An information recommendation method and device, a storage medium, and an electronic device

Abstract

Embodiments of the present application disclose an information recommendation method and device, a storage medium and an electronic device. The method comprises: receiving a target search request input by a user object; determining user attribute information and a historical search and browsing behavior sequence of the user object; determining a user interest representation and a search user representation of the user object based on the user attribute information and the historical search and browsing behavior sequence; and sending the user interest representation, the search user representation and the target search request to an information recommendation engine to perform information recommendation processing on the user object in response to the target search request by the information recommendation engine.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an information recommendation method, apparatus, storage medium, and electronic device. Background Technology

[0002] When users use applications with information search functions (such as search engines), they often receive promotional information pushed by the service provider during the search process. For example, to provide convenience to users, the promotional information may be an advertisement promoting a certain product or service. These promotional messages are of great significance to service providers, users, and the information promoters. Summary of the Invention

[0003] This application provides an information recommendation method, apparatus, storage medium, and electronic device, the technical solutions of which are as follows:

[0004] In a first aspect, embodiments of this application provide an information recommendation method, the method comprising:

[0005] Receive the target search request input by the user object, and determine the user attribute information and historical search browsing behavior sequence for the user object;

[0006] Based on the user attribute information and the historical search and browsing behavior sequence, the user interest representation and search user representation of the user object are determined.

[0007] The user interest representation, the search user representation, and the target search request are sent to the information recommendation engine so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request.

[0008] In one feasible implementation, determining the user interest representation and search user representation of the user object based on the search terms, the user attribute information, and the historical search browsing behavior sequence includes:

[0009] Based on the user attribute information and the historical search and browsing behavior sequence, the user object is processed by an information sequence reasoning model to obtain user interest representation and search user representation.

[0010] In one feasible implementation, the step of using an information sequence reasoning model to perform search interest parsing processing on the user object based on the user attribute information and the historical search browsing behavior sequence to obtain user interest representation and search user representation includes:

[0011] Based on user attribute information and the historical search browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model. The user search representation library includes reference user interest representation and reference search user representation corresponding to the reference user object. The reference user interest representation and reference search user representation are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user object.

[0012] In one feasible implementation, the step of using an information sequence reasoning model to perform search interest parsing processing on the user object based on the user attribute information and the historical search browsing behavior sequence to obtain user interest representation and search user representation includes:

[0013] A global graph of user browsing behavior is established based on the user attribute information and the historical search and browsing behavior sequence. At least one browsing behavior sequence chain for the user object is extracted from the global graph of user browsing behavior. The user attribute information and the browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing and processing, and the user interest representation and search user representation are output.

[0014] In one feasible implementation, the method further includes:

[0015] Obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior.

[0016] The reference user attribute information and the reference browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing processing, and reference user interest representation and reference search user representation are output. Based on the reference user interest representation and the reference search user representation, a user search key value library associated with the information sequence reasoning model is created.

[0017] In one feasible implementation, the step of inputting the reference user attribute information and the reference browsing behavior sequence chain into the information sequence reasoning model for search interest parsing processing and outputting reference user interest representation and reference search user representation includes:

[0018] Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, at least one sample user object's sample training data is selected, and the sample training data is labeled with output result tags, the output result tags including user interest representation tags and search user representation tags;

[0019] Based on the sample training data, the initial information sequence reasoning model is trained for at least one round. During the model training process, the initial information sequence reasoning model determines the predicted output results for the sample training data. The predicted output results include predicted user interest representations and predicted search user representations.

[0020] Based on the predicted output and the output label, the model calculation loss is determined, and the model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss until the initial information sequence reasoning model ends model training, thus obtaining the information sequence reasoning model.

[0021] Based on the reference user attribute information and the reference browsing behavior sequence chain, the information sequence reasoning model is used to determine the reference user interest representation and the reference search user representation.

[0022] In one feasible implementation, the step of establishing a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequence, and extracting at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior, includes:

[0023] The information items in the reference historical search and browsing behavior sequence are determined, and the information items are sorted according to the time order to obtain a node sequence including all information item nodes. The information item edge set including all information item edges is extracted from the reference historical search and browsing behavior sequence with reference to the node sequence. The user edge relationship between all reference user objects and the information item nodes is established, and a user edge relationship set including all the user edge relationships is generated.

[0024] A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set.

[0025] The global graph of reference user browsing behavior is split into sequence chains to obtain at least one reference browsing behavior sequence chain.

[0026] In one feasible implementation, after sending the user interest representation, the search user representation, and the target search request to the information recommendation engine, so that the information recommendation engine performs information recommendation processing on the user object in response to the target search request, the method further includes:

[0027] Collect the user's actual search and browsing behavior sequence for the recommended information items, and determine the user's actual user interest representation based on the actual search and browsing behavior sequence;

[0028] The information sequence reasoning model is calibrated based on the actual user interest representation.

[0029] Secondly, embodiments of this application provide an information recommendation device, the device comprising:

[0030] The request processing module is used to receive the target search request input by the user object and determine the user attribute information and historical search browsing behavior sequence of the user object;

[0031] The representation determination module is used to determine the user interest representation and search user representation of the user object based on the user attribute information and the historical search and browsing behavior sequence.

[0032] The information recommendation module is used to send the user interest representation, the search user representation, and the target search request to the information recommendation engine, so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request.

[0033] In one feasible implementation, the characterization determination module is configured to:

[0034] Based on the user attribute information and the historical search and browsing behavior sequence, the user object is processed by an information sequence reasoning model to obtain user interest representation and search user representation.

[0035] In one feasible implementation, the characterization determination module is configured to:

[0036] Based on user attribute information and the historical search browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model. The user search representation library includes reference user interest representation and reference search user representation corresponding to the reference user object. The reference user interest representation and reference search user representation are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user object.

[0037] In one feasible implementation, the characterization determination module is configured to:

[0038] A global graph of user browsing behavior is established based on the user attribute information and the historical search and browsing behavior sequence. At least one browsing behavior sequence chain for the user object is extracted from the global graph of user browsing behavior. The user attribute information and the browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing and processing, and the user interest representation and search user representation are output.

[0039] In one feasible implementation, the device is further used for:

[0040] Obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior.

[0041] The reference user attribute information and the reference browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing processing, and reference user interest representation and reference search user representation are output. Based on the reference user interest representation and the reference search user representation, a user search key value library associated with the information sequence reasoning model is created.

[0042] In one feasible implementation, the device is further used for:

[0043] Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, at least one sample user object's sample training data is selected, and the sample training data is labeled with output result tags, the output result tags including user interest representation tags and search user representation tags;

[0044] Based on the sample training data, the initial information sequence reasoning model is trained for at least one round. During the model training process, the initial information sequence reasoning model determines the predicted output results for the sample training data. The predicted output results include predicted user interest representations and predicted search user representations.

[0045] Based on the predicted output and the output label, the model calculation loss is determined, and the model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss until the initial information sequence reasoning model ends model training, thus obtaining the information sequence reasoning model.

[0046] Based on the reference user attribute information and the reference browsing behavior sequence chain, the information sequence reasoning model is used to determine the reference user interest representation and the reference search user representation.

[0047] In one feasible implementation, the device is further used for:

[0048] The information items in the reference historical search and browsing behavior sequence are determined, and the information items are sorted according to the time order to obtain a node sequence including all information item nodes. The information item edge set including all information item edges is extracted from the reference historical search and browsing behavior sequence with reference to the node sequence. The user edge relationship between all reference user objects and the information item nodes is established, and a user edge relationship set including all the user edge relationships is generated.

[0049] A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set.

[0050] The global graph of reference user browsing behavior is split into sequence chains to obtain at least one reference browsing behavior sequence chain.

[0051] In one feasible implementation, the device is further used for:

[0052] Collect the user's actual search and browsing behavior sequence for the recommended information items, and determine the user's actual user interest representation based on the actual search and browsing behavior sequence;

[0053] The information sequence reasoning model is calibrated based on the actual user interest representation.

[0054] Thirdly, embodiments of this application provide a computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the above-described method steps.

[0055] Fourthly, embodiments of this application provide an electronic device that may include: a processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and to execute the above-described method steps.

[0056] The beneficial effects of the technical solutions provided in some embodiments of this application include at least the following:

[0057] In one or more embodiments of this application, an electronic device receives a target search request input by a user, determines the user's attribute information and historical search browsing behavior sequence, and accurately constructs and determines the user's interest representation and search user representation by comprehensively utilizing the user's attribute information and historical search browsing behavior. Then, combined with the user's real-time search request, personalized information recommendation processing is performed through an information recommendation engine. This can effectively integrate the user's long-term interests and current search needs to achieve highly personalized promotional information recommendations, improve user satisfaction, and help increase the click-through rate and interaction rate of recommended content, thereby improving the platform's conversion efficiency and user stickiness. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a schematic diagram of a scenario for an information recommendation system provided in an embodiment of this application;

[0060] Figure 2 This is a flowchart illustrating an information recommendation method provided in an embodiment of this application;

[0061] Figure 3 This is a schematic diagram of a search scenario provided in an embodiment of this application;

[0062] Figure 4 This is a schematic diagram illustrating a scenario for recalling promotional information provided in an embodiment of this application;

[0063] Figure 5 This is a schematic diagram illustrating another scenario for recalling promotional information provided in an embodiment of this application;

[0064] Figure 6 This is a schematic diagram of an information sequence reasoning process provided in an embodiment of this application;

[0065] Figure 7 This is a schematic diagram of a scenario for establishing a global graph and sequence chain provided in an embodiment of this application;

[0066] Figure 8 This is a schematic diagram illustrating the process of constructing a user search key-value library according to an embodiment of this application;

[0067] Figure 9 This is a schematic diagram of the training process of an information sequence reasoning model provided in an embodiment of this application;

[0068] Figure 10 This is a schematic diagram illustrating a scenario of information sequence reasoning model processing provided in an embodiment of this application;

[0069] Figure 11 This is a schematic diagram of the structure of an information recommendation device provided in an embodiment of this application;

[0070] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application;

[0071] Figure 13 This is a schematic diagram of the structure of the operating system and user space provided in the embodiments of this application;

[0072] Figure 14 yes Figure 13 Architecture diagram of the Android operating system in China;

[0073] Figure 15 yes Figure 13 Architecture diagram of the iOS operating system. Detailed Implementation

[0074] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0075] In the description of this application, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this application, it should be noted that, unless otherwise expressly specified and limited, "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances. Furthermore, in the description of this application, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist; for example, A and / or B can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0076] In related technologies, displaying recommended promotional information typically requires retrieving the most relevant promotional content from a pool of promotional materials (including search engine data and advertiser-submitted materials) based on the user's search query. For historical users, the retrieved results, aside from the search query itself, cannot recall results not seen in the user's past behavior, resulting in poor timeliness and diversity. For new users, the retrieved results are mostly popular content, failing to meet the user's true needs and lacking personalization. Therefore, the information recommendation methods in these technologies suffer from limitations in meeting user needs and achieving poor recommendation results.

[0077] The present application will now be described in detail with reference to specific embodiments.

[0078] Please see Figure 1 This is a schematic diagram of a scenario for an information recommendation system provided in this specification. Figure 1 As shown, the information recommendation system may include at least a client cluster and a service platform 100.

[0079] The client cluster may include at least one client, such as Figure 1As shown, it specifically includes client 1 corresponding to user 1, client 2 corresponding to user 2, ..., client n corresponding to user n, where n is an integer greater than 0.

[0080] Each client in a client cluster can be an electronic device with communication capabilities, including but not limited to: wearable devices, handheld devices, personal computers, tablets, in-vehicle devices, smartphones, computing devices, or other processing devices connected to a wireless modem. Electronic devices may have different names in different networks, such as: user equipment, access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication equipment, user agent or user device, cellular phone, cordless phone, personal digital assistant (PDA), and electronic devices in 5G networks or future evolved networks.

[0081] The service platform 100 can be a standalone server device, such as a rack-mount, blade, tower, or cabinet-type server device, or a workstation, mainframe, or other hardware device with strong computing power; or it can be a server cluster composed of multiple servers. The servers in the service cluster can be composed in a symmetrical manner, wherein each server is functionally and hierarchically equivalent in the transaction chain, and each server can provide services independently. The independent provision of services can be understood as not requiring the assistance of other servers.

[0082] In one or more embodiments of this specification, the service platform 100 may establish a communication connection with at least one client in the client cluster, and complete the data interaction during the feature processing based on the communication connection.

[0083] It should be noted that the service platform 100 establishes a communication connection with at least one client in the client cluster via a network for interactive communication. This network can be a wireless network or a wired network. Wireless networks include, but are not limited to, cellular networks, wireless LANs, infrared networks, or Bluetooth networks. Wired networks include, but are not limited to, Ethernet, universal serial bus (USB), or controller area networks. In one or more embodiments of the specification, technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network (such as target compressed packets). Furthermore, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.

[0084] The feature processing system embodiments provided in this specification and the information recommendation methods described in one or more embodiments belong to the same concept. The execution entity corresponding to the information recommendation method in one or more embodiments of this specification can be an electronic device, such as the aforementioned service platform 100; the execution entity corresponding to the information recommendation method in one or more embodiments of this specification can also be the electronic device corresponding to the client, depending on the actual application environment. The implementation process of the feature processing system embodiments can be detailed in the following method embodiments, and will not be repeated here.

[0085] In one embodiment, such as Figure 2 As shown, an information recommendation method is proposed, which can be implemented using a computer program and run on an information recommendation device based on the von Neumann architecture. This computer program can be integrated into an application or run as a standalone utility application. The information recommendation device can be an electronic device.

[0086] Specifically, the information recommendation methods include:

[0087] S102: Receive the target search request input by the user object, and determine the user attribute information and historical search browsing behavior sequence for the user object;

[0088] When a user enters a search term "query" into the search engine, the service platform can receive the search request from the user. The search request carries the search term "query" and will then be transmitted to the information recommendation engine, which can obtain the search term carried in the search request.

[0089] Furthermore, after determining the target search request entered by the user, user attribute information and historical search browsing behavior sequences can be obtained, such as basic attributes like age, gender, user ID, geographic location, and occupation. The user ID uses the login account's UID and GUID as unique identifiers. In some embodiments, user attribute information may also include search attribute information corresponding to the user's current target search request. Search attribute information includes search keywords, search context, and immediate search behavior (such as click-through rate and page dwell time). Creatively incorporating the current search attribute information into the user attribute information provides multimodal characteristics, which is beneficial for fully representing user search characteristics and assisting in subsequent accurate information recommendations.

[0090] The historical search and browsing behavior sequence can be understood as recording the historical user's search and browsing behavior based on changes in search and browsing time during the user's historical search and browsing process, extracting the information items involved in each search and browsing step, and the time series of the information items corresponding to the historical user constitutes the historical search and browsing behavior sequence.

[0091] For example, please see Figure 3 , Figure 3 This is a schematic diagram of a search scenario. It obtains user attribute information through a data management service (also known as a DMP service), and retrieves user click logs from search engines and ad click logs from information promotion engines (search engines and information promotion engines can be collectively referred to as information recommendation engines or search engines). These logs are then concatenated using the user's ID. User u i The search, browsing, and click data set can be denoted as , where v k This refers to information items (also known as information entries or information objects) that users perceive during their interaction with the search engine. An information item is a highly abstract representation of the information objects in a user's search behavior, and includes, but is not limited to, the user's search query, Point of Interest (POI), image, and ad title. Represents a timestamp.

[0092] Then, the search, browsing, and click data set is sorted in chronological order to obtain the user's historical search and browsing behavior sequence (which can be called the Session sequence): ,in .

[0093] S104: Determine the user interest representation and search user representation of the user object based on the user attribute information and the historical search and browsing behavior sequence;

[0094] User interest representation can be understood as a user interest feature vector determined for a user object based on user attribute information and historical search and browsing behavior sequences; user interest representation is an abstract description of a user's long-term interests and preferences, and is represented in the form of feature vectors.

[0095] Search user representation can be understood as a user search feature vector determined for a user object, based on user attribute information and historical search browsing behavior sequences. Search user representation mainly focuses on the user's search needs, search intent, user attributes, etc.

[0096] In one or more embodiments of this specification, an information sequence reasoning model can be created and trained based on a machine learning model. Then, in practical applications, based on the user attribute information and the historical search and browsing behavior sequence as model input, the user object is processed by the information sequence reasoning model to obtain user interest representation and search user representation.

[0097] In one feasible implementation, the following approach can be adopted:

[0098] Based on the user attribute information and the historical search and browsing behavior sequence, an information sequence reasoning model is used to analyze the user's search interests, resulting in user interest representations and search user representations. Through this information sequence reasoning model, user interest representations and search user representations based on recent behavior can be inferred from the user's historical behavior.

[0099] For example, based on user attribute information and historical search browsing behavior sequences, an information sequence reasoning model is used to analyze and process user search interests by transforming user interests and search behaviors into vector representations. This vectorized representation helps the information recommendation engine (also known as the information promotion engine) understand and process user data, enabling the recommendation system to more effectively predict user behavior and preferences and accurately recommend promotional information.

[0100] The training process of the information sequence reasoning model is explained below:

[0101] Model creation: Create an initial information sequence reasoning model for the information sequence reasoning scenario based on a machine learning model;

[0102] Sample data acquisition: Acquire a large amount of sample data, which is based on the historical user search click behavior data of the search engine. The sample data includes sample training data of one or more reference user objects (such as reference user attribute information, sample historical search browsing behavior sequence). The sample historical search browsing behavior sequence can also be a browsing behavior sequence chain of a reference user object.

[0103] Sample data annotation: Based on the needs of information sequence reasoning scenarios, an expert service is introduced to manually annotate the sample data with corresponding sample labels. The sample labels (also known as output result labels) include user interest representation labels and search user representation labels.

[0104] Model training process: Input sample data into the initial information sequence reasoning model for at least one round of model training to obtain the prediction output results. The prediction output results include the prediction of user interest representation and the prediction of search user representation. Based on the prediction feature data (such as the prediction of service strategy features, the prediction of service product features, and the prediction of consultation intent features) and the output result labels, the model loss value is determined by the model loss function. Based on the model loss value, the model parameters of the initial information sequence reasoning model are adjusted until the model training termination condition is met to obtain the information sequence reasoning model.

[0105] Optionally, the model's training termination conditions may include, for example, the loss function value being less than or equal to a preset loss function threshold, or the number of iterations reaching a preset threshold. Specific training termination conditions can be determined based on actual circumstances and are not specifically limited here.

[0106] It should be noted that the machine learning models involved in one or more embodiments of this specification include, but are not limited to, fitting one or more of the following machine learning models: Convolutional Neural Network (CNN) model, Deep Neural Network (DNN) model, Recurrent Neural Networks (RNN) model, embedding model, Gradient Boosting Decision Tree (GBDT) model, Logistic Regression (LR) model, etc.

[0107] Optionally, the model loss function can be the hinge loss function, cross-entropy loss function, vector distance loss function, contrastive loss function, etc.

[0108] S106: Send the user interest representation, the search user representation, and the target search request to the information recommendation engine, so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request.

[0109] Integrating user representations and search requests, the user interest representation is a vector reflecting a user's long-term interests and preferences. For example, a user might have a sustained interest in technology, travel, or literature. The search user representation is a vector that focuses more on capturing the user's current search intent and recent behavioral patterns. For example, if a user recently searched for information about healthy eating, that user is likely a young person whose daily activities are in Wuhan. The target search request is the specific search query for this user, containing keywords or phrases entered by the user, such as "low-fat recipes." This information—including the user interest representation, the search user representation, and the target search request—is sent together to the information recommendation engine, which uses this data to perform information recommendation processing for the user.

[0110] Information recommendation engines are typically information recommendation systems trained based on relevant technologies. In this specification, user interest representations and search user representations are creatively introduced as well. These are combined with the original target search request to instruct the information recommendation engine to provide information that best matches the user's personal preferences and immediate needs, thereby improving user satisfaction and the overall performance of the platform.

[0111] In some embodiments, the information recommendation engine first receives the user's interest representation, search user representation, and target search request, and then fuses and analyzes them: The recommendation engine performs fusion analysis on the received data, combining the user's long-term interests (user interest representation) and immediate needs (search user representation and target search request), and applies information recommendation algorithms to select or generate recommended content based on the fused data. The information recommendation engine attempts to find content that best matches the user's current query and historical interests. This content may include promotional information. The final generated recommendation information, including multiple information recommendation items, is displayed to the user, and may take the form of search results, content lists, promotional information, related articles, etc.

[0112] In one or more embodiments of this specification, a cyclic feedback mechanism can be introduced to continuously optimize and calibrate the information sequence reasoning model. That is, after executing the step of sending the user interest representation, the search user representation, and the target search request to the information recommendation engine, so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request, the following approach can be adopted:

[0113] Collect the user's actual search and browsing behavior sequence for the recommended information items, determine the user's actual user interest representation based on the actual search and browsing behavior sequence, and perform model calibration processing on the information sequence reasoning model based on the actual user interest representation.

[0114] To illustrate, after the recommended content is received by the user, the system begins to monitor and record the user's actual interaction data with this content, including: clicks on search results: users click on the search results provided by the recommendation system, which can directly reflect the user's interest in the recommended content; browsing behavior: the user's browsing path, dwell time and interaction actions on the website; feedback behavior: users may provide feedback directly, such as rating, commenting or indicating like / dislike through buttons;

[0115] These actual interaction data, when collected, form a sequence of actual search and browsing behaviors. This sequence can be used as feedback to fine-tune model parameters. Based on the collected search and browsing behavior sequences (e.g., by calling expert services), the user's actual interest representation can be analyzed, and the following steps can be performed:

[0116] Retrain the model: Retrain the information sequence reasoning model using the new actual interest representation as the user dataset, that is, use the actual interest representation as the output label, calculate the model loss value between the current actual interest representation and the user interest representation, and perform model fine-tuning on the information sequence reasoning model based on the model loss value.

[0117] Model fine-tuning: Making subtle adjustments to existing information sequence reasoning models, such as adjusting weights, parameters, and hyperparameters, to optimize model performance.

[0118] Cross-validation: Evaluate the effectiveness of the adjusted model through techniques such as cross-validation to ensure that the new model maintains efficient recommendation performance on different datasets.

[0119] Solution Performance Verification: A user searches for information using the term "hotel." Their recent search history includes the phrases "which hotel is good | hotel booking | hotel facilities." Based on the inference service, this user's interests are captured, resulting in user interest representations and search user representations. These representations, along with the search term "hotel" from the search request, are sent to an information recommendation engine. After competition through promoted information, please refer to [the relevant documentation / reference]. Figure 4 , Figure 4 This is a schematic diagram illustrating a scenario of promotional information recall. The information recommendation engine recalls an advertisement titled "Hotel, Hotel Renovation - Free Quote" and displays it to the user. After the user refreshes the page, please refer to... Figure 5 , Figure 5This is another scenario illustration of promotional information recall. It recalls an advertisement with the title "Ctrip official website is the only one, hotel booking is guaranteed!", which increases the diversity of recall results. Compared with fixed recall results that only apply to the target search terms, it greatly increases the click probability.

[0120] In one or more embodiments of this application, an electronic device receives a target search request input by a user, determines the user's attribute information and historical search browsing behavior sequence, and accurately constructs and determines the user's interest representation and search user representation by comprehensively utilizing the user's attribute information and historical search browsing behavior. Then, combined with the user's real-time search request, personalized information recommendation processing is performed through an information recommendation engine. This can effectively integrate the user's long-term interests and current search needs to achieve highly personalized promotional information recommendations, improve user satisfaction, and help increase the click-through rate and interaction rate of recommended content, thereby improving the platform's conversion efficiency and user stickiness.

[0121] In one feasible implementation, considering the timeliness of information recommendation, it is usually necessary to recommend information quickly. Based on this, the information sequence reasoning model is associated with the user search key value library. After the information sequence reasoning model performs search interest parsing on the recorded reference user objects, the model output results are saved to the user search representation library. Then, in the actual application process, if the current user object is a recorded reference user object, the information sequence reasoning model can directly query the user interest representation and search user representation corresponding to the current user object from the associated user search key value library.

[0122] Specifically, the process of performing search interest parsing on the user object based on the user attribute information and the historical search browsing behavior sequence to obtain user interest representation and search user representation can be carried out in the following manner:

[0123] Based on user attribute information and the historical search and browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model.

[0124] In other words, the information sequence reasoning model first queries the user search key value library to see if the recorded reference user object contains the current user object (the user attribute information contains a user identifier, such as user id). If it exists, the user interest representation and search user representation corresponding to the user object are directly determined in the user search key value library. If it does not exist, the information sequence reasoning model calls the device computing resources to perform search interest parsing processing based on the user attribute information and the historical search browsing behavior sequence, and outputs the user interest representation and search user representation.

[0125] The user search representation library includes reference user interest representations and reference search user representations corresponding to reference user objects. The reference user interest representations and reference search user representations are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user objects.

[0126] In one or more embodiments of this specification, the above-described method can ensure the timeliness of information recommendations, enable rapid information recommendations, help improve user satisfaction, increase user click-through rates and interaction rates with recommended content, thereby improving the platform's conversion efficiency and user stickiness.

[0127] Please refer to Figure 6 , Figure 6 This is a flowchart illustrating information sequence reasoning, specifically:

[0128] S2002: Establish a global graph of user browsing behavior based on the user attribute information and the historical search and browsing behavior sequence;

[0129] To establish a global graph of user browsing behavior, the first step is data integration: collecting user attribute information (such as age, gender, and geographic location) and historical search and browsing behavior data. Then, graph construction is performed: using this user attribute information and the historical search and browsing behavior sequences, a global graph is built. Nodes in this graph can represent individual users, web pages, or other interactive objects, while edges represent interactions between users and these objects (such as browsing and clicking). This graphical structure helps capture the complex relationships and patterns between user behaviors.

[0130] For example, by establishing a global user browsing behavior graph and extracting specific browsing behavior sequence chains from it, combined with user attribute information, and using an information sequence reasoning model for analysis, user interest representations and search user representations are finally generated.

[0131] S2004: Extract at least one browsing behavior sequence chain for the user object from the global user browsing behavior graph, input the user attribute information and the browsing behavior sequence chain into the information sequence reasoning model for search interest parsing processing, and output user interest representation and search user representation.

[0132] In one feasible implementation, the following implementation method may be referred to:

[0133] A2: Determine the information items in the reference historical search and browsing behavior sequence, sort the information items according to the time order to obtain a node sequence including all information item nodes, extract the information item edge set including all information item edges from the reference historical search and browsing behavior sequence according to the node sequence, establish the user edge relationship between all reference user objects and the information item nodes, and generate a user edge relationship set containing all the user edge relationships.

[0134] For example, it can be represented as: ,in The information items are sorted according to time order to obtain the aforementioned node sequence including all information item nodes; using the node sequence as a reference, a set of information item edges including all information item edges is extracted from the reference historical search and browsing behavior sequence, that is, the information items v1...v in the reference historical search and browsing behavior sequence. n Information item edges are extracted in chronological order to obtain the information item edge set: Simultaneously, user edge relationships are established between all reference user objects and the information item nodes, that is, edge relationships are established between users and items, generating a user edge relationship set containing all such user edge relationships. .

[0135] A4: A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set;

[0136] A6: Perform sequence chain splitting on the global graph of the reference user browsing behavior to obtain at least one reference browsing behavior sequence chain.

[0137] Please refer to Figure 7 , Figure 7 This is a schematic diagram of a scenario for establishing a global graph and sequence chains, such as... Figure 7 As shown, assuming A~Z represent different items (information items), the graph connection processing is performed based on the node sequence, the edge set, and the user edge relationship set to connect the nodes with edge connections into a global graph - the reference user browsing behavior global graph. At the same time, the reference user browsing behavior global graph is subjected to sequence chain splitting processing to obtain at least one reference browsing behavior sequence chain.

[0138] For example, such as Figure 7As shown, based on the random walk algorithm: starting from a node in the graph, the user moves to other nodes based on different transition probabilities or rules until a termination condition is met. The path of nodes traversed from the initial node to the end of the walk is the reference browsing behavior sequence chain. For example, starting from node A and passing through nodes B, E, and F until the walk ends, the resulting subsequence is A--B--E--F, and a reference browsing behavior sequence chain is obtained by sampling.

[0139] In one or more embodiments of this specification, a method for constructing a global graph of user browsing behavior and combining it with user attribute information, using an information sequence reasoning model to parse and generate user interest representations and search user representations, can effectively capture and analyze user behavior patterns and interest evolution. This not only improves the accuracy and personalization level of the recommendation system but also enhances its adaptability to changes in user behavior, thereby increasing user satisfaction and platform user retention, ultimately driving platform business growth and optimizing user experience.

[0140] Please refer to Figure 8 , Figure 8 This is a flowchart illustrating the process of building a user-search key-value library. For details, please refer to the following:

[0141] S3002: Obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior.

[0142] For details, please refer to the step explanations in other embodiments of this specification, which will not be repeated here.

[0143] S3004: Input the reference user attribute information and the reference browsing behavior sequence chain into the information sequence reasoning model to perform search interest parsing processing and output the reference user interest representation and the reference search user representation;

[0144] Inputting data into the information sequence reasoning model: First, reference user attribute information and reference browsing behavior sequence chains are input into the information sequence reasoning model. This input data includes the user's basic attributes (such as age, gender, location, etc.) and the user's browsing behavior (such as pages visited, links clicked, browsing order, etc.). This facilitates the model's deeper understanding of user behavior and preferences.

[0145] Output user interest representations and search user representations: The information sequence reasoning model analyzes the input data and outputs two key user representations: reference user interest representations and reference search user representations;

[0146] Reference user interest representation: is a vector that describes a user's long-term interests and preferences.

[0147] Reference search user representation: reflects the user's current search intent and immediate needs.

[0148] S3006: Create a user search key-value library associated with the information sequence reasoning model based on the reference user interest representation and the reference search user representation.

[0149] For example, creating a user search key-value library: Based on the generated reference user interest representation and reference search user representation, an associated user search key-value library is created. The design and functionality of the user search key-value library include:

[0150] Key-value correspondence: Each key corresponds to a user identifier, while the value is the reference user interest representation of the reference user object corresponding to the user identifier and the reference search user representation.

[0151] Dynamic updates: The user search key-value library does not have to be static; it can be updated in real time based on new user data and behavior, thus ensuring that the information in the library always reflects the latest user interests and needs.

[0152] Personalized search optimization: The data in the database is used to optimize the search engine's response, ensuring that search results are more closely aligned with the user's personal interests and immediate needs.

[0153] Considering the timeliness of information recommendation, rapid information recommendation is usually required. Based on this, the above method is used to associate the information sequence reasoning model with the user search key value library. After the information sequence reasoning model performs search interest parsing on the recorded reference user objects, the model output results are saved to the user search representation library. Then, in actual application, if the current user object is a recorded reference user object, the information sequence reasoning model can directly query the user interest representation and search user representation corresponding to the current user object from the associated user search key value library.

[0154] In one or more embodiments of this specification, the application of a user search key-value library can improve the efficiency and accuracy of search engines while enhancing the personalization of recommendation systems. For example, when a user submits a search request, the user search key-value library can be quickly retrieved, thereby providing more relevant and personalized search results and recommendations. A finely correlated user search key-value library with the information sequence reasoning model enables the search and recommendation system to more effectively respond to the specific needs of individual users, thereby improving user satisfaction, increasing user engagement, and ultimately driving the overall success of the platform. Furthermore, this strategy also helps to better manage and utilize user data, bringing higher operational efficiency and data-driven decision support to the platform.

[0155] Please refer to Figure 9 , Figure 9 This is a flowchart illustrating the training process of an information sequence reasoning model. It can be used to build and optimize such models to accurately predict user interest representations and search user representations. Specifically:

[0156] S4002: Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, select sample training data of at least one sample user object, and label the sample training data with output result labels, the output result labels including user interest representation labels and search user representation labels;

[0157] Data preparation and labeling

[0158] Selecting sample training data: A batch of sample user objects is selected from the reference user objects. These sample user objects have corresponding user attribute information and browsing behavior sequence chains. This data is used as the basis for model training.

[0159] Labeling Output Results: Manually or automatically label the training data samples with output results labels, including user interest representation labels and search user representation labels. These labels represent the ideal model output and are used for comparison and error calculation during training.

[0160] S4004: Based on the sample training data, perform at least one round of model training on the initial information sequence reasoning model. During the model training process, determine the prediction output results for the sample training data through the initial information sequence reasoning model. The prediction output results include predicted user interest representations and predicted search user representations.

[0161] Model training

[0162] Initial information sequence reasoning model: Use a predefined model architecture as a starting point. This model may be based on RNN, LSTM, GRU, or other deep learning structures suitable for processing sequence data.

[0163] Model training involves inputting sample data into the initial information sequence reasoning model for at least one round of training to obtain predicted output results. This is achieved by inputting sample training data into the model and generating predicted output results, including predicted user interest representations and search user representations. (The process is repeated twice in the original text.)

[0164] Predicted output: The model's predicted output is a representation of user interests and search intent inferred from the input data based on the current model parameters.

[0165] S4006: Based on the predicted output and the output label, determine the model calculation loss, and use the model calculation loss to adjust the model parameters of the initial information sequence reasoning model until the initial information sequence reasoning model finishes model training, thereby obtaining the information sequence reasoning model.

[0166] Loss Calculation and Parameter Adjustment

[0167] Calculate model loss: Use a loss function (such as mean squared error, cross-entropy, etc.) to calculate the difference between the model's predicted output and the actual labeled output. This loss value measures the model's predictive performance.

[0168] Model parameter tuning: Based on the results of the loss function, the model parameters are adjusted using the backpropagation algorithm to reduce prediction error.

[0169] Training iterations: Repeat this training process until the model's performance reaches a predetermined standard or after a certain number of training epochs. This process may include adjusting training parameters such as the learning rate and batch size.

[0170] End of training and model evaluation

[0171] End of model training: The training process ends when the model error no longer decreases significantly or when the model training termination condition is met.

[0172] Model evaluation: The trained model is evaluated using an independent test dataset to verify its predictive power and generalization ability.

[0173] For example, please see Figure 10 , Figure 10 This is a schematic diagram of an information sequence reasoning model. The graph sequence model obtains a node feature through a gated graph neural network (GGNN) and another node feature through an embedding layer (dense vector representation of other feature information). The node features are connected to obtain a comprehensive node feature representation. Then, an attention mechanism is used to capture long-term interests in the sequence. The user's short-term interests are captured through a pooling layer using the information items recently clicked by the user. Then, a linear transformation layer is used to capture the user's dynamic interests. Finally, the model calculation loss is determined based on the predicted output and the output label by optimizing the cross-entropy loss function. The model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss.

[0174] In one or more embodiments of this specification, the above-described model training method can ensure that the information sequence reasoning model can accurately capture and reflect the user's interests and search behavior. Through fine-tuning and continuous training, the model will become increasingly closer to the complexity of real user behavior, thereby improving the effectiveness of the recommendation system and user satisfaction.

[0175] The following will combine Figure 11 This application provides a detailed description of the information recommendation device provided in its embodiments. It should be noted that... Figure 11 The information recommendation device shown is used to perform the present application. Figures 1-10 The methods shown in the embodiments are illustrated for ease of explanation, showing only the parts relevant to the embodiments of this application. For specific technical details not disclosed, please refer to this application. Figures 1-10 The example shown.

[0176] Please see Figure 11 This diagram illustrates the structure of an information recommendation device according to an embodiment of this application. The information recommendation device 1 can be implemented as all or part of a device through software, hardware, or a combination of both. According to some embodiments, the information recommendation device 1 includes a request processing module 11, a characterization determination module 12, and an information recommendation module 13, specifically used for:

[0177] Request processing module 11 is used to receive the target search request input by the user object, and determine the user attribute information and historical search browsing behavior sequence of the user object;

[0178] The characterization determination module 12 is used to determine the user interest characterization and search user characterization of the user object based on the user attribute information and the historical search and browsing behavior sequence.

[0179] The information recommendation module 13 is used to send the user interest representation, the search user representation and the target search request to the information recommendation engine, so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request.

[0180] Optionally, the characterization determination module 12 is used for:

[0181] Based on the user attribute information and the historical search and browsing behavior sequence, the user object is processed by an information sequence reasoning model to obtain user interest representation and search user representation.

[0182] Optionally, the characterization determination module 12 is used for:

[0183] Based on user attribute information and the historical search browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model. The user search representation library includes reference user interest representation and reference search user representation corresponding to the reference user object. The reference user interest representation and reference search user representation are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user object.

[0184] Optionally, the characterization determination module 12 is used for:

[0185] A global graph of user browsing behavior is established based on the user attribute information and the historical search and browsing behavior sequence. At least one browsing behavior sequence chain for the user object is extracted from the global graph of user browsing behavior. The user attribute information and the browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing and processing, and the user interest representation and search user representation are output.

[0186] Optionally, the device 1 is further used for:

[0187] Obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior.

[0188] The reference user attribute information and the reference browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing processing, and reference user interest representation and reference search user representation are output. Based on the reference user interest representation and the reference search user representation, a user search key value library associated with the information sequence reasoning model is created.

[0189] Optionally, the device 1 is further used for:

[0190] Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, at least one sample user object's sample training data is selected, and the sample training data is labeled with output result tags, the output result tags including user interest representation tags and search user representation tags;

[0191] Based on the sample training data, the initial information sequence reasoning model is trained for at least one round. During the model training process, the initial information sequence reasoning model determines the predicted output results for the sample training data. The predicted output results include predicted user interest representations and predicted search user representations.

[0192] Based on the predicted output and the output label, the model calculation loss is determined, and the model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss until the initial information sequence reasoning model ends model training, thus obtaining the information sequence reasoning model.

[0193] Based on the reference user attribute information and the reference browsing behavior sequence chain, the information sequence reasoning model is used to determine the reference user interest representation and the reference search user representation.

[0194] Optionally, the device 1 is further used for:

[0195] The information items in the reference historical search and browsing behavior sequence are determined, and the information items are sorted according to the time order to obtain a node sequence including all information item nodes. The information item edge set including all information item edges is extracted from the reference historical search and browsing behavior sequence with reference to the node sequence. The user edge relationship between all reference user objects and the information item nodes is established, and a user edge relationship set including all the user edge relationships is generated.

[0196] A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set.

[0197] The global graph of reference user browsing behavior is split into sequence chains to obtain at least one reference browsing behavior sequence chain.

[0198] Optionally, the device 1 is further used for:

[0199] Collect the user's actual search and browsing behavior sequence for the recommended information items, and determine the user's actual user interest representation based on the actual search and browsing behavior sequence;

[0200] The information sequence reasoning model is calibrated based on the actual user interest representation.

[0201] It should be noted that the information recommendation device provided in the above embodiments is only illustrated by the division of the above functional modules when executing the information recommendation method. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the information recommendation device and the information recommendation method embodiments provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0202] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0203] In one or more embodiments of this application, an electronic device receives a target search request input by a user, determines the user's attribute information and historical search browsing behavior sequence, and accurately constructs and determines the user's interest representation and search user representation by comprehensively utilizing the user's attribute information and historical search browsing behavior. Then, combined with the user's real-time search request, personalized information recommendation processing is performed through an information recommendation engine. This can effectively integrate the user's long-term interests and current search needs to achieve highly personalized promotional information recommendations, improve user satisfaction, and help increase the click-through rate and interaction rate of recommended content, thereby improving the platform's conversion efficiency and user stickiness.

[0204] This application also provides a computer storage medium that can store multiple instructions, which are adapted to be loaded and executed by a processor as described above. Figures 1-10 The information recommendation method described in the illustrated embodiment can be found in the following documentation for its specific execution process. Figures 1-10 The specific details of the illustrated embodiments will not be elaborated here.

[0205] This application also provides a computer program product storing at least one instruction, which is loaded and executed by the processor as described above. Figures 1-10 The information recommendation method described in the illustrated embodiment can be found in the following documentation for its specific execution process. Figures 1-10 The specific details of the illustrated embodiments will not be elaborated here.

[0206] Please refer to Figure 12 This diagram illustrates a structural block diagram of an electronic device provided in an exemplary embodiment of this application. The electronic device in this application may include one or more components such as a processor 110, a memory 120, an input device 130, an output device 140, and a bus 150. The processor 110, memory 120, input device 130, and output device 140 may be connected via the bus 150.

[0207] Processor 110 may include one or more processing cores. Processor 110 connects to various parts of the electronic device via various interfaces and lines, and performs various functions and processes data of electronic device 100 by running or executing instructions, programs, code sets, or instruction sets stored in memory 120, and by calling data stored in memory 120. Optionally, processor 110 may be implemented using at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). Processor 110 may integrate one or more of the following: central processing unit (CPU), graphics processing unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem may also not be integrated into processor 110, but may be implemented separately through a communication chip.

[0208] The memory 120 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 120 may include a non-transitory computer-readable storage medium. The memory 120 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 120 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), instructions for implementing the various method embodiments described below, etc. The operating system may be the Android system, including systems deeply developed based on the Android system, the iOS system developed by Apple Inc., including systems deeply developed based on the iOS system, or other systems. The data storage area may also store data created by the electronic device during use, such as phonebook data, audio and video data, chat log data, etc.

[0209] See Figure 13As shown, the memory 120 can be divided into operating system space and user space. The operating system runs in the operating system space, while native and third-party applications run in the user space. To ensure that different third-party applications can achieve good running performance, the operating system allocates corresponding system resources for each application. However, different application scenarios within the same third-party application have different requirements for system resources. For example, in local resource loading scenarios, third-party applications have high requirements for disk read speed; in animation rendering scenarios, third-party applications have high requirements for GPU performance. Since the operating system and third-party applications are independent of each other, the operating system often cannot promptly perceive the current application scenario of a third-party application, resulting in the operating system's inability to adapt system resources accordingly to the specific application scenario of the third-party application.

[0210] In order for the operating system to distinguish the specific application scenarios of third-party applications, it is necessary to establish data communication between the third-party applications and the operating system. This would allow the operating system to obtain the current scenario information of the third-party applications at any time, and then perform targeted system resource adaptation based on the current scenario.

[0211] Taking the Android operating system as an example, the programs and data stored in memory 120 are as follows: Figure 14As shown, the memory 120 can store the Linux kernel layer 320, the system runtime library layer 340, the application framework layer 360, and the application layer 380. The Linux kernel layer 320, system runtime library layer 340, and application framework layer 360 belong to the operating system space, while the application layer 380 belongs to the user space. The Linux kernel layer 320 provides low-level drivers for various hardware components of the electronic device, such as display drivers, audio drivers, camera drivers, Bluetooth drivers, Wi-Fi drivers, and power management. The system runtime library layer 340 provides support for key features of the Android system through several C / C++ libraries. For example, the SQLite library provides database support, the OpenGL / ES library provides 3D graphics support, and the Webkit library provides browser kernel support. The system runtime library layer 340 also provides the Android runtime library, which mainly provides core libraries that allow developers to write Android applications using the Java language. The Application Framework Layer 360 provides various APIs that may be used when building applications. Developers can also use these APIs to build their own applications, such as activity management, window management, view management, notification management, content provider, package management, call management, resource management, and location management. At least one application runs in the Application Layer 380. These applications can be native applications that come with the operating system, such as contacts, SMS, clock, and camera apps; or third-party applications developed by third-party developers, such as games, instant messaging, and photo editing apps.

[0212] Taking the operating system as an example (iOS), the programs and data stored in memory 120 are as follows: Figure 15As shown, the iOS system includes: Core OS layer 420, Core Services layer 440, Media layer 460, and Cocoa Touch layer 480. Core OS layer 420 includes the operating system kernel, drivers, and low-level program frameworks. These low-level program frameworks provide hardware-level functionality for use by the program frameworks located in Core Services layer 440. Core Services layer 440 provides system services and / or program frameworks required by applications, such as Foundation framework, account framework, advertising framework, data storage framework, network connectivity framework, geolocation framework, motion framework, etc. Media layer 460 provides applications with audiovisual interfaces, such as interfaces related to graphics and images, audio technology, video technology, and wireless playback (AirPlay) interfaces. Cocoa Touch layer 480 provides various commonly used interface-related frameworks for application development and is responsible for user touch interaction on electronic devices. Examples include local notification services, remote push services, advertising frameworks, game tool frameworks, message user interface (UI) frameworks, UIKit user interface frameworks, map frameworks, and so on.

[0213] exist Figure 15 The framework shown includes, but is not limited to, the base framework in the core service layer 440 and the UIKit framework in the touchable layer 480. The base framework provides many basic object classes and data types, offering the most basic system services to all applications, and is independent of the UI. The UIKit framework, on the other hand, provides a basic UI class library for creating touch-based user interfaces. iOS applications can use the UIKit framework to provide their UI, thus providing the application's infrastructure for building user interfaces, drawing, handling user interaction events, responding to gestures, and so on.

[0214] The methods and principles for implementing data communication between third-party applications and the operating system in the iOS system can be referenced from the Android system, and will not be elaborated here.

[0215] The input device 130 is used to receive input instructions or data, and includes, but is not limited to, a keyboard, mouse, camera, microphone, or touch device. The output device 140 is used to output instructions or data, and includes, but is not limited to, a display device and a speaker. In one example, the input device 130 and the output device 140 can be combined into a touch screen, which is used to receive touch operations from the user using a finger, stylus, or any suitable object on or near it, and to display the user interface of various applications. The touch screen is usually located on the front panel of the electronic device. The touch screen can be designed as a full-screen, curved screen, or irregularly shaped screen. The touch screen can also be designed as a combination of a full-screen and a curved screen, or a combination of an irregularly shaped screen and a curved screen; this embodiment of the application does not limit this.

[0216] In addition, those skilled in the art will understand that the structure of the electronic device shown in the above figures does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, the electronic device may also include radio frequency circuits, input units, sensors, audio circuits, wireless fidelity (WiFi) modules, power supplies, Bluetooth modules, etc., which will not be described in detail here.

[0217] In the embodiments of this application, the executing entity for each step can be the electronic device described above. Optionally, the executing entity for each step is the operating system of the electronic device. The operating system can be Android, iOS, or other operating systems; this embodiment of the application does not limit this.

[0218] The electronic device in this embodiment may also be equipped with a display device, which can be various devices capable of display functions, such as: cathode ray tube display (CR), light-emitting diode display (LED), electronic ink screen, liquid crystal display (LCD), plasma display panel (PDP), etc. Users can use the display device on the electronic device 101 to view displayed text, images, videos, and other information. The electronic device may be a smartphone, tablet computer, gaming device, AR (Augmented Reality) device, automobile, data storage device, audio playback device, video playback device, laptop, desktop computing device, wearable devices such as electronic watches, electronic glasses, electronic helmets, electronic bracelets, electronic necklaces, electronic clothing, etc.

[0219] exist Figure 12 In the illustrated electronic device, the processor 110 can be used to call the application program stored in the memory 120 and specifically perform the following operations:

[0220] Receive the target search request input by the user object, and determine the user attribute information and historical search browsing behavior sequence for the user object;

[0221] Based on the user attribute information and the historical search and browsing behavior sequence, the user interest representation and search user representation of the user object are determined.

[0222] The user interest representation, the search user representation, and the target search request are sent to the information recommendation engine so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request.

[0223] In one embodiment, the processor 110 performs the following steps when determining the user interest representation and search user representation of the user object based on the search terms, the user attribute information, and the historical search browsing behavior sequence:

[0224] Based on the user attribute information and the historical search and browsing behavior sequence, the user object is processed by an information sequence reasoning model to obtain user interest representation and search user representation.

[0225] In one embodiment, the processor 110, when executing the process of parsing the user object's search interest based on the user attribute information and the historical search browsing behavior sequence using an information sequence reasoning model to obtain a user interest representation and a search user representation, performs the following steps:

[0226] Based on user attribute information and the historical search browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model. The user search representation library includes reference user interest representation and reference search user representation corresponding to the reference user object. The reference user interest representation and reference search user representation are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user object.

[0227] In one embodiment, the processor 110, when executing the process of parsing the user object's search interest based on the user attribute information and the historical search browsing behavior sequence using an information sequence reasoning model to obtain a user interest representation and a search user representation, performs the following steps:

[0228] A global graph of user browsing behavior is established based on the user attribute information and the historical search and browsing behavior sequence. At least one browsing behavior sequence chain for the user object is extracted from the global graph of user browsing behavior. The user attribute information and the browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing and processing, and the user interest representation and search user representation are output.

[0229] In one embodiment, the processor 110 further performs the following steps when executing the information recommendation method:

[0230] Obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior.

[0231] The reference user attribute information and the reference browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing processing, and reference user interest representation and reference search user representation are output. Based on the reference user interest representation and the reference search user representation, a user search key value library associated with the information sequence reasoning model is created.

[0232] In one embodiment, the processor 110 further performs the following steps when executing the method:

[0233] Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, at least one sample user object's sample training data is selected, and the sample training data is labeled with output result tags, the output result tags including user interest representation tags and search user representation tags;

[0234] Based on the sample training data, the initial information sequence reasoning model is trained for at least one round. During the model training process, the initial information sequence reasoning model determines the predicted output results for the sample training data. The predicted output results include predicted user interest representations and predicted search user representations.

[0235] Based on the predicted output and the output label, the model calculation loss is determined, and the model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss until the initial information sequence reasoning model finishes model training, thus obtaining the information sequence reasoning model.

[0236] In one embodiment, the processor 110, when performing the process of establishing a global graph of reference user browsing behavior based on the reference historical search browsing behavior sequence, and extracting at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior, performs the following steps:

[0237] The information items in the reference historical search and browsing behavior sequence are determined, and the information items are sorted according to the time order to obtain a node sequence including all information item nodes. The information item edge set including all information item edges is extracted from the reference historical search and browsing behavior sequence with reference to the node sequence. The user edge relationship between all reference user objects and the information item nodes is established, and a user edge relationship set including all the user edge relationships is generated.

[0238] A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set.

[0239] The global graph of reference user browsing behavior is split into sequence chains to obtain at least one reference browsing behavior sequence chain.

[0240] In one embodiment, after the processor 110 executes the step of sending the user interest representation, the search user representation, and the target search request to the information recommendation engine so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request, it further executes the following steps:

[0241] Collect the user's actual search and browsing behavior sequence for the recommended information items, and determine the user's actual user interest representation based on the actual search and browsing behavior sequence;

[0242] The information sequence reasoning model is calibrated based on the actual user interest representation.

[0243] In one or more embodiments of this application, an electronic device receives a target search request input by a user, determines the user's attribute information and historical search browsing behavior sequence, and accurately constructs and determines the user's interest representation and search user representation by comprehensively utilizing the user's attribute information and historical search browsing behavior. Then, combined with the user's real-time search request, personalized information recommendation processing is performed through an information recommendation engine. This can effectively integrate the user's long-term interests and current search needs to achieve highly personalized promotional information recommendations, improve user satisfaction, and help increase the click-through rate and interaction rate of recommended content, thereby improving the platform's conversion efficiency and user stickiness.

[0244] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0245] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. An information recommendation method, characterized in that, The method includes: Receive the target search request input by the user object, and determine the user attribute information and historical search browsing behavior sequence for the user object; Based on the user attribute information and the historical search and browsing behavior sequence, the user interest representation and search user representation of the user object are determined. The user interest representation, the search user representation, and the target search request are sent to the information recommendation engine so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request. The process of determining the user interest representation and search user representation of the user object based on the user attribute information and the historical search and browsing behavior sequence includes: Obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior. The reference user attribute information and the reference browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing processing, and the output is a reference user interest representation and a reference search user representation. Based on the reference user interest representation and the reference search user representation, a user search key value library associated with the information sequence reasoning model is created. Based on the user attribute information and the historical search browsing behavior sequence, the information sequence reasoning model is used to perform search interest parsing processing on the user object to obtain user interest representation and search user representation. The information sequence reasoning model is used to infer the user interest representation and the search user representation based on the recent behavior of the user object from the user object's historical behavior. The user interest representation is an abstract description of the user object's long-term interests and preferences and is represented in the form of feature vectors. The search user representation represents the user object's search needs, search intent, and user attributes. If the user object is a previously recorded reference user object, then the information sequence reasoning model determines the user interest representation and search user representation corresponding to the user object in the user search key value library.

2. The method according to claim 1, characterized in that, The process of using an information sequence reasoning model to analyze the search interests of the user object based on the user attribute information and the historical search and browsing behavior sequence to obtain user interest representation and search user representation includes: Based on user attribute information and the historical search browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model. The user search representation library includes reference user interest representation and reference search user representation corresponding to the reference user object. The reference user interest representation and reference search user representation are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user object.

3. The method according to claim 1, characterized in that, The process of using an information sequence reasoning model to analyze the search interests of the user object based on the user attribute information and the historical search and browsing behavior sequence to obtain user interest representation and search user representation includes: A global graph of user browsing behavior is established based on the user attribute information and the historical search and browsing behavior sequence. At least one browsing behavior sequence chain for the user object is extracted from the global graph of user browsing behavior. The user attribute information and the browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing and processing, and the user interest representation and search user representation are output.

4. The method according to claim 1, characterized in that, The method further includes: Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, at least one sample user object's sample training data is selected, and the sample training data is labeled with output result tags, the output result tags including user interest representation tags and search user representation tags; Based on the sample training data, the initial information sequence reasoning model is trained for at least one round. During the model training process, the initial information sequence reasoning model determines the predicted output results for the sample training data. The predicted output results include predicted user interest representations and predicted search user representations. Based on the predicted output and the output label, the model calculation loss is determined, and the model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss until the initial information sequence reasoning model finishes model training, thus obtaining the information sequence reasoning model.

5. The method according to claim 1, characterized in that, The step of establishing a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequence, and extracting at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior, includes: The information items in the reference historical search and browsing behavior sequence are determined, and the information items are sorted according to the time order to obtain a node sequence including all information item nodes. The information item edge set including all information item edges is extracted from the reference historical search and browsing behavior sequence with reference to the node sequence. The user edge relationship between all reference user objects and the information item nodes is established, and a user edge relationship set including all the user edge relationships is generated. A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set. The global graph of reference user browsing behavior is split into sequence chains to obtain at least one reference browsing behavior sequence chain.

6. The method according to claim 1, characterized in that, After sending the user interest representation, the search user representation, and the target search request to the information recommendation engine, so that the information recommendation engine performs information recommendation processing on the user object in response to the target search request, the process further includes: Collect the user's actual search and browsing behavior sequence for the recommended information items, and determine the user's actual user interest representation based on the actual search and browsing behavior sequence; The information sequence reasoning model is calibrated based on the actual user interest representation.

7. An information recommendation device, characterized in that, The device includes: The request processing module is used to receive the target search request input by the user object and determine the user attribute information and historical search browsing behavior sequence of the user object; The representation determination module is used to determine the user interest representation and search user representation of the user object based on the user attribute information and the historical search and browsing behavior sequence. The information recommendation module is used to send the user interest representation, the search user representation and the target search request to the information recommendation engine, so that the information recommendation engine can perform information recommendation processing on the user object in response to the target search request; The characterization determination module is specifically used to: obtain reference historical search and browsing behavior sequences and reference user attribute information corresponding to multiple reference user objects; establish a global graph of reference user browsing behavior based on the reference historical search and browsing behavior sequences; and extract at least one reference browsing behavior sequence chain for the reference user object from the global graph of reference user browsing behavior. The reference user attribute information and the reference browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing processing, and the output is a reference user interest representation and a reference search user representation. Based on the reference user interest representation and the reference search user representation, a user search key value library associated with the information sequence reasoning model is created. Based on the user attribute information and the historical search browsing behavior sequence, the information sequence reasoning model is used to perform search interest parsing processing on the user object to obtain user interest representation and search user representation. The information sequence reasoning model is used to infer the user interest representation and the search user representation based on the recent behavior of the user object from the user object's historical behavior. The user interest representation is an abstract description of the user object's long-term interests and preferences and is represented in the form of feature vectors. The search user representation represents the user object's search needs, search intent, and user attributes. If the user object is a previously recorded reference user object, then the information sequence reasoning model determines the user interest representation and search user representation corresponding to the user object in the user search key value library.

8. The apparatus according to claim 7, characterized in that, The characterization determination module is used for: Based on user attribute information and the historical search browsing behavior sequence, the user interest representation and search user representation corresponding to the user object are determined in the user search key value library through the information sequence reasoning model. The user search representation library includes reference user interest representation and reference search user representation corresponding to the reference user object. The reference user interest representation and reference search user representation are saved to the user search representation library after the information sequence reasoning model performs search interest parsing processing on the reference user object.

9. The apparatus according to claim 7, characterized in that, The characterization determination module is used for: A global graph of user browsing behavior is established based on the user attribute information and the historical search and browsing behavior sequence. At least one browsing behavior sequence chain for the user object is extracted from the global graph of user browsing behavior. The user attribute information and the browsing behavior sequence chain are input into the information sequence reasoning model for search interest parsing and processing, and the user interest representation and search user representation are output.

10. The apparatus according to claim 7, characterized in that, The device is also used for: Based on the reference user attribute information and the reference browsing behavior sequence chain of the reference user object, at least one sample user object's sample training data is selected, and the sample training data is labeled with output result tags, the output result tags including user interest representation tags and search user representation tags; Based on the sample training data, the initial information sequence reasoning model is trained for at least one round. During the model training process, the initial information sequence reasoning model determines the predicted output results for the sample training data. The predicted output results include predicted user interest representations and predicted search user representations. Based on the predicted output and the output label, the model calculation loss is determined, and the model parameters of the initial information sequence reasoning model are adjusted using the model calculation loss until the initial information sequence reasoning model ends model training, thus obtaining the information sequence reasoning model. Based on the reference user attribute information and the reference browsing behavior sequence chain, the information sequence reasoning model is used to determine the reference user interest representation and the reference search user representation.

11. The apparatus according to claim 7, characterized in that, The device is also used for: The information items in the reference historical search and browsing behavior sequence are determined, and the information items are sorted according to the time order to obtain a node sequence including all information item nodes. The information item edge set including all information item edges is extracted from the reference historical search and browsing behavior sequence with reference to the node sequence. The user edge relationship between all reference user objects and the information item nodes is established, and a user edge relationship set including all the user edge relationships is generated. A global graph of reference user browsing behavior is obtained by performing graph connection processing based on the node sequence, the edge set, and the user edge relationship set. The global graph of reference user browsing behavior is split into sequence chains to obtain at least one reference browsing behavior sequence chain.

12. The apparatus according to claim 7, characterized in that, The device is also used for: Collect the user's actual search and browsing behavior sequence for the recommended information items, and determine the user's actual user interest representation based on the actual search and browsing behavior sequence; The information sequence reasoning model is calibrated based on the actual user interest representation.

13. A computer storage medium, characterized in that, The computer storage medium stores a plurality of instructions, which are adapted to be loaded by a processor and executed as the method steps of any one of claims 1 to 6.

14. An electronic device, characterized in that, include: A processor and a memory; wherein the memory stores a computer program adapted to be loaded by the processor and executed the method steps as claimed in any one of claims 1 to 6.

Images