A data processing method and related apparatus

Abstract

A data processing method can be applied to the field of artificial intelligence, comprising: obtaining first sequence information; the first sequence information comprises information of a plurality of objects with which a user has interacted; the interaction mode between the user and a second object in the plurality of objects is passive interaction; processing the first sequence information according to a first neural network to obtain an attention prediction value of the user when interacting with the second object; obtaining recommendation information corresponding to the user and the second object through a recommendation model according to the information of the user and the information of the second object; determining a first loss according to the recommendation information and the attention prediction value, wherein the attention prediction value is fused into a loss term corresponding to the second object when determining the first loss. The present application fuses the predicted user attention with the loss term corresponding to the passive interaction sample in the loss function of the downstream recommendation task (for example, weighting, thereby reducing the noise influence of the sample generated by the passive interaction.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more particularly to a data processing method and related apparatus. Background Technology

[0002] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0003] Taking a music recommendation system as an example, the basic operating logic is as follows: Users perform a series of actions on the music display list on the front end, such as actively playing, passively playing, liking, disliking, downloading, and sharing, generating a series of interaction data, which is stored as logs. The music recommendation system uses the data generated by the logs to train a model offline. After the training converges, it outputs a prediction model, which is then deployed online. Based on user characteristics, music characteristics, and contextual information, it provides recommended music results, which the user then interacts with.

[0004] When users interact with the recommended music list, a series of interactive actions occur, including active interactions (such as clicking the "like" or "skip" button) and passive interactions (such as autoplay or not playing). When training the model, the music system needs to construct positive and negative samples based on user feedback actions. The quality of these positive and negative samples significantly impacts the training results. Most existing music recommendation systems construct positive and negative samples based on simple rules during training. For example, they might simply use "liked" songs and autoplaying songs as positive samples, and "skipped" songs and not played songs as negative samples. For positive and negative samples generated by active interactions, the confidence level is relatively high because the user's attention is usually high when making an active interaction; however, for positive and negative samples generated by passive interactions, the confidence level is relatively low because the user's exact attention in these samples is unknown. Therefore, directly using such simply defined samples to train the recommendation model results in performance loss, leading to lower recommendation accuracy. Summary of the Invention

[0005] This application provides a data processing method that can improve the recommendation accuracy of recommendation models.

[0006] In a first aspect, this application provides a data processing method, the method comprising: acquiring first sequence information; the first sequence information including information on multiple objects with which a user has interacted, and the interaction order between the user and the multiple objects, the multiple objects including multiple first objects and second objects, the multiple first objects being objects with which the user interacted before interacting with the second objects; the interaction between the user and the second objects being a passive interaction; processing the first sequence information according to a first neural network to obtain an attention prediction value when the user interacts with the second object; obtaining recommendation information corresponding to the user and the second object through a recommendation model based on the user information and the information of the second object; determining a first loss based on the recommendation information and the attention prediction value, wherein the attention prediction value is fused to a loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model.

[0007] In this embodiment, a neural network can predict the user's attention when interacting with the current object based on the user's interaction sequences with multiple historical objects (including object information, interaction order, and interaction method). The predicted user attention is then fused (e.g., weighted) with the loss term corresponding to passive interaction samples in the loss function of the downstream recommendation task, thereby reducing the noise impact of passive interaction samples. Furthermore, since the data used to predict the attention score is the sequence information of user-object interactions, which can contain information about continuous interactions between the user and objects over a period of time, the attention score predicted based on the sequence information is more accurate. Consequently, the recommendation model trained based on the loss constructed from this attention score has higher recommendation accuracy.

[0008] Optionally, the fusion can be weighted, where the fused second object can be positively correlated with the attention prediction value, that is, the higher the attention prediction value, the higher the corresponding loss term.

[0009] In one possible implementation, the first neural network includes a first feature extraction network and a first task network. The step of processing the first sequence information according to the first neural network to obtain the attention prediction value when the user interacts with the second object includes: processing the first sequence information according to the first feature extraction network to obtain a first state; processing the first state according to the first task network to obtain the attention prediction value when the user interacts with the second object. The method further includes: acquiring second sequence information; the second sequence information includes information indicating the interaction mode between the user and the plurality of first objects, the interaction mode including active interaction and passive interaction; processing the second sequence information according to the second feature extraction network in the second neural network to obtain a second state; the first state and the second state are fused to obtain target information; processing the target information according to the second task network in the second neural network to obtain a propensity score corresponding to the second object; determining a second loss according to the propensity score and the information in the second sequence information indicating the interaction mode corresponding to the second object, the second loss being used to update the first neural network.

[0010] In one possible implementation, the plurality of first objects includes a target object, which is the most recent interaction object between the user and the second object; the first state and the second state are used to fuse to obtain target information, including: the first state, the second state, and information in the second sequence information that indicates the interaction mode corresponding to the target object are used to fuse to obtain target information.

[0011] In one possible implementation, the method further includes: determining a third loss based on the attention prediction value and information in the second sequence information indicating the interaction mode corresponding to the second object, the third loss being used to update the second neural network.

[0012] In one possible implementation, the second loss is used to update the first neural network, including: the second loss is used to update the first neural network while keeping the parameters of the second neural network unchanged.

[0013] In one possible implementation, an unbiased user attention estimator can be obtained by weighting labeled positive samples with bias scores, and an unbiased bias score estimator can be obtained by weighting labeled positive samples with attention scores. The user attention estimator and the bias score estimator are optimized alternately until convergence.

[0014] In one possible implementation, the third loss is used to update the second neural network, including: the third loss is used to update the second neural network while keeping the parameters of the first neural network unchanged.

[0015] In one possible implementation, the object is music or video; the interaction includes at least one of active playback, passive playback, positive evaluation, negative evaluation, download, share, browsing details, and presentation without playback.

[0016] Secondly, this application provides a data processing method, the method comprising:

[0017] Acquire first sequence information and second sequence information; the first sequence information includes information about multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction;

[0018] The first sequence information is processed by the first feature extraction network in the first neural network to obtain the first state;

[0019] Based on the first task network in the first neural network, the first state is processed to obtain the attention prediction value when the user interacts with the second object;

[0020] The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information.

[0021] Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object;

[0022] A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0023] By updating the first and second neural networks in the above manner, a first neural network that meets the training objective can be obtained. The first neural network can be used for user attention corresponding to each sample.

[0024] In one possible implementation, the plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object;

[0025] The first state and the second state are used to fuse and obtain target information, including:

[0026] The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

[0027] In one possible implementation, the method further includes:

[0028] A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0029] Thirdly, this application provides a data processing apparatus, the apparatus comprising:

[0030] The acquisition module is used to acquire first sequence information; the first sequence information includes information about multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects; the interaction mode between the user and the second objects is passive interaction;

[0031] The processing module is used to process the first sequence information according to the first neural network to obtain the attention prediction value when the user interacts with the second object;

[0032] Based on the user's information and the second object's information, recommendation information corresponding to the user and the second object is obtained through a recommendation model;

[0033] A first loss is determined based on the recommendation information and the attention prediction value, wherein the attention prediction value is fused to the loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model.

[0034] In one possible implementation, the first neural network includes a first feature extraction network and a first task network, and the processing module is specifically used for:

[0035] Based on the first feature extraction network, the first sequence information is processed to obtain the first state;

[0036] Based on the first task network, process the first state to obtain the attention prediction value when the user interacts with the second object;

[0037] The processing module is further configured to:

[0038] Acquire second sequence information; the second sequence information includes information for indicating the interaction mode between the user and the plurality of first objects, the interaction mode including active interaction and passive interaction;

[0039] The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information.

[0040] Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object;

[0041] A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0042] In one possible implementation, the plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object;

[0043] The processing module is specifically used for:

[0044] The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

[0045] In one possible implementation, the processing module is further configured to:

[0046] A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0047] In one possible implementation, the processing module is specifically used for:

[0048] The second loss is used to update the first neural network while keeping the parameters of the second neural network unchanged.

[0049] In one possible implementation, the processing module is specifically used for:

[0050] The third loss is used to update the second neural network while keeping the parameters of the first neural network unchanged.

[0051] In one possible implementation, the object is music or video; the interaction includes at least one of active playback, passive playback, positive evaluation, negative evaluation, download, share, browsing details, and presentation without playback.

[0052] Fourthly, this application provides a data processing apparatus, the apparatus comprising:

[0053] The acquisition module is used to acquire first sequence information and second sequence information; the first sequence information includes information on multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction;

[0054] The processing module is used to process the first sequence information based on the first feature extraction network in the first neural network to obtain the first state;

[0055] Based on the first task network in the first neural network, the first state is processed to obtain the attention prediction value when the user interacts with the second object;

[0056] The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information.

[0057] Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object;

[0058] A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0059] In one possible implementation, the plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object;

[0060] The first state and the second state are used to fuse and obtain target information, including:

[0061] The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

[0062] In one possible implementation, the device further includes:

[0063] A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0064] Fifthly, embodiments of this application provide a data processing apparatus, which may include a memory, a processor, and a bus system, wherein the memory is used to store a program, and the processor is used to execute the program in the memory to perform any of the optional methods of the first aspect and any of the optional methods of the second aspect described above.

[0065] In a sixth aspect, embodiments of this application provide a training apparatus, which may include a memory, a processor, and a bus system, wherein the memory is used to store a program, and the processor is used to execute the program in the memory to perform any of the optional methods of the first aspect and any of the optional methods of the second aspect described above.

[0066] In a seventh aspect, embodiments of this application provide a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the first aspect and any optional method described above, as well as the second aspect and any optional method described above.

[0067] Eighthly, embodiments of this application provide a computer program product, including code, which, when executed, is used to perform any of the optional methods described in the first aspect above, and any of the optional methods described in the second aspect above.

[0068] Ninthly, this application provides a chip system including a processor for supporting an execution device or training device in implementing the functions involved in the foregoing aspects, such as transmitting or processing data involved in the foregoing methods; or, information. In one possible design, the chip system further includes a memory for storing program instructions and data necessary for the execution device or training device. The chip system may be composed of chips or may include chips and other discrete devices. Attached Figure Description

[0069] Figure 1 A structural diagram illustrating the main framework of artificial intelligence;

[0070] Figure 2 A schematic diagram of a system architecture provided for an embodiment of this application;

[0071] Figure 3 A schematic diagram of a system architecture provided for an embodiment of this application;

[0072] Figure 4A schematic diagram illustrating a recommendation flow scenario provided in an embodiment of this application;

[0073] Figure 5 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0074] Figure 6 A schematic diagram of a framework provided for an embodiment of this application;

[0075] Figure 7 A schematic diagram of a network structure provided for an embodiment of this application;

[0076] Figure 8 A flowchart illustrating a data processing method provided in an embodiment of this application;

[0077] Figure 9 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application;

[0078] Figure 10 A schematic diagram of an execution device provided in an embodiment of this application;

[0079] Figure 11 A schematic diagram of a training device provided in an embodiment of this application;

[0080] Figure 12 This is a schematic diagram of a chip provided in an embodiment of this application. Detailed Implementation

[0081] The embodiments of the present invention will now be described with reference to the accompanying drawings. The terminology used in the embodiments section is for illustrative purposes only and is not intended to limit the scope of the invention.

[0082] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0083] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0084] First, the overall workflow of the artificial intelligence system is described; please refer to [link / reference]. Figure 1 , Figure 1 The diagram illustrates a structural framework for artificial intelligence (AI). The framework is further elaborated below along two dimensions: the "Intelligent Information Chain" (horizontal axis) and the "IT Value Chain" (vertical axis). The "Intelligent Information Chain" reflects a series of processes from data acquisition to processing. For example, it could be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, and intelligent execution and output. In this process, data undergoes a condensation process of "data—information—knowledge—wisdom." The "IT Value Chain" reflects the value that AI brings to the information technology industry, from the underlying infrastructure of human intelligence and information (provided and processed through technological means) to the industrial ecosystem of the system.

[0085] (1) Infrastructure

[0086] Infrastructure provides computing power to support artificial intelligence systems, enabling communication with the external world and providing support through a basic platform. This communication occurs through sensors; computing power is provided by intelligent chips (hardware acceleration chips such as CPUs, NPUs, GPUs, ASICs, and FPGAs); and the basic platform includes distributed computing frameworks and related platform guarantees and support, which may include cloud storage and computing, interconnected networks, etc. For example, sensors communicate with the outside world to acquire data, and this data is provided to intelligent chips in the distributed computing system provided by the basic platform for computation.

[0087] (2) Data

[0088] The data at the next layer of infrastructure is used to represent the data sources in the field of artificial intelligence. The data involves graphics, images, voice, text, and IoT data from traditional devices, including business data from existing systems and sensor data such as force, displacement, liquid level, temperature, and humidity.

[0089] (3) Data processing

[0090] Data processing typically includes methods such as data training, machine learning, deep learning, search, reasoning, and decision-making.

[0091] Among them, machine learning and deep learning can perform intelligent information modeling, extraction, preprocessing, and training on data, including symbolization and formalization.

[0092] Reasoning refers to the process in which, in a computer or intelligent system, the machine thinks and solves problems by simulating human intelligent reasoning, based on reasoning control strategies and using formalized information. Typical functions include search and matching.

[0093] Decision-making refers to the process of making decisions based on intelligent information after reasoning, and it typically provides functions such as classification, sorting, and prediction.

[0094] (4) General ability

[0095] After the data processing mentioned above, the results of the data processing can be used to form some general capabilities, such as algorithms or a general system, for example, translation, text analysis, computer vision processing, speech recognition, image recognition, etc.

[0096] (5) Smart Products and Industry Applications

[0097] Intelligent products and industry applications refer to products and applications of artificial intelligence systems in various fields. They are the encapsulation of overall artificial intelligence solutions, productizing intelligent information decision-making and realizing practical applications. Their application areas mainly include: intelligent terminals, intelligent transportation, intelligent healthcare, autonomous driving, smart cities, etc.

[0098] This application's embodiments can be applied to the field of information recommendation, including but not limited to e-commerce product recommendations, search engine result recommendations, app store recommendations, music recommendations, and video recommendations. The recommended items in various application scenarios can also be referred to as "objects" for ease of subsequent description. That is, in different recommendation scenarios, the recommended object can be an app, a video, music, or a specific product (such as the presentation interface of an online shopping platform, which displays different products based on different users; this can essentially be presented through the recommendation results of a recommendation model). These recommendation scenarios typically involve user behavior log collection, log data preprocessing (e.g., quantization, sampling), sample set training to obtain a recommendation model, and analysis and processing of the objects (such as apps, music, etc.) involved in the scenarios corresponding to the training sample items based on the recommendation model. For example, if the samples selected in the recommendation model training stage come from the user's operational behavior towards the recommended app in a mobile app store, then the recommendation model trained in this way is applicable to the aforementioned mobile app app store, or can be used for recommending terminal apps in other types of terminal app app stores. The recommendation model will eventually calculate the recommendation probability or score of each object to be recommended. The recommendation system selects the recommendation results according to certain selection rules, such as sorting them according to the recommendation probability or score. The results are then presented to the user through the corresponding application or terminal device. The user interacts with the objects in the recommendation results to generate user behavior logs, etc.

[0099] Reference Figure 4In the recommendation process, when a user interacts with the recommendation system, a recommendation request is triggered. The system inputs this request and its related feature information into the deployed recommendation model, then predicts the click-through rate (CTR) of the user for all candidate items. Subsequently, the candidate items are sorted in descending order based on the predicted CTR and displayed sequentially in different positions as the recommendation result for the user. Users browse the displayed items and perform user actions, such as browsing, clicking, and downloading. These user actions are stored in logs as training data, and the parameters of the recommendation model are periodically updated through an offline training module to improve the model's recommendation performance.

[0100] For example, when a user opens the app store, the recommendation module is triggered. This module predicts the likelihood of the user downloading a given set of candidate apps based on the user's download history, click history, app characteristics, and environmental factors such as time and location. Based on the predictions, the app store displays apps in descending order of probability, thus increasing the likelihood of app downloads. Specifically, apps more likely to be downloaded are listed first, while those less likely are listed last. User behavior is also logged and used to train and update the prediction model's parameters through an offline training module.

[0101] For example, in applications related to lifelong partners, historical data from users across domains such as video, music, and news can be used to construct a cognitive brain, mimicking the mechanisms of the human brain, and build a framework for a lifelong learning system. Lifelong partners can record past events based on system and application data, understand current intentions, predict future actions or behaviors, and ultimately provide intelligent services. In the current first phase, user behavior data (including information such as SMS messages, photos, and email events) obtained from music apps, video apps, and browser apps is used to build a user profile system and implement learning and memory modules based on user information filtering, association analysis, cross-domain recommendation, and causal reasoning to construct a personal knowledge graph for each user.

[0102] The application architecture of this application embodiment will be described next.

[0103] See appendix Figure 2This invention provides a recommendation system architecture 200. A data acquisition device 260 is used to collect samples. A training sample can consist of multiple feature information (or described as attribute information, such as user attributes and item attributes). The feature information can be of various types, specifically including user feature information, object feature information, and tag features. User feature information is used to characterize user characteristics, such as gender, age, occupation, hobbies, etc. Object feature information is used to characterize the features of the objects pushed to the user. Different recommendation systems correspond to different objects, and the types of features to be extracted for different objects are also different. For example, the object features extracted from the training samples of an app market can be the app's name (identifier), type, size, etc. The object features mentioned in the training samples of e-commerce apps can include the product name, category, price range, etc. Tag features are used to indicate whether a sample is positive or negative. Typically, the tag features of a sample can be obtained from the user's actions on the recommended object. Samples where the user has performed an action on the recommended object are positive examples, while samples where the user has not performed an action or has only browsed the recommended object are negative examples. For example, if a user clicks, downloads, or purchases a recommended object, the tag feature is 1, indicating that the sample is a positive example; if the user has not performed any action on the recommended object, the tag feature is 0, indicating that the sample is a negative example. After collection, the samples can be stored in database 230. Some or all of the feature information of the samples in database 230 can also be directly obtained from the client device 240, such as user feature information, user action information on objects (used to determine type identification), and object feature information (such as object identification). The training device 220 trains and obtains the model parameter matrix based on the samples in database 230 to generate the recommendation model 201. The following describes in more detail how the training device 220 trains to obtain the model parameter matrix used to generate the recommendation model 201. The recommendation model 201 can be used to evaluate a large number of objects to obtain the score of each object to be recommended. Furthermore, it can recommend a specified or preset number of objects from the evaluation results of a large number of objects. The calculation module 211 obtains the recommendation results based on the evaluation results of the recommendation model 201 and recommends them to the client device through the I / O interface 212.

[0104] In this embodiment, the training device 220 can select positive and negative samples from the sample set in the database 230 and add them to the training set. Then, the recommendation model is used to train the samples in the training set to obtain the trained recommendation model. The implementation details of the calculation module 211 can be found in [reference needed]. Figure 5 A detailed description of the method embodiments shown.

[0105] After training the model parameter matrix based on the samples, the training device 220 uses it to construct the recommendation model 201 and then sends the recommendation model 201 to the execution device 210. Alternatively, the model parameter matrix can be directly sent to the execution device 210, where the recommendation model is constructed for use in the corresponding system. For example, a recommendation model trained based on video-related samples can be used to recommend videos to users on video websites or apps, while a recommendation model trained based on app-related samples can be used to recommend apps to users in app stores.

[0106] The execution device 210 is equipped with an I / O interface 212 for data interaction with external devices. The execution device 210 can obtain user characteristic information from the client device 240 through the I / O interface 212, such as user identifier, user identity, gender, occupation, and hobbies. This information can also be obtained from the system database. The recommendation model 201 recommends target objects to the user based on the user characteristic information and the characteristic information of the objects to be recommended. The execution device 210 can be located on a cloud server or on the user client.

[0107] The execution device 210 can access data, code, etc., in the data storage system 250, and can also store output data into the data storage system 250. The data storage system 250 can be located within the execution device 210, can be set up independently, or can be located in other network entities; there can be one or multiple such systems.

[0108] The calculation module 211 uses the recommendation model 201 to process the user feature information and the feature information of the object to be recommended. For example, the calculation module 211 uses the recommendation model 201 to analyze and process the user feature information and the feature information of the object to be recommended, thereby obtaining the score of the object to be recommended. The objects to be recommended are sorted according to the score, and the objects ranked higher will be recommended to the client device 240.

[0109] Finally, I / O interface 212 returns the recommendation results to client device 240 and presents them to the user.

[0110] At a deeper level, the training device 220 can generate corresponding recommendation models 201 based on different sample feature information for different targets, so as to provide users with better results.

[0111] It is worth noting that, attached Figure 2 This is merely a schematic diagram of a system architecture provided by an embodiment of the present invention. The positional relationships between the devices, components, modules, etc. shown in the diagram do not constitute any limitation. For example, in the attached diagram... Figure 2In this context, the data storage system 250 is an external memory relative to the execution device 210. In other cases, the data storage system 250 may also be placed within the execution device 210.

[0112] In this embodiment, the training device 220, the execution device 210, and the client device 240 may be three different physical devices. Alternatively, the training device 220 and the execution device 210 may be on the same physical device or a cluster, or the execution device 210 and the client device 240 may be on the same physical device or a cluster.

[0113] See appendix Figure 3 This is a system architecture 300 proposed in an embodiment of the present invention. In this architecture, the execution device 210 is implemented by one or more servers, optionally in conjunction with other computing devices, such as data storage, routers, load balancers, etc. The execution device 210 can be deployed on a single physical site or distributed across multiple physical sites. The execution device 210 can use data in the data storage system 250 or call program code in the data storage system 250 to implement the object recommendation function. Specifically, the information of the object to be recommended is input into the recommendation model. The recommendation model generates an estimated score for each object to be recommended, and then sorts them in descending order of the estimated scores. The object to be recommended is then recommended to the user according to the sorting result. For example, the top 10 objects in the sorting result are recommended to the user.

[0114] The data storage system 250 is used to receive and store the parameters of the recommendation model sent by the training device, as well as the data for storing the recommendation results obtained through the recommendation model. It may also include the program code (or instructions) required for the normal operation of the storage system 250. The data storage system 250 can be a distributed storage cluster consisting of one or more devices deployed outside the execution device 210. In this case, when the execution device 210 needs to use data on the storage system 250, the storage system 250 can send the required data to the execution device 210, and the execution device 210 receives and stores (or caches) the data. Alternatively, the data storage system 250 can be deployed within the execution device 210. When deployed within the execution device 210, the distributed storage system can include one or more storage devices. Optionally, when multiple storage devices exist, different storage devices are used to store different types of data. For example, the model parameters of the recommendation model generated by the training device and the recommendation results obtained through the recommendation model can be stored on two different storage devices.

[0115] Users can interact with execution device 210 by operating their respective user devices (e.g., local device 301 and local device 302). Each local device can represent any computing device, such as a personal computer, computer workstation, smartphone, tablet, smart camera, smart car or other type of cellular phone, media consumption device, wearable device, set-top box, game console, etc.

[0116] Each user's local device can interact with the execution device 210 through a communication network of any communication mechanism / standard. The communication network can be a wide area network, a local area network, a point-to-point connection, or any combination thereof.

[0117] In another implementation, execution device 210 can be implemented by a local device. For example, local device 301 can implement the recommendation function of execution device 210 based on the recommendation model to obtain user feature information and provide recommendation results to the user, or provide services to the user of local device 302.

[0118] Since the embodiments of this application involve a large number of neural network applications, for ease of understanding, the relevant terms and concepts such as neural networks involved in the embodiments of this application will be introduced below.

[0119] 1. Click-through rate (CTR)

[0120] Click probability, also known as click-through rate, refers to the ratio of the number of clicks to the number of impressions of recommended information (e.g., recommended items) on a website or application. Click-through rate is usually an important metric for evaluating recommendation systems.

[0121] 2. Personalized Recommendation System

[0122] A personalized recommendation system refers to a system that analyzes a user's historical data (such as the operation information in the embodiments of this application) using machine learning algorithms, and uses this data to predict new requests and provide personalized recommendation results.

[0123] 3. Offline training

[0124] Offline training refers to a module in a personalized recommendation system that iteratively updates the parameters of the recommendation model according to the algorithm learned by the machine learning, based on the user's historical data (such as the operation information in the embodiments of this application), until the set requirements are met.

[0125] 4. Online Inference

[0126] Online prediction refers to using a model trained offline to predict a user's preference for recommended items in the current context, based on the characteristics of the user, the item, and the context, and to predict the probability that the user will choose the recommended item.

[0127] 5. Selection bias: This mainly comes from user feedback. Because users tend to rate or evaluate only the products or features they are interested in, and rarely mention aspects they are not interested in, this causes non-random missing data.

[0128] 6. Exposure bias: Due to the exposure mechanism of the recommendation system, users can only see a portion of the exposed products and make clicks and other reactions. However, the interactions not included in the data do not necessarily mean that users do not like the products; it may also mean that users are unaware of the products.

[0129] 7. Sample Reweighting: Generally, each sample in a dataset is treated equally, meaning each has a weight of 1. Depending on the specific needs, strategies need to be designed to increase the weight of certain samples and decrease the weight of others.

[0130] For example, Figure 3 This is a schematic diagram of the recommendation system provided in an embodiment of this application. For example... Figure 3 As shown, when a user enters the system, a recommendation request is triggered. The recommendation system inputs this request and its related information (such as the operation information in this embodiment) into the recommendation model, and then predicts the user's selection rate for items within the system. Further, the items are sorted in descending order based on the predicted selection rate or a function based on that selection rate; that is, the recommendation system can display items in different positions sequentially as recommendations to the user. The user browses items in different positions and performs user actions, such as browsing, selecting, and downloading. Simultaneously, the user's actual behavior is stored in a log as training data, and the parameters of the recommendation model are continuously updated through an offline training module to improve the model's predictive performance.

[0131] For example, a user opening the app store on a smart device (e.g., a mobile phone) triggers the app store's recommendation system. The app store's recommendation system predicts the probability of the user downloading each recommended candidate app based on the user's historical behavior logs, such as historical download records and user selection records, as well as the app store's own characteristics, such as environmental features like time and location. Based on the calculation results, the app store's recommendation system can display candidate apps in descending order of predicted probability values, thereby increasing the download probability of candidate apps.

[0132] For example, apps with a predicted high user selection rate can be displayed in the top recommendation positions, while apps with a predicted low user selection rate can be displayed in the bottom recommendation positions.

[0133] The recommended model mentioned above can be a neural network model. The following is an introduction to the relevant terms and concepts of neural networks that may be involved in the embodiments of this application.

[0134] (1) Neural Network

[0135] A neural network can be composed of neural units, which can be defined as a computational unit that takes xs (i.e., input data) and an intercept of 1 as input. The output of this computational unit can be:

[0136]

[0137] Where s = 1, 2, ..., n, where n is a natural number greater than 1, Ws is the weight of xs, and b is the bias of the neural unit. f is the activation function of the neural unit, used to introduce nonlinear characteristics into the neural network to convert the input signal in the neural unit into an output signal. The output signal of this activation function can be used as the input of the next convolutional layer, and the activation function can be the sigmoid function. A neural network is a network formed by connecting multiple of the above-mentioned individual neural units together, that is, the output of one neural unit can be the input of another neural unit. The input of each neural unit can be connected to the local receptive field of the previous layer to extract the features of the local receptive field, which can be a region composed of several neural units.

[0138] (2) Deep Neural Networks

[0139] Deep Neural Networks (DNNs), also known as multilayer neural networks, can be understood as neural networks with many hidden layers, though there's no specific metric for "many." DNNs can be categorized into three layers based on their position: input layers, hidden layers, and output layers. Generally, the first layer is the input layer, the last layer is the output layer, and the layers in between are hidden layers. All layers are fully connected, meaning that any neuron in the i-th layer is connected to any neuron in the (i+1)-th layer. Although DNNs appear complex, the operation of each layer is actually quite simple, resembling a linear relationship as follows: in, It is the input vector. It is the output vector. α is the offset vector, W is the weight matrix (also called coefficients), and α() is the activation function. Each layer is simply an adjustment of the input vector. The output vector is obtained through such a simple operation. Because DNNs have many layers, the coefficients W and the offset vector... The number of these parameters is therefore quite large. The definitions of these parameters in a DNN are as follows: Taking the coefficient W as an example: Assuming a three-layer DNN, the linear coefficient from the 4th neuron in the second layer to the 2nd neuron in the third layer is defined as... The superscript 3 represents the layer number where coefficient W resides, while the subscript corresponds to the output third layer index 2 and the input second layer index 4. In summary, the coefficients from the k-th neuron in layer L-1 to the j-th neuron in layer L are defined as follows: It's important to note that the input layer does not have a W parameter. In deep neural networks, more hidden layers allow the network to better represent complex real-world situations. Theoretically, the more parameters a model has, the higher its complexity and "capacity," meaning it can perform more complex learning tasks. Training a deep neural network is essentially the process of learning the weight matrix, with the ultimate goal of obtaining the weight matrix of all layers in the trained deep neural network (a weight matrix formed by the vectors W from many layers).

[0140] (3) Loss Function

[0141] In training a deep neural network, to ensure the output closely approximates the desired predicted value, we compare the network's prediction with the target value. Based on the difference, we update the weight vector of each layer (usually pre-configuring parameters before the initial update). For example, if the prediction is too high, the weight vector is adjusted to predict a lower value. This adjustment continues until the deep neural network predicts the target value or a value very close to it. Therefore, we need to predefine "how to compare the difference between the predicted and target values," which is the loss function or objective function. These are important equations used to measure the difference between the predicted and target values. Taking the loss function as an example, a higher output value (loss) indicates a greater difference, and training the deep neural network becomes a process of minimizing this loss.

[0142] (4) Backpropagation algorithm

[0143] Backpropagation (BP) can be used during training to correct the parameters in the initial model, thereby reducing the model's error loss. Specifically, forward propagation of the input signal to the output generates error loss; this error loss information is then propagated back to update the parameters in the initial model, leading to convergence of the error loss. The backpropagation algorithm is an error-loss-driven backpropagation process aimed at obtaining optimal model parameters, such as the weight matrix.

[0144] (5) Recurrent Neural Networks (RNN)

[0145] The purpose of RNNs is to process sequential data. In traditional neural network models, the layers are fully connected from the input layer to the hidden layer and then to the output layer, while the nodes within each layer are unconnected. However, this type of ordinary neural network is ineffective for many problems. For example, to predict the next word in a sentence, you generally need to use the preceding words because the words in a sentence are not independent. RNNs are called recurrent neural networks because the current output of a sequence is related to the previous output. Specifically, the network memorizes previous information and applies it to the calculation of the current output. That is, the nodes between hidden layers are no longer unconnected but connected, and the input of the hidden layer includes not only the output of the input layer but also the output of the hidden layer in the previous time step. Theoretically, RNNs can process sequential data of any length.

[0146] Training RNNs is similar to training traditional ANNs (Artificial Neural Networks). It also uses the backpropagation (BP) algorithm, but with a key difference. In RNNs, the parameters W, U, and V are shared when the network is unfolded, unlike in traditional neural networks. Furthermore, in gradient descent, the output at each step depends not only on the current step's network state but also on the states of the network from previous steps. For example, at t=4, it needs to propagate three steps forward, adding various gradients to each of those three steps. This learning algorithm is called Backpropagation ThroughTime (BPTT).

[0147] (6) Residual Network

[0148] In deep neural networks, multiple hidden layers are interconnected layer by layer, for example, the first hidden layer connects to the second, the second to the third, and the third to the fourth (this is a data processing path in the neural network, which can also be figuratively called neural network transmission). Residual networks also have an additional direct connection branch. This direct connection branch connects directly from the first hidden layer to the fourth hidden layer, skipping the processing of the second and third hidden layers and directly transmitting the data from the first hidden layer to the fourth hidden layer for computation. A highway network can be defined as follows: In addition to the aforementioned computational paths and direct connections, a deep neural network also includes a weight acquisition branch. This branch introduces a transformation gate to acquire weight values ​​and outputs a weight value T for subsequent computations in the above computational paths and direct connections.

[0149] In recent years, music streaming services (Spotify, QQ Music, NetEase Cloud Music, Huawei Music, etc.) have become increasingly popular, forming an important part of people's lives and possessing enormous commercial value. Daily playlist recommendations and personalized FM radio are the most important recommendation services, and music recommendation presents unique business-related problems. While training data plays a crucial role, much work focuses on model structure, with little analysis of training data quality and noise issues.

[0150] This application provides a data processing method. The following describes the data processing method provided in the embodiments of this application, taking the model training stage as an example.

[0151] The embodiments of this application can be applied to recommended scenarios, see below. Figure 6 , Figure 6 This is a schematic diagram of an application architecture for a music recommendation scenario, including a displayed music list, user interaction logs with music, an offline training module, and an online prediction module. Figure 5 The corresponding implementation is mainly applied to the offline training module.

[0152] Reference Figure 5 , Figure 5 An embodiment of a data processing method provided in this application is illustrated, such as... Figure 5 As shown, an embodiment of this application provides a data processing method including:

[0153] 501. Obtain first sequence information; the first sequence information includes information on multiple objects with which the user has interacted, and the interaction order between the user and the multiple objects, the multiple objects including multiple first objects and second objects, the multiple first objects being the objects with which the user interacted before interacting with the second objects; the interaction between the user and the second objects is a passive interaction.

[0154] In this embodiment of the application, the execution subject of step 501 can be a terminal device, which can be a portable mobile device, such as, but not limited to, mobile or portable computing devices (such as smartphones), personal computers, server computers, handheld devices (such as tablets) or laptop devices, multiprocessor systems, game consoles or controllers, microprocessor-based systems, set-top boxes, programmable consumer electronics, mobile phones, mobile computing and / or communication devices with wearable or accessory form factors (such as watches, glasses, headphones or earphones), network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0155] In this embodiment of the application, the execution entity of step 501 can be a cloud-side server. For ease of description, the form of the execution entity will not be distinguished below, and all will be described as training devices.

[0156] In one possible implementation, the training device can acquire the user's operation log (or interaction data between the user and multiple objects), which may contain information about the multiple objects with which the user has interacted, the order of the user's interaction with the multiple objects, and the type of interaction between the user and the multiple objects.

[0157] The objects can be physical items or virtual items, such as apps, music, videos, web pages, and news information.

[0158] Taking music as an example, users can perform a series of actions in the music display list on the front end, such as actively playing, passively playing, liking, disliking, downloading, sharing, etc., generating a series of interactive data, which are stored as logs.

[0159] For example, taking music as an example, the training device can obtain the user's operation log when interacting with a music app. For example, the operation log of the user and multiple music (which may include music 1, music 2, and music 3 according to the order of the user's interaction) can be obtained. In the operation log, the user first actively searched for and played music 1, then actively searched for and played music 2, and gave music 2 a negative evaluation (e.g., downvoted). Then the music app automatically played music 3.

[0160] In one possible implementation, the interaction type between the user and the object may include at least one of the following: active playback, passive playback, positive evaluation, negative evaluation, download, sharing, browsing details, and presentation without playback.

[0161] Interaction methods can be divided into active interaction and passive interaction. Active interaction can be understood as using...

[0162] User-initiated interactions (such as clicking "like" or "skip") are highly engaging. Users are highly focused on the interactive object during these interactions, resulting in training samples with higher confidence (because they accurately represent the user's intent). Passive interactions, on the other hand, are interactions performed involuntarily (such as autoplay or not playing). These interactions are less engaging, and users' attention to the interactive object is unknown during these interactions. Consequently, training samples with passive interactions have lower confidence (because they cannot accurately represent the user's intent).

[0163] In this embodiment, the user's attention when interacting with the current object can be predicted by a neural network based on the user's interaction sequence with multiple historical objects (including object information, interaction order, and interaction method, etc.). The predicted user attention is then fused (e.g., weighted) with the loss term corresponding to the passive interaction sample in the loss function of the downstream recommendation task, thereby reducing the noise impact of the passive interaction sample.

[0164] Since the data used to predict attention scores is the sequence information of user and object interactions, which can contain information about the user's continuous interaction with the object over a period of time, the attention scores predicted based on the sequence information are more accurate. Consequently, the recommendation model trained based on the loss constructed from this attention score has higher recommendation accuracy.

[0165] Next, we will introduce sequence information and how to predict a user's current attention based on sequence information.

[0166] In one possible implementation, first sequence information can be obtained; the first sequence information includes information (or features) of multiple objects with which the user has interacted, and the order in which the user interacted with the multiple objects. The multiple objects include multiple first objects and second objects, wherein the multiple first objects are the objects with which the user interacted before interacting with the second objects. The second objects can be the objects whose user attention needs to be predicted.

[0167] For example, the first sequence information can be the features of i musical pieces. Each x represents information about a piece of music. This information could include the music's own attributes (such as music genre, storage space size, duration, and artist information) or the audio-visual information of the user listening to the music.

[0168] In one possible implementation, second sequence information can be obtained, including information indicating the interaction methods between the user and the plurality of first objects, such interaction methods including active and passive interactions. For example, attention tags can be used. Let represent the second sequence information, where ei can be 1 when the interaction mode is active interaction and ei can be 0 when the interaction mode is passive interaction.

[0169] 502. Based on the first neural network, process the first sequence information to obtain the attention prediction value when the user interacts with the second object;

[0170] In one possible implementation, the first neural network includes a first feature extraction network and a first task network. The first feature extraction network can process the first sequence information to obtain a first state; the first task network can process the first state to obtain the attention prediction value when the user interacts with the second object.

[0171] In this embodiment, not only the feature information of the current object (i.e., the second object) is utilized, but also the sequence information of the user's interaction with other objects is also utilized. The first neural network may include a recurrent neural network for feature extraction and a task network (e.g., a multilayer perceptron MLP). That is, the features in the feature sequence Xi can be sequentially passed through a recurrent neural network to model the sequence relationship between objects, thereby obtaining a hidden state containing current and historical information. (First state). Having obtained the implicit state containing current and historical information. Then, the information inside it can be linearly and non-linearly combined using a multilayer perceptron (MLP) (e.g., an MLP with a sigmoid activation function) to finally obtain the desired attention prediction value. (Attention prediction value)

[0172] Taking the daily playlist recommendation scenario in a music app as an example, the training data for this scenario includes seven types of user feedback: five active feedbacks ("like," "dislike," "favorite," "download," and "skip song") and two passive interactions (autoplay and exposure to unplayed content). Passive interactions constitute the majority of these feedback samples, and because user attention may not always be on the songs, these samples contain significant noise. To address this issue, this application proposes an unbiased user attention prediction framework. This framework uses propensity scores for weighting to obtain an unbiased attention score estimator (first neural network) and a propensity score estimator (second neural network), which are then alternately updated to obtain the unbiased attention prediction score. Based on the attention score obtained from this unbiased attention prediction framework, a loss weighting is applied to the passive interaction samples, thereby removing the influence of noise and improving the performance of the recommendation model.

[0173] In one possible implementation, the second sequence information can be processed according to the second feature extraction network in the second neural network to obtain a second state; the first state and the second state are used to fuse to obtain target information; the target information is processed according to the second task network in the second neural network to obtain the tendency score corresponding to the second object.

[0174] To obtain unbiased attention predictions, propensity scores are needed to eliminate bias in loss calculation. The labels in the attention label sequence Ei are sequentially passed through a recurrent neural network (GRU) to model the sequential relationships of attention changes, ultimately yielding a hidden state containing both current and historical attention information. (Second state). Optionally, based on the idea of ​​residuals, attention labels can be... Hidden state and hidden state The information is connected as a whole and then processed through a multilayer perceptron (MLP) with a sigmoid activation function to perform linear and nonlinear interactions to obtain the desired propensity score.

[0175] After obtaining the attention value calculated above and propensity score Then, the loss needs to be calculated based on these two values ​​to achieve an unbiased learning process.

[0176] In one possible implementation, a second loss can be determined based on the propensity score and information in the second sequence information indicating the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0177] In one possible implementation, the plurality of first objects includes a target object, which is the user's most recent interaction object before interacting with the second object; the target information can be obtained by fusing information from the first state, the second state, and the second sequence information that indicates the interaction mode corresponding to the target object.

[0178] In one possible implementation, a third loss can be determined based on the attention prediction value and information in the second sequence information indicating the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0179] In one possible implementation, refer to Figure 7 When updating the unbiased bias score estimator module, the network of the unbiased attention estimator module needs to be frozen; conversely, when updating the unbiased attention estimator module, the network of the unbiased bias score estimator module needs to be frozen. That is, the second loss is used to update the first neural network while keeping the parameters of the second neural network unchanged, and the third loss is used to update the second neural network while keeping the parameters of the first neural network unchanged.

[0180] In one possible implementation, an unbiased user attention estimator can be obtained by weighting labeled positive samples with bias scores, and an unbiased bias score estimator can be obtained by weighting labeled positive samples with attention scores. The user attention estimator and the bias score estimator are alternately optimized until convergence. User attention prediction is a key problem in music recommendation, which can help remove sample noise, but existing user attention learning methods lack effective methods. This application embodiment obtains unbiased user attention by alternately updating two unbiased estimators (a first neural network and a second neural network), which can be used to optimize downstream recommendation tasks (e.g., refer to...). Figure 8 (As shown).

[0181] For example, the following formulas are the loss functions for the propensity score estimator and the attention estimator, respectively.

[0182]

[0183] By updating the first and second neural networks in the above manner, a first neural network that meets the training objective can be obtained. The first neural network can be used for user attention corresponding to each sample.

[0184] 503. Based on the user's information and the second object's information, a recommendation model is used to obtain recommendation information corresponding to the user and the second object.

[0185] It should be understood that there is no explicit temporal sequence between steps 503, 501, and 502, and this application embodiment does not limit this.

[0186] In one possible implementation, recommendation information corresponding to the user and the second object can be obtained through a recommendation model based on the user's information and the information of the second object, which is the feedforward process in the training of the recommendation model.

[0187] The information of the second object can be at least one of the following: object name, size, category, language type, and rating. For example, if the object is music, the category can be rock music, folk music, pop music, etc., and the rating can be a score or comment on the music. This application does not limit the specific type of the object information.

[0188] The user's information can be attributes related to the user's preferences, such as at least one of gender, age, occupation, income, hobbies, and education level. Gender can be male or female, age can be a number between 0 and 100, occupation can be teacher, programmer, chef, etc., hobbies can be basketball, tennis, running, etc., and education level can be primary school, junior high school, high school, university, etc. This application does not limit the specific type of user information.

[0189] In one possible implementation, feature extraction can be performed on the user's information and the information of the second object based on the embedding layer to obtain the feature representation (or embedding representation) corresponding to the user's information and the feature representation (or embedding representation) corresponding to the information of the second object.

[0190] 504. Determine a first loss based on the recommendation information and the attention prediction value, wherein the attention prediction value is fused to the loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model.

[0191] Optionally, the fusion can be weighted, where the fused second object can be positively correlated with the attention prediction value, that is, the higher the attention prediction value, the higher the corresponding loss term.

[0192] In one possible implementation, attention scores can be used to weight the loss of passively interacting samples, thereby removing the influence of noise and improving the performance of the recommendation model.

[0193] For example, the following formula is the loss function for training a recommendation model that uses user attention to weight samples:

[0194]

[0195] The weights for passive interaction samples are obtained by transforming user attention using the following exemplary method:

[0196]

[0197] Compared to existing attention prediction technologies, this application's embodiments model user attention using a reasonable method, learning user attention through data-driven methods, and theoretically guaranteeing unbiasedness. Existing attention prediction technologies are generally based on strong prior assumptions, then sampling and constructing positive and negative samples for training based on these assumptions. This method cannot be universally applied to all different scenarios; once the environment changes, the prior assumptions need to be reset. Furthermore, theoretically, these methods are not unbiased, which inevitably affects the overall system performance. In addition, this application's embodiments not only utilize feature information but also the sequence information between features, fully considering the continuous change in user attention in real-world scenarios, thus more accurately modeling user behavior.

[0198] In music recommendation scenarios, existing technologies cannot provide unbiased and comprehensive modeling of user attention. Most existing solutions employ heuristics to circumvent the unlabeled problem in attention prediction. For example, negative sampling methods sample negative samples based on rules and then use these positive and negative samples to train an attention prediction model. However, negative sampling relies on overly strong assumptions and cannot utilize all samples. Other heuristics include exponential decay functions. These methods are based on prior distributions, assuming that user attention exhibits an exponential decay trend starting from the last active interaction. However, this method relies on prior assumptions and does not align with reality. Therefore, to overcome the drawbacks of overly strong prior assumptions while fully utilizing the sequential information of features, this application proposes an unbiased user attention estimator and an unbiased bias score estimator, respectively. These estimators are alternately optimized using the sequential information of features to model user attention. Furthermore, a model-independent attention-sample weighted recommendation framework is proposed, utilizing attention to remove sample noise and improve the prediction performance of a series of downstream recommendation models.

[0199] This application provides a data processing method, comprising: acquiring first sequence information; the first sequence information including information about multiple objects with which a user has interacted, and the interaction order between the user and the multiple objects, the multiple objects including multiple first objects and second objects, the multiple first objects being objects with which the user interacted before interacting with the second objects; the interaction between the user and the second objects being passive interaction; processing the first sequence information according to a first neural network to obtain a predicted attention value when the user interacts with the second object; obtaining recommendation information corresponding to the user and the second object through a recommendation model based on the user information and the information of the second object; determining a first loss based on the recommendation information and the predicted attention value, wherein the predicted attention value is fused to a loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model. In this application embodiment, based on the user's historical interaction sequence with multiple objects (including object information, interaction order, and interaction method, etc.), a neural network can be used to predict the user's attention when interacting with the current object, and the predicted user attention can be fused (e.g., weighted) with the loss term corresponding to passive interaction samples in the loss function of the downstream recommendation task, thereby reducing the noise impact of samples generated by passive interaction. Furthermore, since the data used to predict the attention score is the sequence information of the user's interaction with the object, the sequence information can contain information about the user's continuous interaction with the object over a period of time. The attention score predicted based on the sequence information is more accurate, and the recommendation model trained based on the loss constructed based on the attention score has higher recommendation accuracy.

[0200] The following is a pseudocode flowchart of an embodiment of this application:

[0201]

[0202] based on Figure 5 The corresponding implementation can obtain a trained recommendation model. The recommendation model can obtain recommendation results based on the user's attribute information and the object's attribute information. Then, when the recommendation information meets the preset conditions, it can determine to recommend items to the user.

[0203] The preset conditions are described below:

[0204] In one possible implementation, when recommending information to a user, the probability of the user selecting multiple items (including items) can be calculated, and the recommendation index of each item for the user can be determined based on the probability of the user selecting multiple items.

[0205] After obtaining the recommendation index of each item for the user, the recommendation index can be sorted, and the M items with the highest recommendation index can be recommended to the user.

[0206] In one possible implementation, a probability threshold can be set, and recommendations can be made to the user when the probability of the user selecting multiple items (including items) is greater than the probability threshold.

[0207] When recommending information, you can present the recommended information to users in the form of a list page, hoping that users will take action.

[0208] Furthermore, taking music as an example, if a user's attention is not high on the music app, the recommendation model obtained using the embodiments of this application can avoid recommending marketing songs or advertisements that require the user's attention. In the latter case, when some users' attention gradually shifts away from the music app, they may be returning to work and using music as background noise. In this situation, it's possible to avoid recommending songs that easily draw the user's attention back, such as rock music, thus improving the user's online experience when interacting with streaming music services.

[0209] To verify the technical effectiveness of the embodiments of this application, two datasets are used to verify the solution of the embodiments of this application: Huawei's industrial dataset (Product) and the 30-Music dataset. The statistical information of the datasets is shown in Table 1, where "M" represents millions and "K" represents thousands.

[0210] Table 1

[0211] Dataset #Conversation #user #song #feature #Feedback Type Product 8.47M 3.75M 1.73M 44 7 30-Music 455K 5.5K 1.99M 12 3

[0212] The evaluation metrics for the experiment were AUC and GAUC, which were compared with several existing attention prediction methods:

[0213] ED (Exponential Decay): Exponential decay

[0214] NDB (Neural Dueling Bandit algorithm): Negative sampling method

[0215] PN: Treat all unlabeled samples as negative samples

[0216] SAR: The PU-learning method assumes that propensity scores are only related to local features and ignores sequence dependencies.

[0217] Experiments were conducted on two datasets, with the DCN-V2 model as the base recommendation model. The experimental results are shown in Table 2, which compares the performance of DCN-V2 and DCN-V2 equipped with different attention prediction models on the product dataset and the 30-Music dataset. "*" indicates the best baseline result, and the performance improvement is statistically significant (t-test, p-value < 0.05). As can be seen from Table 2, the embodiments of this application achieve better results compared to existing attention prediction methods.

[0218] Table 2

[0219]

[0220]

[0221] This application presents a model-agnostic framework that enhances the performance of various base recommendation models. Several commonly used state-of-the-art (SOTA) recommendation models in industry were selected to test the generalization ability of our invention. These models include: FM, Wide&Deep, DeepFM, YoutubeNet, DCN, AutoInt, and DCN-V2. Experimental results are shown in Table 3, which compares the overall performance of the seven base recommendation models trained with and without UAE on the Product and 30-Music datasets. "*" indicates a statistically significant performance improvement equivalent to the base recommendation model (t-test, p-value < 0.05). Table 3 shows that, on both the Product and 30-Music datasets, using our invention's technique (+UAE) effectively improves both AUC and GAUC compared to the baseline recommendation model without UAE. This demonstrates that our invention not only delivers excellent performance but also possesses strong generalization ability, making it suitable for a variety of models.

[0222] Table 3

[0223]

[0224] In addition, this application also provides a data processing method, the method comprising:

[0225] Acquire first sequence information and second sequence information; the first sequence information includes information about multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction;

[0226] The first sequence information is processed by the first feature extraction network in the first neural network to obtain the first state;

[0227] Based on the first task network in the first neural network, the first state is processed to obtain the attention prediction value when the user interacts with the second object;

[0228] The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information.

[0229] Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object;

[0230] A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0231] In one possible implementation, the plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object;

[0232] The first state and the second state are used to fuse and obtain target information, including:

[0233] The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

[0234] In one possible implementation, the method further includes:

[0235] A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0236] Reference Figure 9 , Figure 9 This application provides a schematic diagram of the structure of a data processing apparatus 900, which includes:

[0237] The acquisition module 901 is used to acquire first sequence information; the first sequence information includes information about multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects; the interaction mode between the user and the second objects is passive interaction;

[0238] For a detailed description of the acquisition module 901, please refer to the description of step 501 in the above embodiment, which will not be repeated here.

[0239] Processing module 902 is used to process the first sequence information according to the first neural network to obtain the attention prediction value when the user interacts with the second object;

[0240] Based on the user's information and the second object's information, recommendation information corresponding to the user and the second object is obtained through a recommendation model;

[0241] A first loss is determined based on the recommendation information and the attention prediction value, wherein the attention prediction value is fused to the loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model.

[0242] For a detailed description of the processing module 902, please refer to the descriptions of steps 502, 503, and 504 in the above embodiments, which will not be repeated here.

[0243] In one possible implementation, the first neural network includes a first feature extraction network and a first task network, and the processing module is specifically used for:

[0244] Based on the first feature extraction network, the first sequence information is processed to obtain the first state;

[0245] Based on the first task network, process the first state to obtain the attention prediction value when the user interacts with the second object;

[0246] The processing module is further configured to:

[0247] Acquire second sequence information; the second sequence information includes information for indicating the interaction mode between the user and the plurality of first objects, the interaction mode including active interaction and passive interaction;

[0248] The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information.

[0249] Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object;

[0250] A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0251] In one possible implementation, the plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object;

[0252] The processing module is specifically used for:

[0253] The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

[0254] In one possible implementation, the processing module is further configured to:

[0255] A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0256] In one possible implementation, the processing module is specifically used for:

[0257] The second loss is used to update the first neural network while keeping the parameters of the second neural network unchanged.

[0258] In one possible implementation, the processing module is specifically used for:

[0259] The third loss is used to update the second neural network while keeping the parameters of the first neural network unchanged.

[0260] In one possible implementation, the object is music or video; the interaction includes at least one of active playback, passive playback, positive evaluation, negative evaluation, download, share, browsing details, and presentation without playback.

[0261] This application embodiment also provides a data processing apparatus, the apparatus comprising:

[0262] The acquisition module is used to acquire first sequence information and second sequence information; the first sequence information includes information on multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction;

[0263] The processing module is used to process the first sequence information based on the first feature extraction network in the first neural network to obtain the first state;

[0264] Based on the first task network in the first neural network, the first state is processed to obtain the attention prediction value when the user interacts with the second object;

[0265] The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information.

[0266] Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object;

[0267] A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

[0268] In one possible implementation, the plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object;

[0269] The first state and the second state are used to fuse and obtain target information, including:

[0270] The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

[0271] In one possible implementation, the device further includes:

[0272] A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

[0273] The following describes an execution device provided in an embodiment of this application. Please refer to [link / reference]. Figure 10, Figure 10 This is a schematic diagram of an execution device provided in an embodiment of this application. The execution device 1000 can specifically be a mobile phone, tablet, laptop, smart wearable device, server, etc., and is not limited thereto. Specifically, the execution device 1000 includes: a receiver 1001, a transmitter 1002, a processor 1003, and a memory 1004 (wherein the execution device 1000 may have one or more processors 1003). The processor 1003 may include an application processor 10031 and a communication processor 10032. In some embodiments of this application, the receiver 1001, transmitter 1002, processor 1003, and memory 1004 may be connected via a bus or other means.

[0274] Memory 1004 may include read-only memory and random access memory, and provides instructions and data to processor 1003. A portion of memory 1004 may also include non-volatile random access memory (NVRAM). Memory 1004 stores processor and operation instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof, wherein the operation instructions may include various operation instructions for implementing various operations.

[0275] Processor 1003 controls the operation of the execution device. In specific applications, the various components of the execution device are coupled together through a bus system, which may include not only the data bus, but also power buses, control buses, and status signal buses. However, for clarity, all buses are referred to as the bus system in the diagram.

[0276] The methods disclosed in the embodiments of this application can be applied to or implemented by the processor 1003. The processor 1003 can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above methods can be completed by the integrated logic circuitry in the hardware of the processor 1003 or by instructions in software form. The processor 1003 can be a general-purpose processor, a digital signal processor (DSP), a microprocessor or microcontroller, a vision processing unit (VPU), a tensor processing unit (TPU), or other processors suitable for AI computation. It may further include application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The processor 1003 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory 1004. Processor 1003 reads information from memory 1004 and, in conjunction with its hardware, runs the recommendation model trained based on steps 501 to 504 in the above embodiments.

[0277] Receiver 1001 can be used to receive input digital or character information, and to generate signal inputs related to the settings and function control of the execution device. Transmitter 1002 can be used to output digital or character information through the first interface; transmitter 1002 can also be used to send instructions to the disk group through the first interface to modify the data in the disk group; transmitter 1002 may also include a display device such as a display screen.

[0278] This application also provides a training device; please refer to [link / reference]. Figure 11 , Figure 11This is a schematic diagram of a training device provided in an embodiment of this application. Specifically, the training device 1100 is implemented by one or more servers. The training device 1100 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 1111 (e.g., one or more processors) and memory 1132, and one or more storage media 1130 (e.g., one or more mass storage devices) for storing application programs 1142 or data 1144. The memory 1132 and storage media 1130 can be temporary or persistent storage. The program stored in the storage media 1130 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the training device. Furthermore, the CPU 1111 may be configured to communicate with the storage media 1130 and execute the series of instruction operations in the storage media 1130 on the training device 1100.

[0279] The training device 1100 may also include one or more power supplies 1126, one or more wired or wireless network interfaces 1150, one or more input / output interfaces 1158; or, one or more operating systems 1141, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0280] Specifically, the training device can perform steps 501 to 504 in the above embodiments.

[0281] This application also provides a computer program product that, when run on a computer, causes the computer to perform steps as performed by the aforementioned execution device, or causes the computer to perform steps as performed by the aforementioned training device.

[0282] This application also provides a computer-readable storage medium storing a program for signal processing, which, when run on a computer, causes the computer to perform steps as performed by the aforementioned execution device, or causes the computer to perform steps as performed by the aforementioned training device.

[0283] The execution device, training device, or terminal device provided in this application embodiment can specifically be a chip. The chip includes a processing unit and a communication unit. The processing unit can be, for example, a processor, and the communication unit can be, for example, an input / output interface, pins, or circuits. The processing unit can execute computer execution instructions stored in the storage unit to cause the chip within the execution device to execute the data processing method described in the above embodiments, or to cause the chip within the training device to execute the data processing method described in the above embodiments. Optionally, the storage unit can be a storage unit within the chip, such as a register or cache. Alternatively, the storage unit can be a storage unit located outside the chip within the wireless access device, such as a read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, such as random access memory (RAM).

[0284] For details, please refer to Figure 12 , Figure 12 This is a schematic diagram of a chip provided in an embodiment of this application. The chip can be represented as a neural network processor (NPU) 1200. The NPU 1200 is mounted as a coprocessor on the host CPU, and tasks are assigned by the host CPU. The core part of the NPU is the arithmetic circuit 1203, which is controlled by the controller 1204 to extract matrix data from the memory and perform multiplication operations.

[0285] The NPU 1200 achieves this through the cooperation of its various internal components. Figure 5 The data processing method provided in the described embodiments.

[0286] More specifically, in some implementations, the arithmetic circuitry 1203 within the NPU 1200 includes multiple processing engines (PEs). In some implementations, the arithmetic circuitry 1203 is a two-dimensional pulsating array. The arithmetic circuitry 1203 can also be a one-dimensional pulsating array or other electronic circuitry capable of performing mathematical operations such as multiplication and addition. In some implementations, the arithmetic circuitry 1203 is a general-purpose matrix processor.

[0287] For example, suppose we have an input matrix A, a weight matrix B, and an output matrix C. The arithmetic circuit retrieves the corresponding data of matrix B from the weight memory 1202 and caches it in each PE of the arithmetic circuit. The arithmetic circuit retrieves the data of matrix A from the input memory 1201 and performs matrix operations with matrix B. The partial result or the final result of the obtained matrix is ​​stored in the accumulator 1208.

[0288] Unified memory 1206 is used to store input and output data. Weight data is directly transferred to weight memory 1202 via Direct Memory Access Controller (DMAC) 1205. Input data is also transferred to unified memory 1206 via DMAC.

[0289] BIU stands for Bus Interface Unit, which is used for interaction between the AXI bus and the DMAC and the Instruction Fetch Buffer (IFB) 1209.

[0290] The Bus Interface Unit (BIU) 1210 is used by the instruction fetch memory 1209 to fetch instructions from external memory, and also by the memory access controller 1205 to fetch the original data of the input matrix A or the weight matrix B from external memory.

[0291] The DMAC is mainly used to move input data from external memory DDR to unified memory 1206, or to weight data to weight memory 1202, or to input data to input memory 1201.

[0292] The vector computation unit 1207 includes multiple processing units that further process the output of the computation circuit 1203 when needed, such as vector multiplication, vector addition, exponential operations, logarithmic operations, size comparisons, etc. It is mainly used for computation in non-convolutional / fully connected layers of neural networks, such as batch normalization, pixel-level summation, and upsampling of feature planes.

[0293] In some implementations, the vector computation unit 1207 can store the processed output vector in the unified memory 1206. For example, the vector computation unit 1207 can apply a linear function, or a nonlinear function, to the output of the computation circuit 1203, such as linear interpolation of feature planes extracted by a convolutional layer, or, for example, a vector of accumulated values, to generate activation values. In some implementations, the vector computation unit 1207 generates normalized values, pixel-level summed values, or both. In some implementations, the processed output vector can be used as an activation input to the computation circuit 1203, for example, for use in subsequent layers of the neural network.

[0294] The instruction fetch buffer 1209 connected to the controller 1204 is used to store the instructions used by the controller 1204;

[0295] Unified memory 1206, input memory 1201, weight memory 1202, and instruction fetch memory 1209 are all on-chip memories. External memory is proprietary to this NPU hardware architecture.

[0296] The processor mentioned above can be a general-purpose central processing unit, a microprocessor, an ASIC, or one or more integrated circuits used to control the execution of the above program.

[0297] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0298] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0299] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0300] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A data processing method, characterized in that, The method includes: Obtain first sequence information; the first sequence information includes information about multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects; the interaction mode between the user and the second objects is passive interaction; Based on the first neural network, the first sequence information is processed to obtain the attention prediction value when the user interacts with the second object; Based on the user's information and the second object's information, recommendation information corresponding to the user and the second object is obtained through a recommendation model; A first loss is determined based on the recommendation information and the attention prediction value, wherein the attention prediction value is fused to the loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model.

2. The method according to claim 1, characterized in that, The first neural network includes a first feature extraction network and a first task network. The step of processing the first sequence information according to the first neural network to obtain the attention prediction value when the user interacts with the second object includes: Based on the first feature extraction network, the first sequence information is processed to obtain the first state; Based on the first task network, process the first state to obtain the attention prediction value when the user interacts with the second object; The method further includes: Acquire second sequence information; the second sequence information includes information for indicating the interaction mode between the user and the plurality of first objects, the interaction mode including active interaction and passive interaction; The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information. Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object; A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

3. The method according to claim 2, characterized in that, The plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object; The first state and the second state are used to fuse and obtain target information, including: The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

4. The method according to claim 2, characterized in that, The method further includes: A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

5. The method according to claim 4, characterized in that, The second loss is used to update the first neural network, including: The second loss is used to update the first neural network while keeping the parameters of the second neural network unchanged.

6. The method according to claim 4 or 5, characterized in that, The third loss is used to update the second neural network, including: The third loss is used to update the second neural network while keeping the parameters of the first neural network unchanged.

7. The method according to any one of claims 1 to 5, characterized in that, The object is music or video; the interaction includes at least one of the following: active playback, passive playback, positive evaluation, negative evaluation, download, sharing, browsing details, and presentation without playback.

8. A data processing method, characterized in that, The method includes: Acquire first sequence information and second sequence information; the first sequence information includes information about multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction; The first sequence information is processed according to the first feature extraction network in the first neural network to obtain the first state; Based on the first task network in the first neural network, the first state is processed to obtain the attention prediction value when the user interacts with the second object; The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information. Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object; A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

9. The method according to claim 8, characterized in that, The plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object; The first state and the second state are used to fuse and obtain target information, including: The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

10. The method according to claim 8 or 9, characterized in that, The method further includes: A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

11. A data processing apparatus, characterized in that, The device includes: The acquisition module is used to acquire first sequence information and second sequence information; the first sequence information includes information on multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction; The processing module is used to process the first sequence information according to the first neural network to obtain the attention prediction value when the user interacts with the second object; Based on the user's information and the second object's information, recommendation information corresponding to the user and the second object is obtained through a recommendation model; A first loss is determined based on the recommendation information and the attention prediction value, wherein the attention prediction value is fused to the loss term corresponding to the second object when determining the first loss, and the first loss is used to update the recommendation model.

12. The apparatus according to claim 11, characterized in that, The first neural network includes a first feature extraction network and a first task network, and the processing module is specifically used for: Based on the first feature extraction network, the first sequence information is processed to obtain the first state; Based on the first task network, process the first state to obtain the attention prediction value when the user interacts with the second object; The processing module is further configured to: The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information. Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object; A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

13. The apparatus according to claim 12, characterized in that, The plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object; The processing module is specifically used for: The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

14. The apparatus according to claim 12, characterized in that, The processing module is further configured to: A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

15. The apparatus according to claim 14, characterized in that, The processing module is specifically used for: The second loss is used to update the first neural network while keeping the parameters of the second neural network unchanged.

16. The apparatus according to claim 14 or 15, characterized in that, The processing module is specifically used for: The third loss is used to update the second neural network while keeping the parameters of the first neural network unchanged.

17. The apparatus according to any one of claims 11 to 15, characterized in that, The object is music or video; the interaction includes at least one of the following: active playback, passive playback, positive evaluation, negative evaluation, download, sharing, browsing details, and presentation without playback.

18. A data processing apparatus, characterized in that, The device includes: The acquisition module is used to acquire first sequence information and second sequence information; the first sequence information includes information on multiple objects that the user has interacted with, and the interaction order between the user and the multiple objects, the multiple objects include multiple first objects and second objects, the multiple first objects are the objects that the user interacted with before interacting with the second objects, the second sequence information includes information for indicating the interaction method between the user and the multiple first objects, the interaction method includes active interaction and passive interaction; The processing module is used to process the first sequence information based on the first feature extraction network in the first neural network to obtain the first state; Based on the first task network in the first neural network, the first state is processed to obtain the attention prediction value when the user interacts with the second object; The second sequence information is processed by the second feature extraction network in the second neural network to obtain the second state; the first state and the second state are fused to obtain the target information. Based on the second task network in the second neural network, the target information is processed to obtain the tendency score corresponding to the second object; A second loss is determined based on the propensity score and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the second loss is used to update the first neural network.

19. The apparatus according to claim 18, characterized in that, The plurality of first objects includes a target object, which is the most recently interacted object of the user before interacting with the second object; The first state and the second state are used to fuse and obtain target information, including: The information in the first state, the second state, and the second sequence information used to indicate the interaction mode corresponding to the target object is used to fuse and obtain the target information.

20. The apparatus according to claim 18 or 19, characterized in that, The device further includes: A third loss is determined based on the attention prediction value and the information in the second sequence information that indicates the interaction mode corresponding to the second object, and the third loss is used to update the second neural network.

21. A computing device, characterized in that, The computing device includes a memory and a processor; the memory stores code, and the processor is configured to retrieve the code and execute the method as described in any one of claims 1 to 10.

22. A computer storage medium, characterized in that, The computer storage medium stores one or more instructions that, when executed by one or more computers, cause the one or more computers to perform the method of any one of claims 1 to 10.

23. A computer program product, comprising code, characterized in that, When the code is executed, it is used to implement the method as described in any one of claims 1 to 10.

Images