Method and device for determining recommendation information of application program, equipment, medium and product

Abstract

The invention discloses an application recommendation information determination method and device, equipment, a medium and a product, and relates to the technical field of intelligent recommendation. The method comprises the following steps: acquiring multi-dimensional operation information generated by a target user in a current interaction process; generating a current scene label based on the multi-dimensional operation information, and fusing the current scene label with representation information of the target user to obtain a state feature of the target user; target recommendation information is determined, and a target push message carrying identification information of the target recommendation information and the trigger parameters is sent to terminal equipment matched with the target application program; receiving a target push message, and determining whether a preset trigger condition matched with the trigger parameter exists or not; and if yes, embedding the target recommendation information into the current application interface for display. According to the scheme, the recommendation information matched with the user can be accurately determined in real time and recommended, the user is prevented from being interfered by irrelevant information, and the use experience of the user is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation technology, and in particular to a method, apparatus, device, medium, and product for determining recommendation information for an application. Background Technology

[0002] With the widespread adoption of mobile internet and smart devices, applications have become the primary gateway for users to access services and information. To enhance user experience and improve business conversion efficiency, various applications have generally introduced personalized recommendation mechanisms, pushing information that matches users' interests or current needs in appropriate scenarios.

[0003] Currently, information recommendation is mainly conducted through static methods. This involves building long-term user profiles based on historical user behavior, generating recommendation results offline or at fixed intervals, and displaying them uniformly through preset entry points. This approach lacks dynamic awareness of the user's current actions and interface state, resulting in a disconnect between recommended content and real-time interaction context.

[0004] How to accurately and in real time determine and recommend information that matches users, so as to prevent users from being disturbed by irrelevant information and improve the user experience, is a key research issue in the industry. Summary of the Invention

[0005] This invention provides a method, apparatus, device, medium, and product for determining recommendation information in an application, so as to determine and recommend information that matches the user in real time and accurately, thereby preventing the user from being disturbed by irrelevant information and improving the user experience.

[0006] According to one aspect of the present invention, a method for determining recommendation information for an application is provided, the method comprising:

[0007] In response to the instructions of the target user to operate the target application, acquire multi-dimensional operation information generated by the target user during the current interaction;

[0008] Based on the multi-dimensional operation information, a current scene label corresponding to the target user is generated, and the current scene label is fused with the representation information of the target user to obtain the state characteristics of the target user;

[0009] Determine the target recommendation information that matches the state features, and send a target push message carrying the identification information and trigger parameters to the terminal device adapted to the target application through the message push module integrated in the target application;

[0010] The system receives the target push message and continuously detects the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameter. If there is, the system loads the target recommendation information based on the identification information and embeds the target recommendation information into the current application interface for display.

[0011] According to another aspect of the present invention, an apparatus for determining recommendation information for an application is provided, the apparatus comprising:

[0012] The multi-dimensional operation information acquisition module is used to acquire multi-dimensional operation information generated by the target user during the current interaction process in response to the target user's instruction to operate the target application.

[0013] The state feature determination module is used to generate a current scene label corresponding to the target user based on the multi-dimensional operation information, and to fuse the current scene label with the representation information of the target user to obtain the state features of the target user;

[0014] The recommendation information determination module is used to determine the target recommendation information that matches the state features, and send a target push message carrying the identification information and trigger parameters of the target recommendation information to the terminal device adapted to the target application through the message push module integrated in the target application;

[0015] The recommendation information display module is used to receive the target push message and continuously detect the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameter; if there is, the target recommendation information is loaded based on the identification information and the target recommendation information is embedded in the current application interface for display.

[0016] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0017] At least one processor; and

[0018] A memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the recommendation information determination method for the application described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the recommendation information determination method for an application according to any embodiment of the present invention.

[0021] According to another aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method for determining recommendation information of an application program as described in any embodiment of the present invention.

[0022] The technical solution of this invention, in response to the instructions of a target user to operate a target application, acquires multi-dimensional operation information generated by the target user during the current interaction; this helps to accurately determine the user's state characteristics; based on the multi-dimensional operation information, it generates a current scene tag corresponding to the target user, and fuses the current scene tag with the target user's representation information to obtain the target user's state characteristics; it determines target recommendation information matching the state characteristics, and sends a target push message carrying the identification information and trigger parameters of the target recommendation information to a terminal device adapted to the target application through a message push module integrated in the target application; the target recommendation information is accurately determined based on the state characteristics; the target push message is received, and the target user interface operation sequence is continuously detected to determine whether there is a preset trigger condition matching the trigger parameters; if so, the target recommendation information is loaded based on the identification information and embedded into the current application interface for display. This allows for real-time and accurate determination and recommendation of recommendation information matching the user, preventing the user from being disturbed by irrelevant information and improving the user experience.

[0023] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of a method for determining recommendation information for an application according to Embodiment 1 of the present invention;

[0026] Figure 2 This is a flowchart of a method for determining recommendation information for an application according to Embodiment 2 of the present invention;

[0027] Figure 3 This is a flowchart of a method for determining recommendation information for an application according to Embodiment 3 of the present invention;

[0028] Figure 4This is a schematic diagram of the structure of a device for determining recommendation information of an application according to Embodiment 4 of the present invention;

[0029] Figure 5 This is a schematic diagram of the structure of an electronic device that implements the method for determining recommendation information for applications according to embodiments of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] Example 1

[0033] Figure 1 This is a flowchart of a method for determining recommendation information in an application according to Embodiment 1 of the present invention. This embodiment is applicable to situations where recommendation information matching the user is determined and recommended in real time within an application. This method can be executed by a device for determining recommendation information in the application. This device can be implemented in hardware and / or software and can be configured in electronic devices such as computers, servers, or tablet computers. Figure 1 As shown, the method includes the following steps 110-140.

[0034] Step 110: In response to the target user's instruction to operate the target application, obtain multi-dimensional operation information generated by the target user during the current interaction.

[0035] The target user can be a user of any terminal device (e.g., a smartphone, wearable device, or tablet computer), which has different applications deployed on it. In this embodiment, these are referred to as target applications. For example, the target application can be a consulting application, a reading application, a financial application, or a chat application, etc., and this embodiment does not limit it.

[0036] Optionally, in this embodiment, after receiving the target user's operation instruction on the target application, such as the target user opening the target application, entering a certain function page, or triggering an interactive control, various structured operation events generated by the user in this session and their associated device operating environment parameters can be obtained in real time. For example, this may include user interface interaction behaviors (e.g., page navigation, button clicks, swiping, form input, etc.), changes in business process status (e.g., order creation, payment interruption, etc.), and terminal context information (e.g., geographical location, network type, battery level, foreground / background status, etc.).

[0037] Furthermore, the acquired data can be standardized and encapsulated through a unified data pipeline to form high-dimensional, heterogeneous, but time-aligned multi-dimensional operational information. This multi-dimensional operational information fully depicts the user's operational intent and usage context at a specific moment, and can provide basic data support for building real-time, accurate state characteristics. The entire process does not rely on network requests or server-side computation, ensuring low latency and strong privacy protection.

[0038] It should be noted that in this embodiment, the acquisition of multi-dimensional operation information is only done after user authorization, and the acquisition method complies with the relevant provisions of national laws and regulations.

[0039] Step 120: Generate a current scene label corresponding to the target user based on multi-dimensional operation information, and fuse the current scene label with the target user's representation information to obtain the target user's state characteristics.

[0040] Optionally, in this embodiment, after obtaining multi-dimensional operation information, a current scene label corresponding to the target user can be generated based on the multi-dimensional operation information, and the determined current scene label can be fused with the target user's representation information to obtain the target user's state characteristics.

[0041] Among them, the target user's representation information can be a set of structured features extracted from the target user's multi-round operation information generated during the target application's historical use. It can be used to reflect the user's stable operation habits and preferences formed in long-term interaction.

[0042] In one optional implementation of this embodiment, after obtaining multi-dimensional operation information, this information can be further parsed, and the parsing result mapped into a dense vector. This vector serves as a numerical representation of the current scene label, effectively capturing the user's immediate intent and interaction context in this session. Simultaneously, pre-built and stored representation information of the target user is loaded. Furthermore, the dense vector corresponding to the current scene label and the static feature vector corresponding to the representation information can be dimensionally aligned, and their weights dynamically adjusted based on the timeliness of the current context. A unified state feature vector is generated through weighted fusion. This state feature retains the stability reflected in historical operations while incorporating the transience reflected in the current interaction.

[0043] In one example of this embodiment, a user continuously performs the following actions in a health management application: after opening the application, they directly enter the exercise record page, view yesterday's running data, then click to start a new exercise but do not select an activity, briefly linger on multiple exercise type icons, and then exit the page. The health management system should simultaneously detect that the current time is 6:00 AM, the device is connected to headphones via Bluetooth, the network is cellular data, and the battery level is above 50%. This multi-dimensional operational information is encoded into a dynamic context vector, representing the current scenario of "having the intention to exercise in the morning but hesitating to make a decision." Simultaneously, it can retrieve the representational information formed by the user's historical jogging behavior from Monday to Friday mornings over the past two weeks; the fused state characteristics reflect a high probability of a morning run need.

[0044] Step 130: Determine the target recommendation information that matches the state characteristics, and send a target push message carrying the identification information and trigger parameters to the terminal device adapted to the target application through the message push module integrated in the target application.

[0045] Optionally, in this embodiment, after determining the state characteristics of the target user, a target push message carrying the identification information and trigger parameters of the target recommendation information can be sent to the terminal device adapted to the target application through the message push module.

[0046] In one optional implementation of this embodiment, after determining the state characteristics of the target user, these characteristics can be used as the basis for querying and matching calculations can be performed in a local or remote candidate recommendation information database. For example, the state characteristics can be normalized and converted into a fixed-dimensional query vector, which is then compared with the pre-generated semantic vectors of each candidate recommendation information to measure similarity. A candidate set with relevance scores higher than a preset threshold is then selected, and the optimal item is chosen as the target recommendation information.

[0047] Once the target recommendation information is determined, its unique identifier and trigger parameters associated with the current scene context can be extracted. These trigger parameters define the operational conditions and time window that must be met for subsequent display. Furthermore, the message push module integrated within the target application can be invoked to encapsulate the identifier, trigger parameters, and interface anchor position into a target push message according to a preset protocol, and then write this message to the local message queue.

[0048] Step 140: Receive the target push message and continuously detect the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameters; if so, load the target recommendation information based on the identification information and embed the target recommendation information into the current application interface for display.

[0049] Optionally, in this embodiment, after the target recommendation information has been selected and encapsulated into a target push message, the application does not immediately display the content. Instead, it waits to see if the user's subsequent actions meet the preset triggering conditions. Each time the user performs an action on the interface, such as clicking a button or entering a page, the type and time of this action can be compared with the triggering parameters in the previously received push message. If the current action matches the type of action required by the triggering parameters and occurs within the allowed time window, the triggering condition is considered met. At this point, the application quickly loads the matched target recommendation information based on the identification information carried in the push message and embeds it as a native component into the corresponding area of ​​the current screen according to the specified interface anchor point position.

[0050] In the example above, based on the determined state characteristics, a target recommendation message can be determined: "We detected that you often run in the morning. Do you want to start today's jogging plan with one click?", and it is encapsulated as a target push message, which includes the identification information "MORNING_RUN_TIP_01", the trigger parameter "when the user re-enters the 'Start a New Exercise' page and the operation occurs within 90 seconds of this session", and the interface anchor "top prompt bar of the exercise type selection area".

[0051] The push notification can be temporarily stored locally within the application. After 60 seconds, the user clicks the "Start a New Exercise" entry from the homepage again, re-entering the exercise type selection page. Upon this page transition, the application recognizes the action as entering the "Start a New Exercise" page and confirms that the event occurred within the 90-second window, meeting the trigger conditions defined in the push notification. Based on the identifier "MORNING_RUN_TIP_01," corresponding recommended content can be loaded, and a lightweight prompt card can be embedded above the list of exercise type icons, displaying the text: "Detected that you often run in the morning, would you like to start today's jogging plan with one click?", along with "Start Now" and "Remind Me Later" buttons. This process neither interrupts the user's workflow nor fails to provide precise guidance at a high-intent moment, achieving context-aware dynamic recommendation display.

[0052] This embodiment's solution, in response to a target user's instruction to operate the target application, acquires multi-dimensional operation information generated by the target user during the current interaction; this helps to accurately determine the user's state characteristics; based on the multi-dimensional operation information, it generates a current scene tag corresponding to the target user, and fuses the current scene tag with the target user's representation information to obtain the target user's state characteristics; it determines target recommendation information matching the state characteristics, and sends a target push message carrying the identification information and trigger parameters of the target recommendation information to a terminal device adapted to the target application through a message push module integrated in the target application; the target recommendation information is accurately determined based on the state characteristics; the target push message is received, and the target user interface operation sequence is continuously monitored to determine whether there is a preset trigger condition matching the trigger parameters; if so, the target recommendation information is loaded based on the identification information and embedded into the current application interface for display. This allows for real-time and accurate determination and recommendation of user-matched information, preventing users from being disturbed by irrelevant information and improving the user experience.

[0053] Example 2

[0054] Figure 2 This is a flowchart of a method for determining recommendation information for an application according to Embodiment 2 of the present invention. This embodiment is a further refinement of the above technical solution, and the technical solution in this embodiment can be combined with various optional solutions in one or more of the above embodiments. Figure 2 As shown, the method may include the following steps 210-260:

[0055] Step 210: By integrating the event tracking component into the target application, obtain in real time the operation events triggered by the user in the application interface, as well as the terminal device operating environment parameters adapted to the target application.

[0056] The operation events include at least one of the following: page navigation events, control interaction events, and business process state change events. Page navigation events include: page load completion events and page exit events. Control interaction events include: button click events, swipe operation events, and form input completion events. Business process state change events include: business process start events, business process interruption events, and business process successful completion events. Operating environment parameters include: location information, network connection type, device battery status, and application foreground running indicator.

[0057] Optionally, in this embodiment, after receiving the target user's operation instruction for the target application, the data tracking component integrated within the target application can automatically start and run continuously. This component is embedded in the application's interface framework and business logic layer in a non-intrusive manner, and can capture various operation events triggered by the user on the graphical interface in real time without interfering with normal functions. These events include, but are not limited to, page loading completion, page exit, button clicks, gesture swipes, and form input completion. Simultaneously, the component also collects terminal device environmental parameters closely related to the current application's operation, covering geographical location information, network connection type, device remaining battery status, and whether the application is active in the foreground. All collected data is structured and encapsulated according to a preset data model and stamped with a unified timestamp, forming highly timely and multi-dimensional operation information.

[0058] The solution in this embodiment integrates a data tracking component within the application, enabling real-time and seamless synchronous collection of interface interaction events and terminal operating environment parameters during user operations. This not only comprehensively captures the user's immediate behavioral intent and context but also ensures high data timeliness and strong consistency. Since all information is obtained locally on the device side, there is no need for frequent reporting to the server, effectively reducing network overhead and response latency. At the same time, it enhances user privacy and security, providing a high-quality, multi-dimensional input foundation for subsequent accurate and low-interference scenario-based recommendations.

[0059] Step 220: Generate current scene tags corresponding to the target user based on multi-dimensional operation information.

[0060] Optionally, in this embodiment, generating a current scene tag corresponding to the target user based on multi-dimensional operation information may include: inputting multi-dimensional operation information into a preset scene recognition rule engine; when the operation event types contained in the continuously received sequence of operation events match the event type set defined by the target scene template, the occurrence order of each operation event conforms to the temporal constraint relationship specified by the target scene template, and the time interval between any two adjacent operation events does not exceed the time window threshold set by the target scene template, activating the current scene tag corresponding to the target scene template; assigning an initial timeliness weight to the current scene tag and starting a timed decay mechanism so that the initial timeliness weight decreases over time according to a preset decay function.

[0061] The scene recognition rule engine is configured with multiple scene templates. Each scene template defines a set of operation event types, the temporal constraint relationship between each operation event, and the maximum time window threshold.

[0062] In an optional implementation of this embodiment, after obtaining the multi-dimensional operation information, the multi-dimensional operation information can be further input into a preset scene recognition rule engine. This engine is configured with multiple target scene templates, each corresponding to a recognizable user usage scenario, and explicitly defines three core matching conditions: The first condition is the event type set, i.e., which specific types of operation events the scenario must include. For example, entering a product details page, clicking "add to cart," or viewing coupons. This condition is satisfied only when all event types appearing in the user's operation sequence completely cover all elements in the set. The second condition is a temporal constraint relationship, i.e., these events must occur sequentially according to the order specified in the template, and skipping intermediate steps or executing in reverse order is not allowed. For example, entering the page first and then clicking "press," but not the other way around. The third condition is a time window threshold, i.e., the time interval between any two adjacent operation events must not exceed the maximum allowed duration set by the template, for example, 30 seconds or 60 seconds, to ensure that the entire operation sequence has behavioral coherence and contextual relevance.

[0063] When a user's actual action sequence simultaneously meets the above three conditions, it can be determined that the target scene template has been successfully matched, and its corresponding current scene tag is activated. Furthermore, an initial timeliness weight can be assigned to this tag, and a timed decay mechanism can be initiated, causing this weight to gradually decrease over time according to a preset decay function, thereby ensuring that the scene tag only affects subsequent recommendation logic within its validity period.

[0064] The solution in this embodiment introduces a multi-dimensional matching mechanism based on event type sets, temporal constraints, and time window thresholds. This mechanism can accurately identify the specific usage scenarios of users in the application, effectively distinguish similar but different operation sequences, and significantly improve the accuracy and timeliness of scenario determination. At the same time, by assigning initial timeliness weights to activated scenario tags and implementing dynamic decay, it ensures that contextual information naturally weakens over time, avoiding expired states from interfering with subsequent decisions. This provides the recommendation system with high-fidelity and highly timely real-time contextual input.

[0065] Step 230: Obtain historical operation log data generated by the target user within a preset historical time period; perform data cleaning and feature normalization processing on the historical operation log data to obtain a standardized operation feature set; determine a set of static attribute labels based on the standardized operation feature set; determine the set of static attribute labels as the representation information of the target user, and store the representation information.

[0066] The historical operation log data includes page access sequences, business operation event streams, and interaction feedback records.

[0067] In this embodiment, the preset historical time period can be the past three days, the past week, or the past month, etc., and this embodiment does not limit it.

[0068] Optionally, in this embodiment, historical operation log data accumulated by the target user within a preset historical time period can be obtained from local storage or the server. This data fully records the user's past page access sequences, business operation event flows, and interactive feedback behaviors on recommended content in the application. Furthermore, the obtained raw logs can be cleaned to remove invalid, duplicate, or abnormal records, and the retained valid operation items can be subjected to feature extraction and normalization processing. For example, different page paths can be mapped to a unified category code, operation frequencies can be converted into standardized numerical ranges, and feedback behaviors such as clicking or ignoring can be quantified into response intensity indicators, thereby forming a standardized operation feature set with consistent structure and uniform scale.

[0069] Furthermore, by using clustering, rule induction, or statistical analysis methods, a set of static attribute tags that can stably reflect users' long-term behavioral tendencies can be extracted from the standardized operational feature set. These tags could include high-frequency usage periods, preferred functional modules, typical task paths, and consistent response patterns to specific types of content. This set of static attribute tags can then be persistently stored as representation information of the target user, and subsequently fused with real-time scenario tags to construct user state features that combine stability and dynamism.

[0070] The solution in this embodiment systematically cleans, normalizes, and abstracts features from user historical operation log data to construct a stable and structured set of static attribute tags as user representation information, effectively extracting the habitual behavior patterns and preference tendencies formed by users in long-term use.

[0071] Step 240: Fuse the current scene label with the target user's representation information to obtain the target user's state characteristics.

[0072] Optionally, in this embodiment, fusing the current scene label with the target user's representation information to obtain the target user's state features may include: performing structured encoding on the current scene label to generate a dynamic context feature vector; performing dimensional alignment processing on each static attribute label in the representation information to generate a static feature vector; determining the weighted combination result of the dynamic context feature vector and the static feature vector according to a preset fusion weight strategy; and outputting the weighted combination result as the target user's state features.

[0073] The preset fusion weight strategy can dynamically allocate the fusion ratio of dynamic context feature vectors and static feature vectors based on the timeliness weight of the current scene label. For example, the timeliness weight can be directly used as the dynamic context weight coefficient, while the static user weight coefficient is set to one minus the timeliness weight, thereby ensuring that the sum of the two is one, and making the state features focus more on the real-time context when the user's behavioral intent is strong, and gradually return to long-term preference dominance after the scene popularity decays.

[0074] In an optional implementation of this embodiment, the currently identified scene tags can be structured and encoded into dense vectors of fixed dimensions. These vectors can numerically represent the specific contextual state of the user in the current interaction, forming a dynamic context feature vector. Simultaneously, multiple static attribute tags contained in the representation information determined in the above steps are uniformly mapped and aligned in dimensions. For example, discrete tags can be converted into numerical vectors of the same dimension using a predefined embedding table or hash encoding, thereby generating a static feature vector with the same dimension as the dynamic context feature vector. Further, according to a preset fusion weight strategy, the weight coefficients of the dynamic context feature vector and the static feature vector can be dynamically calculated based on the timeliness intensity of the current scene. The weight of the dynamic part increases with the freshness of the scene, while the weight of the static part decreases accordingly. Further, the two feature vectors are multiplied by their corresponding weight coefficients and then summed element-wise to obtain a weighted combination result that integrates immediate intent and long-term preferences. Finally, this combination result is determined as the target user's state feature and output as the core input basis for the subsequent recommendation matching stage.

[0075] The solution in this embodiment transforms the current scene label and user representation information into dimension-aligned dynamic context feature vectors and static feature vectors, respectively, and then combines them in a weighted manner according to a preset fusion weight strategy. This achieves the organic fusion of real-time interaction intent and long-term behavioral preferences, which helps to significantly improve the accuracy and adaptability of subsequent recommendation matching.

[0076] Optionally, in this embodiment, determining the weighted combination result of the dynamic context feature vector and the static feature vector according to a preset fusion weight strategy may include: obtaining the timeliness weight associated with the current scene label; calculating the dynamic context weight coefficient based on the timeliness weight, and setting the static weight coefficient as the complement of the dynamic context weight coefficient; multiplying the dynamic context feature vector by the dynamic context weight coefficient to obtain the first weighted feature vector; multiplying the static feature vector by the static weight coefficient to obtain the second weighted feature vector; and adding the first weighted feature vector and the second weighted feature vector element by element to generate the weighted combination result.

[0077] In an optional implementation of this embodiment, during the fusion of the current scene tag and the target user's representation information, a timeliness weight associated with the current scene tag can be obtained. This weight reflects the freshness and contextual relevance of the current scene. Further, a dynamic context weight coefficient can be calculated based on this timeliness weight, and the static weight coefficient can be set to one minus the dynamic context weight coefficient, ensuring that the sum of the two is one. Each element of the dynamic context feature vector is multiplied by the dynamic context weight coefficient to generate a first weighted feature vector; simultaneously, each element of the static feature vector is multiplied by the static weight coefficient to generate a second weighted feature vector; finally, the first weighted feature vector and the second weighted feature vector are added element-wise along the same dimension to obtain the final weighted combination result. This result retains the stability of the user's long-term behavior pattern while incorporating the immediacy of the current interaction intent, serving as the target user's state feature for subsequent recommendation matching.

[0078] The solution in this embodiment achieves refined modeling of user state by dynamically allocating the fusion ratio of dynamic context and static features based on timeliness weights and generating combined features by weighted addition of elements. It can highlight the dominant role of the current interaction signal when the user is in a high-intent real-time scenario, and smoothly regress the stable influence of long-term behavioral preferences when the scenario popularity decreases, effectively balancing the response sensitivity and decision robustness of the recommendation system.

[0079] Step 250: Determine the target recommendation information that matches the state characteristics, and send a target push message carrying the identification information and trigger parameters to the terminal device adapted to the target application through the message push module integrated in the target application.

[0080] Step 260: Receive the target push message and continuously detect the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameters; if so, load the target recommendation information based on the identification information and embed the target recommendation information into the current application interface for display.

[0081] The technical solution of this embodiment collects multi-dimensional operation information of the user's current interaction process in real time on the user side, and dynamically generates accurate current scene labels based on this information. Then, it integrates these labels with representation information reflecting the user's long-term behavioral preferences to effectively construct state features that are both timely and stable. This not only overcomes the response lag problem caused by the reliance on historical data in traditional recommendation systems, but also avoids the risk of misjudgment caused by the lack of context depth in pure real-time judgment. As a result, it significantly improves the ability to perceive the user's immediate intent and the scene adaptability of recommendation decisions.

[0082] Example 3

[0083] Figure 3 This is a flowchart of a method for determining recommendation information for an application according to Embodiment 3 of the present invention. This embodiment is a further refinement of the above technical solution, and the technical solution in this embodiment can be combined with various optional solutions in one or more of the above embodiments. Figure 3 As shown, the method may include the following steps 310-360:

[0084] Step 310: In response to the target user's instruction to operate the target application, obtain multi-dimensional operation information generated by the target user during the current interaction.

[0085] Step 320: Generate a current scene label corresponding to the target user based on multi-dimensional operation information, and fuse the current scene label with the target user's representation information to obtain the target user's state characteristics.

[0086] Step 330: Determine the target recommendation information that matches the state features.

[0087] Optionally, in this embodiment, determining the target recommendation information matching the state features may include: normalizing the state features, scaling the feature values ​​of each dimension to a preset numerical range to generate standardized state features; converting the standardized state features into a fixed-dimensional query vector; selecting candidate recommendation information from a preset recommendation strategy library and determining the recommendation semantic vector corresponding to each candidate recommendation information; the recommendation semantic vector is a fixed-dimensional vector generated by hashing the keywords and category tags of the recommendation content; determining the similarity score between the query vector and each recommendation semantic vector, and determining the target recommendation information from the candidate recommendation information based on the similarity score.

[0088] The preset recommendation strategy library can be a structured data set stored on the target application's local or remote server. It contains multiple enabled candidate recommendation information, each of which is associated with a unique identifier, content metadata, and a pre-calculated recommendation semantic vector. The content metadata includes the title, category tags, keywords, applicable scenario description, and resource address of the recommended content. The recommendation semantic vector is generated by hashing the keywords and category tags and accumulating weights in a fixed-dimensional space, which is used to support efficient similarity matching with the query vector.

[0089] In one optional implementation of this embodiment, after obtaining the target user's state features, they can be normalized by linearly scaling the feature values ​​of each dimension to a preset uniform numerical range, such as 0-1, to eliminate computational bias caused by differences in feature scales and generate standardized state features. Further, these standardized state features can be converted into fixed-dimensional query vectors through projection or truncation padding to ensure that their semantic representations are in the same vector space as the semantic representations of all candidate options in the recommendation library. Further, all available candidate recommendation information can be read from a preset recommendation strategy library, and a corresponding recommendation semantic vector can be generated for each candidate. This vector is a fixed-dimensional dense vector obtained by hashing or aggregating the textual metadata such as keywords, category tags, and business attributes contained in the recommendation content. The similarity score between the query vector and each recommendation semantic vector is further calculated, and all candidate recommendation information is sorted according to the similarity score. The candidate with the highest score exceeding a preset threshold is selected as the target recommendation information.

[0090] In this embodiment, by standardizing the state features and transforming them into query vectors, and performing efficient similarity matching with the recommendation semantic vectors, rapid filtering and accurate positioning of massive candidate recommendation information are achieved. This not only avoids the latency and resource overhead caused by real-time calls to complex models or frequent access to the server, but also effectively captures the semantic relationship between the recommended content and the user's state through structured semantic vector expression.

[0091] In one optional implementation of this embodiment, obtaining candidate recommendation information from a preset recommendation strategy library and determining the recommendation semantic vector corresponding to each candidate recommendation information may include:

[0092] Read all enabled candidate recommendation information from the recommendation strategy library; each candidate recommendation information is associated with structured metadata; for each candidate recommendation information, parse the structured metadata, perform hash operations on each text field in the structured metadata to generate the corresponding integer index; initialize an all-zero vector according to the preset vector dimension length; take the modulo of the integer index with the vector dimension length to obtain the target dimension position, and accumulate the preset weight value at the target dimension position to obtain the recommendation semantic vector.

[0093] Optionally, in this embodiment, all enabled candidate recommendation information can be read from a preset recommendation strategy library. Each candidate recommendation information is associated with structured metadata, which includes text fields such as title, category tags, keywords, and applicable scenario descriptions. For each candidate recommendation information, each text field in its structured metadata is parsed sequentially, and a hash operation is performed on each text field to generate a corresponding integer index. Further, a zero-based vector can be initialized according to a preset vector dimension length; each integer index is moduloed by the vector dimension length to obtain the corresponding target dimension position, and a preset weight value is accumulated at that target dimension position; if multiple fields are mapped to the same dimension, the value of that dimension is the sum of the accumulated results; the resulting non-zero sparse vector is the recommendation semantic vector of the candidate recommendation information.

[0094] The solution in this embodiment efficiently converts the structured metadata of candidate recommendation information into fixed-dimensional recommendation semantic vectors, significantly reducing storage overhead and computational complexity. It effectively balances vectorization efficiency, memory usage, and matching accuracy, providing a reliable foundation for subsequent low-latency, high-concurrency similarity retrieval.

[0095] Optionally, in this embodiment, determining the similarity score between the query vector and each recommended semantic vector, and determining the target recommended information from the candidate recommended information based on the similarity score, may include: normalizing the query vector and each recommended semantic vector respectively, and calculating the cosine similarity between the normalized query vector and each recommended semantic vector to obtain each similarity score; comparing each similarity score with a preset matching threshold, filtering out candidate recommended information with similarity scores higher than the preset matching threshold, and forming a set of effective recommended information; if the set of effective recommended information is not empty, then selecting the candidate recommended information with the highest similarity score from the set of effective recommended information as the target recommended information.

[0096] In one optional implementation of this embodiment, the query vector can be normalized to convert it into a unit-length vector. Furthermore, the recommendation semantic vector corresponding to each candidate recommendation information in the recommendation strategy library is also normalized, ensuring all vectors are in a unified metric space. Based on this, the cosine similarity between the normalized query vector and each normalized recommendation semantic vector is calculated sequentially. Specifically, this is achieved by multiplying the corresponding elements of each dimension of the two vectors and summing the results to obtain a numerical value reflecting the degree of semantic matching, i.e., a similarity score. After all calculations are completed, each similarity score is compared with a preset matching threshold, and only candidate recommendation information with scores higher than the threshold is retained, forming a set of effective recommendation information.

[0097] Furthermore, if the set of valid recommendation information is not empty, the candidate recommendation information with the highest similarity score is selected as the target recommendation information; if the set of valid recommendation information is empty, it is determined that there is not enough matching content at present, and the subsequent push process can be terminated.

[0098] The solution in this embodiment uses normalized vectors and cosine similarity for filtering to ensure that only candidate recommendation information that is highly relevant to the user's current state can enter the subsequent process, effectively avoiding unnecessary push logic triggered by low-relevance content; at the same time, it selects the candidate with the highest score as the target recommendation information, providing clear and high-confidence input for subsequent identifier extraction, trigger parameter binding and embedded display.

[0099] Step 340: Send a target push message carrying the identification information and trigger parameters of the target recommendation information to the terminal device adapted to the target application through the message push module integrated in the target application.

[0100] Optionally, in this embodiment, sending a target push message carrying the identification information and trigger parameters of the target recommendation information to a terminal device adapted to the target application through the message push module integrated in the target application may include: obtaining the identification information of the target recommendation information, the trigger parameters associated with the current scene tag, and the anchor point information of the display interface; encapsulating the identification information and the trigger parameters into a target push message according to a preset message protocol; temporarily storing the target push message in a local message queue through the message push module, and starting a delay detection task.

[0101] Among them, the delay detection task continuously detects the user operation event stream within the delay period of the recommendation information display, so as to be used for subsequent trigger condition matching.

[0102] In an optional implementation of this embodiment, after determining that the target recommendation message has been obtained, the unique identifier information of the recommendation information, the trigger parameters bound to the current scene tag, and the display anchor point information used for interface positioning can be further obtained. Furthermore, the above data can be organized and encapsulated into a structured push message body according to a preset message protocol to form the target push message.

[0103] In this embodiment, the generated target push message is not immediately triggered for display. Instead, it is temporarily stored in a local message queue by the message push module integrated into the target application. At the same time, a delay detection task is started. This task continuously captures the user's subsequent operation event stream within the preset delay time for displaying recommended information, including interactive behaviors such as page jumps, control clicks, and changes in business status, providing real-time input basis for the next stage to determine whether the triggering conditions are met.

[0104] The solution in this embodiment manages pending tasks through a local message queue, avoiding frequent thread creation or blocking of the main thread, thus improving application performance and stability. The delay detection task runs only within a limited time window, ensuring timely response to subsequent user operations while preventing long-term resource occupation.

[0105] Step 350: Receive the target push message and continuously detect the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameters.

[0106] Optionally, in this embodiment, after receiving the target push message determined above, the target application can continue to detect the target user interface operation sequence and determine whether there is a preset triggering condition that matches the triggering parameters based on the detected target user interface operation sequence.

[0107] In one optional implementation of this embodiment, receiving a target push message and continuously detecting the target user interface operation sequence to determine whether there is a preset trigger condition matching the trigger parameters may include: parsing the target push message and extracting the trigger parameters, which include the target operation event type and the recommendation information display delay duration; continuously acquiring the target user's operation event stream during the target application's operation, the operation event stream containing multiple operation events arranged in chronological order, each operation event carrying an event type identifier and an occurrence timestamp; extracting a subsequence of operation events within the most recent preset duration from the operation event stream; determining whether there is at least one operation event in the subsequence whose event type identifier is consistent with the target operation event type, and whose occurrence timestamp is no more than a time window threshold from the current system time; if so, determining that the preset trigger condition matching the trigger parameters is met.

[0108] Optionally, in this embodiment, after receiving the aforementioned target push message, the target application can parse the target push message and extract predefined trigger parameters from it. In this embodiment, the target operation event type can be used to identify the type of operation expected to be performed by the user, and the recommendation information display delay duration represents the time interval between meeting the trigger conditions and actually displaying the recommended content.

[0109] While the target application is running in the foreground, a stream of user-generated action events can be continuously monitored. This stream consists of multiple action events arranged chronologically, each carrying an event type identifier and a timestamp. To focus on recent user behavior, a subsequence of action events within a preset time period closest to the current system time can be dynamically extracted from this stream.

[0110] Furthermore, the subsequence of operation events can be traversed to determine if there exists at least one operation event whose event type identifier matches the extracted target operation event type, and the difference between the timestamp of the operation event and the current system time does not exceed a preset time window threshold. If an operation event that meets both of the above conditions exists, it is determined that the preset triggering condition matching the triggering parameters has been met, and the delayed display process of recommendation information can be started accordingly.

[0111] The solution in this embodiment effectively avoids accidental or expired triggering by combining semantic matching of event types with timeliness constraints, thereby improving the contextual relevance of the recommendation display and the user experience.

[0112] Step 360: Load target recommendation information based on the identification information, and embed the target recommendation information into the current application interface for display.

[0113] Optionally, in this embodiment, loading target recommendation information based on identification information and embedding the target recommendation information into the current application interface for display may include: obtaining recommendation resource data of the target recommendation information from a local recommendation cache or a remote recommendation resource server according to the identification information, wherein the recommendation resource data includes the Uniform Resource Locator (URL) of the recommendation material and the rendering template identifier; parsing the rendering template identifier and loading the corresponding lightweight recommendation rendering component; locating the target container view in the view hierarchy of the current application interface according to the display interface anchor information carried in the target push message; passing the URL of the recommendation material to the recommendation rendering component, which generates a recommendation view instance; and adding the recommendation view instance to the target container view to complete the embedded display of the target recommendation information in the current application interface.

[0114] In this embodiment, the display of target recommendation information is not achieved through static hard-coding, but rather through a dynamic resource loading and templated rendering mechanism. The identifier information serves as a unique index for the recommended content, efficiently locating the corresponding recommendation resource data between local caches or remote servers, avoiding duplicate network requests and improving response speed. The recommendation resource data contains two core elements: first, the URL of the recommended material, i.e., an accessible network address pointing to the actual image, video, or structured content; and second, the rendering template identifier. This identifier does not directly define the interface style but maps to a type of lightweight rendering component pre-registered on the client. Each component encapsulates layout strategies, interaction logic, and data binding rules for specific business scenarios. The display interface anchor point information is a logical marker pre-embedded during the application interface design phase, used to accurately locate container nodes where recommended content can be inserted within complex view hierarchies, ensuring that the recommended view is embedded in a position that meets user experience expectations. Furthermore, by having the material URL processed by the corresponding rendering component, the component internally completes resource retrieval, content parsing, and view construction, generating an independent recommendation view instance, which is then mounted to the target container, achieving seamless integration of the recommended content and the native interface.

[0115] In one optional implementation of this embodiment, upon receiving a push message containing recommendation intent, corresponding recommendation resource data can be retrieved from a local cache or a remote recommendation resource server based on the identification information carried therein. Further, a lightweight recommendation rendering logic matching the rendering template identifier can be loaded; simultaneously, based on the display interface anchor point information specified in the push message, the target container view for carrying the recommendation content is located in the view hierarchy of the current application interface. Further, the Uniform Resource Locator (URL) of the recommendation material is passed to the loaded rendering logic, which initiates a network request to obtain the actual material and generates a specific recommendation view instance. Finally, the recommendation view instance can be added to the target container view, enabling the recommendation content to be embedded and displayed at a specified location on the application interface, without requiring page redirection or reloading of the main interface.

[0116] The solution in this embodiment combines identifier-driven resource acquisition, templated rendering logic, and interface anchor point positioning mechanism to achieve accurate, dynamic, and embedded display of recommended content in the application interface, thereby improving the user experience.

[0117] To better understand the recommendation information determination method for the application involved in this embodiment, an example of advertising is used below for illustration. The main steps include: continuously collecting user operation data from multiple sources such as mobile device sensors, positioning modules, and user operation logs; and using a stream processing engine to analyze events such as financial operations, geographical location changes, and page browsing trajectories in real time. Based on this, combining the user's long-term profile and the current session context, fine-grained real-time scenarios are dynamically constructed. For example, when it is detected that a user is in a shopping mall area and continuously browsing electronic product detail pages, an interest scenario for electronic products in the mall is generated. Further, based on this scenario, the most relevant advertising content is matched from the advertising strategy library, and a silent message carrying personalized parameters is sent to the terminal through the integrated push notification SDK. This message may not display a notification bar alert, but it contains triggering conditions and advertising configuration information. When the user performs a specific operation within the application or meets preset interaction conditions, the client parses the parameters in the message and dynamically loads and displays the corresponding advertising pop-up or animation at a specified interface anchor point. Simultaneously, the user's subsequent interactions are fed back to the server in real time to verify the advertising effect and provide feedback for optimizing the scenario recognition model and delivery strategy. The entire process achieves a high degree of coupling between the advertising content and the user's current context.

[0118] Example 4

[0119] Figure 4 This is a schematic diagram of the structure of a device for determining recommendation information for an application according to Embodiment 4 of the present invention. Figure 4 As shown, the device 400 for determining the recommendation information of an application includes: a multi-dimensional operation information acquisition module 410, a state feature determination module 420, a recommendation information determination module 430, and a recommendation information display module 440.

[0120] The multi-dimensional operation information acquisition module 410 is used to acquire multi-dimensional operation information generated by the target user during the current interaction in response to the target user's instruction to operate the target application.

[0121] The state feature determination module 420 is used to generate a current scene label corresponding to the target user based on multi-dimensional operation information, and to fuse the current scene label with the target user's representation information to obtain the target user's state features;

[0122] The recommendation information determination module 430 is used to determine the target recommendation information that matches the state features, and send a target push message carrying the identification information and trigger parameters of the target recommendation information to the terminal device adapted to the target application through the message push module integrated in the target application;

[0123] The recommendation information display module 440 is used to receive target push messages and continuously detect the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameters; if so, it loads target recommendation information based on the identification information and embeds the target recommendation information into the current application interface for display.

[0124] The solution in this embodiment can determine and recommend information that matches the user in real time and accurately, so as to prevent the user from being disturbed by irrelevant information and improve the user experience.

[0125] In an optional implementation of this embodiment, the multi-dimensional operation information acquisition module 410 is used to acquire, in real time, operation events triggered by the user in the application interface and terminal device operating environment parameters adapted to the target application through the tracking component integrated into the target application.

[0126] The operation events include at least one of the following: page navigation events, control interaction events, and business process state change events;

[0127] Page navigation events include: page load complete event and page exit event;

[0128] Control interaction events include: button click events, swipe operation events, and form input completion events;

[0129] Business process status change events include: business process start events, business process interruption events, and business process successful completion events;

[0130] Operating environment parameters include: location information, network connection type, device battery status, and application foreground running status.

[0131] In an optional implementation of this embodiment, the state feature determination module 420 includes a scene label determination submodule, which is used to input multi-dimensional operation information into a preset scene recognition rule engine; wherein, the scene recognition rule engine is configured with multiple scene templates, and each scene template defines a set of operation event types, the temporal constraint relationship between each operation event, and the maximum time window threshold;

[0132] When the types of operation events contained in the continuously received sequence of operation events match the set of event types defined by the target scene template, the order of occurrence of each operation event conforms to the temporal constraints specified by the target scene template, and the time interval between any two adjacent operation events does not exceed the time window threshold set by the target scene template, the current scene label corresponding to the target scene template is activated.

[0133] Assign an initial timeliness weight to the current scene label and start a timed decay mechanism so that the initial timeliness weight decreases over time according to a preset decay function.

[0134] In an optional implementation of this embodiment, the device for determining the recommendation information of the application further includes: a characterization information determination submodule, used to obtain historical operation log data generated by the target user within a preset historical time period; the historical operation log data includes page access sequences, business operation event streams, and interaction feedback records;

[0135] Perform data cleaning and feature normalization on historical operation log data to obtain a standardized operation feature set;

[0136] Determine the set of static attribute labels based on the standardized operational feature set;

[0137] The set of static attribute tags is determined as the representation information of the target user, and the representation information is stored.

[0138] In an optional implementation of this embodiment, the state feature determination module 420 is specifically used to perform structured encoding on the current scene label to generate a dynamic context feature vector;

[0139] Perform dimension alignment on each static attribute label in the representation information to generate a static feature vector;

[0140] Based on the preset fusion weight strategy, the weighted combination result of the dynamic context feature vector and the static feature vector is determined;

[0141] The weighted combination result is determined as the state characteristics of the target user and output.

[0142] In an optional implementation of this embodiment, the state feature determination module 420 includes: a weighted combination result determination submodule, used to obtain the timeliness weight associated with the current scene label;

[0143] The dynamic context weight coefficient is calculated based on the timeliness weight, and the static weight coefficient is set as the complement of the dynamic context weight coefficient;

[0144] Multiply the dynamic context feature vector by the dynamic context weight coefficient to obtain the first weighted feature vector;

[0145] Multiplying the static feature vector by the static weight coefficient yields the second weighted feature vector;

[0146] The first weighted feature vector and the second weighted feature vector are added element by element to generate a weighted combination result.

[0147] In an optional implementation of this embodiment, the recommendation information determination module 430 is specifically used to normalize the state features, scale the feature values ​​of each dimension to a preset numerical range, and generate standardized state features.

[0148] Convert standardized state features into a fixed-dimensional query vector;

[0149] The system retrieves candidate recommendation information from a pre-defined recommendation strategy library and determines the corresponding recommendation semantic vector for each candidate recommendation. The recommendation semantic vector is a fixed-dimensional vector generated by hashing the keywords and category tags of the recommended content.

[0150] Determine the similarity score between the query vector and each recommended semantic vector, and determine the target recommendation information from the candidate recommendation information based on the similarity score.

[0151] In an optional implementation of this embodiment, the recommendation information determination module 430 is further configured to read all enabled candidate recommendation information from the recommendation strategy library; wherein each candidate recommendation information is associated with structured metadata;

[0152] For each candidate recommendation, the structured metadata is parsed, and a hash operation is performed on each text field in the structured metadata to generate the corresponding integer index.

[0153] Initialize a vector of all zeros based on the preset vector dimension length;

[0154] The integer index is modulo the length of the vector dimension to obtain the target dimension position, and a preset weight value is accumulated at the target dimension position to obtain the recommended semantic vector.

[0155] In an optional implementation of this embodiment, the recommendation information determination module 430 is further configured to normalize the query vector and each recommendation semantic vector respectively, and calculate the cosine similarity between the normalized query vector and each recommendation semantic vector respectively to obtain each similarity score;

[0156] Each similarity score is compared with a preset matching threshold, and candidate recommendation information with similarity scores higher than the preset matching threshold is selected to form an effective recommendation information set;

[0157] If the set of valid recommendation information is not empty, then the candidate recommendation information with the highest similarity score is selected from the set of valid recommendation information as the target recommendation information.

[0158] In an optional implementation of this embodiment, the recommendation information determination module 430 includes a message push submodule, which is used to obtain the identification information of the target recommendation information, the trigger parameters associated with the current scene tag, and the anchor point information of the display interface;

[0159] The identification information and trigger parameters are encapsulated into a target push message according to a preset message protocol;

[0160] The message push module temporarily stores the target push message in the local message queue and starts a delay detection task;

[0161] Among them, the delay detection task continuously detects the user operation event stream within the delay period of the recommendation information display, so as to be used for subsequent trigger condition matching.

[0162] In an optional implementation of this embodiment, the recommendation information display module 440 is specifically used to parse the target push message and extract the trigger parameters therein. The trigger parameters include the target operation event type and the recommendation information display delay duration.

[0163] During the execution of the target application, continuously acquire the target user's operation event stream. The operation event stream contains multiple operation events arranged in chronological order, and each operation event carries an event type identifier and an occurrence timestamp.

[0164] Extract the subsequence of operation events within the most recent preset time period from the operation event stream;

[0165] Determine whether there exists at least one operation event in the operation event subsequence whose event type identifier is consistent with the target operation event type, and the timestamp of the operation event is no more than the time window threshold from the current system time;

[0166] If it exists, then it is determined that the preset triggering conditions that match the triggering parameters are met.

[0167] In an optional implementation of this embodiment, the recommendation information display module 440 is further specifically used to obtain recommendation resource data of the target recommendation information from a local recommendation cache or a remote recommendation resource server according to the identification information. The recommendation resource data includes the recommendation material URL and the rendering template identifier.

[0168] Parse the rendering template identifier and load the corresponding lightweight recommended rendering component;

[0169] Based on the display interface anchor information carried in the target push message, locate the target container view in the view hierarchy of the current application interface;

[0170] The URL of the recommended material is passed to the recommendation rendering component, which then generates a recommended view instance.

[0171] Add the recommended view instance to the target container view to complete the embedded display of the target recommendation information in the current application interface.

[0172] The application recommendation information determination device provided in the embodiments of the present invention can execute the application recommendation information determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0173] The acquisition, storage, use, and processing of data in this application comply with relevant national laws and regulations. It should be noted that certain software, components, models, and other existing industry solutions may be mentioned in the embodiments of this application. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used the relevant content of such solutions.

[0174] Example 5

[0175] Figure 5 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0176] like Figure 5 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded into the RAM 13 from the storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0177] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0178] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the recommendation information determination method for an application.

[0179] In some embodiments, the application recommendation information determination method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the application recommendation information determination method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the application recommendation information determination method by any other suitable means (e.g., by means of firmware).

[0180] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0181] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0182] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, erasable programmable read-only memory (EPROM), optical fibers, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0183] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a cathode ray tube (CRT) or liquid crystal display (LCD) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0184] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0185] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system. It addresses the shortcomings of traditional physical hosts and Virtual Private Servers (VPS) in terms of management difficulty and weak business scalability.

[0186] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0187] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0188] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements a database detection method as provided in any embodiment of this application.

[0189] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LANs or WANs—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0190] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0191] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for determining recommendation information in an application, characterized in that, The method includes: In response to the instructions of the target user to operate the target application, acquire multi-dimensional operation information generated by the target user during the current interaction; Based on the multi-dimensional operation information, a current scene label corresponding to the target user is generated, and the current scene label is fused with the representation information of the target user to obtain the state characteristics of the target user; Determine target recommendation information that matches the state features, and send a target push message carrying the identification information and trigger parameters to a terminal device adapted to the target application through the message push module integrated in the target application; The system receives the target push message and continuously detects the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameter. If there is, the system loads the target recommendation information based on the identification information and embeds the target recommendation information into the current application interface for display.

2. The method for determining recommendation information for an application according to claim 1, characterized in that, The step of responding to the target user's instruction to operate the target application and acquiring multi-dimensional operation information generated by the target user during the current interaction includes: By integrating the event tracking component into the target application, the operation events triggered by the user in the application interface and the terminal device operating environment parameters adapted to the target application can be obtained in real time. The operation events include at least one of the following: page navigation events, control interaction events, and business process state change events; The page navigation events include: page loading complete event and page exit event; The control interaction events include: button click events, swipe operation events, and form input completion events; The business process status change events include: business process start event, business process interruption event, and business process successful completion event; The operating environment parameters include: location information, network connection type, device battery status, and application foreground running identifier.

3. The method for determining recommendation information for an application according to claim 1 or 2, characterized in that, The step of generating the current scene tag corresponding to the target user based on the multi-dimensional operation information includes: The multi-dimensional operation information is input into a preset scene recognition rule engine; wherein, the scene recognition rule engine is configured with multiple scene templates, and each scene template defines a set of operation event types, the temporal constraint relationship between each operation event, and the maximum time window threshold; When the types of operation events contained in the continuously received sequence of operation events match the set of event types defined by the target scene template, the order of occurrence of each operation event conforms to the temporal constraint relationship specified by the target scene template, and the time interval between any two adjacent operation events does not exceed the time window threshold set by the target scene template, the current scene tag corresponding to the target scene template is activated. An initial timeliness weight is assigned to the current scene label, and a timed decay mechanism is started so that the initial timeliness weight decreases over time according to a preset decay function.

4. The method for determining recommendation information for an application according to claim 1, characterized in that, Before fusing the current scene label with the target user's representation information, the method further includes: Acquire historical operation log data generated by the target user within a preset historical time period; the historical operation log data includes page access sequences, business operation event flows, and interaction feedback records. Data cleaning and feature normalization are performed on the historical operation log data to obtain a standardized operation feature set; A set of static attribute labels is determined based on the standardized operational feature set; The set of static attribute tags is determined as the representation information of the target user, and the representation information is stored.

5. The method for determining recommendation information for an application according to claim 1, characterized in that, The step of fusing the current scene label with the target user's representation information to obtain the target user's state features includes: The current scene labels are structured and encoded to generate dynamic context feature vectors; Perform dimension alignment processing on each static attribute label in the representation information to generate a static feature vector; Based on a preset fusion weighting strategy, the weighted combination result of the dynamic context feature vector and the static feature vector is determined; The weighted combination result is determined as the state feature of the target user and output.

6. The method for determining recommendation information for an application according to claim 5, characterized in that, The step of determining the weighted combination result of the dynamic context feature vector and the static feature vector according to the preset fusion weight strategy includes: Obtain the timeliness weight associated with the current scene label; The dynamic context weight coefficient is calculated based on the timeliness weight, and the static weight coefficient is set as the complement of the dynamic context weight coefficient. Multiplying the dynamic context feature vector by the dynamic context weight coefficient yields the first weighted feature vector; Multiplying the static feature vector by the static weight coefficient yields the second weighted feature vector; The first weighted feature vector and the second weighted feature vector are added element by element to generate the weighted combination result.

7. The method for determining recommendation information for an application according to claim 1, characterized in that, The determination of target recommendation information matching the state features includes: The state features are normalized by scaling the feature values ​​of each dimension to a preset range to generate standardized state features. The standardized state features are converted into a query vector of fixed dimensions; The system retrieves candidate recommendation information from a pre-defined recommendation strategy library and determines the corresponding recommendation semantic vector for each candidate recommendation information. The recommendation semantic vector is a fixed-dimensional vector generated by hashing the keywords and category tags of the recommended content. The similarity score between the query vector and each of the recommended semantic vectors is determined, and the target recommended information is determined from the candidate recommended information based on the similarity score.

8. The method for determining recommendation information for an application according to claim 7, characterized in that, The step of obtaining candidate recommendation information from a preset recommendation strategy library and determining the recommendation semantic vector corresponding to each candidate recommendation information includes: Read all enabled candidate recommendation information from the recommendation strategy library; each candidate recommendation information is associated with structured metadata; For each candidate recommendation, the structured metadata is parsed, and a hash operation is performed on each text field in the structured metadata to generate a corresponding integer index; Initialize a vector of all zeros based on the preset vector dimension length; The integer index is modulo the length of the vector dimension to obtain the target dimension position, and a preset weight value is accumulated at the target dimension position to obtain the recommended semantic vector.

9. The method for determining recommendation information for an application according to claim 7, characterized in that, The step of determining the similarity score between the query vector and each of the recommended semantic vectors, and determining the target recommended information from the candidate recommended information based on the similarity score, includes: The query vector and each of the recommended semantic vectors are normalized respectively, and the cosine similarity between the normalized query vector and each recommended semantic vector is calculated to obtain the similarity score. The similarity scores are compared with a preset matching threshold, and candidate recommendation information with similarity scores higher than the preset matching threshold is selected to form an effective recommendation information set. If the set of valid recommendation information is not empty, then the candidate recommendation information with the highest similarity score is selected from the set of valid recommendation information as the target recommendation information.

10. The method for determining recommendation information for an application according to claim 1, characterized in that, The step of sending a target push message carrying the identification information and trigger parameters of the target recommendation information to a terminal device adapted to the target application through the message push module integrated in the target application includes: Obtain the identification information of the target recommendation information, the trigger parameters associated with the current scene tag, and the anchor point information of the display interface; The identification information and the triggering parameters are encapsulated into a target push message according to a preset message protocol; The message push module temporarily stores the target push message in a local message queue and starts a delay detection task. The delay detection task continuously detects the user operation event stream within the delay period of the recommendation information display, so as to be used for subsequent trigger condition matching.

11. The method for determining recommendation information for an application according to claim 1, characterized in that, The step of receiving the target push message and continuously detecting the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameters includes: The target push message is parsed, and the triggering parameters are extracted. The triggering parameters include the target operation event type and the delay duration for displaying the recommendation information. During the operation of the target application, the operation event stream of the target user is continuously acquired. The operation event stream contains multiple operation events arranged in chronological order, and each operation event carries an event type identifier and an occurrence timestamp. Extract the subsequence of operation events within the most recent preset time period from the operation event stream; Determine whether there exists at least one operation event in the subsequence of operation events, whose event type identifier is consistent with the target operation event type, and the timestamp of the operation event occurring is no more than a time window threshold from the current system time; If it exists, then it is determined that the preset triggering condition matching the triggering parameter is satisfied.

12. The method for determining recommendation information for an application according to claim 1, characterized in that, The step of loading the target recommendation information based on the identification information and embedding the target recommendation information into the current application interface for display includes: Based on the identification information, the recommendation resource data of the target recommendation information is obtained from the local recommendation cache or the remote recommendation resource server. The recommendation resource data includes the Uniform Resource Locator URL of the recommendation material and the rendering template identifier. Parse the rendering template identifier and load the corresponding lightweight recommended rendering component; Based on the display interface anchor information carried in the target push message, locate the target container view in the view hierarchy of the current application interface; The URL of the recommended material is passed to the recommendation rendering component, which then generates a recommended view instance. Add the recommended view instance to the target container view to complete the embedded display of the target recommendation information in the current application interface.

13. A device for determining recommendation information for an application, characterized in that, include: The multi-dimensional operation information acquisition module is used to acquire multi-dimensional operation information generated by the target user during the current interaction process in response to the target user's instruction to operate the target application. The state feature determination module is used to generate a current scene label corresponding to the target user based on the multi-dimensional operation information, and to fuse the current scene label with the representation information of the target user to obtain the state features of the target user; The recommendation information determination module is used to determine the target recommendation information that matches the state features, and send a target push message carrying the identification information and trigger parameters of the target recommendation information to the terminal device adapted to the target application through the message push module integrated in the target application; The recommendation information display module is used to receive the target push message and continuously detect the target user interface operation sequence to determine whether there is a preset trigger condition that matches the trigger parameter; If it exists, the target recommendation information is loaded based on the identification information, and the target recommendation information is embedded in the current application interface for display.

14. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, which enables the at least one processor to perform the recommendation information determination method of the application according to any one of claims 1-12.

15. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method for determining recommendation information of the application program as described in any one of claims 1-12.

16. A computer program product comprising a computer program that, when executed by a processor, implements a method for determining recommendation information for an application according to any one of claims 1-12.

Images