Data processing method and system

Abstract

Embodiments of the present specification provide a data processing method and system. In the method, when a data processing system detects that a target user performs a target behavior in a target scene, the data processing system generates a user understanding request and acquires a user portrait of a target user corresponding to a target user group grid based on the user understanding request. Subsequently, the data processing system combines scene description information in the target scene, fuses and reasons the user portrait reflecting long-term commonality and the scene description information reflecting real-time context, thereby generating user understanding information of the target user in the target scene, determining a target supply object in supply information of the target scene based on the user understanding information, and feeding back content corresponding to the target supply object to the target user.

Description

Technical Field

[0001] This specification relates to the field of data processing technology, and in particular to a data processing method and system. Background Technology

[0002] With the rapid development of artificial intelligence technology, especially with the support of large language models and multimodal reasoning models, enterprises are increasingly demanding a deep understanding and real-time response to massive amounts of user data. In scenarios such as personalized recommendations, enterprises need to build user profiles based on user data to support intelligent and personalized service delivery.

[0003] In related technologies, structured data and manually generated rules are mainly used to extract user tags for each user and construct static user profiles by matching fixed profile templates based on rules. However, when faced with complex requirements such as massive amounts of unstructured data, high real-time requirements, and adaptation to multiple scenarios, these technologies often exhibit problems such as limited data processing capabilities, high response latency, and untimely profile updates.

[0004] Therefore, there is an urgent need for a user understanding solution that can balance depth of understanding, computational efficiency, response speed, and cross-scenario adaptability.

[0005] The information in the background section is merely information known only to the inventor and does not imply that such information had entered the public domain before the date of this application, nor does it imply that it can be considered prior art in this disclosure. Summary of the Invention

[0006] This specification provides a data processing method and system that can be applied to application scenarios that require accurate and efficient personalized recommendations or contextualized service pushes to users on a large scale.

[0007] Firstly, this specification provides a data processing method, the method comprising: generating a user understanding request in response to a target user's target behavior in a target scenario; obtaining a user profile of a target user's corresponding target user grid based on the user understanding request, the user profile being generated based on common characteristics of users within the target user grid; obtaining scenario description information associated with the target behavior in the target scenario; generating user understanding information of the target user in the target scenario based on the user profile and the scenario description information; and determining a target supply object in the supply information of the target scenario based on the user understanding information, and feeding back the content corresponding to the target supply object to the target user.

[0008] In some embodiments, the scene description information includes one or more of the following: the target user's role information in the target scene, the supply information of the target scene, the target user's behavior information in the target scene, and the task information corresponding to the target scene.

[0009] In some embodiments, before obtaining the user profile of the target user corresponding to the target user's audience grid based on the user understanding request, the method further includes: obtaining user data corresponding to all users in the entire domain, performing cluster analysis on the users in the entire domain based on the similarity of the user data to obtain multiple preset audience grids, wherein the similarity of user data in the same preset audience grid satisfies a preset condition; and for each preset audience grid, generating a user profile of the preset audience grid based on the common features of the user data in the preset audience grid.

[0010] In some embodiments, generating a user profile for a preset population grid based on the common characteristics of user data in the preset population grid includes: obtaining supply data corresponding to all supply objects; performing cluster analysis on the all supply objects based on the supply data to obtain multiple preset supply grids; determining a corresponding supply profile for each preset supply grid, wherein supply objects in the same preset supply grid have common characteristics; and determining a user profile corresponding to the preset population grid based on the common characteristics of user data in the preset population grid and the supply profile corresponding to each preset supply grid.

[0011] In some embodiments, determining the user profile corresponding to the preset population grid based on the common features of user data in the preset population grid and the supply profile corresponding to each preset supply grid includes: determining a target behavior dataset in the user data contained in the preset population grid; and determining the user profile corresponding to the preset population grid based on the target behavior dataset, the common features of user data in the preset population grid, and the supply profile corresponding to each preset supply grid.

[0012] In some embodiments, determining the target behavior dataset from the user data contained in the preset crowd grid includes: sampling and / or aggregating the user behavior data contained in the preset crowd grid to obtain the target behavior dataset.

[0013] In some embodiments, the method further includes: triggering an update of the user profile corresponding to the preset population grid at a preset period; determining that an update event of a target dimension has occurred, and triggering an update of the user profile corresponding to the preset population grid, wherein the target dimension includes one or more of the following: the target behavior dataset corresponding to the preset population grid has shifted; the behavior of multiple users included in the preset population grid has changed.

[0014] In some embodiments, determining that an update event of the target dimension has occurred triggers an update of the user profile corresponding to the preset audience grid, which includes: determining that an update event of the target dimension has occurred triggers an update of all data of the user profile corresponding to the preset audience grid; or, determining that an update event of the target dimension has occurred triggers an update of the data of the target dimension of the user profile corresponding to the preset audience grid.

[0015] In some embodiments, determining that an update event of the target dimension has occurred and triggering an update of the user profile corresponding to the preset audience grid includes: detecting whether an update event of the target dimension has occurred at a detection period corresponding to the target dimension; and triggering an update of the user profile corresponding to the preset audience grid when it is determined that an update event of the target dimension has occurred.

[0016] In some embodiments, generating user understanding information of the target user in the target scenario based on the user profile and the scenario description information includes: generating user understanding information of the target user in the target scenario using a preset inference model based on the user profile and the scenario description information.

[0017] In some embodiments, the user understanding request includes: the identifier of the target user; obtaining the user profile of the target user corresponding to the target audience grid based on the user understanding request includes: determining whether a corresponding target audience grid exists in a preset audience grid according to the identifier of the target user; if it exists, obtaining the user profile corresponding to the target audience grid; if it does not exist, adding a new target audience grid corresponding to the target user, and generating the user profile corresponding to the target audience grid according to the user data corresponding to the target user.

[0018] Secondly, this specification also provides a data processing system, including at least one storage medium and at least one processor, wherein the at least one storage medium stores at least one instruction set for data processing; the at least one processor is communicatively connected to the at least one storage medium, wherein the at least one processor reads the at least one instruction set during operation and executes the method described in any of the first aspects above according to the instructions of the at least one instruction set.

[0019] As can be seen from the above technical solutions, the data processing method and system provided in this specification, when the data processing system detects that a target user is performing a target behavior in a target scenario, generates a user understanding request and obtains a user profile of the target user's corresponding target audience grid based on the user understanding request. Subsequently, the data processing system combines the scenario description information in the target scenario with the scenario description information reflecting long-term commonalities to perform fusion reasoning, thereby generating user understanding information of the target user in the target scenario. Furthermore, based on the user understanding information, the data processing system determines the target supply object from the supply information in the target scenario and feeds back the corresponding target supply object to the target user. The above data processing method, by introducing a target audience grid, transforms the complex feature calculation for a single user into an efficient retrieval of the user profile of a pre-generated target audience grid, reducing computational overhead and response latency, enabling the data processing system to handle high-concurrency requests from massive numbers of users. Furthermore, the aforementioned data processing method collaboratively analyzes static user profiles that reflect the long-term interests of all users in the target population grid with dynamic scene description information that reflects the immediate context. This allows the generated user understanding information to capture both stable user preferences and adapt to subtle changes in the target scene, thereby achieving cross-scene adaptability while ensuring depth of understanding.

[0020] The data processing methods and other functions of the system provided in this specification are partially listed in the following description. The inventive aspects of the data processing methods and systems provided in this specification can be fully explained through practice or use of the methods, apparatus, and combinations described in the detailed examples below. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram illustrating an application scenario of a data processing system provided according to an embodiment of this specification is shown. Figure 2 A schematic diagram of the hardware structure of a computing device provided according to some embodiments of this specification is shown; Figure 3 A schematic flowchart of a data processing method according to an embodiment of this specification is shown; Figure 4 A schematic diagram of a process for determining user understanding information according to an embodiment of this specification is shown; Figure 5A method for generating a user profile according to an embodiment of this specification is illustrated; Figure 6 A schematic diagram of the process for determining a user profile is shown in one embodiment of this specification; Figure 7 This specification illustrates a method for generating a user profile according to another embodiment; and Figure 8 A method for updating a user profile according to an embodiment of this specification is shown. Detailed Implementation

[0023] The following description provides specific application scenarios and requirements for this specification, intended to enable those skilled in the art to make and use the contents of this specification. Various partial modifications to the disclosed embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments and applications without departing from the spirit and scope of this specification. Therefore, this specification is not limited to the embodiments shown, but rather to the widest scope consistent with the claims.

[0024] The terminology used herein is for the purpose of describing particular exemplary embodiments only and is not restrictive. For example, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. When used in this specification, the terms “comprising,” “including,” and / or “containing” mean that the associated integers, steps, operations, elements, and / or components are present, but do not exclude the presence of one or more other features, integers, steps, operations, elements, components, and / or groups, or that other features, integers, steps, operations, elements, components, and / or groups may be added to the system / method.

[0025] Considering the following description, these and other features of this specification, as well as the operation and function of the related components of the structure, and the economy of assembly and manufacture of the parts, can be significantly improved. All of these form part of this specification with reference to the accompanying drawings. However, it should be clearly understood that the drawings are for illustrative and descriptive purposes only and are not intended to limit the scope of this specification. It should also be understood that the drawings are not drawn to scale.

[0026] The flowcharts used in this specification illustrate operations implemented according to some embodiments of this specification. It should be clearly understood that the operations in the flowcharts may not be implemented in a sequential order. Instead, the operations may be implemented in reverse order or simultaneously. Furthermore, one or more additional operations may be added to the flowcharts. One or more operations may be removed from the flowcharts.

[0027] In this specification, "X includes at least one of A, B, or C" means that X includes at least A, or X includes at least B, or X includes at least C. That is, X may include only one of A, B, and C, or any combination of A, B, and C, as well as other possible content / elements. The arbitrary combination of A, B, and C can be A, B, C, AB, AC, BC, or ABC.

[0028] In this specification, unless explicitly stated otherwise, the relationships between structures can be direct or indirect. For example, when describing "A is connected to B," unless it is explicitly stated that A and B are directly connected, it should be understood that A can be directly connected to B or indirectly connected to B. Similarly, when describing "A is on top of B," unless it is explicitly stated that A is directly above B (AB is adjacent and A is above B), it should be understood that A can be directly above B or indirectly above B (AB is separated by other elements, and A is above B). And so on.

[0029] It should be noted that the user data obtained in this manual is authorized by the user and does not involve user privacy.

[0030] For ease of description, the terms that will appear later in this manual will be explained first.

[0031] User profile: Based on multi-dimensional data and information such as user product usage, behavior trajectory, attribute characteristics, needs and preferences, and consumption habits, a typical and representative virtual user figurative model is constructed through collection, organization, analysis and modeling. It is not a real replica of a single user, but a common extraction and comprehensive description of a target user group with similar behaviors, needs and characteristics.

[0032] Crowd Grid: A system that divides a population into several structured, locationable, and quantifiable grid units based on multi-dimensional characteristics such as space, time, attributes, and behavior, forming a gridded, refined, and manageable population distribution system. By labeling and data-driven characterizing the size, structure, preferences, needs, and activity levels of the population within each grid, it enables precise stratification, regional positioning, dynamic monitoring, and refined operation of the population.

[0033] In this specification, the Large Language Model (LLM) may also be referred to simply as the Large Model. A Large Language Model is a natural language processing model based on deep learning techniques, typically with billions to hundreds of billions or even more parameters, possessing powerful language understanding and generation capabilities. Large Language Models can employ the Transformer architecture or its variants (such as GPT, BERT, etc.), which utilizes an attention mechanism to globally model sequential data, efficiently handling long-distance dependencies and thus performing excellently in natural language tasks. Large Language Models learn the statistical features and semantic relationships of language through pre-training on large-scale corpora, giving them excellent generalization capabilities. The core capabilities of Large Language Models include, but are not limited to: understanding contextual semantics, generating coherent and grammatically correct text, performing logical reasoning, and handling multi-task scenarios. Its usage typically includes two modes: direct inference and fine-tuning. In direct inference mode, the user guides the Large Language Model to generate specific outputs by designing prompts. Cue words can be task descriptions or instructions in text form, used to stimulate the semantic understanding and generation capabilities of large language models. In fine-tuning mode, large language models are further trained on small-scale datasets in specific domains to optimize their performance on specific tasks. The powerful generalization ability and flexibility of large language models make them an important tool in the field of artificial intelligence, providing efficient and accurate solutions for automated text generation and understanding.

[0034] In some embodiments, large language models can also understand and generate data from other modalities (such as visual and audio data). In this case, large language models can also be called multimodal large language models (MLLMs). MLLMs provide a richer and more natural interactive experience by integrating multiple types of input and output, such as text, images, and sound. The core advantage of MLLMs lies in their ability to process and understand information from different modalities and fuse this information to accomplish complex tasks. For example, the Vision Language Model (VLM) discussed in this specification is a branch of MLLMs; a VLM can analyze an image and generate descriptive text. In other examples, MLLMs can also generate corresponding images or videos based on text descriptions. This cross-modal understanding and generation capability makes MLLMs widely applicable in multiple fields.

[0035] It should be noted that the key technologies of large language models can be found in the detailed description in the paper "A Survey of Large Language Models" (paper number: arXiv:2303.18223v16, published on March 11, 2025, public link: https: / / doi.org / 10.48550 / arXiv.2303.18223), and will not be repeated here.

[0036] The technical solutions provided in this specification are applicable to large-scale personalized recommendation and user understanding scenarios. They aim to provide a user understanding method and system based on target audience grids for reasoning. This method can achieve efficient, accurate, real-time, and scenario-adaptive understanding and behavior prediction of large-scale user groups, thereby improving the effectiveness of personalized recommendations and user experience.

[0037] In related technologies, recommender systems require real-time feature extraction and intent inference from comprehensive user behavior data. Due to the massive volume and complex dimensions of comprehensive user data, it is typically stored in distributed databases. This necessitates loading the user's historical behavioral characteristics, profile information, and real-time context data from the database into the computing engine each time a real-time recommendation is made. However, because the data transmission rate between the database and the computing engine is constrained by network bandwidth, the large amount of data transmission and feature computation results in high latency and resource consumption, thus impacting the real-time response efficiency and user experience of the recommender system.

[0038] To address this, this specification provides a hierarchical reasoning method for user understanding, which can be executed by a data processing system. This system can cluster user data from across the entire user base into multiple predefined user groups based on common behavioral characteristics, and generate a user profile for each user group, representing the common characteristics of the user groups within that group. During the user's real-time request phase, the data processing system responds to the user's target behavior in the target scenario. After generating a user understanding request, it can directly obtain the user profile of the target user's corresponding target user group based on the request, and integrate it with the scenario description information of the target scenario to generate user understanding information for the target user in the target scenario. Based on this user understanding information, the system can then determine the target supply object corresponding to the target user among the supply objects in the target scenario.

[0039] In this way, the data processing system can perform both the time-consuming full feature calculation and deep inference processes through preprocessing, such as offline preprocessing. During the user's real-time request phase (online phase), only low-latency cache reading and lightweight fusion calculation are required, avoiding repeated processing and transmission of the original massive data for each request. This reduces the computational overhead and data access latency of real-time inference, and improves the real-time response efficiency and service throughput of personalized recommendations in large-scale user scenarios.

[0040] It should be noted that the above description of application scenarios is only one of the many usage scenarios provided in this specification. Those skilled in the art should understand that when the data processing methods and systems provided in this specification are applied to other usage scenarios, their implementation methods and technical effects are similar.

[0041] Figure 1 A schematic diagram of an application scenario 100 of a data processing system 130 provided according to an embodiment of this specification is shown.

[0042] like Figure 1 As shown, application scenario 100 includes user terminal 110, database 120 and data processing system 130.

[0043] User terminal 110 can be a smartphone, tablet, personal computer, smart wearable device, etc., on which various applications (such as e-commerce applications, content platforms, social software, etc.) are installed. Users interact with the applications through user terminal 110 and generate behavioral data.

[0044] refer to Figure 1 The data processing system 130 can be deployed on a device or cluster of devices with data processing capabilities. For example, the data processing system 130 can be deployed on physical devices such as servers, server clusters, and cloud servers. In this case, the physical device corresponding to the data processing system 130 can store data or instructions for executing the data processing methods described in this specification, and can execute or be used to execute the data or instructions. In some embodiments, the data processing system 130 is responsible for receiving user understanding requests, executing processing logic, and returning processing results. For example, if the target scenario is a recommendation scenario and the target behavior is browsing behavior, the data processing system 130 responds to the browsing behavior of the target user, obtains user understanding information of the target user in the current recommendation scenario based on the browsing behavior of the target user, and then determines at least one target supply object from the supply information of the recommendation scenario.

[0045] The data processing system 130 is also responsible for executing the hierarchical inference process, including steps such as population grid matching, supply grid enhancement, user profile inference, and generation of contextualized user understanding information. The data processing system 130 can be deployed on a server or cloud computing platform, supporting high concurrency and low latency real-time inference requests.

[0046] Database 120 is used to store global user data, global supply data, user grids and supply grids, user profiles, scenario description templates, model parameters, etc. Database 120 may include relational databases, Not Only SQL (NoSQL) databases, distributed file systems, vector databases, etc., supporting efficient access and updates of massive amounts of data.

[0047] Figure 2 A schematic diagram of the hardware structure of a computing device 200 according to some embodiments of this specification is shown. This computing device 200 can be used as... Figure 1 The data processing system 130 is described in some embodiments. When the data processing system 130 employs a device cluster, the computing device 200 can be any one of the devices in the data processing system 130.

[0048] like Figure 2 As shown, the computing device 200 includes at least one storage medium 230 and at least one processor 220. In some embodiments, the computing device 200 may further include an internal communication bus 210. In some embodiments, the computing device 200 may further include a communication port 250. In some embodiments, the computing device 200 may further include I / O components 260.

[0049] The internal communication bus 210 can connect different system components, including storage medium 230 and processor 220. I / O component 260 supports input / output between computing device 200 and other components.

[0050] Communication port 250 is used for data communication between computing device 200 and the outside world. For example, computing device 200 can connect to a network through communication port 250.

[0051] Storage medium 230 may include a data storage device. The data storage device may be a non-transitory storage medium or a temporary storage medium. For example, the data storage device may include one or more of a disk 232, a read-only storage medium (ROM) 234, or a random access storage medium (RAM) 236. Storage medium 230 also includes at least one instruction set stored in the data storage device. The instruction set is computer program code, which may include programs, routines, objects, components, data structures, procedures, modules, etc., that execute the data processing methods provided in this specification.

[0052] At least one processor 220 is communicatively connected to at least one storage medium 230 via an internal communication bus 210. The at least one processor 220 is used to execute at least one instruction set. When the data processing system 130 is running, the at least one processor 220 reads at least one instruction set and executes the data processing methods provided in this specification according to the instructions of the at least one instruction set.

[0053] Processor 220 can execute all the steps included in the data processing method. Processor 220 can be in the form of one or more processors. Processor 220 can issue execution instructions. Processor 220 may include one or more hardware processors, such as microcontrollers, microprocessors, reduced instruction set computers (RISC), application-specific integrated circuits (ASICs), application-specific instruction set processors (ASIPs), central processing units (CPUs), graphics processing units (GPUs), physical processing units (PPUs), microcontroller units, digital signal processors (DSPs), field-programmable gate arrays (FPGAs), advanced RISC machines (ARMs), programmable logic devices (PLDs), any circuit or processor capable of performing one or more functions, or any combination thereof.

[0054] For illustrative purposes only, only one processor 220 is shown in the accompanying drawings of the computing device 200. However, it should be noted that the computing device 200 may also include multiple processors. Therefore, the operation and / or method steps disclosed herein may be executed by a single processor or by multiple processors in combination, as described herein. For example, if processor 220 of the computing device 200 in this specification executes steps A and B, it should be understood that steps A and B may also be executed jointly or separately by two different processors 220 (e.g., a first processor executes step A, a second processor executes step B, or the first and second processors jointly execute steps A and B).

[0055] Figure 3 A schematic flowchart of a data processing method according to an embodiment of this specification is shown; this data processing method P300 can be executed by a data processing system 130. Figure 3 As shown, the method P300 provided in this specification may include S310-S370, wherein: S310: In response to the target user's target behavior in the target scenario, generate a user understanding request.

[0056] The target user is a specific user within the overall user base. The target scenario refers to the application environment or interface in which the target user is currently located, such as an e-commerce homepage, product details page, short video feed, search page, or customer service chat window. The target behavior refers to the specific action triggered by the target user within the target scenario, such as clicking, browsing, searching, adding to cart, making a payment, commenting, or sharing.

[0057] When a target user performs a target action in a target scenario, the data processing system captures the target action and its corresponding context information, and generates a user understanding request. The user understanding request may include one or more of the following: user ID, scene ID, action type, and related context parameters (such as timestamp, location information, device information, etc.).

[0058] S330: Based on the user understanding request, obtain the user profile of the target user corresponding to the target user's target population grid, wherein the user profile is generated based on the common characteristics of users within the target population grid.

[0059] After receiving a user's understanding request, the data processing system can retrieve the target user's target audience grid based on the request. The target audience grid is a grouping of users across the entire domain based on the similarity of their behavioral characteristics. Users within each grid share highly similar behavioral patterns, interests, or demographic attributes. Therefore, all users within each grid can share the same user profile, which reflects the common characteristics of the group rather than a precise profile of a single user.

[0060] In some embodiments, the user understanding request may include: the identifier of the target user. Based on the identifier of the target user, the data processing system determines the target audience grid corresponding to the target user's identifier within a preset audience grid, and obtains the user profile corresponding to the target audience grid.

[0061] As an example, the user profiles corresponding to the target audience grids are pre-generated and configured with a mapping table. This mapping table records the mapping relationship between each user's user identifier and the identifier of their respective audience grid. During use, the data processing system can query this mapping table using the target user's identifier to obtain the identifier of their respective audience grid, thereby determining the target audience grid among multiple preset audience grids. Of course, the above mapping relationship can also be stored using structured storage, tables, or other forms.

[0062] S350: Obtain scene description information associated with the target behavior in the target scene.

[0063] Scene description information is a semantic and structured description of the target scene, used to describe the context of the target scene, such as: the target user's role information in the target scene, the supply information of the target scene, the target user's behavior information in the target scene (including current behavior information and historical behavior information), and the task information corresponding to the target scene, etc.

[0064] Scene description information can be static and determined in a predefined way, or it can be dynamically generated.

[0065] Role information describes the role a target user plays in a target scenario. The same user may have different roles in different scenarios. For example, when browsing products, the target user may be a "buyer"; when browsing information, the target user may be a "browser"; and when asking customer service questions, the target user may be a "consultant." Role information can be inferred from the scenario type and the target user's behavioral sequence.

[0066] Supply information for a target scenario describes the set of supply objects available for user selection or interaction within that scenario, along with their characteristics. For example, in a product recommendation scenario, supply information could include the product identifier, category, price, and inventory of the candidate product pool. In a content recommendation scenario, supply information could include the identifier, tags, duration, and theme of candidate videos.

[0067] The target user's behavioral information in the target scenario is used to describe the target user's current behavior, the sequence of previous and subsequent behaviors, the intensity of the behavior, the duration of the stay, and the depth of the interaction. For example, "The target user browsed 3 sporting goods in the past 5 minutes, clicked on one of them to view details, and stayed for more than 30 seconds."

[0068] Task information corresponding to the target scenario describes the task objectives that the data processing system or the target user expects to complete in the target scenario. For example, "Task: Recommend products that users may be interested in to improve click-through rate and conversion rate"; or, "Task: Identify the user's current inquiry intent and provide accurate answers".

[0069] Scene description information can be dynamically assembled by the data processing system based on preset scene templates, or it can be generated by retrieving from the scene knowledge base based on scene identifiers and behavior types.

[0070] S370: Based on the user profile and the scene description information, generate user understanding information of the target user in the target scene.

[0071] User understanding information is a comprehensive judgment of the target user's immediate needs, intentions, preferences, emotional state, etc. in the target scenario. It is a bridge connecting the static user profile and the dynamic scenario context.

[0072] In some embodiments, the data processing system may use a preset inference model to generate user understanding information of the target user in the target scenario based on the user profile and the scenario description information.

[0073] As an example, the pre-defined inference model can be a lightweight inference model or a large language model. In other words, the process by which the data processing system generates user-understood information is a lightweight inference process. Its inputs are pre-calculated user profiles (reflecting long-term common characteristics) and real-time collected scene description information (reflecting the current specific scene), and its output is contextualized, task-oriented user-understood information.

[0074] Figure 4 A schematic diagram of a process for determining user understanding information according to an embodiment of this specification is shown, such as... Figure 4 As shown, after obtaining the user profile of the target user, the data processing system can determine the role information of the target user in the target scenario based on the target scenario, and obtain the behavioral information (scenario behavior), the supply information (scenario supply), and the task information (scenario-based task) corresponding to the target scenario. All of the above information is input into the preset inference model as scenario description information to obtain the user understanding information output by the preset inference model.

[0075] In some embodiments, the data processing system may employ a preset inference model to generate user understanding information of the target user in the target scenario based on the user profile and the scenario description information. The preset inference model may be a pre-trained lightweight inference model, a lightweight large language model, a neural network classifier, a rule engine, or a hybrid inference system, etc. The preset inference model is trained or configured to integrate contextual information from the user profile and scenario description information and output structured user understanding information. This user understanding information may be in the form of labels or vectors.

[0076] For example, in e-commerce recommendation scenarios: Input user profile: {"Interests": ["Sports Gear", "Outdoor Travel"], "Consumption Level": "Medium-High", "Active Time": "Evening"]} Input scenario description information: {"Scenario": "Product search page", "User role": "Searcher", "Behavior": "Search keyword "hiking poles"", "Task": "Recommend related products"} The user-understood information output by the pre-defined inference model might be: {"Immediate Intent": "Buy trekking poles", "Preference Features": "Lightweight, carbon fiber, foldable", "Price Sensitivity": "Medium", "Decision Stage": "Comparison and Selection"} User understanding information is usually more specific, timely, and relevant to the current task than user profiles, thus providing data support for the selection of subsequent target supply objects.

[0077] S390: Based on the user understanding information, determine the target supply object from the supply information of the target scenario, and feed back the content corresponding to the target supply object to the target user.

[0078] After obtaining user understanding information, the data processing system filters, sorts, or rearranges the supply objects in the target scenario based on the user understanding information to determine the target supply objects (such as goods, content, services, videos, advertisements, etc.) corresponding to the user understanding request.

[0079] In some embodiments, the method by which the data processing system determines the target supply object may include: matching features in the user-understood information with features of the supply object based on recommendation algorithms such as collaborative filtering, content matching, or vector similarity calculation.

[0080] Alternatively, the data processing system can directly query the supply tag index by understanding the intent tags in the user's information, and quickly recall relevant supplies.

[0081] Alternatively, the data processing system can use user-understood information as input features of the ranking model to score and rank the candidate supply list, and select the top N candidate supply objects as target supply objects, where N is an integer greater than or equal to 1.

[0082] Once the target audience is identified, the data processing system sends feedback to the target users, for example, through the user's terminal interface. As examples, feedback could take the form of a product list, video stream, image and text cards, chat replies, or advertising banners; the specific form depends on the scenario.

[0083] In summary, the data processing method and system provided in this specification involve the data processing system generating a user understanding request when it detects a target user performing a target behavior in a target scenario. Based on this request, the system obtains a user profile from the target user's corresponding target audience grid. Subsequently, the system combines the scenario description information with the user profile reflecting long-term commonalities and the scenario description information reflecting real-time context to generate user understanding information for the target user in the target scenario. Furthermore, based on this user understanding information, the system identifies the target supply object from the supply information in the target scenario and provides feedback to the target user. This data processing method, by introducing a target audience grid, transforms the complex feature calculations for individual users into efficient retrieval of user profiles from a pre-generated target audience grid, reducing computational overhead and response latency, and enabling the data processing system to handle high-concurrency requests from a massive number of users. Furthermore, the aforementioned data processing method collaboratively analyzes static user profiles that reflect the long-term interests of all users in the target population grid with dynamic scene description information that reflects the immediate context. This allows the generated user understanding information to capture both stable user preferences and adapt to subtle changes in the target scene, thereby ensuring depth of understanding while achieving cross-scene adaptability.

[0084] Figure 5 This specification illustrates a method for generating a user profile according to an embodiment of the present invention, such as... Figure 5 As shown, the data processing system can acquire user data corresponding to all users across the entire domain, perform cluster analysis on the users across the entire domain based on the similarity of the user data, and obtain multiple preset user grids. The similarity of user data in the same preset user grid meets preset conditions. For each preset user grid: the data processing system generates a user profile for the preset user grid based on the common features of the user data in the preset user grid.

[0085] For each user grid, the data processing system can generate a user profile that represents the common characteristics of all users in the preset user grid, based on the common features (or representative sample data) of the user data in the preset user grid. Continuing... Figure 5 As shown, for each preset population grid, the data processing system can use the first major language model to analyze and process the common features of user data in the preset population grid in order to obtain the user profile corresponding to the preset population grid.

[0086] The term "full-domain user" refers to all users collected from various scenarios, terminal devices, and time periods. User data includes, but is not limited to: demographic data (grade, region, occupation, etc.), behavioral data (browsing, clicks, searches, purchases, favorites, sharing, comments, logins, duration, etc.), transaction data (order amount, frequency, category, payment method, coupon usage, etc.), content interaction data (articles read, videos watched, audio listened to, creators followed, etc.), device and network data (device model, operating system, network type, IP address, etc.), and survey and feedback data (ratings, reviews, customer service conversations, satisfaction surveys, etc.).

[0087] Before clustering, the data processing system can preprocess the user data to transform it into user feature vectors suitable for clustering algorithms. Preprocessing can include one or more of the following methods: cleaning, denoising, normalization, and feature extraction. Feature extraction methods can be based on statistical indicators (such as the number of purchases in the past 30 days), embedding models (such as user vectors trained based on behavioral sequences), or manually defined feature combinations.

[0088] Clustering algorithms for clustering users across the entire user base can include K-means, DBSCAN, hierarchical clustering, spectral clustering, or neural network-based clustering methods. During the clustering process, the target number of clusters K (number of grids) can be determined using metrics such as the elbow rule, silhouette coefficient, and clustering stability; alternatively, the number of grids can be preset. Each clustering result represents a user grid, and each user grid has a unique corresponding grid identifier.

[0089] After clustering is completed, the data processing system can establish and maintain a mapping table between user identifiers and population grid identifiers, which facilitates real-time querying of the population grids corresponding to different users.

[0090] This approach allows the data processing system to replace individual user understanding with an understanding of a pre-defined population grid, reducing the number of inferences required. This eliminates the need for the system to create and store a separate user profile for each user, thus reducing resource consumption in profile storage and computation. Furthermore, by determining subsequent user understanding information based on the user profile corresponding to the pre-defined population grid, the system ensures that the determined user understanding information has certain group commonalities and is less susceptible to the influence of short-term noisy behavior from individual users, thereby providing stable results.

[0091] Furthermore, the user profiles of the preset audience grids reflect the statistical characteristics of those grids. Once the audience grids and corresponding user profiles are generated, the data processing system can store them for reuse in other audiences or target scenarios, thereby reducing the amount of data processing required in subsequent use.

[0092] Figure 6 This specification illustrates a flowchart of a user profile determination process according to an embodiment, as shown below. Figure 6 As shown, the user understanding request includes: the identifier of the target user; the data processing system can determine whether a corresponding target audience grid exists in a preset audience grid based on the identifier of the target user. If it exists, the user profile corresponding to the target audience grid is obtained. If it does not exist, a new target audience grid corresponding to the target user is added, and a user profile corresponding to the target audience grid is generated based on the user data corresponding to the target user.

[0093] The above processing method not only ensures that a corresponding user profile can be generated for each new user in a timely manner, but also provides an immediately available structured profile foundation for subsequent services such as content recommendation through the profile management method of the population grid. This improves the response speed and service adaptability of the data processing system to new users, while also adding new data dimensions to the continuous improvement of the population grid.

[0094] Figure 7 This specification illustrates a method for generating a user profile according to another embodiment, such as... Figure 7 As shown, the data processing system can acquire supply data corresponding to all supply objects across the entire domain. Subsequently, based on this supply data, the data processing system performs cluster analysis on the all-domain supply objects to obtain multiple preset supply grids, and determines a corresponding supply profile for each preset supply grid. Supply objects within the same preset supply grid share common characteristics. Furthermore, based on the common characteristics of user data in the preset population grids and the supply profile corresponding to each preset population grid, the data processing system determines the user profile corresponding to each preset population grid.

[0095] Continue as Figure 7 As shown, for each supply grid, the data processing system can use a second major language model to generate a corresponding supply profile based on all supply data within the preset supply grid. The first and second major language models can be the same major language model or different major language models.

[0096] In this context, "full-domain supply data" refers to all object data that can be interacted with by users, such as product libraries, content libraries, service lists, and advertising material libraries. Supply data typically includes multi-dimensional information such as supply identifiers, categories, tags, attributes, descriptive text, images, prices, and inventory.

[0097] Similar to user clustering, supply clustering aims to group supply objects based on their attributes, functions, content, audience, and other dimensions, forming multiple supply grids. All supply objects within each grid share a high degree of similarity and can share the same supply profile. This profile reflects the common characteristics of the grid group, rather than a precise portrait of a single supply object. After the supply grids and their corresponding supply profiles are generated, the data processing system can store them for later reuse in other supply objects or target scenarios, thereby reducing the amount of data processing required in subsequent use. The supply profile can include category tags, price ranges, style characteristics, target audience, and usage scenarios.

[0098] As an example, in an e-commerce scenario, a supply grid might be "affordable skincare products," "small home appliances," "outdoor sports equipment," or "gaming equipment." Each supply grid has a corresponding supply profile. For example, the supply profile for the "affordable skincare products" grid might be: {"Category": "Beauty and skincare," "Price range": "Mid-to-low," "Efficacy": "Moisturizing"}.

[0099] The establishment of a supply grid simplifies the subsequent understanding of user behavior. Specifically, when a user interacts with a target supply, the data processing system can obtain the supply profile of the target supply grid corresponding to the target supply object without separately analyzing the relevant information of the target supply object. The relevant information corresponding to the supply profile is then used as the supply object. By directly referencing the supply profile of the supply grid to which the target supply object belongs, the computational load in real-time inference is reduced, meeting the performance requirements of high-concurrency scenarios. This achieves the generalization and compression of supply features for massive supply objects, avoiding the resource consumption problems caused by processing fine-grained attributes of individual supplies in real-time computation, and reducing feature dimensionality and computational complexity.

[0100] The data processing system combines the supply profile corresponding to the target supply object interacting with the user with the common features of user data in a pre-defined population grid. This supplements the common features of user data with the supply understanding information contained in the supply profile, thereby enhancing the user profile based on the supply profile. Furthermore, by dividing the population grid, supply grid, and acquiring scene description information, the data processing system obtains relevant information about the target user from three dimensions: people, supply, and scene. This makes the user understanding information determined based on these three pieces of information more accurate. Moreover, user profiles and supply profiles can be generated offline for basic reasoning, while scene-related information can be acquired through lightweight online analysis. This combination of offline and online methods not only reduces the amount of data for online reasoning but also solves the problems of real-time response and cross-scene reuse.

[0101] In some embodiments, the data processing system can determine a target behavior dataset from the user data contained in the preset population grid. Then, based on the target behavior dataset, the common features of the user data in the preset population grid, and the supply profile corresponding to each preset supply grid, the data processing system determines the user profile corresponding to the preset population grid.

[0102] Since a pre-defined user grid may contain a large amount of user and behavioral data, directly processing the entire dataset is computationally expensive and inefficient. Therefore, the data processing system can sample and / or aggregate user behavior data within the pre-defined user grid to obtain the target behavior dataset. By sampling and / or aggregating, the data processing system can reduce the amount of data while preserving the group characteristics corresponding to the pre-defined user grid, resulting in a streamlined and representative target behavior dataset, thus improving the efficiency of user profile generation.

[0103] As an example, methods for sampling user behavior data may include: random sampling (randomly selecting a certain proportion of users or behavior records from a preset user grid), stratified sampling (sampling proportionally after stratifying users according to their activity level, behavior type, etc.), time window sampling (selecting behavior data within a preset time period only), and key behavior sampling (focusing on retaining conversion-related behaviors such as purchases and payments, while filtering browsing behaviors).

[0104] As an example, methods for sampling user behavior data may include: behavior statistics aggregation (summing, averaging, etc., the number of interactions, amounts, etc. of multiple users with the same supply), sequence pattern aggregation (extracting common behavior sequence patterns and replacing the original sequence with representative sequences), and feature distribution aggregation (calculating the distribution of behavior features, such as category distribution and time distribution, and replacing the original records with distribution parameters).

[0105] Specifically, for a target audience grid, the data processing system extracts historical behavior records (such as clicks, purchases, browsing, etc.) of all users within the preset audience grid to obtain a target behavior data set. Each behavior record in the target behavior data set is associated with a supply object identifier. The data processing system can map the supply object identifier to the corresponding supply grid, thereby obtaining the supply profile of that supply grid. Then, the data processing system performs statistical analysis and feature fusion on the supply profiles corresponding to all behaviors within the preset audience grid.

[0106] As an example, the data processing system can statistically analyze the interaction frequency of supply objects (supply grids) under various categories within a pre-defined audience grid, identifying high-frequency categories as interest tags corresponding to the pre-defined audience grid; or, analyze the distribution of supply price ranges to infer the purchasing power of the audience corresponding to the pre-defined audience grid; or, extract keywords from supply description text to form a set of preferred keywords corresponding to the pre-defined audience grid; or, based on behavioral sequence patterns, identify the decision-making path or content consumption habits of the audience corresponding to the pre-defined audience grid. Ultimately, the data processing system integrates the above analytical results into a structured user profile, representing the common characteristics of the pre-defined audience grid.

[0107] In some embodiments, since user behavior and scene environment are both dynamically changing, the division of the crowd grid and the content of the corresponding user profile also need to be updated regularly or irregularly to maintain the timeliness and accuracy of the preset crowd grid and the corresponding user profile.

[0108] In some embodiments, the data processing system updates the user profiles corresponding to preset audience grids in two ways: a full update (updating all data of the user profiles corresponding to the preset audience grids) or an incremental update (updating the data of the target dimension of the user profiles corresponding to the preset audience grids). In the full update method, the data processing system can use the latest global user data and supply data to re-cluster the audience grids and generate corresponding user profiles for each newly generated audience grid. Because full updates are computationally expensive, they are typically performed periodically (e.g., monthly, quarterly, semi-annually). In the incremental update method, the data processing system can regenerate user profiles only for the specific audience grid that triggered the update, based on the latest behavioral data and supply profiles of that specific audience grid, without changing the preset audience grid divisions. Incremental updates are low-cost and fast-responding, and are therefore generally used to handle update events with high real-time requirements.

[0109] Taking incremental updates of user profiles as an example, that is, only the user profiles corresponding to a preset user grid are updated. The data processing system can trigger the update of the user profiles corresponding to the preset user grid at a preset period. Alternatively, the data processing system can determine that an update event of a target dimension has occurred, triggering the update of the user profiles corresponding to the preset user grid. The target dimension includes one or more of the following: the target behavior dataset corresponding to the preset user grid has shifted; the behavior of multiple users included in the preset user grid has changed.

[0110] The data processing system ensures the timeliness and accuracy of user profiles through the aforementioned dual update mechanism. Specifically, the system can periodically trigger updates to user profiles corresponding to preset user groups within a pre-defined timeframe. This periodic triggering method automatically triggers an update to the user profile at the pre-defined timeframe (e.g., weekly, monthly, quarterly), regardless of whether the user profile has changed. This ensures that all user groups can be updated within the pre-defined time window (pre-defined timeframe), avoiding the problem of outdated and invalid profiles due to prolonged periods without updates.

[0111] Alternatively, the data processing system can trigger updates to user profiles based on update events. For example, the system can respond to sudden changes in group behavior (shifts in the target behavior dataset or movement of multiple users' behaviors) to update the user profiles corresponding to preset user grids. This event-triggered update method only updates the user profiles of the user grids that have changed, avoiding frequent full scans of all user grids and reducing computational overhead.

[0112] In some embodiments, the data processing system may determine that an update event of the target dimension has occurred, triggering an update of all data of the user profile corresponding to the preset audience grid; or, the data processing system may determine that an update event of the target dimension has occurred, triggering an update of the data of the target dimension of the user profile corresponding to the preset audience grid.

[0113] The occurrence of update events can be detected in real time by the data processing system, or it can be detected periodically by the data processing system according to a preset detection cycle. The specific detection method of update events can be flexibly adjusted according to user needs and is not limited to the embodiments given above.

[0114] In other words, when an update event is detected in the target dimension, the data processing system can trigger the regeneration of the entire user profile corresponding to the preset audience grid, that is, update the data of all dimensions in the user profile corresponding to the preset audience grid. For example, if an update event is triggered in the "Outdoor Sports Enthusiasts" grid due to a shift in the target behavior dataset, the data processing system can retrieve the latest behavior data of users in the "Outdoor Sports Enthusiasts" grid and recalculate all features in the user profile to generate a completely new user profile. That is, regardless of which dimension in the preset audience grid triggers the update, the data processing system will generate a completely new user profile corresponding to the preset audience grid. The above update method has a simple triggering logic, reducing the complexity of the update.

[0115] Alternatively, when an update event for the target dimension is detected, the data processing system can perform a partial update only on the target dimension that triggered the update within a preset user profile grid, while keeping other dimensions of the user profile within the corresponding user profile grid unchanged. Taking the "Outdoor Sports Enthusiasts" grid as an example, where an update event is triggered due to a shift in the target behavior dataset, the data processing system only recalculates the user profile portion related to the target behavior dataset, while other parts of the user profile remain unchanged. This update method only updates user profile content directly related to the target dimension, avoiding redundant calculations of irrelevant features.

[0116] In some embodiments, the data processing system may detect whether an update event of the target dimension has occurred at a detection period corresponding to the target dimension; when it is determined that an update event of the target dimension has occurred, it triggers an update of the user profile corresponding to the preset population grid.

[0117] For example, the detection cycle for shifts in the target behavior dataset corresponding to a preset audience grid can be set to once a week: the data processing system calculates the current behavior data distribution weekly, and if the change exceeds a threshold compared to the previous week, it determines that the target behavior dataset has shifted. Alternatively, the detection cycle for abnormal user behavior can be set to once a day: the data processing system scans the behavior of users within the preset audience grid daily, and if it detects group anomalies (such as a large number of users simultaneously searching for a product), it determines that abnormal user behavior has occurred. This detection method based on the detection cycle corresponding to unread targets means that the detection of update events is not real-time but periodically polled, thereby reducing the resource consumption caused by real-time computation.

[0118] Figure 8 This specification illustrates a method for updating a user profile according to an embodiment, such as... Figure 8 As shown, the data processing system can trigger an update of the user profile corresponding to a preset user profile grid when a preset period is reached. It can also trigger an update of the user profile corresponding to the preset user profile grid when the target behavior dataset corresponding to the preset user profile grid shifts. Furthermore, it can trigger an update of the user profile corresponding to the preset user profile grid when the behavior of multiple users included in the preset user profile grid changes.

[0119] As an example, when the preset period for the user profile corresponding to the preset audience grid is reached, the data processing system can determine the user profile corresponding to the preset audience grid based on the redefined target behavior dataset, the preset audience grid, and each supply profile. Alternatively, when the target behavior dataset shifts—that is, when the overall behavior patterns, preference characteristics, or interaction patterns of the user group undergo systematic changes (e.g., due to seasonal activities, product feature iterations, hot events, etc., leading to significant differences in the distribution of core indicators such as user clicks, purchases, and dwell time)—the data processing system can redefine the target behavior data set corresponding to the preset audience grid and determine the user profile corresponding to the preset audience grid based on the redefined target behavior dataset, the preset audience grid, and each supply profile. Or, if the behavior of users included in the preset audience grid changes—for example, if a user grid originally corresponding to camping enthusiasts suddenly includes many users who prefer games and exhibit abnormal behavior—the data processing system can re-divide the audience grid to re-group users with similar preferences and behavioral habits and redefine the user profile corresponding to each preset audience grid.

[0120] In some embodiments, in order to ensure that users’ online use is not affected, the steps of updating user profiles can all be triggered offline, thereby ensuring that users always have available user profiles while using the service online, and users can directly obtain the corresponding user profiles.

[0121] To improve the accuracy and adaptability of personalized recommendation results in real-time scenarios, the data processing system needs to determine the user profile of the target user grid based on the common behavioral characteristics of users within the target user grid. Then, based on the user profile and scenario description information associated with the target behavior, it generates user understanding information for the user in the target scenario. Subsequently, based on the user understanding information, the data processing system determines the target supply object from the supply information of the target scenario. In other words, the embodiments in this specification, through a hierarchical approach of "user-supply-scenario" and gridded clustering (user grid) user understanding construction, achieve the determination of the target supply object for the user in the current scenario. See the following description for details: The embodiments in this specification construct a hierarchical reasoning user understanding system based on comprehensive user behavior data and supply data. User behavior data can be abstracted into a "user-supply-scenario" triple structure: a user interacts with an item in a specific scenario (Scene). Based on this structure, the data processing system first clusters user behavior features to divide them into preset user groups and then clusters supply attributes to divide them into preset supply groups. Next, the data processing system aggregates behavioral data from the same preset user group to generate corresponding user profiles. During the real-time request phase, the data processing system combines the user profile of the target user's target user group with the scenario description information of the current scenario to generate user understanding information for the target scenario, and then determines the target supply object from the supply information of the target scenario based on this user understanding information. Although "profile" is interpreted here as an abstract representation of user characteristics, it does not mean that it only reflects static, long-term interests. Rather, it can evolve dynamically by injecting real-time contextual information. For example, in the process of e-commerce recommendation, a user's long-term profile reflects their category preferences and spending power, while current contextual information (such as browsing pages, search terms, time periods, etc.) can instantly adjust their intent judgment and preference weights, thereby achieving the fusion reasoning of "long-term memory" and "short-term context".

[0122] This specification, in another aspect, provides a computer-readable non-transitory storage medium storing at least one instruction set of executable instructions for data processing. When the at least one instruction set is executed by a processor, it instructs the processor to implement the steps of the data processing method P300 described herein. In some possible embodiments, various aspects of this specification may also be implemented as a program product comprising program code. When the program product is run on the data processing system 130, the program code causes the data processing system 130 to perform the steps of the method P300 described herein. The program product for implementing the above method may employ a portable compact disc read-only memory (CD-ROM) containing program code and may run on the data processing system 130. However, the program product of this specification is not limited thereto. In this specification, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system. The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A readable storage medium may also be any readable medium other than a readable storage medium that can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device. Program code contained on a readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination of the foregoing. Program code for performing the operations described herein may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar programming languages.

[0123] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0124] In summary, after reading this detailed disclosure, those skilled in the art will understand that the foregoing detailed disclosure may be presented by way of example only and may not be restrictive. Although not explicitly stated herein, those skilled in the art will understand that this specification requires various reasonable changes, improvements, and modifications to the embodiments. These changes, improvements, and modifications are intended to be made by this specification and are within the spirit and scope of the exemplary embodiments described herein.

[0125] Furthermore, certain terms in this specification have been used to describe embodiments of this specification. For example, "an embodiment," "an embodiment," and / or "some embodiments" mean that a particular feature, structure, or characteristic described in connection with that embodiment may be included in at least one embodiment of this specification. Therefore, it is to be emphasized and understood that two or more references to "an embodiment" or "an embodiment" or "alternative embodiment" in various parts of this specification do not necessarily refer to the same embodiment. Moreover, specific features, structures, or characteristics may be suitably combined in one or more embodiments of this specification.

[0126] It should be understood that in the foregoing description of the embodiments in this specification, various features are combined in a single embodiment, drawing, or description for the purpose of simplifying the description and to aid in understanding a feature. However, this does not mean that the combination of these features is necessary, and those skilled in the art, upon reading this specification, may readily identify some of the devices as separate embodiments. That is, the embodiments in this specification can also be understood as an integration of multiple secondary embodiments. And the content of each secondary embodiment is valid even if it contains fewer than all the features of a single foregoing disclosed embodiment.

[0127] Every patent, patent application, publication of a patent application, and other material, such as articles, books, specifications, publications, documents, and literature (excluding any related historical examination documents), cited in this disclosure is incorporated herein for all purposes, including, for example, in the specification and claims of this disclosure. However, in the event of any inconsistency or conflict between the descriptions, definitions, and / or terms used in the foregoing and those used in this disclosure, the descriptions, definitions, and / or terms used in this disclosure shall prevail.

[0128] Finally, it should be understood that the embodiments disclosed herein are illustrative of the principles of the embodiments described in this specification. Other modified embodiments are also within the scope of this specification. Therefore, the embodiments disclosed in this specification are merely examples and not limitations. Those skilled in the art can implement the applications described in this specification using alternative configurations based on the embodiments in this specification. Therefore, the embodiments in this specification are not limited to the embodiments precisely described in the applications.

Claims

1. A data processing method, wherein, The method includes: In response to the target user's target behavior in the target scenario, generate a user-understanding request; Based on the user understanding request, a user profile of the target user corresponding to the target user's target audience grid is obtained, and the user profile is generated based on the common characteristics of users within the target audience grid; Obtain scene description information associated with the target behavior in the target scene; Based on the user profile and the scene description information, user understanding information of the target user in the target scene is generated; and Based on the user understanding information, the target supply object is determined from the supply information of the target scenario, and the corresponding content of the target supply object is fed back to the target user.

2. The method according to claim 1, wherein, The scenario description information includes one or more of the following: the target user's role information in the target scenario, the supply information of the target scenario, the target user's behavior information in the target scenario, and the task information corresponding to the target scenario.

3. The method according to claim 1, wherein, Before obtaining the user profile of the target user's target audience grid based on the user understanding request, the method further includes: Obtain user data corresponding to all users across the entire domain; perform cluster analysis on the users across the entire domain based on the similarity of the user data to obtain multiple preset user grids; the similarity of user data within the same preset user grid satisfies preset conditions; and For each of the preset population grids, a user profile for the preset population grid is generated based on the common characteristics of the user data in the preset population grid.

4. The method according to claim 3, wherein, The step of generating a user profile for the preset user grid based on the common features of user data in the preset user grid includes: Obtain supply data corresponding to all supply objects across the entire domain; based on the supply data, perform cluster analysis on the all-domain supply objects to obtain multiple preset supply grids; and determine a corresponding supply profile for each preset supply grid. Supply objects within the same preset supply grid share common characteristics; and Based on the common characteristics of user data in the preset population grid and the supply profile corresponding to each preset supply grid, the user profile corresponding to the preset population grid is determined.

5. The method according to claim 4, wherein, The step of determining the user profile corresponding to the preset population grid based on the common features of user data in the preset population grid and the supply profile corresponding to each preset supply grid includes: Determine the target behavior dataset from the user data contained in the preset crowd grid; and Based on the common features of the target behavior dataset, the user data in the preset population grid, and the supply profile corresponding to each preset supply grid, the user profile corresponding to the preset population grid is determined.

6. The method according to claim 5, wherein, The step of determining the target behavior dataset from the user data contained in the preset population grid includes: The user behavior data contained within the preset population grid is sampled and / or aggregated to obtain the target behavior dataset.

7. The method according to claim 3, wherein, The method further includes: The user profile corresponding to the preset population grid is updated at a preset periodic interval. When an update event for a target dimension occurs, it triggers an update of the user profile corresponding to the preset audience grid, wherein the target dimension includes one or more of the following: The target behavior dataset corresponding to the preset crowd grid has shifted; The behavior of multiple users included in the preset user grid has changed.

8. The method according to claim 7, wherein, When the update event for the determined target dimension occurs, it triggers an update of the user profile corresponding to the preset audience grid, including: Upon the occurrence of an update event in the target dimension, all data in the user profile corresponding to the preset audience grid are updated; or, Once an update event for the target dimension is determined, the data for the target dimension of the user profile corresponding to the preset audience grid is updated.

9. The method according to claim 7, wherein, When the update event for the determined target dimension occurs, it triggers an update of the user profile corresponding to the preset audience grid, including: Using the detection period corresponding to the target dimension, detect whether the update event of the target dimension has occurred; When an update event for the target dimension is determined to occur, the user profile corresponding to the preset population grid is updated.

10. The method according to claim 1, wherein, The step of generating user understanding information for the target user in the target scenario based on the user profile and the scenario description information includes: Based on the user profile and the scene description information, a preset inference model is used to generate user understanding information of the target user in the target scene.

11. The method according to claim 1, wherein, The user understanding request includes: the identifier of the target user; The process of obtaining the user profile of the target user's target audience grid based on the user understanding request includes: Based on the identifier of the target user, determine whether there is a corresponding target audience grid in the preset audience grid; If it exists, then obtain the user profile corresponding to the target audience grid; If it does not exist, then add a target audience grid corresponding to the target user, and generate a user profile corresponding to the target audience grid based on the user data corresponding to the target user.

12. A data processing system, comprising: At least one storage medium storing at least one instruction set for data processing; as well as At least one processor is communicatively connected to the at least one storage medium, wherein the at least one processor reads the at least one instruction set during operation and executes the method according to any one of claims 1-11 as instructed by the at least one instruction set.

Images