User interest mining method and system, electronic device, and medium

Abstract

Embodiments of the present disclosure relate to a user interest mining method and system, an electronic device, and a medium. The method comprises: splitting a behavior sequence of a user; extracting a post sequence of an object user in a social relationship sequence of the user; extracting four sequences of post ID, post category, post time, and post author ID for each of the split sequence and the extracted post sequence; inputting the four sequences of post ID, post category, post time, and post author ID extracted for each sequence into a first deep neural network to obtain a group of vectors A for each sequence; performing an attention mechanism operation on the group of vectors A for each sequence by taking the current training post as input to obtain an attentioned vector A'; inputting the vector A' into a training model to obtain an estimated score for the current training post; and adjusting the training model according to the estimated score, and taking the adjusted training model as a user interest mining model.

Description

Technical Field

[0001] This disclosure relates to the fields of machine learning and personalized recommendation, and more specifically, to methods, systems, electronic devices, and media for mining user interests. Background Technology

[0002] In the recommendation field, extracting interests based on user behavior sequences is a common method. Alibaba's Click-Through Rate (CTR) prediction algorithm plays a crucial role in recommendation systems. Among them, Deep Interest Network (DIN) and Deep Interest Evolution Network (DIEN) are commonly used CTR algorithms in recommendation systems.

[0003] In the field of recommender systems, it's not enough to simply recommend things users are interested in; the goal is to maximize purchases. Therefore, it's necessary to optimize both CTR and Conversion Rate (CVR) models simultaneously. Alibaba proposed the ESMM model, offering a solution that optimizes both pCTR and pCTCVR. However, ESMM is a typical shared bottom model, meaning the two models share some parameters. A significant problem with multi-task learning is its poor performance when the two sub-tasks are very different. Google, building on the MoE model, proposed the Multi-gate Mixture of Experts (MMoE) model, which can simultaneously capture both the correlations and differences between sub-tasks.

[0004] However, current DIN-like methods can only handle single sequences, and the capture of short-term and long-term interests is unstable, failing to meet the needs of current business scenarios. This invention proposes to expand the information of sequences based on different types of user behavior, and to model short-term and long-term sequences separately, enabling the acquisition of stable short-term and long-term interests. Summary of the Invention

[0005] This summary section is provided to briefly introduce the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0006] According to some embodiments of this disclosure, a user interest mining method is provided, comprising: splitting a user's behavior sequence; extracting the posting sequence of object users from the user's social relationship sequence; extracting four sequences—post ID, post category, posting time, and post author ID—for each of the split sequences and the extracted posting sequences; inputting the four sequences—post ID, post category, posting time, and post author ID—extracted for each sequence into a first deep neural network to obtain a set of vectors A for each sequence; using the currently trained post as input, performing an attention mechanism operation on the set of vectors A for each sequence to obtain an attention-adjusted vector A'; inputting vector A' into a training model to obtain a predicted score for the currently trained post; and adjusting the training model according to the predicted score, and using the adjusted training model as a user interest mining model.

[0007] According to some embodiments of this disclosure, a user interest mining system is provided, comprising: a splitting unit configured to split a user's behavior sequence; a first extraction unit configured to extract posting sequences of target users from the user's social relationship sequence; a second extraction unit configured to extract four sequences—post ID, post category, posting time, and post author ID—for each of the split sequences and the extracted posting sequences; a vector merging unit configured to input the four sequences—post ID, post category, posting time, and post author ID—extracted for each sequence into a first deep neural network to obtain a set of vectors A for each sequence; an attention mechanism unit configured to take the currently trained post as input and perform an attention mechanism operation on the set of vectors A for each sequence to obtain an attended vector A'; a prediction scoring unit configured to input vector A' into a training model to obtain a predicted score for the currently trained post; and an adjustment unit configured to adjust the training model according to the predicted score and use the adjusted training model as a user interest mining model.

[0008] According to some embodiments of the present disclosure, an electronic device is provided, including: a memory; and a processor coupled to the memory, the processor being configured to perform the methods of any of the embodiments described in the present disclosure based on instructions stored in the memory.

[0009] According to some embodiments of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, performs the methods of any of the embodiments described in the present disclosure.

[0010] According to some embodiments of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, performs the methods of any of the embodiments described in the present disclosure.

[0011] Other features, aspects, and advantages of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0012] Preferred embodiments of the present disclosure are described below with reference to the accompanying drawings. The accompanying drawings, which are included to provide a further understanding of the present disclosure, and which, together with the following detailed description, are incorporated in and form a part of this specification and are used to explain the present disclosure. It should be understood that the drawings described below only relate to some embodiments of the present disclosure and are not intended to limit the present disclosure. In the drawings:

[0013] Figure 1 A schematic flowchart of a training method for a user interest mining system according to an exemplary embodiment of the present disclosure is shown.

[0014] Figure 2 A schematic block diagram of a training method for a user interest mining system according to an exemplary embodiment of the present disclosure is shown.

[0015] Figure 3 A schematic block diagram of a user interest mining system according to an exemplary embodiment of the present disclosure is shown.

[0016] Figure 4 A schematic block diagram of an electronic device according to an exemplary embodiment of the present disclosure is shown.

[0017] Figure 5 A block diagram illustrating an example structure of a computer system that may be employed in an exemplary embodiment of the present disclosure is shown.

[0018] It should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not necessarily drawn to actual scale. The same or similar reference numerals are used in the various drawings to denote the same or similar parts. Therefore, once an item is defined in one drawing, it may not be discussed further in subsequent drawings. Detailed Implementation

[0019] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. However, it is obvious that the described embodiments are only some embodiments of this disclosure, and not all embodiments. The following description of the embodiments is merely illustrative and is in no way intended to limit this disclosure or its application or use. It should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein.

[0020] It should be understood that the various steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect. Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of components and steps set forth in these embodiments should be interpreted as merely exemplary and do not limit the scope of this disclosure.

[0021] As used in this disclosure, the term "comprising" and its variations are open-ended terms that include at least the following elements / features but do not exclude other elements / features, i.e., "including but not limited to". Furthermore, as used in this disclosure, the term "including" and its variations are open-ended terms that include at least the following elements / features but do not exclude other elements / features, i.e., "including but not limited to". Therefore, "comprising" and "including" are synonymous. The term "based on" means "at least partially based on".

[0022] Throughout this specification, the terms "one embodiment," "some embodiments," or "embodiment" mean that a specific feature, structure, or characteristic described in connection with an embodiment is included in at least one embodiment of this disclosure. For example, the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; and the term "some embodiments" means "at least some embodiments." Furthermore, the appearance of the phrases "in one embodiment," "in some embodiments," or "in an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment, but may refer to the same embodiment.

[0023] It should be noted that the concepts of "first," "second," etc., used in this disclosure are used only to distinguish different devices, modules, or units, and are not intended to define the order of functions performed by these devices, modules, or units or their interdependencies. Unless otherwise specified, the concepts of "first," "second," etc., are not intended to imply that the objects described herein must be in a given temporal, spatial, rank, or any other given order.

[0024] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0025] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0026] The embodiments of this disclosure are described in detail below with reference to the accompanying drawings; however, this disclosure is not limited to these specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. Furthermore, in one or more embodiments, specific features, structures, or characteristics can be combined in any suitable manner that will be apparent to those skilled in the art from this disclosure.

[0027] Existing DIN-like methods typically process individual sequences within a user's behavior sequence. This invention proposes two novel approaches. One approach involves splitting various user behavior sequences into sequences including, but not limited to, clicks, likes, follows, and comments. Simultaneously, behaviors occurring within a predetermined timeframe (e.g., 1 hour) are labeled as short-term behaviors, while other behaviors are labeled as long-term behaviors. For each individual sequence, four sequences are extracted: post ID, post category, post time, and post author ID. These four sequences are then processed through a fully connected layer (i.e., the input weights are dynamically adjusted using a DNN) and merged into a single vector parameter.

[0028] Another approach is to incorporate users' social relationship sequences, including but not limited to private chat and following sequences. A posting sequence is formed by extracting posting data from each user's private chat and following sequences. Similarly, actions posted within a predetermined time (e.g., 1 hour) are labeled as short-term actions, while other actions are labeled as long-term actions. For each posting sequence, four sequences are extracted: post ID, post category, post posting time, and post author ID. These four sequences are then passed through a fully connected layer (i.e., the input DNN dynamically adjusts the weights) and merged into a single vector parameter.

[0029] The vector parameters of the above sequence are extracted using an attention mechanism similar to DIN to obtain the attention-enhanced vector parameters. These attention-enhanced vector parameters are then input into a training model composed of MMOE and DNN for training, resulting in the model's predicted score.

[0030] Furthermore, this invention proposes to process short-term and long-term interests separately within the sequence and distinguish between social and content interests. Specifically, parameters belonging to short-term behavior and parameters belonging to long-term behavior are extracted from the merged vector parameters. Then, it is determined whether the post corresponding to the parameter is a social post or a content post. If it is a social post, a fully connected layer (i.e., DNN) for social interest is applied to obtain short-term or long-term social interest vector parameters; if it is a content post, a fully connected layer (i.e., DNN) for content interest is applied to obtain short-term or long-term content interest vector parameters. The attention-adjusted vector parameters and the short-term / long-term social / content interest vector parameters are then input into a training model composed of MMOE and DNN for training to obtain the model's predicted score.

[0031] The following is for reference. Figures 1 to 4 This paper details a user interest mining system and a training method for the user interest mining system according to exemplary embodiments of the present invention.

[0032] Figure 1 A flowchart 100 is shown of a training method (also referred to as a user interest mining method) for a user interest mining system according to an exemplary embodiment of the present invention.

[0033] like Figure 1 As shown, in step S101, the user's behavior sequence is segmented. It should be understood that the user's behavior sequence can be segmented into sequences including, but not limited to, clicks, likes, follows, and comments. According to a preferred embodiment of the present invention, the user's behavior sequence is segmented into click, like, follow, and / or comment sequences.

[0034] In step S102, the posting sequence of the target user in the user's social relationship sequence is extracted. According to an embodiment of the present invention, the social relationship sequence includes, but is not limited to, user private chat sequence and user following sequence.

[0035] In step S103, for each sequence in the split sequence and the extracted posting sequence, four sequences are extracted: post ID, post category, posting time, and post author ID. According to a preferred embodiment of the present invention, when the user's behavior sequence is split into click, like, follow, and / or comment sequences, four sequences—post ID, post category, posting time, and post author ID—are extracted for each of these sequences.

[0036] In step S104, the four sequences extracted for each sequence—post ID, post category, posting time, and post author ID—are input into the first deep neural network to obtain a set of vector parameters A for each sequence. The deep neural network is a known network and will not be described in detail here. It should be understood that the four sequences extracted for each sequence—post ID, post category, posting time, and post author ID—can be processed through the same deep neural network or different deep neural networks.

[0037] In step S105, the currently trained post is used as input, and an attention mechanism operation is performed on a set of vector parameters A for each sequence to obtain the attention-adjusted vector parameters A'.

[0038] In step S106, the vector parameter A' is input into the training model to obtain the predicted score for the currently trained post.

[0039] According to an embodiment of the present invention, the training model consists of a multi-expert gating network and a second deep neural network. Therefore, step S106, inputting the vector parameter A' into the training model to obtain the estimated score for the currently trained post, may include inputting the vector A' into the multi-expert gating network to obtain an array, and inputting the array into the second deep neural network to obtain the estimated score. It should be understood that the working principles of the multi-expert gating network and the deep neural network are known in the art and will not be elaborated here.

[0040] In step S107, the training model is adjusted according to the estimated score, and the adjusted training model is used as the user interest mining model. It should be understood that adjusting the parameters of the training model according to the estimated score, for example through a backpropagation algorithm, to make the subsequent estimated scores more accurate is known in the art and will not be elaborated here.

[0041] Although not shown in the figure, the user interest mining method according to the present invention also includes processing of short-term and long-term interests. According to embodiments of the present invention, Figure 1 The method flowchart 100 also includes step S108 (not shown), which marks the behavior within a predetermined time in each of the split sequences as short-term behavior and other behaviors as long-term behavior, and marks the behavior of the posting time within the predetermined time in the posting sequence of the object user in the social relationship sequence as short-term behavior and other behaviors as long-term behavior.

[0042] According to an embodiment of the present invention, Figure 1 The method flowchart 100 also includes step S109 (not shown), which extracts vector parameters B and C for short-term behavior and long-term behavior respectively from a set of vector parameters A for each sequence.

[0043] According to an embodiment of the present invention, Figure 1The method flowchart 100 also includes step S110 (not shown), which determines whether the post in vector B is a social post or a content post. If it is a social post, it is connected to a deep neural network of social interests to obtain a short-term social interest vector B1. If it is a content post, it is connected to a deep neural network of content interests to obtain a short-term content interest vector B2.

[0044] According to an embodiment of the present invention, Figure 1 The method flowchart 100 also includes step S111 (not shown), which determines whether the post in vector C is a social post or a content post. If it is a social post, it is connected to a deep neural network of social interests to obtain a long-term social interest vector C1. If it is a content post, it is connected to a deep neural network of content interests to obtain a long-term content interest vector C2.

[0045] According to an embodiment of the present invention, Figure 1 The method flowchart 100 also includes step S112 (not shown), which inputs vectors A', B1, B2, C1 and C2 into the training model to obtain the predicted score for the currently trained post.

[0046] According to an embodiment of the present invention, Figure 1 The method flowchart 100 also includes step S113 (not shown), adjusting the training model according to the estimated score, and using the adjusted training model as the user interest mining model.

[0047] refer to Figure 2 This illustrates a schematic block diagram of a user interest mining method according to an exemplary embodiment of the present invention. Figure 2As shown, the user's behavior sequence is broken down into user click sequence, user like sequence, user follow sequence, and user comment sequence. Each sequence is further extracted with four sequences: post ID, post category, post time, and post author. Simultaneously, the posting data of the target users in the user private chat sequence and the user follow sequence are extracted, and four sequences (post ID, post category, post time, and post author) are extracted from this posting data. The four sequences extracted from each sequence are each input into a deep neural network (DNN) to obtain a set of vector parameters A. All vector parameters A, along with the currently trained post pid embedding (post ID vector representation), undergo an attention mechanism to obtain vector parameters A'. Simultaneously, vector parameters B and C, representing short-term and long-term behaviors respectively, are extracted from vector parameters A. Furthermore, by connecting to DNNs for social interests (not shown) and content interests (DNNs), short-term social interest B1, long-term social interest C1, short-term content interest B2, and long-term content interest C2 are obtained respectively. Then, A', B1, C1, B2, and C2 are input together into an MMOE structure and a deep neural network (DNN) to train a model that obtains a predicted score for the currently trained post pid emb. Finally, based on the difference between the predicted score and the label, the training model is adjusted using a backpropagation algorithm to further improve it.

[0048] The user interest mining training method proposed in this disclosure mines information from multiple user sequences, fuses sequences of multiple behaviors and multiple dimensions, and extracts long-term and short-term interest sequences. It also integrates user interest extraction based on social relationship (following and private chat) sequences, which helps the training model to predict user preferences more accurately. This can improve the user's dwell time in the recommendation square, and to a certain extent ensure that short-term user interests are responded to in a timely manner and long-term interests are not lost.

[0049] Figure 3 A block diagram of a user interest mining system 300 according to an exemplary embodiment of the present disclosure is shown.

[0050] like Figure 3 As shown, the user interest mining system 300 includes a segmentation unit 31 configured to segment user behavior sequences. It should be understood that user behavior sequences can be segmented into sequences including, but not limited to, clicks, likes, follows, and comments. According to a preferred embodiment of the present invention, user behavior sequences are segmented into click, like, follow, and / or comment sequences.

[0051] like Figure 3 As shown, the user interest mining system 300 includes a first extraction unit 32, configured to extract the posting sequence of target users from the user's social relationship sequence. According to an embodiment of the present invention, the social relationship sequence includes, but is not limited to, user private chat sequences and user following sequences.

[0052] like Figure 3 As shown, the user interest mining system 300 includes a second extraction unit 33, configured to extract four sequences—post ID, post category, posting time, and post author ID—for each sequence in the split sequence and the extracted posting sequence. According to a preferred embodiment of the present invention, when the splitting unit 31 is configured to split the user's behavior sequence into click, like, follow, and / or comment sequences, the second extraction unit 33 is configured to extract four sequences—post ID, post category, posting time, and post author ID—for each sequence in the click, like, follow, comment, and / or posting sequence.

[0053] like Figure 3 As shown, the user interest mining system 300 includes a vector merging unit 34, configured to input the four sequences extracted for each sequence—post ID, post category, posting time, and post author ID—into a first deep neural network to obtain a set of vectors A for each sequence. Here, the deep neural network is a known network and will not be described in detail. It should be understood that the four sequences—post ID, post category, posting time, and post author ID—extracted for each sequence can be processed through the same deep neural network or different deep neural networks.

[0054] like Figure 3 As shown, the user interest mining system 300 includes an attention mechanism unit 35, which is configured to take the currently trained posts as input and perform attention mechanism operations on a set of vectors A for each sequence to obtain the attention vector A'.

[0055] like Figure 3 As shown, the user interest mining system 300 includes a prediction scoring unit 36, configured to input vector A' into the training model to obtain a prediction score for the currently trained post.

[0056] It should be understood that the prediction training model consists of a multi-expert gating network and a second deep neural network. For example, according to one embodiment of the present invention, the prediction scoring unit 36 ​​can be further configured to input vector A' into the multi-expert gating network to obtain an array, and input the array into the second deep neural network to obtain the predicted score. It should be understood that the working principles of the multi-expert gating network and the deep neural network are known in the art and will not be described in detail here.

[0057] like Figure 3As shown, the user interest mining system 300 includes an adjustment unit 37, configured to adjust the training model according to the estimated score, and use the adjusted training model as the user interest mining model. It should be understood that adjusting the parameters of the training model according to the estimated score, for example through a backpropagation algorithm, to make subsequent estimated scores more accurate is known in the art and will not be elaborated here.

[0058] Although not shown, the user interest mining system 300 also includes a labeling unit 38 (not shown), configured to label behaviors within a predetermined time period in each split sequence as short-term behaviors and other behaviors as long-term behaviors, and to label behaviors within a predetermined time period in the posting sequence of the target user in the social relationship sequence as short-term behaviors and other behaviors as long-term behaviors.

[0059] The user interest mining system 300 also includes a third extraction unit 39 (not shown), configured to extract vectors B and C for short-term behavior and long-term behavior respectively from a set of vectors A for each sequence.

[0060] The user interest mining system 300 also includes a judgment and processing unit 40 (not shown), configured to: determine whether the post in vector B is a social post or a content post; if it is a social post, it is connected to the third deep neural network of social interests to obtain a short-term social interest vector B1; if it is a content post, it is connected to the fourth deep neural network of content interests to obtain a short-term content interest vector B2; and determine whether the post in vector C is a social post or a content post; if it is a social post, it is connected to the third deep neural network of social interests to obtain a long-term social interest vector C1; if it is a content post, it is connected to the fourth deep neural network of content interests to obtain a long-term content interest vector C2.

[0061] According to an embodiment of the present invention, the estimated scoring unit 36 ​​can also be configured to input vectors A', B1, B2, C1 and C2 into the training model to obtain an estimated score for the currently trained post.

[0062] According to an embodiment of the present invention, the adjustment unit 37 may also be configured to adjust the training model according to the estimated score, and use the adjusted training model as a user interest mining model.

[0063] The user interest mining system provided in the embodiments of this disclosure can implement the user interest mining training method provided in the above embodiments. Its implementation principle and technical effect are similar, and will not be described again here.

[0064] The user interest mining system disclosed in this application can serve various scenarios, such as news, short videos, and advertisements, to improve the user's dwell time in the recommendation square, and to a certain extent ensure that users' short-term interests are responded to in a timely manner and that long-term interests are not lost.

[0065] Some embodiments of this disclosure also provide an electronic device. Figure 4 Block diagrams of some embodiments of the electronic device 4 of this disclosure are shown. This electronic device can be used to implement the methods described according to any embodiment of this disclosure.

[0066] For example, in some embodiments, the electronic device 4 can be various types of devices, including but not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. For example, the electronic device 4 may include a display panel for displaying data and / or execution results utilized according to the scheme of this disclosure. For example, the display panel can be of various shapes, such as a rectangular panel, an elliptical panel, or a polygonal panel. Furthermore, the display panel can be not only a planar panel but also a curved panel, or even a spherical panel.

[0067] like Figure 4 As shown, the electronic device 4 of this embodiment includes a memory 41 and a processor 42 coupled to the memory 41. It should be noted that... Figure 4 The components of the electronic device 4 shown are merely exemplary and not limiting; the electronic device 4 may have other components as needed for the actual application. The processor 42 can control other components in the electronic device 4 to perform the desired functions.

[0068] In some embodiments, memory 41 is used to store one or more computer-readable instructions. When processor 42 executes the computer-readable instructions, the computer-readable instructions are executed by processor 42 to implement the method according to any of the above embodiments. For specific implementations and related explanations of the various steps of the method, please refer to the above embodiments; repeated details will not be elaborated here.

[0069] For example, processor 42 and memory 41 can communicate with each other directly or indirectly. For example, processor 42 and memory 41 can communicate via a network. The network can include wireless networks, wired networks, and / or any combination of wireless and wired networks. Processor 42 and memory 41 can also communicate with each other via a system bus, which is not limited in this disclosure.

[0070] For example, processor 42 can be embodied in various suitable processors, processing devices, such as central processing unit (CPU), graphics processing unit (GPU), network processor (NP), etc.; it can also be digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The central processing unit (CPU) can be an x86 or ARM architecture, etc. For example, memory 41 can include any combination of various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Memory 41 can include, for example, system memory, which stores, for example, the operating system, application programs, boot loader, database, and other programs. Various application programs and various data can also be stored in the storage medium.

[0071] Furthermore, according to some embodiments of this disclosure, various operations / processes according to this disclosure, when implemented via software and / or firmware, can be transmitted from a storage medium or network to a computer system with a dedicated hardware architecture, such as... Figure 5 The computer system 500 shown is equipped with the programs that constitute the software. When various programs are installed, the computer system is able to perform various functions, including functions such as those described above. Figure 5 A block diagram of an example structure of a computer system that may be employed according to embodiments of the present disclosure is shown.

[0072] exist Figure 5 In this configuration, the Central Processing Unit (CPU) 501 performs various processes based on a program stored in the Read-Only Memory (ROM) 502 or a program loaded from the Storage Section 508 into the Random Access Memory (RAM) 503. The RAM 503 also stores data required as needed when the CPU 501 performs various processes. The CPU is merely an example; it can also be other types of processors, such as the various processors described above. The ROM 502, RAM 503, and Storage Section 508 can be various forms of computer-readable storage media, as described below. It should be noted that although... Figure 5 The image shows ROM 502, RAM 503 and storage device 508, but one or more of them may be combined or located in the same or different memory or storage modules.

[0073] CPU 501, ROM 502 and RAM 503 are connected to each other via bus 504. Input / output interface 505 is also connected to bus 504.

[0074] The following components are connected to the input / output interface 505: input section 506, such as a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output section 907, including displays such as cathode ray tube (CRT), liquid crystal display (LCD), speakers, vibrators, etc.; storage section 508, including hard disks, magnetic tapes, etc.; and communication section 509, including network interface cards such as LAN cards, modems, etc. Communication section 509 allows communication processing via a network such as the Internet. It is easy to understand that, although... Figure 5 The various devices or modules in the computer system 500 shown communicate via bus 504, but they may also communicate via a network or other means, wherein the network may include a wireless network, a wired network, and / or any combination of wireless and wired networks.

[0075] As needed, drive 510 is also connected to input / output interface 505. Removable media 511, such as disks, optical disks, magneto-optical disks, semiconductor memories, etc., are installed on drive 510 as needed, so that computer programs read from them can be installed into storage section 508 as needed.

[0076] When the above series of processes are implemented through software, the program constituting the software can be installed from a network such as the Internet or from a storage medium such as removable media 511.

[0077] According to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the CPU 501, it performs the functions defined in the methods of embodiments of this disclosure.

[0078] It should be noted that, in the context of this disclosure, a computer-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, capable of transmitting, propagating, or transmitting a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0079] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0080] In some embodiments, a computer program is also provided, comprising: instructions that, when executed by a processor, cause the processor to perform the method of any of the above embodiments. For example, the instructions may be embodied in computer program code.

[0081] In embodiments of this disclosure, computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include, but are not limited to, object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network (including a local area network (LAN) or a wide area network (WAN)), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0082] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0083] The modules, components, or units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the modules, components, or units do not necessarily constitute a limitation on the module, component, or unit itself.

[0084] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary hardware logic components that can be used include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0085] The above description is merely an embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0086] Many specific details are set forth in the description provided herein. However, it is understood that embodiments of this disclosure may be practiced without these specific details. In other instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of the description.

[0087] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0088] While specific embodiments of this disclosure have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A method for mining user interests, comprising: Break down the user's behavior sequence; Extract the posting sequence of the target user from the user's social relationship sequence; For each sequence in the split sequence and the extracted posting sequence, extract four sequences: post ID, post category, posting time, and post author ID. The four sequences extracted for each sequence—post ID, post category, posting time, and post author ID—are input into the first deep neural network to obtain a vector A for each sequence. Using the current training post as input, perform an attention mechanism operation on a set of vectors A for each sequence to obtain the attention-adjusted vector A'; Input vector A' into the training model to obtain the predicted score for the currently trained post; as well as The training model is adjusted based on the predicted score, and the adjusted training model is used as the user interest mining model. The method further includes: In each of the split sequences, behaviors within a predetermined time period are marked as short-term behaviors, and other behaviors are marked as long-term behaviors. In the posting sequence of the target user in the social relationship sequence, behaviors with posting time within a predetermined time period are marked as short-term behaviors, and other behaviors are marked as long-term behaviors. Extract vectors B and C for short-term and long-term behaviors from a set of vectors A for each sequence, respectively; Determine whether the post in vector B is a social post or a content post. If it is a social post, connect it to the third deep neural network of social interests to obtain the short-term social interest vector B1. If it is a content post, connect it to the fourth deep neural network of content interests to obtain the short-term content interest vector B2. Determine whether the post in vector C is a social post or a content post. If it is a social post, connect it to the third deep neural network of social interests to obtain the long-term social interest vector C1. If it is a content post, connect it to the fourth deep neural network of content interests to obtain the long-term content interest vector C2. Inputting vectors A', B1, B2, C1, and C2 into the training model yields a predicted score for the currently trained post; and The training model is adjusted based on the estimated score, and the adjusted training model is used as the user interest mining model.

2. The user interest mining method according to claim 1, wherein the training model consists of a multi-expert gating network and a second deep neural network, and wherein inputting vector A' into the training model to obtain a predicted score for the currently trained post includes: Inputting vector A' into a multi-expert gating network yields an array; as well as The array is then input into a second deep neural network to obtain the predicted score.

3. The user interest mining method according to any one of claims 1-2, wherein splitting the user's behavior sequence includes splitting the user's behavior sequence into click, like, follow and / or comment sequences, and wherein the social relationship sequence includes user private chat sequence and user follow sequence.

4. A user interest mining system, comprising: The split unit is configured to split the user's behavior sequence; The first extraction unit is configured to extract the posting sequence of the target user from the user's social relationship sequence; The second extraction unit is configured to extract four sequences for each sequence in the split sequence and the extracted posting sequence: post ID, post category, posting time, and post author ID. The vector merging unit is configured to input four sequences extracted for each sequence—post ID, post category, posting time, and post author ID—into the first deep neural network to obtain a set of vectors A for each sequence; The attention mechanism unit is configured to take the currently trained post as input and perform an attention mechanism operation on a set of vectors A for each sequence to obtain the attention vector A'. The scoring prediction unit is configured to input vector A' into the training model to obtain a predicted score for the currently trained post; The adjustment unit is configured to adjust the training model according to the estimated score and use the adjusted training model as a user interest mining model. The tagging unit is configured to tag behaviors within a predetermined time period in each split sequence as short-term behaviors and other behaviors as long-term behaviors, and to tag behaviors within a predetermined time period in the posting sequence of the target user in the social relationship sequence as short-term behaviors and other behaviors as long-term behaviors. The third extraction unit is configured to extract vectors B and C for short-term and long-term behaviors respectively from a set of vectors A for each sequence; as well as The judgment and processing unit is configured as follows: Determine whether the post in vector B is a social post or a content post. If it is a social post, connect it to the third deep neural network of social interests to obtain the short-term social interest vector B1. If it is a content post, connect it to the fourth deep neural network of content interests to obtain the short-term content interest vector B2. The algorithm determines whether a post in vector C is a social post or a content post. If it's a social post, it's fed into a third deep neural network for social interests to obtain a long-term social interest vector C1; if it's a content post, it's fed into a fourth deep neural network for content interests to obtain a long-term content interest vector C2. The predicted score unit is further configured to input vectors A', B1, B2, C1, and C2 into the training model to obtain a predicted score for the currently trained post; and The adjustment unit is also configured to adjust the training model based on the estimated score and use the adjusted training model as a user interest mining model.

5. The user interest mining system according to claim 4, wherein the training model consists of a multi-expert gating network and a second deep neural network, and wherein the prediction scoring unit is further configured as follows: Inputting vector A' into a multi-expert gating network yields an array; and The array is then input into a second deep neural network to obtain the predicted score.

6. The user interest mining system according to any one of claims 4-5, wherein splitting the user's behavior sequence includes splitting the user's behavior sequence into click, like, follow and / or comment sequences, and wherein the social relationship sequence includes user private chat sequence and user follow sequence.

7. An electronic device, comprising: Memory; and A processor coupled to the memory, the memory storing instructions that, when executed by the processor, cause the electronic device to perform the method according to any one of claims 1-3.

8. A computer-readable storage medium having a computer program stored thereon, the computer program implementing the method according to any one of claims 1-3 when executed by a processor.

Images