Information recommendation method, device, electronic device and storage medium
By obtaining user behavior parameters, determining user clusters and adjusting the number of recommended information, the problem that traditional information recommendation systems cannot accurately match user interests is solved, efficient and accurate information recommendations are achieved, and user experience is improved.
Patent Information
- Application Number
- CN202010768083.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-08-03
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2040-12-06
AI Technical Summary
Traditional information recommendation systems are based on extensive classified label information, and cannot accurately match users' real interests, affecting user experience.
By obtaining the behavioral parameter information of the target user, determining the user cluster information, obtaining candidate recommendation information, and adjusting the number of recommended information through the matching information information recommendation model to achieve matching information and users.
It improves the accuracy and timeliness of information recommendations, improves user experience, and enhances the quality of recommendations.
Smart Images

Figure CN111859149B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to information processing technology, and in particular to an information recommendation method, device, electronic device and storage medium. Background Art
[0002] In traditional technology, various information recommendation systems can use collaborative filtering to recommend corresponding information to users. Collaborative filtering is widely used in various information recommendation systems as an effective recommendation method. However, this recommendation based on user interests uses user historical behavior to establish user interest scores in specific categories and tags. When recalling news information, if the tag information in the news information hits the corresponding user interest tag, the information will be recalled. However, this type of representation method is relatively rough, and the classification tag information needs to be set in advance and updated in time, and its content representation ability is limited. Class tag information needs to be set in advance and updated in time, and it is impossible to recommend matching news information based on the user's real interests, which affects the user's experience. Summary of the invention
[0003] In view of this, the embodiments of the present invention provide an information recommendation method, device, electronic device and storage medium. The technical solution of the embodiments of the present invention is implemented as follows:
[0004] The present invention provides a method for recommending information, the method comprising:
[0005] In response to the information recommendation request, obtaining the behavior parameter information of the target user;
[0006] Based on the behavior parameter information of the target user, determining information of different user clusters to which the target user belongs;
[0007] Based on the different user cluster information, obtaining candidate recommended information corresponding to each user cluster information in the information source;
[0008] Determining the information to be recommended that matches the target user based on the candidate recommended information corresponding to each user cluster information;
[0009] According to the behavior parameter information of the target user, through the information recommendation model matching the target user, the number of candidate recommended information corresponding to different user cluster information in the information to be recommended is adjusted to achieve matching of the information to be recommended with the target user.
[0010] The embodiment of the present invention further provides a news information recommendation device, comprising:
[0011] An information transmission module, used to obtain the behavior parameter information of the target user in response to the information recommendation request;
[0012] An information processing module, used to determine information of different user clusters to which the target user belongs based on the behavior parameter information of the target user;
[0013] The information processing module is used to obtain candidate recommended information corresponding to each user cluster information in the information source based on the different user cluster information;
[0014] The information processing module is used to determine the information to be recommended that matches the target user based on the candidate recommended information corresponding to each user cluster information;
[0015] The information processing module is used to adjust the number of candidate recommended information corresponding to different user cluster information in the information to be recommended according to the behavior parameter information of the target user and through an information information recommendation model matching the target user, so as to match the information to be recommended with the target user.
[0016] In the above scheme,
[0017] The information processing module is used to obtain original data sets of different dimensions corresponding to the target user;
[0018] The information processing module is used to determine the user behavior parameter information matching the target user according to the original data sets of different dimensions;
[0019] The information processing module is used to perform dynamic clustering rule classification processing on the user behavior parameter information to determine the identifiers of different user cluster information;
[0020] The information processing module is used to determine the different user cluster information to which the target user belongs based on the identifiers of the different user cluster information.
[0021] In the above scheme,
[0022] The information processing module is used to determine the user cluster identifier corresponding to each user cluster based on the different user cluster information;
[0023] The information processing module is used to determine the content identifier of each piece of information in the information source;
[0024] The information processing module is used to obtain candidate recommended information corresponding to each user cluster information in the information source according to the matching result between the content identifier and the user cluster identifier.
[0025] In the above scheme,
[0026] The information processing module is used to determine a first joint feature vector and a second joint feature vector that match the behavior parameter information of the target user;
[0027] The information processing module is used to process the first joint feature vector and the second joint feature vector through the information information recommendation model to determine the optional range of information corresponding to different user clusters in the display interface of the target user, wherein the optional range of information information includes the number of information corresponding to different user clusters;
[0028] The information processing module is used to determine the browsing times of the information respectively corresponding to different user clusters in the display interface of the target user based on the information recommendation model;
[0029] The information processing module is used to determine to adjust the number of candidate recommended information corresponding to different user cluster information in the information to be recommended based on the number of views of the information corresponding to the different user clusters.
[0030] In the above scheme,
[0031] The information processing module is used to cross-sort the browsing times of the news information corresponding to the different user clusters;
[0032] The information processing module is used to determine the proportion parameters corresponding to different user clusters based on the cross-ranking result of the browsing times;
[0033] The information processing module is used to determine the amount of news information to be displayed in the display interface of the target user;
[0034] The information processing module is used to adjust the number of candidate recommended information corresponding to different user cluster information in the information to be recommended based on the number of information to be displayed in the display interface of the target user and the proportion parameters corresponding to the same user clusters.
[0035] In the above scheme, the device also includes:
[0036] A training module, used to obtain a training sample set, wherein the training sample set includes a multi-dimensional state feature vector and the number of information corresponding to different user clusters;
[0037] The training module is used to process the training sample set through the information recommendation model to determine the initial parameters of the information recommendation model;
[0038] The training module is used to process different training samples in the training sample set through the information information recommendation model in response to the initial parameters of the information information recommendation model to determine the update parameters of the information information recommendation model;
[0039] The training module is used to iteratively update the parameters of the information information recommendation model through the training sample set according to the update parameters of the information information recommendation model, so that the information information recommendation model can determine the number of views of the information corresponding to different user clusters.
[0040] In the above scheme,
[0041] The training module is used to substitute different training samples in the training sample set into the loss function corresponding to the multi-layer perceptron network of the information recommendation model;
[0042] The training module is used to determine the parameters of the multilayer perceptron network in the information information recommendation model corresponding to the loss function when the convergence condition is met as the update parameters of the information information recommendation model.
[0043] In the above scheme,
[0044] The training module is used to determine a dynamic noise threshold that matches the video information recommendation process when the news information recommendation model is applied to the video information recommendation process;
[0045] The training module is used to perform denoising on the first training sample set according to the dynamic noise threshold to form a second training sample set matching the dynamic noise threshold.
[0046] In the above scheme,
[0047] The training module is used to determine a fixed noise threshold corresponding to the news information recommendation process when the news information recommendation model is applied to the news information recommendation process, and to denoise the first training sample set according to the fixed noise threshold to form a second training sample set matching the fixed noise threshold.
[0048] In the above scheme,
[0049] The information processing module is used to detect the operation information of the target user when outputting the information to be recommended to the target user;
[0050] The information processing module is used to determine the click rate and browsing time of the information to be recommended based on the detected operation information of the target user.
[0051] An embodiment of the present invention further provides an electronic device, the electronic device comprising:
[0052] A memory for storing executable instructions;
[0053] The processor is used to implement the previous information recommendation method when running the executable instructions stored in the memory.
[0054] An embodiment of the present invention further provides a computer-readable storage medium storing executable instructions, wherein the executable instructions implement the preceding information recommendation method when executed by a processor.
[0055] The embodiments of the present invention have the following beneficial effects:
[0056] The present invention obtains the behavior parameter information of the target user in response to the information recommendation request; determines the different user cluster information to which the target user belongs based on the behavior parameter information of the target user; obtains the candidate recommended information corresponding to each user cluster information in the information source based on the different user cluster information; determines the information to be recommended that matches the target user based on the candidate recommended information corresponding to each user cluster information; and adjusts the number of candidate recommended information corresponding to different user cluster information in the information to be recommended according to the behavior parameter information of the target user through an information recommendation model that matches the target user, thereby achieving the matching of the information to be recommended with the target user, while enhancing the accuracy and timeliness of the information recommendation, effectively improving the quality of the information recommendation, and enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] Figure 1 A schematic diagram of a usage scenario of the information recommendation method provided by an embodiment of the present invention;
[0058] Figure 2 A schematic diagram of the structure of an information recommendation device provided by an embodiment of the present invention;
[0059] Figure 3 An optional flowchart of the information recommendation method provided by an embodiment of the present invention;
[0060] Figure 4 This is a schematic diagram of an optional model structure of an information recommendation model in an embodiment of the present invention;
[0061] Figure 5 An optional flowchart of the information recommendation method provided by an embodiment of the present invention;
[0062] Figure 6A schematic diagram of an application environment of a training method for an information recommendation model according to an embodiment of the present invention;
[0063] Figure 7 An optional workflow of the information recommendation method provided in the embodiment of the present invention;
[0064] Figure 8 An optional workflow of the information recommendation method provided in the embodiment of the present invention;
[0065] Fig. 9 A schematic diagram of the working process of the information recommendation method provided by an embodiment of the present invention;
[0066] Fig.10 A schematic diagram of the training process of the information recommendation method provided by an embodiment of the present invention;
[0067] Fig.11 Schematic diagram of the display effect of the information recommendation method in an embodiment of the present invention. DETAILED DESCRIPTION
[0068] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present invention. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.
[0069] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.
[0070] The relevant data collection and processing in the embodiments of this application should be strictly in accordance with the requirements of relevant national laws and regulations when applied in examples, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of authorization of laws and regulations and the personal information subject.
[0071] Before further describing the embodiments of the present invention in detail, the nouns and terms involved in the embodiments of the present invention are described. The nouns and terms involved in the embodiments of the present invention are subject to the following interpretations.
[0072] 1) In response, it is used to indicate the conditions or states on which the executed operations depend. When the dependent conditions or states are met, one or more operations executed may be in real time or have a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations executed are executed.
[0073] 2) News information: various forms of information available on the Internet, such as video files, multimedia information, news information, etc. presented on the client or smart devices.
[0074] 3) Clustering: According to the neighborhood density of n data or the distance between data, the data is aggregated into N categories, where N is much smaller than n.
[0075] 4) Model training, multi-classification learning of image datasets. The model can be built using deep learning frameworks such as Tensor Flow and torch, and uses multiple layers of neural network layers such as CNN to form a multi-classification model. The input of the model is a three-channel or original channel matrix formed by reading the image through tools such as openCV. The model output is a multi-classification probability, and the web page category is finally output through algorithms such as softmax. During training, the model approaches the correct trend through objective functions such as cross entropy.
[0076] 5) Neural Network (NN), Artificial Neural Network (ANN), referred to as neural network or quasi-neural network, in the field of machine learning and cognitive science, is a mathematical model or computational model that imitates the structure and function of biological neural networks (the central nervous system of animals, especially the brain) and is used to estimate or approximate functions.
[0077] 6) Token, word unit, before any actual processing of the input text, it needs to be segmented into language units such as words, punctuation marks, numbers or pure alphanumeric characters. These units are called word units.
[0078] 7) Softmax, the normalized exponential function, is a generalization of the logic function. It can "compress" a K-dimensional vector containing any real number into another K-dimensional real vector, so that each element is between [0, 1] and the sum of all elements is 1.
[0079] 8) Word segmentation: Use Chinese word segmentation tools to segment Chinese text and obtain a set of fine-grained words. Stop words: characters or words that have no contribution to the semantics of the text or whose contribution can be ignored.
[0080] 9) Cluster: A division of all data obtained through clustering algorithms. Its characteristics are that the data in the same cluster have the same characteristics (high data density and small spacing), and the data in different clusters are obviously different. Among them, the interests of different users in the user cluster are the same or similar. For example, in the use scenario of news recommendation, it can be a sports news user cluster, an entertainment news user cluster, and an economic news user cluster. The entertainment news user cluster can also include: Mainland entertainment news user cluster, Hong Kong and Taiwan entertainment news user cluster, and European and American entertainment news user cluster, and can be marked according to the user cluster type divided by the user.
[0081] Figure 1 A schematic diagram of a usage scenario of the information recommendation method provided by an embodiment of the present invention, see Figure 1 The terminal (including the terminal 10-1 and the terminal 10-2) is provided with a client of software capable of displaying corresponding different information, such as a client or plug-in for video playback. The user can obtain different information (such as different short video information or news information) and display it through the corresponding client; the terminal is connected to the server 200 through the network 300, and the network 300 can be a wide area network or a local area network, or a combination of the two, and a wireless link is used to realize data transmission.
[0082] As an example, the server 200 is used to deploy a corresponding information recommendation model to implement the information information recommendation method provided by the present invention, or to deploy an information information recommendation device to implement the information information recommendation method. Specifically, the information information recommendation process includes: in response to an information information recommendation request, obtaining behavior parameter information of a target user; based on the behavior parameter information of the target user, determining different user cluster information to which the target user belongs; based on the different user cluster information, obtaining candidate recommended information information corresponding to each user cluster information in an information information source; based on the candidate recommended information information corresponding to each user cluster information, determining information to be recommended that matches the target user; according to the behavior parameter information of the target user, through an information information recommendation model that matches the target user, adjusting the number of candidate recommended information information corresponding to different user cluster information in the information to be recommended, so as to match the information to be recommended with the target user, and displaying and outputting the information to be recommended that matches the target user through a terminal (terminal 10-1 and / or terminal 10-2). Taking short video information as an example, the information recommendation model provided by the present invention can be applied to short video playback. In short video playback, different short video information from different data sources is usually processed, and finally the corresponding different information and the corresponding short video recommendation process are presented on the user interface UI (User Interface). The accuracy and timeliness of the characteristics of different information directly affect the user experience. The background database of video playback receives a large amount of video data from different sources every day. The different information information obtained for recommending information to target users can also be called by other applications (for example, the recommendation results of the short video recommendation process are migrated to the long video recommendation process or the news recommendation process). Of course, the information recommendation model matching the corresponding target user can also be migrated to different video recommendation processes (for example, web video recommendation process, mini program video recommendation process or long video client video recommendation process).
[0083] Among them, the information recommendation method provided in the embodiment of the present application is based on artificial intelligence. Artificial Intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that the machines have the functions of perception, reasoning and decision-making.
[0084] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0085] In the embodiments of the present application, the artificial intelligence software technologies mainly involved include the above-mentioned speech processing technology and machine learning. For example, it may involve speech recognition technology (Automatic Speech Recognition, ASR) in speech technology (Speech Technology), including speech signal preprocessing (Speech signal preprocessing), speech signal frequency domain analysis (Speech signal frequency analyzing), speech signal feature extraction (Speech signal feature extraction), speech signal feature matching / recognition (Speech signal feature matching / recognition), speech training (Speech training), etc.
[0086] For example, it may involve machine learning (ML), which is a multi-disciplinary interdisciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning usually includes technologies such as deep learning. Deep learning includes artificial neural networks, such as convolutional neural networks (CNN), recurrent neural networks (RNN), deep neural networks (DNN), etc.
[0087] It can be understood that the information recommendation model training method and voice processing provided in this application can be applied to smart devices (Intelligent device). The smart device can be any device with information display function, such as smart terminals, smart home devices (such as smart speakers, smart washing machines, etc.), smart wearable devices (such as smart watches), in-vehicle intelligent central control systems (displaying information to users through small programs that perform different tasks) or AI smart medical devices (displaying treatment cases by displaying information), etc.
[0088] The structure of the information recommendation device of the embodiment of the present invention is described in detail below. The information recommendation device can be implemented in various forms, such as a dedicated terminal with an information recommendation processing function, or a server provided with an information recommendation device processing function, such as the preceding Figure 1 Server 200 in. Figure 2 The schematic diagram of the structure of the information recommendation device provided by the embodiment of the present invention can be understood as follows: Figure 2 Only an exemplary structure of the information recommendation device is shown, not all structures, and can be implemented as needed. Figure 2 Partial or complete structure shown.
[0089] The information recommendation device provided in the embodiment of the present invention includes: at least one processor 201, a memory 202, a user interface 203 and at least one network interface 204. The various components in the information recommendation device are coupled together through a bus system 205. It can be understood that the bus system 205 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 205 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, in Figure 2 Various buses are labeled as bus system 205 .
[0090] The user interface 203 may include a display, a keyboard, a mouse, a trackball, a click wheel, keys, buttons, a touch pad or a touch screen.
[0091] It is understood that the memory 202 can be a volatile memory or a non-volatile memory, and can also include both volatile and non-volatile memories. The memory 202 in the embodiment of the present invention can store data to support the operation of the terminal (such as 10-1). Examples of these data include: any computer program for operating on the terminal (such as 10-1), such as an operating system and an application. Among them, the operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application can include various applications.
[0092] In some embodiments, the information recommendation device provided by the embodiment of the present invention may be implemented in a combination of software and hardware. As an example, the information recommendation device provided by the embodiment of the present invention may be a processor in the form of a hardware decoding processor, which is programmed to execute the training method of the video information recommendation model provided by the embodiment of the present invention. For example, the processor in the form of a hardware decoding processor may be one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs), or other electronic components.
[0093] As an example of an information information recommendation device provided by an embodiment of the present invention being implemented by a combination of software and hardware, the information information recommendation device provided by an embodiment of the present invention can be directly embodied as a combination of software modules executed by a processor 201, and the software module can be located in a storage medium, and the storage medium is located in the memory 202. The processor 201 reads the executable instructions included in the software module in the memory 202, and combines with necessary hardware (for example, including the processor 201 and other components connected to the bus 205) to complete the training method of the video information recommendation model provided by the embodiment of the present invention.
[0094] As an example, the processor 201 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where the general-purpose processor can be a microprocessor or any conventional processor, etc.
[0095] As an example of hardware implementation of the information recommendation device provided in the embodiment of the present invention, the device provided in the embodiment of the present invention can be directly executed by a processor 201 in the form of a hardware decoding processor, for example, one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs) or other electronic components to implement the training method of the video information recommendation model provided in the embodiment of the present invention.
[0096] The memory 202 in the embodiment of the present invention is used to store various types of data to support the operation of the information recommendation device. Examples of such data include: any executable instructions for operating on the information recommendation device, such as executable instructions, and the program for implementing the training method of the video information recommendation model in the embodiment of the present invention can be included in the executable instructions.
[0097] In other embodiments, the information recommendation device provided by the embodiments of the present invention may be implemented in software. Figure 2 The information recommendation device stored in the memory 202 is shown, which may be software in the form of a program and a plug-in, and includes a series of modules. As an example of a program stored in the memory 202, an information recommendation device may be included, and the information recommendation device includes the following software modules:
[0098] Information transmission module 2081 and information processing module 2082. When the software modules in the information recommendation device are read into the RAM by the processor 201 and executed, the training method of the video information recommendation model provided by the embodiment of the present invention will be implemented, wherein the functions of each software module in the information recommendation device include: an information transmission module, which is used to obtain the behavior parameter information of the target user in response to the information recommendation request;
[0099] The information processing module 2081 is used to determine information of different user clusters to which the target user belongs based on the behavior parameter information of the target user;
[0100] The information processing module 2082 is used to obtain candidate recommended information corresponding to each user cluster information in the information source based on the different user cluster information;
[0101] The information processing module 2082 is used to determine the information to be recommended that matches the target user based on the candidate recommended information corresponding to each user cluster information;
[0102] The information processing module 2082 is used to adjust the number of candidate recommended information corresponding to different user cluster information in the information to be recommended according to the behavior parameter information of the target user and through an information information recommendation model matching the target user, so as to match the information to be recommended with the target user.
[0103] Combination Figure 2 The information recommendation device shown illustrates the information recommendation method provided by the embodiment of the present invention, see Figure 3 , Figure 3 This is an optional flow chart of the information recommendation method provided by the embodiment of the present invention. It can be understood that: Figure 3 The steps shown can be executed by various electronic devices running the information recommendation device, for example, a dedicated terminal, a server or a server cluster with the information recommendation device, wherein the dedicated terminal with the information recommendation device can be the preceding sequence Figure 2 The electronic device with information recommendation device in the embodiment shown is as follows. Figure 3 The steps shown are explained.
[0104] Step 301: The information recommendation device receives an information recommendation request sent by a terminal.
[0105] Step 302: The information recommendation device obtains the behavior parameter information of the target user in response to the information recommendation request.
[0106] In some embodiments of the present invention, different program components can be used to collect various user behaviors matched by the corresponding client, and the original log of user behavior data can be effectively extracted, such as the user's device number (user account), information type, information browsing time, and information browsing completeness parameters. Among them, the user's historical click behavior and the corresponding information browsing time will be recorded through the subscription service and stored in Redis. When the user request arrives, the online recommendation system will pull the corresponding user's historical click behavior.
[0107] Step 303: The news information recommendation device determines information of different user clusters to which the target user belongs based on the behavior parameter information of the target user.
[0108] In some embodiments of the present invention, based on the behavior parameter information of the target user, determining the different user cluster information to which the target user belongs can be achieved in the following manner:
[0109] Obtain original data sets of different dimensions corresponding to the target user; determine user behavior parameter information matching the target user based on the original data sets of different dimensions; perform dynamic clustering rule classification processing on the user behavior parameter information to determine the identifiers of different user cluster information; determine the different user cluster information to which the target user belongs based on the identifiers of the different user cluster information. Among them, taking news information recommendation as an example, the personalized recommendation of news information is to recommend news information of interest to the target user based on the interest characteristics and reading behavior of the target user. Through the information information recommendation method provided in this application, the personalized recommendation system for news information usually includes four modules: user portrait module, recall module, CTR estimation module, and rearrangement module. After the recall module screens the roughly sorted news information, it outputs it to the click-through-rate (CTR) estimation module. The CTR estimation module performs fine sorting on these recalled news information, and then passes through the rearrangement module, and finally recommends it to the user. By determining the user behavior parameter information matching the target user through the original data sets of different dimensions, the user cluster information corresponding to the target user can be determined, wherein one target user can belong to multiple different user clusters. For example, the same target user can belong to a sports user cluster, an entertainment user cluster, and a social user cluster respectively, and further recommend matching information based on different user clusters.
[0110] Step 304: The information recommendation device obtains candidate recommended information corresponding to each user cluster information from the information source based on the different user cluster information.
[0111] In some embodiments of the present invention, based on the different user cluster information, obtaining the candidate recommended information corresponding to each user cluster information in the information source can be achieved in the following manner:
[0112] Based on the different user cluster information, determine the user cluster identifier corresponding to each user cluster; determine the content identifier of each piece of information in the information source; and obtain the candidate recommended information corresponding to each piece of user cluster information in the information source according to the matching result between the content identifier and the user cluster identifier. Among them, the different information stored in the information source carries the corresponding content identifier, for example, by classifying or labeling the video, to make relevant video recommendations. For example, the content identifiers in the information source can be divided into sports, entertainment, society, etc., so as to recommend the corresponding video according to the user cluster information of each user. A short video of a sports event or news information in the sports identifier can be pushed to the target user whose user cluster is the sports user cluster.
[0113] Step 305: The information recommendation device determines the information to be recommended that matches the target user based on the candidate recommended information corresponding to each user cluster information.
[0114] Step 306: The information recommendation device adjusts the number of candidate recommended information corresponding to different user cluster information in the information to be recommended according to the behavior parameter information of the target user and through an information recommendation model matching the target user.
[0115] In this way, the information to be recommended can be matched with the target user. Furthermore, when the information to be recommended is output to the target user, the operation information of the target user can be detected. Based on the detected operation information of the target user, the click-through rate and browsing time of the information to be recommended can be determined, so as to continuously adjust the information to be recommended corresponding to different user clusters determined by the information recommendation model during the process of the user using the news client or short video client.
[0116] In some embodiments of the present invention, according to the behavior parameter information of the target user, the number of candidate recommended information corresponding to different user cluster information in the information to be recommended is adjusted by an information information recommendation model matching the target user, which can be achieved by:
[0117] Determine a first joint feature vector and a second joint feature vector that match the behavior parameter information of the target user; process the first joint feature vector and the second joint feature vector through the information information recommendation model to determine the optional range of information information corresponding to different user clusters in the display interface of the target user, wherein the optional range of information information includes the number of information information corresponding to different user clusters; based on the information information recommendation model, determine the number of views of information information corresponding to different user clusters in the display interface of the target user; based on the number of views of information information corresponding to different user clusters, determine to adjust the number of candidate recommended information information corresponding to different user cluster information in the information to be recommended. Among them, the information information recommendation model matched with the target user provided in the present application can be a reinforcement learning model DDQN (Deep Q-Learning), so as to realize the determination of the corresponding information information to be recommended by combining the perception ability of deep learning (Deep Learning) and the decision-making ability of reinforcement learning (Reinforcemen Learning) through deep reinforcement learning.
[0118] Among them, reference Figure 4 , Figure 4This is an optional model structure diagram of the information recommendation model in the embodiment of the present invention. Reinforcement learning is to learn an optimal strategy that allows the agent to take actions according to the current state in a specific environment to obtain the maximum reward. Reinforcement learning can be simply achieved through<A,S,R,P> Four-tuple modeling. A represents Action, which is the action issued by the Agent; State is the state of the world that the Agent can perceive; Reward is a real value, representing a reward or punishment; P is the world that the Agent interacts with. In the embodiment of the present application, A is the recommended information, S is the current state, R is the recommended benefit value Reward of the recommended information, and P is the recommendation prediction model.
[0119] <A,S,R,P> The influence relationship between the four tuples is as follows: Action space: A, that is, all actions A constitute the action space. State space: S, that is, all states S constitute the state space. Reward: R: S×A×S'→R, that is, in the current state S, after executing action A, the current state becomes S', and the reward R corresponding to action A is obtained. Transition: P: S×A→S', that is, in the current state S, after executing action A, the current state becomes S,.
[0120] In the process of adjusting the number of candidate recommended information corresponding to different user cluster information in the recommended information through the information recommendation model, after obtaining the state and reward of the environment feedback, the corresponding action is performed; for the environment, after accepting the action performed by the subject, the state o(t+1) and reward r(t+1) of the environment feedback are output. Among them, action is the number of articles belonging to each cluster displayed per screen, reward is the number of pv / vv obtained as a result, the first joint feature vector old_state is the feature representation of the environment and the user before taking the action, new_state is the feature representation of the environment and the user after taking the action, is_terminal is whether it is the last refreshed news, and the information recommendation model adopts the multi-layer perceptron structure MLP. In the process of transforming the space, the first joint feature vector and the second joint feature vector can be converted into representations in the corresponding mapping space respectively through the multi-layer perceptron (MLP), wherein the output of the last fully connected layer in the information recommendation model is action and aciton score.
[0121] In some embodiments of the present invention, based on the number of views of the information information respectively corresponding to the different user clusters, determining to adjust the number of candidate recommended information information corresponding to the different user cluster information in the information to be recommended can be achieved in the following manner:
[0122] The browsing times of the information information corresponding to the different user clusters are cross-sorted; based on the cross-sorting results of the browsing times, the proportion parameters corresponding to the different user clusters are determined; the number of information to be displayed in the display interface of the target user is determined; based on the number of information to be displayed in the display interface of the target user and the proportion parameters corresponding to the user clusters, the number of candidate recommended information information corresponding to the different user cluster information in the information to be recommended is adjusted. In particular, since the size of the user's terminal interface is not unique, the number of recommended information information presented in the terminal interface also changes with the size of the terminal interface. With the page sliding instruction corresponding to the recommended information information display page, after the recommended information information display page slides, the displayed recommended information will be removed from the terminal interface, and the new recommended information will enter the terminal interface. When the target user belongs to different user clusters, the proportion parameters corresponding to the different user clusters are determined by the cross-sorting results of the number of browsing times; for example, when the target user belongs to the sports user cluster, the entertainment user cluster, and the social user cluster, the cross-sorting results are determined to be 10% for the sports user cluster, 70% for the entertainment user cluster, 10% for the social user cluster, and 10% for the economic user cluster. The number of information to be displayed in the display interface of the target user is 20 news information. When the information information recommendation method provided in the present application is not adopted, 5 news information corresponding to each type of user cluster can be displayed. However, through the information information recommendation method provided in the present application, the information related to the entertainment user cluster can be adjusted to 14, and the remaining user cluster information can be adjusted to 2. As a result, users can more efficiently obtain information that suits their interests.
[0123] Before deploying the trained information recommendation model, you need to train the information recommendation model. Figure 5 , Figure 5 This is an optional flow chart of the information recommendation method provided by the embodiment of the present invention. It can be understood that: Figure 5 The steps shown can be executed by various electronic devices running the information recommendation device, for example, a dedicated terminal, a server or a server cluster with the information recommendation device, wherein the dedicated terminal with the information recommendation device can be the preceding sequence Figure 2 The electronic device with information recommendation device in the embodiment shown is as follows. Figure 5 The steps shown are explained.
[0124] Step 501: Acquire a training sample set, wherein the training sample set includes a multi-dimensional state feature vector and the number of news information corresponding to different user clusters.
[0125] In some embodiments of the present invention, when the information recommendation model is applied to the video information recommendation process, a dynamic noise threshold matching the video information recommendation process is determined; and the first training sample set is denoised according to the dynamic noise threshold to form a second training sample set matching the dynamic noise threshold. For example, in the process of recommending different short videos to users in the short video process, the short video playback interface can be displayed in the corresponding APP, or can be triggered by the WeChat applet (the information recommendation model can be encapsulated in the corresponding APP after training or saved in the WeChat applet in the form of a plug-in). With the continuous development and increase of short video application products, the carrying capacity of video information is far greater than that of text information. Different types of short videos in the short video server can be continuously recommended to users through the corresponding application program. In this training process, in the use environment of triggering short video recommendation through the WeChat applet, the dynamic noise threshold matching the use environment of the information recommendation model needs to be smaller than the dynamic noise threshold of directly recommending short videos to users in the short video client.
[0126] In some embodiments of the present invention, when the information recommendation model is applied to the news information recommendation process, a fixed noise threshold corresponding to the news information recommendation process is determined, and the first training sample set is denoised according to the fixed noise threshold to form a second training sample set matching the fixed noise threshold. Wherein, when the information recommendation model is solidified in the corresponding hardware mechanism (such as a news reading terminal, an e-book terminal, a financial news terminal), and the use environment is to push different news information to the user through the news reading terminal or the e-book terminal, the fixed noise threshold corresponding to the fixed information recommendation model can effectively improve the training speed of the information recommendation model and reduce the waiting time of the user. Wherein, when the noise is fixed in the use environment, the training sample set can come from the historical data of the target user, and the historical recommended information browsing data can be the recommended information viewing behavior data generated when the recommended information was recommended to the target user, which can be extracted from the historical browsing log. Here, the historical recommended information browsing data can be all the historical recommended information browsing data; it can also take into account the timeliness of the behavior data, and only include the historical recommended information browsing data within a preset time period, such as the historical recommended information browsing data within a week. Taking the financial news terminal as an example, the corresponding financial news user cluster can be further subdivided into a local financial news user cluster, a stock financial news user cluster, and a futures financial news user cluster, and can be marked according to the user cluster classification set by the user.
[0127] Step 502: Process the training sample set through the information recommendation model to determine the initial parameters of the information recommendation model.
[0128] Step 503: In response to the initial parameters of the information recommendation model, different training samples in the training sample set are processed by the information recommendation model to determine the update parameters of the information recommendation model.
[0129] In some embodiments of the present invention, in response to the initial parameters of the information information recommendation model, different training samples in the training sample set are processed by the information information recommendation model to determine the update parameters of the information information recommendation model, which can be achieved in the following manner:
[0130] Substitute different training samples in the training sample set into the loss function corresponding to the multilayer perceptron network of the information information recommendation model; determine the parameters of the multilayer perceptron network in the information information recommendation model corresponding to the convergence condition when the loss function meets the convergence condition as the update parameters of the information information recommendation model.
[0131] Step 504: According to the update parameters of the information information recommendation model, the parameters of the information information recommendation model are iteratively updated through the training sample set, so that the information information recommendation model can determine the browsing times of the information information corresponding to different user clusters.
[0132] The following takes the video news information recommendation scenario in the short video playback interface as an example to illustrate the training method of the information recommendation model provided by the embodiment of the present invention, wherein: Figure 6 FIG. 1 is a schematic diagram of an application environment of a training method for an information recommendation model according to an embodiment of the present invention, wherein Figure 6 As shown, the video news information playback interface can be displayed in the corresponding APP, or it can be triggered by the WeChat applet (the information recommendation model can be encapsulated in the corresponding APP after training or saved in the WeChat applet in the form of a plug-in). With the continuous development and increase of short video application products, the carrying capacity of video news information is far greater than that of text information. Video news information can be continuously recommended to users through the corresponding application. For example, the "Take a look" entrance included in the discovery page of the WeChat application, or the audio recommendation entrance of the audio application, or the video recommendation entrance of the video application, or the live broadcast recommendation entrance of the live broadcast application, etc. When the target terminal runs the target application according to the user operation and controls the target application to display the application page including the trigger entrance for triggering the opening of the recommended content display page, the trigger operation of the trigger entrance can be detected. When the trigger operation corresponding to the trigger entrance is generated, a recommendation request is sent to the server, and after receiving the recommended content fed back by the server in response to the recommendation request, the recommended content is displayed in the recommended content display page in the recommended order.
[0133] refer to Figure 7 and Figure 8 , Figure 7 The information recommendation method provided in the embodiment of the present invention is an optional workflow. In the information recommendation method provided in the present application, personalized news recommendation can be divided into two stages: recall and sorting. The two stages perform their respective duties and complete different tasks, and their respective focuses are also different. Specifically: In the recall stage, the main task is to filter out important content. The focus of the work is how to quickly and effectively extract content that a large number of users may be interested in from a large amount of news. The difficulty lies in the matching problem of massive data (directly facing a large amount of news and a large number of users). The complexity of the calculation is an issue that must be considered. The focus of the sorting side is to comprehensively and accurately estimate the degree of interest of a large number of users in a small amount of recalled news, that is, CTR. Figure 8The information recommendation method provided in the embodiment of the present invention provides an optional workflow. In personalized news recommendation, personalized news recall must first be implemented. Specifically, in personalized news recommendation recall, it is usually necessary to implement the recall of news that best suits the interests of each user. However, in scenarios with a large number of users, clustering can be used to cluster users with the same or similar interests into a cluster, and then recommend the same news to people in this cluster. Due to the variability of user interests, a user may belong to multiple clusters at the same time. However, since it is not known in advance which cluster the user is more interested in, when calculating the user's interest in multiple clusters, the overall interest characteristics of the user are calculated in an equal-weighted average manner. This will cause the following problems:
[0134] 1) If the recommended content is too average, the proportion of the recalled content of the cluster that the user is really interested in may be too small, and it is difficult to be ranked in the front, so that the user misses the content that he is really interested in, which is not conducive to user experience. 2) If the recommended content is too biased towards the recall of a certain cluster, the content recommended to the user may become narrower and narrower, and the content diversity is poor, so that the user needs to re-search for video news information in related fields, which increases the complexity of the user's operation.
[0135] To solve the above defects, refer to Fig. 9 , Fig. 9 The following is a schematic diagram of the working process of the information recommendation method provided by an embodiment of the present invention, wherein: Fig.10 The training process diagram of the information recommendation method provided by the embodiment of the present invention is shown below. Fig. 9 The data structure diagram of the information recommendation method shown in the figure illustrates the working process of the information recommendation model provided by the present invention, which specifically includes the following steps:
[0136] Step 901: Obtain a video news information training set and input corresponding training parameters.
[0137] The input training parameters include: the number of iterations T, the state feature dimension n, the action set A, the step size α, the decay factor γ, the exploration rate ε, the network Q of the current information recommendation model, the network Q′ of the target information recommendation model, the number of samples of batch gradient descent m, and the network parameter update frequency C of the target information recommendation model. The output parameters are the network parameters of the information recommendation model
[0138] Step 902: Train the information recommendation model to determine the parameters of the information recommendation model.
[0139] Among them, reference Fig.10During the training process, first randomly initialize the values Q corresponding to all states and actions. Randomly initialize all parameters w of the network of the current information recommendation model, and initialize the parameters w′=w of the network Q′ of the target information recommendation model. Clear the set D of experience playback.
[0140] Then execute for i from 1 to T to iterate, which specifically includes the following steps:.
[0141] a) Initialize S as the first state of the current state sequence and obtain the corresponding feature vector φ(S).
[0142] b) Use φ(S) as input in the network of the information recommendation model to obtain the Q value output corresponding to all actions of the network of the information recommendation model. Use the ∈-greedy algorithm to select the corresponding action A in the current Q value output.
[0143] c) Execute the current action A in state S, obtain the feature vector φ(S′) and reward φ(S′) and reward R(S) corresponding to the new state S′, and whether the state is terminated is_end
[0144] d) Store the five-tuple {φ(S), A, R, φ(S′), is_end} into the experience replay set D
[0145] e) S = S′
[0146] f) Sample m samples {φ(S j ), Aj, Rj, φ(S′ j )), is_endj}, j = 1, 2, ... m, calculate the current target Q value yj by formula 1,
[0147]
[0148] g) Use mean square error loss function Update all the parameters w of the network of the information recommendation model through the gradient back propagation of the neural network
[0149] h) If T%C=1, update the network parameter w′=w of the target information recommendation model
[0150] i) If S′ is in the terminal state, the current round of iteration is completed, otherwise go to step b)
[0151] The Q value calculations in steps f and g in the above steps also need to be obtained through network calculations of the information recommendation model.
[0152] Furthermore, in order to facilitate function convergence, the exploration rate ∈ can become smaller as the iteration proceeds, thereby improving the training speed of the model.
[0153] Step 903: Deploy the trained news information recommendation model.
[0154] Specifically, the trained information recommendation model can be deployed in the server, encapsulated in the corresponding APP, or saved in the WeChat applet in the form of a plug-in.
[0155] Step 904: adjusting the quantity of information corresponding to different user cluster information in the recommended information through the information recommendation model.
[0156] refer to Fig.11 , Fig.11 The present invention is a schematic diagram of the display effect of the information recommendation method in the embodiment of the present invention, wherein the size of the user's terminal interface is not unique, and the number of recommended information presented in the terminal interface also changes with the size of the terminal interface. The information to be recommended can be matched with the target user, in accordance with the different user clusters to which the user belongs, while enhancing the accuracy and timeliness of the information recommendation, effectively improving the quality of the information recommendation, and enhancing the user experience.
[0157] Step 905: Output the video news information after the quantity is adjusted.
[0158] Beneficial technical effects:
[0159] 1) Compared with the conventional technology, the present application obtains the behavior parameter information of the target user in response to the information recommendation request; determines the different user cluster information to which the target user belongs based on the behavior parameter information of the target user; obtains the candidate recommended information information corresponding to each user cluster information in the information source based on the different user cluster information; determines the information to be recommended that matches the target user based on the candidate recommended information information corresponding to each user cluster information; and adjusts the number of candidate recommended information information corresponding to different user cluster information in the information to be recommended according to the behavior parameter information of the target user through the information recommendation model that matches the target user, thereby matching the information to be recommended with the target user, and enhancing the accuracy and timeliness of the information recommendation, effectively improving the quality of the information recommendation, and further overcoming the defect that the conventional technology simply uses the relevance of the information content for average recommendation and cannot take into account the user's feelings, thereby improving the user experience in different recommendation scenarios.
[0160] The above descriptions are merely embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for recommending information, characterized in that: The method comprises: In response to the information recommendation request, obtaining the behavior parameter information of the target user; Based on the behavior parameter information of the target user, determining information of different user clusters to which the target user belongs; Based on the different user cluster information, obtaining candidate recommended information corresponding to each user cluster information in the information source; Based on the candidate recommended information information corresponding to each user cluster information, determine the information to be recommended that matches the user cluster information to which the target user belongs, wherein the information to be recommended is used to recommend to the user corresponding to the user cluster information; According to the behavior parameter information of the target user, through an information information recommendation model matching the target user, based on the number of views of the information corresponding to different user clusters in the display interface of the target user, the number of candidate recommended information corresponding to different user cluster information in the information to be recommended is adjusted to achieve matching of the information to be recommended with the target user.
2. The method according to claim 1, characterized in that The determining, based on the behavior parameter information of the target user, information of different user clusters to which the target user belongs includes: Acquire original data sets of different dimensions corresponding to the target user; Determining user behavior parameter information matching the target user according to the original data sets of different dimensions; Performing dynamic clustering rule classification processing on the user behavior parameter information to determine the identifiers of different user cluster information; Based on the identifiers of the different user cluster information, the different user cluster information to which the target user belongs is determined.
3. The method according to claim 1, characterized in that The acquiring, based on the different user cluster information, candidate recommended information corresponding to each user cluster information in the information source includes: Based on the different user cluster information, determining the user cluster identifier corresponding to each user cluster; Determining the content identifier of each piece of information in the information source; According to the matching result between the content identifier and the user cluster identifier, candidate recommended information corresponding to each user cluster information is obtained from the information source.
4. The method according to claim 1, characterized in that The step of adjusting the number of candidate recommended information corresponding to different user cluster information in the information to be recommended based on the behavior parameter information of the target user, using an information recommendation model matching the target user, and based on the number of views of the information respectively corresponding to different user clusters in the display interface of the target user, includes: Determine a first joint feature vector and a second joint feature vector that match the behavior parameter information of the target user; The first joint feature vector and the second joint feature vector are processed by the information information recommendation model to determine the selectable range of information corresponding to different user clusters in the display interface of the target user, wherein the selectable range of information information includes the number of information corresponding to different user clusters; Based on the information recommendation model, determining the number of views of the information respectively corresponding to different user clusters in the display interface of the target user; Based on the number of views of the information information respectively corresponding to the different user clusters, it is determined to adjust the number of candidate recommended information information corresponding to the different user cluster information in the information to be recommended.
5. The method according to claim 4, characterized in that The determining, based on the number of views of the information information respectively corresponding to the different user clusters, to adjust the number of candidate recommended information information corresponding to the different user cluster information in the information to be recommended includes: Cross-ranking the browsing times of the information information respectively corresponding to the different user clusters; Based on the cross-ranking results of the browsing times, determining the proportion parameters corresponding to different user clusters; Determine the amount of information to be displayed in the display interface of the target user; Based on the amount of information to be displayed in the display interface of the target user and the ratio parameters corresponding to the user clusters respectively, the amount of candidate recommended information corresponding to different user cluster information in the information to be recommended is adjusted.
6. The method according to claim 1, characterized in that The method further comprises: Acquire a training sample set, wherein the training sample set includes a multi-dimensional state feature vector and the amount of information corresponding to different user clusters; Processing the training sample set through an information recommendation model to determine initial parameters of the information recommendation model; In response to the initial parameters of the information recommendation model, different training samples in the training sample set are processed by the information recommendation model to determine the update parameters of the information recommendation model; According to the update parameters of the information information recommendation model, the parameters of the information information recommendation model are iteratively updated through the training sample set, so that the information information recommendation model can determine the browsing times of the information corresponding to different user clusters.
7. The method according to claim 6, characterized in that The step of processing different training samples in the training sample set by the information information recommendation model in response to the initial parameters of the information information recommendation model to determine the update parameters of the information information recommendation model includes: Substituting different training samples in the training sample set into the loss function corresponding to the multilayer perceptron network of the information recommendation model; Determine that when the loss function satisfies the convergence condition, the parameters of the multilayer perceptron network in the information information recommendation model corresponding to the information information recommendation model are used as update parameters of the information information recommendation model.
8. The method according to claim 6, characterized in that The method further comprises: When the news information recommendation model is applied to a video information recommendation process, determining a dynamic noise threshold that matches the video information recommendation process; The first training sample set is denoised according to the dynamic noise threshold to form a second training sample set matching the dynamic noise threshold.
9. The method according to claim 6, characterized in that The method further comprises: When the news information recommendation model is applied to the news information recommendation process, a fixed noise threshold corresponding to the news information recommendation process is determined, and the first training sample set is denoised according to the fixed noise threshold to form a second training sample set that matches the fixed noise threshold.
10. The method according to claim 1, characterized in that The method further comprises: When outputting the information to be recommended to the target user, detecting operation information of the target user; Based on the detected operation information of the target user, the click rate and browsing time of the information to be recommended are determined.
11. An information recommendation device, characterized in that: The device comprises: An information transmission module, used to obtain the behavior parameter information of the target user in response to the information recommendation request; An information processing module, used to determine information of different user clusters to which the target user belongs based on the behavior parameter information of the target user; The information processing module is used to obtain candidate recommended information corresponding to each user cluster information in the information source based on the different user cluster information; The information processing module is used to determine the information to be recommended that matches the user cluster information to which the target user belongs based on the candidate recommended information corresponding to each user cluster information, wherein the information to be recommended is used to recommend to the user corresponding to the user cluster information; The information processing module is used to adjust the number of candidate recommended information corresponding to different user cluster information in the information to be recommended based on the behavior parameter information of the target user, through an information information recommendation model matching the target user, and based on the number of views of the information corresponding to different user clusters in the display interface of the target user, so as to match the information to be recommended with the target user.
12. The device according to claim 11, characterized in that The information processing module is used to obtain original data sets of different dimensions corresponding to the target user; The information processing module is used to determine the user behavior parameter information matching the target user according to the original data sets of different dimensions; The information processing module is used to perform dynamic clustering rule classification processing on the user behavior parameter information to determine the identifiers of different user cluster information; The information processing module is used to determine the different user cluster information to which the target user belongs based on the identifiers of the different user cluster information.
13. An electronic device, characterized in that: The electronic device comprises: A memory for storing executable instructions; A processor, configured to implement the information recommendation method according to any one of claims 1 to 10 when running the executable instructions stored in the memory.
14. A computer-readable storage medium storing executable instructions, characterized in that: When the executable instructions are executed by the processor, the information recommendation method according to any one of claims 1 to 10 is implemented.
