple sequences contained in each desensitized user operation behavior data. The data statistics module 304 can determine the set of key identifiers contained under each topic of interest based on different topics of interest as categories, so as to obtain a mapping relationship between each topic of interest and the set of key identifiers contained in the topic of interest. For example, following the example in Table 6, after the summary statistics of the data statistics module 304, the mapping relationship of Table 7 is obtained, as follows:

Table 7

physical education	Finance	current affairs
basketball	real economy	new crown
https://china.nba.com/

Further, the data analysis module 305 can train the topic interest model after obtaining the statistical summary result of the desensitized user operation behavior data. Exemplarily, the construction of the topic interest model may be implemented by an algorithm such as a latent Dirichlet allocation (LDA) topic model algorithm, which is not limited in this application.

The topic interest model can generate topic interest tables, user interest topic radar charts, etc. to determine user interests. The topic interest table can be in the form shown in Table 7, each column represents an interest topic, and each column contains one or more key identifiers, wherein the key identifiers included in each column of interest topics are processed from mass desensitization It is obtained by analyzing the user operation behavior data. In addition, each column of interest topics can also be associated with a corresponding weight value, and the weight value represents the user's interest in the interest topic, where the weight value can be obtained through statistics and analysis of a large number of desensitized user operation behavior data Obtained, for example, the weight value corresponding to the interest topic that appears more frequently in the desensitized user operation behavior data is larger. It can be understood that the larger the weight value associated with the interest topic, the greater the interest degree of most users in the interest topic.

Alternatively, FIG. 5 is an example diagram of a radar chart of user interest topics shown in an embodiment of the present application. It is assumed that the data analysis module 305 learns a large number of desensitized data by analyzing the desensitized user operation behavior data. In the user operation behavior data, the interest topic "sports" is the most searched and browsed interest topic by users, followed by "current affairs", and less "financial and economics", and then generate the user operation shown in Figure 5 that can reflect the desensitization process. A radar chart of the behavior's level of interest in a topic of interest.

In stage 6, the data analysis module 305 on the server side sends the topic interest table to the data calibration module 302 on the terminal device side, so as to implement content recommendation based on user interests.

In an optional embodiment, the topic interest table generated by the data analysis module 305 on the server side can be used by the data calibration module 302 on the terminal device side to preprocess the user operation behavior data to obtain the preprocessed user operation data. behavioral data. As described in the previous section 2, the data calibration module 302 on the terminal device side can combine the topic interest table obtained from the server in the process of preprocessing the collected desensitized user operation behavior data. Therefore, it can be realized that the desensitized user operation behavior data after preprocessing can include key identifiers and interest topics, thereby obtaining more accurate desensitized user operation behavior data.

In another optional embodiment, the topic interest table generated by the data analysis module 305 on the server side can also be used by the privacy protection module 303 on the terminal device side to perform a differential privacy algorithm on the preprocessed user operation behavior data. processing to obtain desensitized user operation behavior data. As described in Section 3 above, the privacy protection module 303 of the terminal device can randomly replace the content in the preprocessed user operation behavior data based on the topic interest table, so as to protect the privacy of the user operation behavior data .

In another optional embodiment, the topic interest table generated by the data analysis module 305 on the server side can be used by the real-time recommendation module 306 on the terminal device side to perform real-time content recommendation, such as the content introduced in the following stage 7 part. , which will not be described in detail here.

In stage 7, the real-time recommendation module 306 on the terminal device determines user interests according to the user operation behavior data collected in real time by the data collection module 301 and the topic interest table obtained from the server side, and performs real-time recommendation according to the user interests.

Exemplarily, the real-time recommendation by the terminal device may include the following scenarios:

Scenario 1. The user opens the application as a new user. In this scenario, the terminal device does not store the user's historical user interests for the application.

During implementation, after the terminal device detects the user operation behavior of the user entering the application program, and does not detect other user operation behaviors of the user in the application program, it can be carried out according to the weight value of the interest topic included in the topic interest table. Content recommendation. For example, assuming that the terminal device detects that the user has opened the browser, and before detecting the user's user operation behavior in the browser, the subject interest table contains the interest topics of "sports", "current affairs" and "finance and finance", then the terminal device The relevant content of these interest topics can be obtained from the server side, and recommended through the information flow on the home page of the browser, and if the weight value of the interest topic is larger, the proportion of the recommended relevant content will be higher. Exemplarily, the topic interest table is as shown in the following table 8a:

Table 8a

physical education	Finance	current affairs
basketball	real economy	new crown
https://china.nba.com/

Further, if the terminal device detects the user's real-time operation behavior in the application program, it can determine the user's interest in combination with the user's real-time operation behavior and the subject interest table obtained from the server side, and then perform real-time operation according to the user's interest. recommend. For example, the user's real-time operation behavior is that the user enters the search term "Pitaya" in the browser, and the terminal device generates the user's user information according to the key identifier "Pitaya" and its corresponding interest topic "Fruit" together with the topic interest table interest (as shown in Table 8b below).

During specific implementation, the terminal device may associate different weight values for the topic of interest according to the type of user operation behavior. For example, since the search operation can better reflect the user's personal interests, a higher weight value may be assigned to the interest topic "fruit", so that the terminal device recommends a higher proportion of "fruit" related content. For another example, the terminal device detects that the user clicks on a certain information stream in the process of browsing the information stream on the home page interface, and the terminal device determines that the key identifier contained in the information stream is "Leo" and the corresponding interest topic is "Constellation". , you can add it to user interests. Since the information flow browsing operation is generally expressed as the user's immediate interest, a lower weight can be assigned to the interest topic "constellation", so that the terminal device recommends a lower proportion of "constellation" related content.

It should be noted that, as the number of operations performed by the user in the browser increases, the weight of the corresponding topic of interest may also be updated according to the number of operations performed by the user. For example, if the user browses the content related to "constellation" for many times in the browser subsequently, the weight value allocated to the "constellation" can be increased as the number of times the user browses increases. Among them, user interests can be reflected in the form of personalized topic interest tables, as shown in Table 8b below:

Table 8b

fruit	physical education	Finance	current affairs	constellation
pitaya	basketball	real economy	new crown	Leo
	https://china.nba.com/

The subject interest table shown in Table 8b above is only a possible example, and is not used to limit the embodiment of the user's interest. During the implementation of this application, the interest topics shown in Table 8b can also be sorted from left to right according to the weight value, and the key identifiers included in each interest topic can also be sorted from top to bottom according to the weight value. That is to say, it can be understood that the weight value associated with the topic of interest "fruit" in Table 8b is currently the largest, so the terminal device displays the highest proportion of the content recommended for "fruit".

In addition, after the terminal device detects that the user leaves the application this time, the user interest obtained according to the real-time operation behavior of the user this time can also be stored as the user's historical user interest, which can be used as the user's next entry into the application. References to determine user interests.

Scenario 2. The user opens the application as an old user. In this scenario, the terminal device generally stores the user's historical user interests for the application. It should be noted that, if the terminal device detects that the user is an old user, but does not store the user's historical user interests, the user interests can also be determined according to the implementation manner described in the foregoing scenario 1.

In implementation, after the terminal device detects the user's instruction to start the application, but before detecting other user operation behaviors in the browser, the terminal device can combine historical user interests and the topic interest table obtained from the server side to perform content recommendation. At this time, the user interest determined by the terminal device can be determined through the personalized topic interest table shown in Table 8b. For example, "fruit" and "constellation" in Table 8b are historical user interests, while "Sports" in Table 8b , "Finance", and "Current Affairs" are obtained from the subject interest table obtained from the server side.

Further, if the terminal device detects the user's real-time operation behavior in the application program, it can update the user's interest in combination with the user's real-time operation behavior and the above-mentioned personalized theme interest table, and then perform real-time recommendation according to the updated user interest. . For example, if the terminal device detects from the user's real-time operation behavior that the number of users' browsing of the topic interest "constellation" has increased significantly, and the relevant associated identifiers browsed are also "capricornus" and "horoscope", the terminal device is added to "constellation" The assigned weight value, and the key ID under "Constellation" is updated. The personalized topic interest table corresponding to the updated user interests can be shown in Table 8c below:

Table 8c

constellation	fruit	physical education	Finance	current affairs
Leo	pitaya	basketball	real economy	new crown
Capricornus		https://china.nba.com/
Horoscope

It should be noted that, after the terminal device determines the user's interest, the recommended related content and the manner of acquiring the related content are not limited. For example, the terminal device may search for the recommended content corresponding to the user interest from the locally cached content, or may also obtain the recommended content corresponding to the user interest from a content providing server that provides the recommended content corresponding to the user interest Wait.

Through the content recommendation method provided by the present application, one can avoid the cold start problem of the information flow of the application program on the terminal device. The cold start problem of the information flow of the application generally occurs in the scenario where the user opens the application for the first time. Since the terminal device does not store historical user interests, the terminal device cannot recommend the information flow. With the method provided in this application, the terminal device can determine the user's real-time interest according to the topic interest table obtained from the server side and in combination with the user's real-time operation behavior, etc., so that content can be recommended according to the user's real-time interest of the terminal device.

Second, when the present application is implemented, the real-time interests of users are generated on the terminal device side. Compared with the related art, after the user interests are generated on the server side according to big data, the user interests generated on the server side are directly recommended to different users. Interest content, the content recommendation method based on user interests provided by the present application can update user interests in time according to the user's real-time operation, thereby ensuring that the recommended content can better reflect the content that the user is currently interested in.

In order to better understand the overall flow of the method provided by the present application, the method provided by the present application will be further introduced below with reference to FIG. 6 . The method provided in this application can be mainly divided into two parts: the sampling part and the recommendation part. The terminal device may collect each operation of the user to obtain user operation behavior data. Then, the terminal device processes the user operation behavior data and sends it to the server side, so that the server side generates a topic interest table according to a large amount of collected user operation behavior data after desensitization processing. In addition, the server may also send a topic interest table to the terminal device, so that the terminal device can perform content recommendation according to the topic interest table. FIG. 6 is a schematic flowchart of a content recommendation based on user interests provided by an embodiment of the present application, including the following steps:

1. Sampling part

S601. The terminal device detects the user's instruction to start the target application, and starts the target application. Optionally, the user entering the target application may be the user clicking on the application icon on the main interface of the terminal device, or the user may wake up the target application through voice, or the user may also use any display interface on the terminal device. The target application can be quickly entered into the target application, etc., which is not limited in this application. For example, assuming that the target application is a browser, the terminal device detects the user operation behavior of the user clicking the browser icon, and can refer to the content shown in 1 in FIG. 7; The wake-up word of the browser, or the terminal device receives the user's click on the shortcut entry identifier of the browser contained in the drop-down interface, etc., you can refer to the content shown in 2 in FIG.

S602. The terminal device collects at least one operation data performed by the user on the target application. Exemplarily, as shown in FIG. 4 , after the terminal device enters the target application program, the user's operation behavior is collected in real time.

S603. When the terminal device detects the user's instruction to quit the target application, the terminal device closes the target application. Exemplarily, when the user exits the target application, the user may close the display interface of the target application and return to the main display interface of the terminal device; or the user may close the running of the target application through background cleaning; or it may be The application does not limit the forced exit of the target application due to the program unresponsiveness, etc.

S604. The terminal device stores at least one operation data of the processing procedures of S601 to S603 as a set of user operation behavior data.

S605: The terminal device preprocesses the user operation behavior data, and obtains preprocessed user operation behavior data whose sequence value is a preset value. Exemplarily, due to the multiple possibilities of user operation behaviors, each group of user operation behavior data may have different sequence lengths, and the terminal device may sample user operation behavior data of different sequence lengths based on a preset value. Specifically, if the sequence length of the user operation behavior data is less than the preset value, supplement the user operation behavior data with a default user operation behavior, where the default user operation behavior may be obtained from the subject interest table, or is pre-defined, and obtains the user operation behavior data of the specified sequence length. If the sequence length of the user operation behavior data is greater than the preset value, random truncation and sampling processing is performed on the user operation behavior data to obtain user operation behavior data with a specified sequence length. In this way, by preprocessing the sequence length of the user operation behavior data, it is possible to better avoid the failure to fully reflect the user operation behavior due to the short sequence of the user operation behavior data, and the long sequence of the user operation behavior data. The problem of sampling sample data is too large. It should be noted that the preset value may be customized by the terminal device, or obtained based on historical experience, or determined according to other rules, which is not limited in this application.

S606 , the terminal device performs desensitization processing on the preprocessed user operation behavior data to obtain desensitized user operation behavior data. In a possible design, the terminal device can randomly replace the content contained in the preprocessed user operation behavior data according to the topic interest table, so that the interest characteristics of the user operation behavior can be blurred to a certain extent. In another possible design, the terminal device may also perform user information stripping on the obfuscated user operation behavior data to obtain desensitized user operation behavior data, thereby avoiding leakage of user privacy.

The subject interest table involved in the implementation of the above S605 and S606 may be stored after the terminal device obtains it from the server side. Therefore, S6050 is located before S605 and S606, but the execution sequence between S6050 and S601 to S604 is not limited.

S6050. The terminal device acquires the topic interest table generated on the server side. Optionally, the server side may automatically send the topic interest table to the terminal device side in real time, or periodically, or after the topic interest table is updated. Alternatively, the terminal device may also send the request information to the server side, and the server sends the latest topic interest table to the terminal device after receiving the request information.

S607. The terminal device uploads the desensitized user operation behavior data to the server side. Considering that the computing power of the terminal device side is limited, and the server can train the topic interest model based on more user operation behavior data, it is helpful for training to obtain a more accurate and comprehensive topic interest table. When the application is implemented, the desensitized user operation behavior data obtained after the terminal device performs the processing on the user operation behavior data such as S605 and S606 is uploaded to the server. With the embodiments provided in this application, the terminal device side processes the collected user operation behavior data to obtain desensitized user operation behavior data, and then uploads it to the server side. Therefore, the server side cannot collect the user's private data, so The security of user operation behavior data can be improved.

S608. The server performs statistical summary on the desensitized user operation behavior data uploaded by one or more terminal devices. The server may be connected to one or more terminal devices, and FIG. 6 only takes one terminal device as an example for description, and the interactions between other terminal devices and the server are similar. Exemplarily, through the statistical summary of the desensitized user operation behavior data, currently popular interest topics and key identifiers included in each interest topic can be obtained.

S609: The server generates or updates a topic interest table based on the user operation behavior data after desensitization processing after statistical aggregation. Exemplarily, the server may use the statistical aggregated user operation behavior data as a training sample to perform unsupervised learning on user interests, thereby obtaining a topic interest table that can reflect the mapping relationship between interest topics and key identifiers. The topic interest table may include multiple interest topics and key identifiers included in each interest topic. In addition, the weight value of each interest topic and the weight value of each key identifier included in each interest topic can also be associated in the topic interest table, and the weight value reflects the interest topics and key identifiers that most users are interested in. For example, the interest topic "sports" contained in the topic interest table has the largest weight value, indicating that most users are currently interested in the sports topic; further, the key identifier "basketball" contained in the "sports" topic has the largest weight value, It means that most current users are more interested in basketball keywords.

S610. The server sends the topic interest table to the one or more terminal devices.

The recommended part

S611. The terminal device detects the user's instruction to start the target application, and starts the target application. For example, as shown in FIG. 7 , the terminal device detects a user operation behavior of the user opening the browser again. It should be noted that, for a clearer understanding of this application, S601-S603 and S611-S613 are respectively used to represent the processing performed by the terminal device in response to user operations in two different scenarios, wherein S601-S603 are collecting user operations in real time. While performing content recommendation based on user interests, S611-S 6113 can also perform real-time collection of user operation behaviors while performing content recommendation based on user interests.

S612. The terminal device determines the user's interest. Exemplarily, the determination of the user interest by the terminal device may include the following possible scenarios:

Scenario A: The user opens the browser for the first time. In such a scenario, since there is no historical user interest in the browser, the terminal device can recommend the information flow in the browser homepage interface for the user according to the topic interest table, that is, display the first recommendation interface. For example, according to the aforementioned Table 7, the topics of interest acquired by the terminal device from the server side include "sports", "financial and economics" and "current affairs", and the key identifiers contained in "sports" include basketball, https://china.nba.com/, The key identifiers included in "Finance" include the real economy, and the key identifiers included in "Current Affairs" include the new crown. When the user opens the browser for the first time, the browser homepage interface can be as shown in Figure 8a. According to the content shown in Figure 8a, it can be obtained that when the user opens the browser for the first time, the content displayed on the homepage interface of the browser includes the recommended information flow of the basketball association-homepage related to the key identifier "basketball", and "https://" china.nba.com/” related to the NBA China official website, articles related to the “real economy”, and hot interpretation articles related to the “new crown”; and, by sliding down the home page of the browser, users can also It is possible to browse to more relevant recommended content (not shown in FIG. 8a ) related to the interest topics contained in the topic interest table.

Scenario B: It is not the first time that the user opens the browser, and before the user's operation behavior is performed. In such a scenario, historical user interests may be stored in the browser, and the terminal device may recommend the information flow in the browser homepage interface for the user according to the historical user interests and the topic interest table. For example, in combination with the aforementioned subject interest table obtained as shown in Table 8b, in addition to including "sports", "financial affairs" and "current affairs" as included in the aforementioned scenario A, it can also include historical user interests "fruit", "constellation", At this time, when the user opens the browser, the browser home page interface may be as shown in FIG. 8b.

Scenario C: After the user opens the browser for the first time and performs some user actions. In such a scenario, the terminal device can recommend the information flow in the browser homepage interface for the user according to the user's real-time operation behavior, historical user interest and topic interest table, that is, the first recommendation interface is displayed. For example, in combination with the aforementioned table 8c, according to the updated personalized topic interest table obtained by the user's real-time operation behavior, it can be obtained that the user is more interested in the topic interest "constellation", then in the content recommendation of the browser home page interface, "" If the proportion of content related to "Constellation" increases, the browser home page interface after the terminal device refreshes according to the updated personalized topic interest table may be as shown in Figure 8c. Since the updated personalized topic interest table indicates that users are more interested in "constellations", the proportion of recommended content related to "constellations" in the browser homepage interface increases, and it is higher in the browser homepage interface. s position.

When the above example is implemented, the second recommendation interface may be triggered to display after the terminal device detects an instruction of the user to refresh the first recommendation interface. For example, in an application, when the user swipes down from the top of the terminal device, it means that the user wants to refresh the current interface. At this time, the terminal device can display the recommended content corresponding to the updated user interests on the terminal On the device, it can also be understood as displaying the second recommendation interface.

It should be noted that a user's interest does not specifically refer to a certain interest. It can represent a collection of multiple interest topics, and each interest topic is associated with a weight value. The weight value is used to reflect the degree of user interest. The greater the interest topic weight value , which can indicate that the user is more interested in the topic of interest.

S613. The terminal device recommends content according to the user's interests. Optionally, the terminal device may obtain the relevant content corresponding to the user's interests from the local cache, or may also send an obtaining request to the server to obtain the relevant content corresponding to the user's interests from the server.

Based on the same technical concept, FIG. 9 shows a terminal device 900 provided by an embodiment of the present application. The terminal device 900 includes one or more processors 901 ; one or more memories 902 ; a communication interface 903 , and one or more computer programs 904 . The communication interface 903 is used to implement communication with other devices (such as terminal devices), for example, the communication interface may be a transceiver. wherein the one or more computer programs 904 are stored in the aforementioned memory 902 and configured to be executed by the one or more processors 901, the one or more computer programs 904 include instructions that can be used to perform the following steps ,include:

Collect multiple user operation behavior data input by the user when the user uses the target application program one or more times in the set duration; perform desensitization processing on the multiple collected user operation behavior data, and the desensitization processing is to desensitize all the collected user operation behavior data. Filter out the privacy data related to the user in the user operation behavior data; send the desensitized multiple user operation behavior data to the server, so that the server analyzes the desensitized multiple user operation behavior data, Obtain the subject interest table of the user using the target application.

Exemplarily, the collecting a plurality of user operation behavior data input by the user when the user uses the target application program at one time is specifically implemented as starting the target application program when an instruction of the user to start the target application program is detected; After the target application is started, collect at least one operation data performed by the user on the target application; when detecting the user's instruction to exit the target application, close the target application; At least one operation data collected from the start-up to the shutdown process of the target application is stored as a set of user operation behavior data.

Optionally, performing desensitization processing on the multiple user operation behavior data collected is specifically implemented as, for one or more user operation behavior data in the multiple user operation behavior data, based on a differential privacy algorithm. Perform random replacement of user operation behavior data under the same interest topic, and the interest topic is determined according to the topic interest table; strip the user information contained in the plurality of user operation behavior data.

In a possible embodiment, before randomly replacing the user operation behavior data under the same interest topic based on the differential privacy algorithm, the sequence length of each user operation behavior data is determined; The operation behavior data is truncated and compensated to obtain the user operation behavior data of the specified sequence length.

The truncation and compensation processing is performed on the user operation behavior data according to the preset value to obtain the user operation behavior data of the specified sequence length, which is specifically implemented as follows: if the sequence length of the user operation behavior data is less than the preset value, Supplementing the user operation behavior data with pre-defined user operation behavior data of a target length to obtain user operation behavior data with a specified sequence length; if the sequence length of the user operation behavior data is greater than the preset value, the The user operation behavior data is truncated to the target length to obtain the user operation behavior data of the specified sequence length; wherein the target length is the absolute value of the difference between the sequence length of the user operation behavior data and a preset value.

Based on the same technical concept, FIG. 9 may also provide a server 900 provided in this embodiment of the present application. The server 900 includes one or more processors 901 ; one or more memories 902 ; a communication interface 903 , and one or more computer programs 904 . The communication interface 903 is used to implement communication with other devices (such as terminal devices), for example, the communication interface may be a transceiver. wherein the one or more computer programs 904 are stored in the aforementioned memory 902 and configured to be executed by the one or more processors 901, the one or more computer programs 904 include instructions that can be used to perform the following steps ,include:

Receive multiple desensitized user operation behavior data sent by one or more terminal devices; the desensitized user operation behavior data is collected by the one or more terminal devices once the user or the user has a set duration. Multiple user operation behavior data input by the user when the target application is used multiple times, and obtained by desensitizing the collected user operation behavior data; the desensitization process is to desensitize the user operation behavior data Filter out the privacy data involved in the user; analyze multiple user operation behavior data after the desensitization process to obtain the subject interest table of the user using the target application; send the subject interest table to the one or more terminal devices.

Exemplarily, the desensitization-processed plurality of user operation behavior data is analyzed to obtain the subject interest table of the user using the target application, which is specifically implemented as: The user operation behavior data is input into a pre-built topic interest model to perform unsupervised learning on the desensitized multiple user operation behavior data; a topic interest table output by the pre-built topic interest model is obtained.

Based on the same technical concept, FIG. 9 may also provide a terminal device 900 provided in an embodiment of the present application. The terminal device 900 includes one or more processors 901 ; one or more memories 902 ; a communication interface 903 , and one or more computer programs 904 . The communication interface 903 is used to implement communication with other devices (such as terminal devices), for example, the communication interface may be a transceiver. wherein the one or more computer programs 904 are stored in the aforementioned memory 902 and configured to be executed by the one or more processors 901, the one or more computer programs 904 include instructions that can be used to perform the following steps ,include:

Receive the subject interest table sent by the server, where the subject interest table is obtained by the server analyzing multiple user operation behavior data after desensitization processing; the user operation behavior data after desensitization processing is one or more The terminal device collects multiple user operation behavior data input by the user when the user uses the target application program one or more times in the set duration, and desensitizes the collected multiple user operation behavior data; The desensitization process is to filter out the private data related to the user in the user operation behavior data; when detecting the user's instruction to start the target application, start the target application and display the first recommendation The first recommendation interface includes at least one item of recommended content; the at least one item of recommended content is determined according to the topic interest table.

Exemplarily, the starting the target application and displaying the first recommendation interface is specifically implemented as: starting the target application; displaying the first recommendation interface after starting the target application; One or more interest topics included in the user interest are taken as user interests, at least one recommended content is obtained according to the user interest, and the obtained at least one recommended content is displayed in the first recommendation interface; wherein, each of the interests The topic has an associated weight value, and the greater the weight value associated with the interest topic, the higher the proportion of the recommended content including the related content of the interest topic.

In a possible embodiment, after the target application is started and the first recommendation interface is displayed, one or more user operation behavior data input by the user when the user uses the target application is received and collected; When the user refreshes the instruction of the first recommendation interface, a second recommendation interface is displayed; the recommended content included in the second recommendation interface is determined according to the one or more user operation behavior data and the topic interest table .

In a possible design, the displaying of the second recommendation interface is specifically implemented as determining one or more corresponding interest topics according to the one or more user operation behavior data, and assigning an association to each of the interest topics take one or more interest topics corresponding to the user operation behavior data and one or more of the interest topics included in the topic interest table as the user interest, and obtain at least one recommendation according to the user interest content, and display at least one item of recommended content obtained in the second recommendation interface; wherein, each of the interest topics included in the topic interest table has an associated weight value, and the greater the weight value associated with the interest topic , the higher the proportion of content related to the topic of interest included in the recommended content.

In a possible design, the acquiring at least one piece of the recommended content according to the user's interest is specifically implemented as searching for the recommended content corresponding to the user's interest from locally cached content; The content providing server of the recommended content corresponding to the user's interest acquires the recommended content corresponding to the user's interest.

From the description of the above embodiments, those skilled in the art can clearly understand that for the convenience and brevity of the description, only the division of the above functional modules is used as an example for illustration. In practical applications, the above functions can be allocated as required. It is completed by different functional modules, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. For the specific working process of the system, apparatus and unit described above, reference may be made to the corresponding process in the foregoing method embodiments, and details are not described herein again.

Each functional unit in each of the embodiments of the embodiments of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, and the computer software products are stored in a storage The medium includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: flash memory, removable hard disk, read-only memory, random access memory, magnetic disk or optical disk and other media that can store program codes.

The above are only specific implementations of the embodiments of the present application, but the protection scope of the embodiments of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed in the embodiments of the present application shall be covered by this within the protection scope of the application examples. Therefore, the protection scope of the embodiments of the present application should be subject to the protection scope of the claims.

Claims (17)

1. A method for recommending content based on user interests, characterized in that, when applied to a terminal device, the method includes:

   Collect multiple user operation behavior data entered by the user when the user uses the target application one or more times during the set time period;

   performing desensitization processing on the plurality of collected user operation behavior data, where the desensitization processing is to filter out private data involving the user in the user operation behavior data;

   Sending the multiple user operation behavior data after desensitization processing to the server, so that the server analyzes the multiple user operation behavior data after the desensitization processing, and obtains the subject interest table of the user using the target application program.
2. The method according to claim 1, wherein collecting a plurality of user operation behavior data input by the user when the user uses the target application program once, comprising:

   When detecting the user's instruction to start the target application, start the target application;

   After the target application is started, collect at least one operation data performed by the user on the target application;

   When detecting the user's instruction to quit the target application, close the target application;

   At least one operation data collected from the start-up to the shutdown process of the target application is stored as a set of user operation behavior data.
3. The method according to claim 1 or 2, wherein performing desensitization processing on the plurality of collected user operation behavior data, comprising:

   For one or more user operation behavior data in the plurality of user operation behavior data, random replacement of user operation behavior data under the same interest topic is performed based on a differential privacy algorithm, and the interest topic is determined according to the topic interest table;

   Strip the user information contained in the plurality of user operation behavior data.
4. The method according to claim 3, wherein, before the random replacement of user operation behavior data under the same topic of interest based on a differential privacy algorithm, the method further comprises:

   determining the sequence length of each of the user operation behavior data;

   The user operation behavior data is truncated and compensated according to a preset value to obtain user operation behavior data with a specified sequence length.
5. The method according to claim 4, wherein the user operation behavior data is truncated and compensated according to a preset value to obtain the user operation behavior data of a specified sequence length, comprising:

   If the sequence length of the user operation behavior data is less than the preset value, supplementing the user operation behavior data with pre-defined user operation behavior data of a target length to obtain user operation behavior data with a specified sequence length;

   If the sequence length of the user operation behavior data is greater than the preset value, truncate the target length of the user operation behavior data to obtain the user operation behavior data of the specified sequence length;

   The target length is the absolute value of the difference between the sequence length of the user operation behavior data and a preset value.
6. A method for recommending content based on user interests, characterized in that, when applied to a server, the method includes:

   Receive multiple desensitized user operation behavior data sent by one or more terminal devices; the desensitized user operation behavior data is collected by the one or more terminal devices once the user or the user has a set duration. Multiple user operation behavior data input by the user when the target application is used multiple times, and obtained by desensitizing the collected user operation behavior data; the desensitization process is to desensitize the user operation behavior data Filtering out the privacy data of the user involved in;

   Analyzing a plurality of user operation behavior data after the desensitization process, to obtain the subject interest table of the user using the target application;

   Sending the topic interest list to the one or more terminal devices.
7. The method according to claim 6, characterized in that, the process of analyzing a plurality of user operation behavior data after the desensitization process to obtain a subject interest table of the user using the target application program, comprising:

   Inputting the desensitized multiple user operation behavior data into a pre-built topic interest model to perform unsupervised learning on the desensitized multiple user operation behavior data;

   A topic interest table output by the pre-built topic interest model is obtained.
8. A method for recommending content based on user interests, characterized in that, when applied to a terminal device, the method includes:

   Receive the subject interest table sent by the server, where the subject interest table is obtained by the server analyzing multiple user operation behavior data after desensitization processing; the user operation behavior data after desensitization processing is one or more The terminal device collects multiple user operation behavior data input by the user when the user uses the target application program one or more times in the set duration, and desensitizes the collected multiple user operation behavior data; The desensitization processing is to filter out the private data involving the user in the user operation behavior data;

   When detecting the user's instruction to start the target application, start the target application and display a first recommendation interface, where the first recommendation interface includes at least one recommended content; the at least one recommended content is Determined according to the subject interest table.
9. The method according to claim 8, wherein the starting the target application and displaying the first recommendation interface comprises:

   start the target application;

   Displaying a first recommendation interface after starting the target application;

   Taking one or more interest topics included in the topic interest table as a user interest, acquiring at least one recommended content according to the user interest, and displaying the acquired at least one recommended content in the first recommendation interface; Wherein, each of the interest topics has an associated weight value, and the greater the associated weight value of the interest topic, the higher the proportion of the recommended content including the related content of the interest topic.
10. The method according to claim 8, wherein after the target application is started and the first recommendation interface is displayed, the method further comprises:

    Receive and collect one or more user operation behavior data input by the user when the user uses the target application;

    When detecting the user's instruction to refresh the first recommendation interface, a second recommendation interface is displayed; the recommended content included in the second recommendation interface is based on the one or more user operation behavior data and the topic interest table is determined.
11. The method according to claim 10, wherein the displaying the second recommendation interface comprises:

    Determine one or more corresponding interest topics according to the one or more user operation behavior data, and assign an associated weight value to each of the interest topics;

    Take one or more interest topics corresponding to the user operation behavior data and one or more of the interest topics included in the topic interest table as the user interest, obtain at least one recommended content according to the user interest, and use the The obtained at least one recommended content is displayed in the second recommendation interface; wherein, each of the interest topics included in the topic interest table has an associated weight value, and the greater the weight value associated with the interest topic, the higher the recommended weight value. The higher the proportion of content related to the topic of interest contained in the content.
12. The method according to claim 9 or 11, wherein the acquiring at least one piece of the recommended content according to the user's interests comprises:

    Find recommended content corresponding to the user's interests from locally cached content; and/or,

    The recommended content corresponding to the user's interest is acquired from a content providing server that provides the recommended content corresponding to the user's interest.
13. A terminal device, comprising: one or more processors; one or more memories;

    the one or more memories for storing one or more computer programs and data information; wherein the one or more computer programs comprise instructions;

    When executed by the one or more processors, the instructions cause the terminal device to perform the method as claimed in any one of claims 1 to 5, or to perform the method as claimed in any one of claims 8 to 12. method described.
14. A server, characterized by comprising: one or more processors; one or more memories;

    the one or more memories for storing one or more computer programs and data information; wherein the one or more computer programs comprise instructions;

    The instructions, when executed by the one or more processors, cause the server to perform the method of claim 6 or 7.
15. A communication system, characterized by comprising the terminal device as claimed in claim 13 and the server as claimed in claim 14 .
16. A computer-readable storage medium, characterized in that it includes a computer program or instruction, which, when the computer program or instruction is run on a computer, causes the method according to any one of claims 1 to 5 to be performed, or The method of claim 6 or 7 is performed, or the method of any one of claims 8 to 12 is performed.
17. A graphical user interface on a terminal device, characterized in that the terminal device has a display screen, one or more memories, and one or more processors, and the one or more processors are used to execute the One or more computer programs in one or more memories, said graphical user interface comprising said terminal device performing the method of any one of claims 1 to 5 or performing any one of claims 8 to 12 The graphical user interface displayed when the method described in the item is used.

Images