Model training method, data recommendation method, device, electronic device and storage medium

Abstract

The present disclosure provides a model training method, data recommendation method, device, electronic device, and storage medium, relating to the field of artificial intelligence technology, particularly deep learning, natural language processing, data recommendation, and other fields. A specific implementation scheme is as follows: obtaining target behavior data from a first feed information stream and a first channel in the first feed information stream; obtaining target resource data from the first feed information stream and a first channel in the first feed information stream; obtaining sample data based on the target behavior data and the target resource data; and performing cross-domain, multi-perspective model training based on the sample data to obtain a trained first target model. The present disclosure improves the accuracy of model training.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to the fields of deep learning, natural language processing, and data recommendation. Background Art

[0002] In the field of data recommendation, the client can recommend various types of data of interest to users, for example, recommending an article or playlist of interest to users. The client can push various types of data of interest to users in the form of feed information flow.

[0003] With the development of technology, AI can accelerate data processing, and its application scenarios are diverse. For example, AI can be used in hardware and software design for image processing, graphic processing, video processing, and other data recommendation tasks to improve data processing speed and accuracy. In the data recommendation field, a model with high data accuracy is needed. Summary of the Invention

[0004] The present disclosure provides a model training method, device, electronic device and storage medium.

[0005] According to one aspect of the present disclosure, a model training method is provided, comprising:

[0006] Acquire target behavior data from a first feed information stream and a first channel in the first feed information stream;

[0007] Acquire target resource data from the first feed information stream and a first channel in the first feed information stream;

[0008] Obtaining sample data according to the target behavior data and the target resource data;

[0009] A cross-domain and multi-perspective model training is performed based on the sample data to obtain a trained first target model.

[0010] According to another aspect of the present disclosure, a data recommendation method is provided, comprising:

[0011] Get the touch operation on the first channel in the first feed information stream;

[0012] In response to the touch operation, resource data matching the touch operation is obtained according to the first target model obtained by the above-mentioned model training method.

[0013] According to another aspect of the present disclosure, there is provided a model training device, comprising:

[0014] a first data acquisition unit, configured to acquire target behavior data from a first feed information stream and a first channel in the first feed information stream;

[0015] a second data acquisition unit, configured to acquire target resource data from the first feed information stream and the first channel in the first feed information stream;

[0016] a sample data determining unit, configured to obtain sample data according to the target behavior data and the target resource data;

[0017] The model training unit is used to perform cross-domain and multi-perspective model training based on the sample data to obtain a trained first target model.

[0018] According to another aspect of the present disclosure, a data recommendation device is provided, comprising:

[0019] An operation acquisition unit, configured to acquire a touch operation on a first channel in a first feed information stream;

[0020] An operation response unit is used to respond to the touch operation and obtain resource data matching the touch operation according to the first target model obtained by the above-mentioned model training method.

[0021] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0022] at least one processor; and

[0023] a memory communicatively connected to the at least one processor; wherein,

[0024] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method provided by any embodiment of the present disclosure.

[0025] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, where the computer instructions are used to enable the computer to execute the method provided by any one of the embodiments of the present disclosure.

[0026] According to another aspect of the present disclosure, a computer program product is provided, comprising computer instructions, which implement the method provided in any embodiment of the present disclosure when executed by a processor.

[0027] Using the present disclosure, target behavior data can be obtained from a first feed information stream and a first channel within the first feed information stream, and target resource data can be obtained from the first feed information stream and a first channel within the first feed information stream. Sample data is obtained based on the target behavior data and target resource data, and cross-domain, multi-perspective model training is performed based on the sample data to obtain a trained first target model, thereby improving the accuracy of model training.

[0028] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are used to better understand the present invention and do not constitute a limitation of the present invention.

[0030] Figure 1 is a schematic diagram of a distributed cluster processing scenario according to an embodiment of the present disclosure;

[0031] Figure 2 is a flowchart of a model training method according to an embodiment of the present disclosure;

[0032] Figure 3 is a schematic diagram of a feed flow in an example according to an embodiment of the present disclosure;

[0033] Figure 4 is a schematic diagram of a feed flow in another example according to an embodiment of the present disclosure;

[0034] Figure 5 is a structural diagram of a double-tower model in an example according to an embodiment of the present disclosure;

[0035] Figure 6 is a flowchart of a data recommendation method according to an embodiment of the present disclosure;

[0036] Figure 7 1 is a schematic diagram of the structure of a model training device according to an embodiment of the present disclosure;

[0037] Figure 8 This is a schematic diagram of the structure of the data recommendation device according to an embodiment of the present disclosure.

[0038] Figure 9 It is a block diagram of an electronic device used to implement the model training method or data recommendation method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0039] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0040] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The term "at least one" in this article means any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C, can mean including any one or more elements selected from the set consisting of A, B, and C. The terms "first" and "second" in this article refer to multiple similar technical terms and distinguish them, and do not mean to limit the order or to limit to only two. For example, the first feature and the second feature refer to two categories / two features. The first feature can be one or more, and the second feature can also be one or more.

[0041] In addition, numerous specific details are provided in the following detailed description to better illustrate the present disclosure. Those skilled in the art will appreciate that the present disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art are not described in detail in order to highlight the main points of the present disclosure.

[0042] Feed information flow is a data form that continuously provides content to users. It is a resource aggregator composed of multiple content sources. Users can actively subscribe to the message source and push data content to users.

[0043] The feed can be the feed in the client's main feed interface, or it can be the feed for one or more specific channels in the client's main feed interface. For newly launched channels, such as women's columns or sports, the feed in the main feed interface is often used as sample data for model training. For example, a large amount of behavioral data based on active users directly loaded from the main feed interface can be used as sample data. This differs from the user distribution and resource distribution for one or more specific channels, resulting in low model accuracy when trained based on this sample data. For example, in the field of data recommendation, low model accuracy also leads to low data recommendation accuracy based on this model. For example, for one or several specific channels, an inverted index is established based on the tags and categories of all resources in the current scenario to recommend data through inverted retrieval based on user interests. This is useful in the initial stage when data collection is required, but in the long run, the accuracy of data recommendation achieved through the inverted index is also low. In the scenario of matching polysemous words, the "apple" in fruit is different from the "apple" in mobile phone, which can easily introduce bad cases; in the scenario of matching synonyms, it is difficult to cover and is completely unable to express the semantic similarity of user behavior (for example, if a user clicks on news about celebrity A, he is likely to be interested in other news about celebrity B).

[0044] In summary, in order to improve the accuracy of model training, it is necessary to expand the current sample data and consider introducing more sample data for one or several specific channels.

[0045] According to an embodiment of the present disclosure, Figure 1 This is a schematic diagram of a distributed cluster processing scenario according to an embodiment of the present disclosure. The distributed cluster system is an example of a cluster system. It exemplifies how the distributed cluster system can be used for model training. The present disclosure is not limited to model training on a single machine or multiple machines. The use of distributed processing can further improve the accuracy of model training. Figure 1As shown, the distributed cluster system includes multiple nodes (such as server cluster 101, server 102, server cluster 103, server 104, server 105, and server 105 can also be connected to electronic devices, such as mobile phone 1051 and desktop computer 1052). Multiple nodes, as well as multiple nodes and connected electronic devices can jointly execute one or more model training tasks. Optionally, multiple nodes in the distributed cluster system can adopt a data-parallel model training method, and multiple nodes can perform model training tasks based on the same training method to better train the model; if multiple nodes in the distributed cluster system adopt a model-parallel model training method, multiple nodes can perform model training tasks based on different training methods to better train the model. Optionally, after each round of model training is completed, data exchange (such as data synchronization) can be performed between multiple nodes.

[0046] According to an embodiment of the present disclosure, a model training method is provided. Figure 2 This is a flow chart of a model training method according to an embodiment of the present disclosure. The method can be applied to a model training device. For example, the device can be deployed in a terminal or server or other processing device in a single machine, multi-machine or cluster system to implement model training and other processing. The terminal can be a user equipment (UE), a mobile device, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. In some possible implementations, the method can also be implemented by a processor calling computer-readable instructions stored in a memory. Figure 2 As shown, this method is applied to Figure 1 Any node or electronic device (such as a mobile phone or desktop computer) in the cluster system shown includes:

[0047] S201: Acquire target behavior data from a first feed information stream and a first channel in the first feed information stream.

[0048] S202: Acquire target resource data from the first feed information stream and the first channel in the first feed information stream.

[0049] S203: Obtain sample data according to the target behavior data and the target resource data.

[0050] S204: Perform cross-domain and multi-perspective model training based on the sample data to obtain a trained first target model.

[0051] In one example of S201-S204, target behavior data can be obtained from a first feed information stream and a first channel within the first feed information stream (e.g., a women's column, sports, news, etc.) (for example, a user clicks on an article about clothing combinations, a sports video, a news report, etc., and can then obtain knowledge related to the clothing combinations, play the sports video, or read the news). Target resource data can also be obtained from the first feed information stream and the first channel within the first feed information stream (the target resource data can be in a variety of forms, such as graphic information, dynamic video information, short videos of 5-10 minutes, or short videos of less than 1 minute). Sample data is obtained based on the target behavior data and the target resource data, and cross-domain, multi-perspective model training is performed based on the sample data to obtain a trained first target model.

[0052] Using the disclosed embodiments, target behavior data can be obtained from a first feed information stream and a first channel within the first feed information stream, as well as target resource data. Sample data is generated based on the target behavior data and target resource data, and cross-domain, multi-perspective model training is performed based on the sample data to obtain a trained first target model, thereby improving the accuracy of model training.

[0053] In one embodiment, obtaining target behavior data from a first feed information stream and a first channel within the first feed information stream includes: obtaining first behavior data matching a target user within the first feed information stream, determining all user behavior data matching the first channel within the first feed information stream as second behavior data, and obtaining target behavior data based on the first behavior data and the second behavior data. This embodiment combines multiple types of behavior data from the first feed information stream and the first channel within the first feed information stream, resulting in target behavior data that more accurately describes user behavior.

[0054] In one embodiment, obtaining target resource data from a first information feed and a first channel within the first information feed includes: obtaining first resource data from the first information feed that matches the target resource data, determining resource data associated with all user behavior data in the first information feed that matches the first channel as second resource data, and obtaining the target resource data based on the first resource data and the second resource data. This embodiment combines multiple resource data from the first information feed and the first channel within the first information feed, resulting in target resource data that more accurately describes the resource type that matches the user behavior.

[0055] In one embodiment, obtaining first behavior data matching a target user in a first feed information stream includes obtaining the first behavior data matching the target user in the first feed information stream based on user characteristics. The user characteristics include at least one of user ID, user gender, user age, user city, user historical behavior, and user scenario. In this embodiment, the first feed information stream (main feed) is filtered based on matching user characteristics to obtain the first behavior data matching the target user in the first feed information stream. The first behavior data matching the target user is then filtered from all user behavior data based on the user characteristics.

[0056] In one embodiment, all user behavior data matching a first channel in a first feed information stream is determined as second behavior data, including: in response to a touch operation on the first channel, switching from a first feed information stream interface to which the first feed information stream belongs to a second feed information stream interface to which the first channel belongs, and determining all user behavior data matching the first channel collected based on the second feed information stream interface as the second behavior data. In this embodiment, for one or more specific first channels (such as women's columns, sports, news, etc.), all user behavior data matching the first channel collected after switching from the first feed information stream (main feed) interface to the feed information stream interface containing the first channel is considered second behavior data.

[0057] Figure 3-4 is a schematic diagram of a feed flow in an example according to an embodiment of the present disclosure, such as Figure 3 As shown, it includes: a main feed flow interface 301, which includes multiple channels. Clicking on the "Women's Column" channel switches to the feed flow interface 303 where the channel is located. The feed flow interface 303 where the channel is located includes data of various resource types recommended to the user, which can be graphic information, dynamic video information, short videos of 5-10 minutes, short videos of less than 1 minute, etc. Figure 4 As shown, the main feed stream interface 401 includes multiple channels and multiple feed information streams 403 of the main feed stream interface 401. Click the "Women's Column" channel to switch to the feed stream interface 404 where the channel is located. The feed stream interface 404 where the channel is located includes data of various resource types recommended to users, which can be graphic information, dynamic video information, short videos of 5-10 minutes, short videos within 1 minute, etc.

[0058] In one embodiment, obtaining first resource data that matches target resource data in the first feed information stream includes: obtaining first resource data that matches target resource data in the first feed information stream based on resource characteristics. The resource characteristics include at least one of user ID, resource type, and the type of content provider that produces the resource. In this embodiment, the first feed information stream (main feed) is filtered based on the matching of resource characteristics (resource characteristics can be represented by an identifier distributed on the current first channel), thereby obtaining first resource data that matches the target user in the first feed information stream, and then filtering out first resource data that matches the target resource from all resource data related to user behavior based on the resource characteristics. The content provider that produces the resource, for example, an original article, can be the author of the original article.

[0059] In one embodiment, resource data associated with all user behaviors matching a first channel in a first feed information stream is determined as second resource data, including: responding to a touch operation on the first channel, switching from a first feed information stream interface to which the first feed information stream belongs to a second feed stream interface to which the first channel belongs, obtaining all user behavior data matching the first channel collected based on the second feed stream interface, and determining the resource data associated with all user behavior data matching the first channel as the second resource data. Using this embodiment, for one or several specific first channels (such as women's columns, sports, news, etc.), for all users of the first channel, after switching from the first feed information stream (main feed) interface to the feed stream interface containing the channel, the resource data associated with all user behavior data matching the first channel collected is the second resource data.

[0060] In one embodiment, cross-domain, multi-perspective model training is performed based on sample data to obtain a trained first target model, including: inputting the sample data into a dual-tower model, and performing cross-domain, multi-perspective model training using binary cross-entropy as a loss function to obtain the trained first target model. In this embodiment, the sample data is input into the dual-tower model. In the dual-tower model, the target behavior data in the sample data can be used as the user tower, and the target resource data in the sample data can be used as the resource tower. A cross-entropy loss function is calculated based on the user tower and the resource tower to obtain the cross-entropy loss function. Model training is performed based on the cross-entropy loss function to obtain the trained first target model. Because the sample data is a combination of multiple behavioral data from a first feed information stream and a first channel within the first feed information stream, the target behavior data obtained thereby more accurately describes user behavior, and the sample data is combined with multiple resource data from the first feed information stream and the first channel within the first feed information stream. The target resource data obtained thereby more accurately describes the resource type that matches the user behavior. Therefore, the sample data is more accurate, and the first target model obtained through model training using the dual-tower model and the cross-entropy loss function has a high model accuracy.

[0061] Figure 5 This is a schematic diagram of the structure of a dual-tower model in an example according to an embodiment of the present disclosure, comprising multiple user towers (e.g., multiple user towers derived from the target behavior data) and multiple resource towers (e.g., multiple resource towers derived from the target resource data). Each user tower and each resource tower comprises multiple layers (e.g., multiple convolutional layers derived from multiple convolution processes). It should be noted that, for example, the multiple resource types recommended to users include graphic information, dynamic video information, short videos of 5-10 minutes, and short videos of less than 1 minute. For the four main resource types currently in the first channel, four resource towers can be constructed, and a single "user-resource" sample data entry can only activate the user tower and the corresponding resource tower.

[0062] Furthermore, during model training, the above-mentioned target behavior data and target resource data can be used as positive samples in the sample data, and combined with the user's touch operation (such as click operation) to obtain the final positive sample. The positive samples of other users randomly sampled are used as negative samples relative to the current user. After sample expansion, the sample data includes both positive samples and negative samples, which can further improve the accuracy of model training.

[0063] Furthermore, the user features and model parameters output by the above model can be saved in the cluster with timeliness to ensure that the trained model is applied to the application scenario of the current channel and matches the application scenario of the current channel, thereby improving the accuracy of data recommendation.

[0064] Furthermore, resource vector indexes can be constructed to obtain four types of resource indexes. During the model's use, for the "user-resource recall" scenario, the user characteristics of the current user are obtained through the above model. The four types of resource indexes are matched based on the user characteristics of the current user to obtain the resource type that matches the current user.

[0065] According to an embodiment of the present disclosure, a data recommendation method is provided. Figure 6 This is a flow chart of a data recommendation method according to an embodiment of the present disclosure. The method can be applied to a data recommendation device. For example, the device can be deployed in a terminal or server or other processing device in a single machine, multi-machine or cluster system to implement model training and other processing. The terminal can be a user equipment (UE), a mobile device, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. In some possible implementations, the method can also be implemented by a processor calling computer-readable instructions stored in a memory. Figure 6 As shown, it is applied to terminal equipment, including:

[0066] S601: Acquire a touch operation on a first channel in a first feed information stream.

[0067] S602: In response to the touch operation, obtain resource data matching the touch operation according to the first target model obtained according to any one of the above embodiments.

[0068] In an example of S601-S602, a touch operation on the first channel in the first feed information flow can be obtained. In response to the touch operation, the first target model obtained according to any of the above embodiments (such as the first target model obtained by training the model through the dual-tower model and based on the cross-entropy loss function) obtains resource data matching the touch operation (such as graphic and text information, dynamic video information, short videos of 5-10 minutes, small videos within 1 minute, etc.).

[0069] By adopting the embodiment of the present disclosure, the first target model obtained by training the above-mentioned model can be used to recommend corresponding resources to users on the client based on the output results of the first target model (such as user features), thereby realizing user-resource recall. Thus, data recommendation is performed through the trained first target model, thereby improving the accuracy of data recommendation.

[0070] According to an embodiment of the present disclosure, a model training device is provided. Figure 7 Schematic diagram of the structure of the model training device according to an embodiment of the present disclosure. Figure 7As shown, the model training device 700 includes: a first data acquisition unit 701, used to obtain target behavior data from a first feed information stream and a first channel in the first feed information stream; a second data acquisition unit 702, used to obtain target resource data from the first feed information stream and the first channel in the first feed information stream; a sample data determination unit 703, used to obtain sample data based on the target behavior data and the target resource data; a model training unit 704, used to perform cross-domain and multi-perspective model training based on the sample data to obtain a trained first target model.

[0071] In one embodiment, the first data acquisition unit is used to obtain first behavior data matching a target user in a first feed information stream; determine all user behavior data matching a first channel in the first feed information stream as second behavior data; and obtain the target behavior data based on the first behavior data and the second behavior data.

[0072] In one embodiment, the second data acquisition unit is used to obtain first resource data that matches the target resource data in the first feed information stream; determine the resource data associated with all user behavior data that matches the first channel in the first feed information stream as second resource data; and obtain the target resource data based on the first resource data and the second resource data.

[0073] In one embodiment, the first data acquisition unit is used to obtain first behavior data matching the target user in the first feed information flow based on user characteristics; wherein the user characteristics include: user ID, user gender, user age, user city, user historical behavior, and user scenario. At least one of the following.

[0074] In one embodiment, the first data acquisition unit is used to respond to a touch operation on the first channel, switch from a first feed stream interface to which the first feed information stream belongs to a second feed stream interface to which the first channel belongs; and determine all user behavior data matching the first channel collected based on the second feed stream interface as the second behavior data.

[0075] In one embodiment, the second data acquisition unit is used to acquire first resource data matching the target resource data in the first feed information flow based on resource characteristics; wherein the resource characteristics include: at least one of: user ID, resource type, and type of content provider that produces the resource.

[0076] In one embodiment, the second data acquisition unit is used to respond to a touch operation on the first channel, switch from the first feed stream interface to which the first feed information stream belongs to the second feed stream interface to which the first channel belongs; obtain all user behavior data matching the first channel collected based on the second feed stream interface; and determine the resource data associated with all user behavior data matching the first channel as the second resource data.

[0077] In one embodiment, the model training unit is used to input the sample data into the twin-tower model, perform the cross-domain multi-perspective model training using binary cross entropy as a loss function, and obtain the trained first target model.

[0078] In one embodiment, the model training unit is used to use the target behavior data in the sample data as the user tower and the target resource data in the sample data as the resource tower in the dual-tower model; calculate the binary cross entropy based on the user tower and the resource tower to obtain a cross entropy loss function; perform model training based on the cross entropy loss function to obtain the trained first target model.

[0079] According to an embodiment of the present disclosure, a data recommendation device is provided. Figure 8 FIG. 1 is a schematic diagram of the structure of a data recommendation device according to an embodiment of the present disclosure. Figure 8 As shown, the data recommendation device 800 includes: an operation acquisition unit 801, used to obtain a touch operation on a first channel in a first feed information flow; an operation response unit 802, used to respond to the touch operation, and obtain resource data matching the touch operation according to a first target model obtained according to any of the above-mentioned model training methods.

[0080] In one embodiment, the resource data includes at least one of graphic information, dynamic video information, short video information, and mini video information.

[0081] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0082] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0083] Figure 9A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0084] like Figure 9 As shown, electronic device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. RAM 903 may also store various programs and data required for the operation of electronic device 900. Computing unit 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to bus 904.

[0085] Multiple components in the electronic device 900 are connected to the I / O interface 905, including an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a magnetic disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the electronic device 900 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0086] The computing unit 901 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as the model training method or the data recommendation method. For example, in some embodiments, the model training method or the data recommendation method may be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the model training method or the data recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the model training method or the data recommendation method in any other appropriate manner (for example, by means of firmware).

[0087] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0088] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0089] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0090] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0091] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0092] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0093] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions of this disclosure can be achieved, and this document is not limited here.

[0094] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A model training method, comprising: Acquire target behavior data from a first feed information stream and a first channel in the first feed information stream; wherein the first channel is a newly launched channel; Acquiring target resource data from the first feed information stream and the first channel in the first feed information stream, including: acquiring first resource data in the first feed information stream that matches the target resource data; determining resource data associated with all user behavior data in the first feed information stream that matches the first channel as second resource data; and obtaining the target resource data based on the first resource data and the second resource data; Obtaining sample data according to the target behavior data and the target resource data; Performing cross-domain and multi-perspective model training based on the sample data to obtain a trained first target model; The acquiring target behavior data from the first feed information stream and the first channel in the first feed information stream includes: Obtaining the first behavior data matching the target user in the first feed information flow; Determining all user behavior data matching the first channel in the first feed information flow as second behavior data; The target behavior data is obtained according to the first behavior data and the second behavior data.

2. The method according to claim 1, wherein The obtaining of first behavior data matching the target user in the first feed information flow includes: Obtaining first behavior data matching the target user in the first feed information flow based on user characteristics; The user characteristics include at least one of user ID, user gender, user age, user city, user historical behavior, and user scenario.

3. The method according to claim 1, wherein The step of determining all user behavior data matching the first channel in the first feed information flow as second behavior data includes: In response to a touch operation on the first channel, switching from a first feed stream interface to which the first feed information stream belongs to to a second feed stream interface to which the first channel belongs; All user behavior data collected based on the second feed stream interface and matching the first channel are determined as the second behavior data.

4. The method according to claim 1, wherein The obtaining of first resource data matching the target resource data in the first feed information flow includes: Acquire first resource data matching the target resource data in the first feed information flow according to the resource characteristics; The resource characteristics include at least one of: user ID, resource type, and type of content provider that produces the resource.

5. The method according to claim 1, wherein The step of determining resource data associated with all user behaviors matching the first channel in the first feed information flow as second resource data includes: In response to a touch operation on the first channel, switching from a first feed stream interface to which the first feed information stream belongs to to a second feed stream interface to which the first channel belongs; Obtaining all user behavior data matching the first channel collected based on the second feed stream interface; Resource data associated with all user behavior data matching the first channel is determined as the second resource data.

6. The method according to any one of claims 1 to 5, wherein The cross-domain and multi-perspective model training is performed based on the sample data to obtain a trained first target model, including: The sample data is input into the dual-tower model, and the cross-domain multi-perspective model training is performed using the binary cross entropy as the loss function to obtain the trained first target model.

7. The method according to claim 6, wherein: The step of inputting the sample data into the twin-tower model and performing the cross-domain multi-perspective model training using the binary cross entropy as the loss function to obtain the trained first target model includes: In the dual-tower model, the target behavior data in the sample data is used as the user tower, and the target resource data in the sample data is used as the resource tower; Calculate the binary cross entropy based on the user tower and the resource tower to obtain a cross entropy loss function; Model training is performed according to the cross entropy loss function to obtain the trained first target model.

8. A data recommendation method, comprising: Get the touch operation on the first channel in the first feed information stream; In response to the touch operation, resource data matching the touch operation is obtained according to the first object model obtained according to any one of claims 1 to 7.

9. The method according to claim 8, wherein The resource data includes at least one of graphic information, dynamic video information, short video information, and mini video information.

10. A model training device comprising: A first data acquisition unit is configured to acquire target behavior data from a first feed information stream and a first channel in the first feed information stream; wherein the first channel is a newly launched channel; a second data acquisition unit configured to acquire target resource data from the first feed information stream and the first channel in the first feed information stream; the second data acquisition unit configured to: acquire first resource data in the first feed information stream that matches the target resource data; determine resource data associated with all user behavior data in the first feed information stream that matches the first channel as second resource data; and obtain the target resource data based on the first resource data and the second resource data; a sample data determining unit, configured to obtain sample data according to the target behavior data and the target resource data; A model training unit, configured to perform cross-domain and multi-perspective model training based on the sample data to obtain a trained first target model; Wherein, the first data acquisition unit is used to: Obtaining the first behavior data matching the target user in the first feed information flow; Determining all user behavior data matching the first channel in the first feed information flow as second behavior data; The target behavior data is obtained according to the first behavior data and the second behavior data.

11. The device according to claim 10, wherein The first data acquisition unit is configured to: Obtaining first behavior data matching the target user in the first feed information flow based on user characteristics; The user characteristics include at least one of user ID, user gender, user age, user city, user historical behavior, and user scenario.

12. The device according to claim 10, wherein The first data acquisition unit is configured to: In response to a touch operation on the first channel, switching from a first feed stream interface to which the first feed information stream belongs to to a second feed stream interface to which the first channel belongs; All user behavior data collected based on the second feed stream interface and matching the first channel are determined as the second behavior data.

13. The device according to claim 10, wherein The second data acquisition unit is configured to: Acquire first resource data matching the target resource data in the first feed information flow according to the resource characteristics; The resource characteristics include at least one of: user ID, resource type, and type of content provider that produces the resource.

14. The device according to claim 10, wherein The second data acquisition unit is configured to: In response to a touch operation on the first channel, switching from a first feed stream interface to which the first feed information stream belongs to a second feed stream interface to which the first channel belongs; Obtaining all user behavior data matching the first channel collected based on the second feed stream interface; Resource data associated with all user behavior data matching the first channel is determined as the second resource data.

15. The device according to any one of claims 10 to 14, wherein: The model training unit is used to: The sample data is input into the dual-tower model, and the cross-domain multi-perspective model training is performed using the binary cross entropy as the loss function to obtain the trained first target model.

16. The device according to claim 15, wherein The model training unit is used to: In the dual-tower model, the target behavior data in the sample data is used as the user tower, and the target resource data in the sample data is used as the resource tower; Calculate the binary cross entropy based on the user tower and the resource tower to obtain a cross entropy loss function; Model training is performed according to the cross entropy loss function to obtain the trained first target model.

17. A data recommendation device, comprising: An operation acquisition unit, configured to acquire a touch operation on a first channel in a first feed information stream; An operation response unit is used to respond to the touch operation and obtain resource data matching the touch operation according to the first target model obtained according to any one of claims 1 to 7.

18. The device according to claim 17, wherein The resource data includes at least one of graphic information, dynamic video information, short video information, and mini video information.

19. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

20. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 7.

21. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images