Content recommendation method and device and electronic equipment

Abstract

The invention discloses a content recommendation method and device and electronic equipment, and belongs to the technical field of electronic equipment.The method comprises the steps that a user intention is determined according to user input content; obtaining a user scene state and an external event corresponding to the user intention; and pushing a first recommendation content based on the external event when it is monitored that the user scene state meets a preset state triggering condition and the external event meets a preset event triggering condition.

Description

Technical Field

[0001] This application belongs to the field of electronic equipment technology, specifically relating to a content recommendation method, apparatus, and electronic equipment. Background Technology

[0002] With the rapid development of artificial intelligence technology, especially natural language processing and large-scale modeling, intelligent assistant applications have become deeply integrated into users' daily lives and work scenarios. To lower the barrier to entry and enhance the interactive experience, these applications generally integrate content recommendation functions to help users achieve their goals more efficiently by proactively and timely providing information or service guidance.

[0003] In related technologies, the recommendation methods for intelligent assistant applications mainly adopt random recommendation mechanisms, which suffer from insufficient recommendation accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a content recommendation method, apparatus, and electronic device that can improve the accuracy of recommended content.

[0005] In a first aspect, embodiments of this application provide a content recommendation method, the method comprising: Determine the user's intent based on the user's input; Obtain the user scenario state and external events corresponding to the user intent; If the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions, the first recommended content is pushed based on the external event.

[0006] Secondly, embodiments of this application provide a content recommendation device, the device comprising: The determination module is used to determine the user's intent based on the user's input. The acquisition module is used to acquire the user scenario state and external events corresponding to the user intent; The push module is used to push first recommended content based on the external event when the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions.

[0007] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the content recommendation method as described in the first aspect.

[0008] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the content recommendation method as described in the first aspect.

[0009] Fifthly, embodiments of this application provide a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the steps of the content recommendation method as described in the first aspect.

[0010] In this embodiment of the application, the user intent is determined based on the user input; the user scenario state and external events corresponding to the user intent are obtained; and when the user scenario state meets the preset state triggering conditions and the external events meet the preset event triggering conditions, the first recommended content is pushed based on the external events.

[0011] As can be seen, in this embodiment, by parsing user intent and continuously tracking their user scenario state, the recommended content is ensured to be closely related to the current scenario, thus improving the accuracy of the recommended content. Simultaneously, proactively monitoring external events associated with intent can trigger warnings or suggestions immediately when changes in the objective environment affect decision-making, achieving a shift from passive response to proactive care. Furthermore, the recommendation action is triggered only when the monitored user scenario state meets preset state triggering conditions and the external event meets preset event triggering conditions, ensuring both accuracy and timeliness. Attached Figure Description

[0012] Figure 1 This is one of the flowcharts of a content recommendation method provided in the embodiments of this application; Figure 2 This is a second flowchart of a content recommendation method provided in the embodiments of this application; Figure 3 This is an example diagram of a content recommendation method provided in an embodiment of this application; Figure 4 This is a structural block diagram of a content recommendation device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of an electronic device that implements an embodiment of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0014] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0015] The content recommendation method provided in this application will be described in detail below with reference to the accompanying drawings, specific embodiments, and application scenarios.

[0016] Figure 1 This is one of the flowcharts of a content recommendation method provided in the embodiments of this application, such as... Figure 1 As shown, the method may include the following steps: step 101, step 102 and step 103.

[0017] In step 101, the user intent is determined based on the user input.

[0018] In this embodiment of the application, user input refers to the original information that a user inputs into an electronic device in the form of text, voice, or images, expressing their needs or intentions. For example, a query text "7-day Bali trip guide for National Day", a voice command "Book me a flight to Shanghai", or an image containing travel destination information.

[0019] In this embodiment of the application, user intent is an abstract representation of the goal or core need that the user hopes to achieve through input content, and it is the logical starting point for subsequent intelligent recommendations.

[0020] For example, if a user inputs "7-day Bali trip guide during National Day", it can be determined that the user's intent is "to seek travel planning advice for Bali during National Day".

[0021] In this embodiment, by identifying the user's intent from the user's input, the user's fundamental goal can be understood, thereby transforming the subsequent recommendation action from random information distribution into a service with a clear goal orientation.

[0022] In step 102, the user scenario state and external events corresponding to the user intent are obtained.

[0023] In this embodiment of the application, the user scenario state refers to procedural data that is related to the user's intent and reflects the user's current task progress or real-time situation.

[0024] In some embodiments, for travel planning intent, the user scenario status includes at least one of the following: transportation ticket booking status, accommodation booking status, document processing status, and user geographical location.

[0025] In this embodiment, by verifying the aforementioned concrete business status, the system can accurately pinpoint the user's current decision-making stage (e.g., query only, already booked, or about to travel), thereby ensuring that the recommended content closely matches the user's actual progress and avoiding unnecessary alerts before the user has finalized their itinerary. Simultaneously, this design allows recommendations to directly link to the user's next action; for example, providing visa application guidance when "tickets have been purchased but visas have not been obtained," and providing real-time local information when "departure has commenced." By tracking the complete status chain, the intelligent assistant can provide coherent and precise proactive services covering the entire process from pre-trip planning to itinerary execution.

[0026] In this application embodiment, external events refer to objective event information that is associated with the user's intent and is provided in real time or periodically by external data sources (such as meteorological bureaus, airlines, and news organizations).

[0027] In some embodiments, external events may include at least one of the following: weather warning events, flight change events, traffic control events, and safety notice events provided in real time by external data sources.

[0028] In this embodiment, by clearly defining the scope of information monitored by the intelligent assistant, it can focus on key information that directly impacts user decisions, is objectively verifiable, and is timely, ensuring the value and credibility of the recommended content from the source. By setting differentiated triggering and response logic for different event types (e.g., security events have the highest priority, flight changes must be combined with the user's itinerary), the intelligent assistant can achieve refined scenario responses, specifically meeting the core needs of users in different situations. Simultaneously, by connecting with clearly defined external data sources (e.g., meteorological bureaus, airlines, traffic management departments), the accuracy and real-time nature of event information are guaranteed, providing a data foundation for the intelligent assistant to provide reliable decision-making basis at critical moments, thereby upgrading recommendations from subjective guesswork to precise services based on objective facts.

[0029] For example, in response to a user's intention to "seek travel planning advice for Bali during the National Day holiday", the intelligent assistant needs to know whether the user has already booked flights and hotels (user scenario status), and at the same time needs to pay attention to whether there is a sudden typhoon warning for the destination they plan to visit (external event).

[0030] In this embodiment, a dual-source information foundation for recommendation decision-making is constructed: the real-time progress of internal users and the dynamic changes of the external environment. This enables the intelligent assistant to have scene awareness capabilities, and it is no longer responding to user queries in isolation, but can make optimal decisions in the complete user-environment context.

[0031] In step 103, if the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions, the first recommended content is pushed based on the external event.

[0032] In this embodiment, after acquiring the user scenario state and external events corresponding to the user's intent, the intelligent assistant continuously or periodically monitors the user scenario state data and external event sources. This continuous monitoring mechanism, on the one hand, frees the recommendation service from dependence on single user sessions, enabling long-term, uninterrupted proactive monitoring of users' long-term tasks (such as a trip from planning to completion), ensuring that key decision-making points are not missed due to the user exiting the application. On the other hand, it can perceive the dynamic changes in the user's state and the external environment in real time. Whether the user completes a hotel reservation or a sudden security incident occurs at the destination, the intelligent assistant can capture these changes immediately and reassess the triggering conditions. Ultimately, this ensures the high reliability and timeliness of the recommendation service, ensuring that decision support is proactively provided to the user at the most suitable time (when the user's state is ready and the external event occurs).

[0033] In this embodiment of the application, the preset state triggering condition refers to a set of logical judgment rules defined based on user scenario state data associated with user intent.

[0034] In some embodiments, for travel planning intent, the user scenario state includes at least one of the following: transportation ticket booking status, accommodation booking status, document processing status, and user geographical location. The state triggering condition is a logical condition defined based on the combination of values ​​for multiple state items in the user scenario state. This condition is not a simple judgment of a single state item, but a logical expression that comprehensively evaluates the combination of values ​​for multiple state items, used to accurately determine whether the user has entered a specific task stage or situation highly relevant to the recommendation service.

[0035] For example, for the intent of "travel-guide-overseas travel", its status triggering condition may be defined as: (transportation booking status = "flight booked") AND (accommodation booking status = "hotel booked") AND (document processing status = "visa not applied for"). This condition means that the status prerequisite for triggering the recommendation is only met when the user has booked flights and hotels but has not yet applied for a visa.

[0036] In this embodiment, the preset state triggering condition, by combining multiple state items, can precisely identify the user's current task stage (such as "intention inquiry," "trip confirmation," or "pre-trip preparation"), thereby ensuring that the recommendation service intervenes accurately at the appropriate time. This design tightly binds the recommended content to the user's current actual task progress, effectively avoiding invalid pushes when the user only has an intention or the trip has ended, thus improving the relevance and usability of the recommendations. Furthermore, this condition exists in the form of configurable logical rules, allowing operators to flexibly adjust it according to their strategies (e.g., changing the condition from "visa not applied for" to "visa being processed"), or defining entirely new state combinations for different user intentions, all without modifying the core system code.

[0037] In this embodiment, the preset event triggering conditions refer to a set of filtering and matching criteria pre-set to select external events that are relevant to the user's intent and have recommendation value. The preset event triggering conditions define the type, level, region, and time of occurrence of external events that are of interest.

[0038] For example, given the user input "7-day Bali trip guide during National Day holiday", the intelligent assistant determines that the user's intent is "to seek travel planning suggestions for Bali during National Day holiday". The event trigger condition might be instantiated as: (Event type = "weather warning") AND (Event level) [“Orange Alert”, “Red Alert”] AND (Event Location = “Bali”) AND (Event Time Window overlaps with “10.1-10.7”) This condition is used to accurately monitor whether medium-to-high level weather warnings are issued for Bali during the user’s travel period.

[0039] In this embodiment, by dynamically binding the event location and time window with the user's intent, personalized and targeted monitoring of external events is achieved. This allows the intelligent assistant to focus only on the spatiotemporal range directly related to the user's specific task, avoiding information overload from the outset. Furthermore, by pre-setting event types and levels, the intelligent assistant can automatically filter out a large amount of low-relevance or low-urgency general information, ensuring that the events ultimately triggered for recommendation are all of high value, high timeliness, and have a substantial impact on the user's decision-making. In addition, this templated event triggering condition design allows a set of general rules (such as "weather warning rules") to serve a large number of users' different task instances (such as traveling to different destinations). The intelligent assistant automatically instantiates and executes monitoring based on each user's parameters, reducing the complexity of rule maintenance and the consumption of computing resources.

[0040] In summary, the preset state trigger conditions and preset event trigger conditions set thresholds from two dimensions: internal user state and external environment changes. This ensures that recommendations are triggered only at the intersection of the user entering the relevant scenario and the external environment exhibiting relevant dynamics, thus achieving accurate and timely recommendations.

[0041] In this embodiment of the application, the generated recommended content can be delivered to the user through in-app notifications, push notifications, or other means.

[0042] For example, the smart assistant monitors that the user has already booked flights and hotels (status condition met), and at the same time, it receives a typhoon orange warning issued at the destination (event condition met). At this point, a recommendation is triggered, and a safety reminder notification is generated based on the warning event and pushed to the user.

[0043] In this embodiment, recommendations are made only when the user has entered a relevant scenario (such as having booked a trip) and relevant risks / opportunities arise externally. On the one hand, this filters out the vast majority of invalid push notifications, improving the accuracy of recommendations and user acceptance; on the other hand, it can proactively intervene at critical moments that affect user decisions, providing information that is timely and has action guidance value, directly assisting users in optimizing decisions or avoiding risks.

[0044] As can be seen from the above embodiments, this embodiment ensures that the recommended content is closely related to the current scenario by parsing the user's intent and continuously tracking their user scenario state, thus improving the accuracy of the recommended content. Simultaneously, proactively monitoring external events related to intent can trigger warnings or suggestions immediately when changes in the objective environment affect decision-making, achieving a shift from passive response to proactive care. Furthermore, the recommendation action is only triggered when the monitored user scenario state meets preset state triggering conditions and the external event meets preset event triggering conditions, ensuring both accuracy and timeliness.

[0045] Figure 2 This is a second flowchart of a content recommendation method provided in an embodiment of this application, such as... Figure 2 As shown, the method may include the following steps: step 201, step 202 and step 203.

[0046] In step 201, the atomic intent corresponding to the user input content is determined, and the key parameters associated with the atomic intent are extracted.

[0047] In this embodiment, an atomic intent is a standardized, structured, and minimally granular implementation of user intent. It refers to a standardized and indivisible semantic description of basic user needs, typically defined using a three-level classification system of "vertical category-scenario-intent" to ensure its accuracy and reusability. For example, "travel-guide-overseas travel" is an atomic intent.

[0048] In this application embodiment, key parameters, also known as associated parameters, refer to specific information entities or attributes closely related to the atomic intent, used to instantiate and refine the specific content of the intent, such as time, location, person, amount, etc.

[0049] For example, when a user inputs "7-day Bali trip guide during National Day", a specific intent recognition model determines that the corresponding atomic intent is "travel-guide-overseas travel", and extracts the key parameters as "destination = Bali, travel time = October 1 to October 7".

[0050] In this embodiment, unstructured natural language input is precisely converted into structured semantic units that can be accurately manipulated and matched by machines, namely atomic intent tags and key-value pairs of key parameters. This provides a unified and reliable data foundation for subsequent steps (such as determining the states and events to be monitored in step 202). Specifically, atomic intents, as standardized demand identifiers, are directly associated with predefined user scenario state structures; while key parameters, as specific values, are used to dynamically populate and instantiate variables (such as destination and time) in external event triggering conditions. This quantitative and structured processing provides a prerequisite for achieving subsequent accurate and automated recommendation decisions.

[0051] In some embodiments, step 201 can be implemented using a pre-trained intent and parameter parsing model. Accordingly, step 201 may specifically include the following sub-step: sub-step 2011.

[0052] In sub-step 2011, the user input content, input intent, and parameter parsing model are processed, and the model outputs the atomic intent and key parameters corresponding to the user input content.

[0053] In this embodiment, the intent and parameter parsing model is a specially trained machine learning model capable of deep semantic understanding of input natural language text and outputting structured intent classification results and parameter extraction results. Preferably, this model is constructed based on a pre-trained language model (such as various large language models) using a supervised fine-tuning (SFT) method.

[0054] For example, given the input "7-day Bali trip guide during National Day", the intent and parameter parsing model can determine its atomic intent as "tourism-guide-overseas travel" and extract the key parameters as "destination = Bali, travel time = October 1 to October 7, duration = 7 days".

[0055] To better support the accuracy of the intent and parameter parsing model, this application also proposes a detailed model training and construction method, which may include the following steps: 1) Constructing the training dataset: The acquired training dataset contains a large amount of vertical text data collected from mainstream platforms and social media, and non-vertical text is included as negative samples to enhance the robustness of the model. Each training data point consists of text and its corresponding structured annotation. The annotation content strictly follows the three-level classification system of "vertical category-scenario-intent" to label user intent tags, and uses the BIO sequence labeling format to label associated parameter entity tags.

[0056] For example, the intent labeling results are: "Tourism" as a first-level vertical category, "Travel Guide" as a second-level scenario, and "Overseas Travel" as a third-level atomic intent. The final labeling results of the training data can be represented as: [ { First-level vertical category: [Tourism] Level 2 Scene: [Strategy Guide], Level 3 User Intent: [Overseas Travel] Core parameters: { Destination: Bali Travel period: October 1st to October 7th } } ].

[0057] 2) Supervised Model Fine-Tuning: A pre-trained language model is fine-tuned using the dataset constructed above. During training, the dataset is proportionally divided into training, validation, and test sets. Hierarchical label encoding is used for atomic intents, and BIO label encoding is used for key parameters.

[0058] 3) Training configuration: A combined loss function for multi-task learning is adopted: Total loss = a × Intent loss + (1-a) × Parameter loss; where the intent loss is the cross-entropy loss of the three-level classification, the parameter loss is the conditional random field loss, and the weight 'a' is used to balance the two classes of tasks. Optimizer: The initial learning rate is set to 3e-5, and weight decay is applied to the unbiased parameters of the model.

[0059] 4) Training control: Set reasonable batch size and training rounds, and prevent overfitting or underfitting by monitoring the training set loss and the intent classification F1 value and parameter extraction F1 value on the validation set, so as to ensure model performance. The intent and parameter parsing model constructed in the above manner has a powerful semantic generalization ability. Even when faced with diverse user natural language expressions that are not preset in the rule base (such as "How to arrange a week trip to Bali during the National Day holiday"), the model can still accurately normalize them into standard atomic intents (such as "travel-guide-overseas travel") and extract precise key parameters, which fundamentally ensures the accuracy and reliability of state determination and event monitoring in subsequent steps.

[0060] In step 202, the user scenario state that needs to be monitored is determined based on atomic intent; and the external events that need to be monitored are determined based on key parameters.

[0061] In this embodiment, a "user scenario state data" structure associated with different atomic intents is predefined. For example, the state structure associated with the atomic intent "travel-guide-overseas travel" is {transportation booking status, accommodation booking status, document processing status}. The intelligent assistant indexes this structure based on the atomic intent, thereby determining which data sources need to be consulted for which state information.

[0062] In this embodiment, extracted key parameter values ​​can be used to instantiate (populate) variables in the dynamic recommendation rule template. For example, the variables "event region = {{destination}}" and "event time window = {{travel time}}" in the rule are replaced with specific "event region = Bali" and "event time window = 10.1-10.7", thereby generating a concrete external event query instruction for the user's current trip.

[0063] For example, based on the atomic intent "travel-guide-overseas travel", the intelligent assistant determines that it is necessary to monitor the user's "flight booking status" and "hotel booking status". Based on the key parameters "destination = Bali, travel time = October 1-7", the intelligent assistant determines that it is necessary to query the meteorological bureau API for weather warning events for "Bali during October 1-7".

[0064] In this embodiment, atomic intents serve as a standardized semantic bridge, effectively decoupling the varied user natural language input from the relatively stable business logic (such as predefined state structures and rule templates). This allows operators to manage recommendation strategies by flexibly configuring the mapping relationship between intents and rules without modifying the underlying code. Simultaneously, the introduction of key parameters enables dynamic personalization, ensuring that the monitoring scope of external events (such as specific destinations and time windows) is entirely determined by the specific content of the user's input, thereby generating highly customized monitoring tasks for each user.

[0065] In step 203, if the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions, the first recommended content is pushed based on the external event.

[0066] In this embodiment, the design of the state triggering condition is closely related to the atomic intent. Each atomic intent (such as "travel-guide-overseas travel") predefines its corresponding user scenario state data structure (e.g., including transportation ticket booking status, accommodation booking status, document processing status, and user geographical location). The state triggering condition is not a judgment on a single state, but a logical expression defined based on the combination of values ​​of multiple state items associated with the atomic intent.

[0067] For example, if the atomic intent is "Travel - Guide - Overseas Travel", its state triggering condition can be defined as: (Transportation ticket booking status = "Flight booked") AND (Accommodation booking status = "Hotel booked") AND (Document processing status = "Visa not applied for"). This logical combination precisely defines the specific user task stage required to trigger the recommendation.

[0068] In this embodiment, based on the combination of state conditions of atomic intents, it is possible to finely distinguish the different stages of the intent lifecycle of the user (such as "intent query", "trip confirmation" or "pre-trip preparation"), ensuring that the recommendation intervenes accurately at the most relevant time; through the strong binding of atomic intents and state structures, the recommended content is deeply associated with the user's current task scenario, avoiding invalid pushes when the user has not yet entered the scenario corresponding to the atomic intent; the state conditions based on atomic intents are highly configurable, and operators can flexibly define the state combination logic for different atomic intents according to business needs without modifying the underlying system architecture, thus improving the operability of the recommendation strategy.

[0069] In some embodiments, external events include at least one of the following: weather warning events, flight change events, traffic control events, and safety notice events provided in real time by external data sources; wherein, the value of the event region in the event triggering conditions is set based on the destination parameter in the key parameters.

[0070] In this embodiment, the core implementation mechanism of the event triggering conditions is key parameter-driven event monitoring templates. Specifically, the intelligent assistant predefines a general event triggering condition template, which includes parameterized fields such as "event region = {{destination}}" and "event time window = {{travel time}}". When the intelligent assistant extracts key parameters from the user's input (such as "destination = Bali" and "travel time = October 1st to October 7th"), it automatically substitutes these specific values ​​into the template to generate a completely personalized monitoring condition instance for the user's current task.

[0071] In this embodiment, through the aforementioned dynamic binding mechanism based on key parameters, external event monitoring achieves precise spatiotemporal association with the user's specific task scenario. This allows the intelligent assistant to track events that strictly correspond to the user's key parameters (such as destination and travel time), effectively filtering out interference from irrelevant information sources. Furthermore, by combining the "event type" and "event level" filtering logic in the event triggering conditions with the geographical scope and / or time window defined by the key parameters, the intelligent assistant can perform step-by-step value filtering and purification of massive amounts of external information, ensuring that the events ultimately triggered for recommendation simultaneously meet the requirements of relevance, high timeliness, and high importance. In addition, this parameterized design transforms the configuration method of event monitoring rules from customized writing for a single user to template-based automatic instantiation for key parameter combination categories. Based on this, a universal event monitoring rule template can adaptively generate personalized monitoring instances serving different users through dynamic input of key parameters, reducing the complexity of rule configuration, maintenance, and management while maintaining recommendation accuracy.

[0072] In this embodiment, the atomic intents and associated key parameters obtained from parsing user input can be persistently stored as personal recommendation memory data. This memory data is bound to the user identifier and records the user's specific intents and task constraints. Using this data, the intelligent assistant can asynchronously and continuously execute monitoring tasks associated with this memory data in the background even after the user ends the current session; that is, periodically checking whether the corresponding user scenario state and external events meet the triggering conditions. This mechanism enables recommendation services to break through the temporal and spatial limitations of a single conversation, achieving long-term, uninterrupted proactive monitoring and timely intervention for users' long-term tasks (such as the entire process of a trip from planning to completion), thereby constructing a long-term recommendation mechanism with memory capabilities.

[0073] In this embodiment, within the aforementioned long-term recommendation mechanism, the intelligent assistant can implement refined recommendation rhythm control by maintaining the recommendation time and recommendation frequency fields in the user's recommendation memory data. Specifically, the intelligent assistant's preset recommendation frequency strategy can be manifested as rule constraints, such as "a single user can recommend a maximum of 3 times within 24 hours" or "the interval between two adjacent recommendations is no less than 3 hours." Before executing a recommendation, the intelligent assistant verifies whether the interval between the current time and the last recommendation time meets the minimum interval requirement, and whether the daily recommendation frequency has reached its limit. This dual verification mechanism ensures a reasonable distribution of recommendation actions over time, enabling timely access to users when information is updated, while effectively preventing information overload and user disturbance caused by high-frequency or short-interval pushes. This ensures service proactivity while optimizing user experience and the long-term efficiency of system resources.

[0074] As can be seen from the above embodiments, in this embodiment, by introducing structured parsing of atomic intents and key parameters, vague user needs are transformed into precise and operable instructions, thereby accurately driving subsequent status monitoring and event attention. This not only has high feasibility and scalability, but also provides technical support for realizing the evolution of intelligent assistants from immediate response to long-term care.

[0075] In some embodiments provided in this application, step 103 or step 203 may specifically include the following sub-steps: when there are multiple different types of external events, and all of them meet their respective event triggering conditions, a target event is determined from the multiple external events according to the urgency of each external event and the degree of matching between each external event and the user's scenario state; and the first recommended content is pushed based on the target event.

[0076] In this embodiment, when the intelligent assistant simultaneously acquires multiple eligible external events of different types for the same user task (such as a trip) within the same monitoring period, it does not simply stack or randomly select one for recommendation, but introduces an intelligent optimization decision-making process. This decision-making process is mainly based on a comprehensive evaluation of two core dimensions: first, the urgency of the event, which is usually determined by the event type (e.g., security event > weather event > traffic event > promotional information) and the event level (e.g., red alert > orange alert); second, the degree of matching between the event and the current user scenario state, that is, the actual impact of the event information on the user in the current specific state and its relevance to action.

[0077] For example, the smart assistant monitors two newly occurring events for a user who has booked flights and hotels and is about to travel to Bali: Event A (the meteorological department issues a "red typhoon warning") and Event B (the local tourism bureau issues a "limited-time discount on tickets to popular attractions"), both of which meet their respective event triggering conditions.

[0078] From an urgency perspective, a red typhoon warning involves personal safety and is far more urgent than information about attraction discounts. From the perspective of relevance to the user's situation, the user's itinerary is already set and imminent; the typhoon warning directly threatens the safety and feasibility of their trip, making it a highly relevant information. While ticket discounts are related, they are relatively experience-enhancing information and lack the urgency to address safety risks.

[0079] Based on the above analysis, the intelligent assistant will prioritize identifying event A (typhoon red alert) as the target event and generate safety reminders and travel adjustment suggestions to push to the user, rather than prioritizing the push of promotional information.

[0080] As can be seen, in this embodiment, the selection mechanism ensures that, given limited user attention and push opportunities, the intelligent assistant always prioritizes delivering the most valuable and urgent information. This effectively avoids decision-making interference caused by information overload or a lack of prioritization when multiple events occur concurrently, maximizing the information utility of the recommended content. Simultaneously, by considering the matching degree with the user's real-time state, the recommendations are made more personalized and action-oriented, further enhancing the intelligent assistant's decision support capabilities and user trust in complex scenarios.

[0081] In some embodiments provided in this application, in order to achieve the above... Figure 1 and Figure 2 The method shown can construct and maintain a pool of recommendation elements. This pool of recommendation elements is based on "atomic intents". Key parameters The recommended rule is to organize memory using triples as the basic storage unit. Figure 1 In this embodiment, the user intent determined in step 101 and the operation of obtaining state and event in step 102 can be specifically implemented based on querying the element pool: that is, using the user intent (or the further parsed atomic intent and key parameters) as an index, retrieving matching recommendation rules from the pool, and then determining the specific user scenario state and external events to be monitored based on the rules. Figure 2 In this embodiment, the association is more direct: the atomic intent and key parameters output in step 201 serve as the first two items of the triple, directly used to retrieve the corresponding recommendation rule (the third item of the triple) from the element pool, thereby precisely driving the process of "determining the states and events to be monitored" in step 202. It is evident that the recommendation element pool and its triple structure... Figure 1 and Figure 2 The content recommendation method shown provides a configurable, scalable strategy knowledge base and a unified data interaction interface.

[0082] In some embodiments provided in this application, in addition to employing Figure 1 or Figure 2 The dynamic recommendation mechanism shown in this application embodiment can also include a static recommendation mechanism.

[0083] In this embodiment, the static recommendation mechanism is implemented based on pre-configured static recommendation rules, which directly establish a deterministic mapping relationship between user intent and recommended content. Its triggering logic is direct and efficient: once user input is parsed and matched to a specific user intent (or atomic intent), the intelligent assistant immediately retrieves the static rule associated with that intent and directly presents the corresponding recommended content to the user.

[0084] For example, when a user enters "7-day Bali trip guide for National Day" and the intent is parsed as "travel - guide - overseas travel", the intelligent assistant will trigger both static and dynamic paths simultaneously. While the dynamic path is monitored over a long period, the static path will immediately recommend general content strongly related to this intent, such as "Bali visa application guide", "comparison of popular airfares", or "local customs and information", and will be directly displayed in the intelligent assistant's response interface.

[0085] As can be seen, in this embodiment, the introduction of a static recommendation mechanism, together with the dynamic recommendation mechanism, constitutes a complete service system. This provides a basic recommendation guarantee with zero latency, high determinism, and full coverage, ensuring that users can immediately obtain valuable guidance information after any query, thus improving the service's immediate responsiveness and basic user experience satisfaction. Simultaneously, its simple and reliable characteristics reduce reliance on real-time data and complex calculations, effectively distributing system load and complementing the dynamic recommendation mechanism in terms of service breadth, depth, and resource efficiency.

[0086] In some embodiments provided in this application, the content recommendation method provided in this application also includes a closed-loop feedback and self-optimization mechanism.

[0087] In this embodiment, after pushing recommended content to the user, the intelligent assistant collects user feedback data on the recommended content through preset interaction interfaces, such as satisfaction rating, usefulness judgment, or negative feedback attribution options. This feedback data is recorded in a structured manner and associated with specific recommendation rules, recommended content, and user identifiers.

[0088] For example, the intelligent assistant can periodically aggregate and analyze collected user feedback data to pinpoint specific strategies for recommendations that fail to meet expectations. For instance, if a high percentage of negative feedback for a dynamic "weather warning" rule indicates "inappropriate timing," and analysis reveals this is due to the warning's issuance time being too close to the user's travel time, the intelligent assistant can automatically (or, after confirmation by operations personnel) adjust the "event time window" parameter in the event triggering condition associated with that rule from the original "{{travel time}}" to "{{3 days before travel time}} to {{travel time}}," thereby achieving earlier warnings of potential risks. Similarly, for static recommended content, if its "average satisfaction rating" consistently falls below a preset threshold within a continuous statistical period, the intelligent assistant can automatically reduce the content's display priority in the recommendation list or remove it entirely, thus achieving continuous optimization of the recommended content.

[0089] As can be seen, in this embodiment of the application, by introducing a feedback closed-loop mechanism, the intelligent assistant has the ability to self-evolve and continuously optimize. It drives the recommendation rules (including state triggering conditions, event triggering conditions and content generation strategies) to be continuously iterated and optimized through real user feedback signals, thereby improving the accuracy of recommended content.

[0090] In some embodiments provided in this application, the content recommendation method provided is as described above. Figure 1 or Figure 2 Based on the illustrated embodiment, a social or collaborative recommendation mechanism may also be included. Specifically, the method further includes the following steps: obtaining authorized personal recommendation memory data of a second user; wherein the second user has an association relationship with the first user currently requesting recommendations, the association relationship including relationships such as family members, travel companions, etc., predefined by the electronic device or declared by the user; when generating recommendation content, supplementing or optimizing the recommendation content generated for the first user by combining the second user's personal recommendation memory data.

[0091] In this embodiment, "authorized" means that the acquisition and use of the second user's personal data have been preceded by the user's explicit consent and authorization, and the entire process strictly complies with data privacy and security regulations. When recommending content to the first user, the intelligent assistant checks their social graph to verify whether the associated second user has authorized the sharing of relevant types of memory data, such as travel memories, and reads the corresponding data after obtaining valid authorization.

[0092] For example, a first user (father) searches for "family-friendly hotels in Sanya," and a second user (mother), a family member, authorizes the sharing of travel memories. The smart assistant reads the mother's past memories of "Sanya" and "family travel," discovering that she highly rated a certain hotel and showed interest in children's activities. The smart assistant uses the second user's memory information as a basis for decision-making; for example, in the hotel recommendation list, it prioritizes the hotels highly rated by the mother and adds a "family recommended" tag; when recommending attractions, it adds, "Your travel companion's mother just went there last week and said it's suitable for traveling with children."

[0093] As can be seen, in this embodiment of the application, by introducing a collaborative recommendation mechanism based on authorized social relationships, the dimensions of the recommendation service are expanded from individual behavior to social networks. By integrating the real experiences and states of related users, the credibility and reference value of the recommended content are improved.

[0094] In summary, the embodiments of this application enable users who have already made a decision to make a better decision through the accurate recommendations of intelligent assistant applications, avoiding the risk that the original decision cannot be successfully executed due to external changes or lack of information. For example Figure 3As shown, when a user clicks on "7-Day Bali Trip Guide for National Day", the user will receive news push notifications or weather warnings about risks in Bali, allowing them to change their itinerary in a timely manner and make the best decision.

[0095] The content recommendation method provided in this application can be executed by a content recommendation device. This application uses the execution of the content recommendation method by a content recommendation device as an example to illustrate the content recommendation device provided in this application.

[0096] Figure 4 This is a structural block diagram of a content recommendation device provided in an embodiment of this application, such as... Figure 4 As shown, the content recommendation device 400 may include: a determination module 401, an acquisition module 402, and a push module 403; The determining module 401 is used to determine the user's intent based on the user's input. The acquisition module 402 is used to acquire the user scenario state and external events corresponding to the user intent; The push module 403 is used to push first recommended content based on the external event when the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions.

[0097] As can be seen from the above embodiments, this embodiment ensures that the recommended content is closely related to the current scenario by parsing the user's intent and continuously tracking their user scenario state, thus improving the accuracy of the recommended content. Simultaneously, proactively monitoring external events related to intent can trigger warnings or suggestions immediately when changes in the objective environment affect decision-making, achieving a shift from passive response to proactive care. Furthermore, the recommendation action is only triggered when the monitored user scenario state meets preset state triggering conditions and the external event meets preset event triggering conditions, ensuring both accuracy and timeliness.

[0098] Optionally, as an embodiment, the determining module 401 is specifically used to determine the atomic intent corresponding to the user input content, and to extract key parameters associated with the atomic intent; The acquisition module 402 is specifically used to determine the user scenario state to be monitored based on the atomic intent; and to determine the external events to be monitored based on the key parameters.

[0099] Optionally, as an embodiment, the user scenario state includes at least one of the following: transportation ticket booking status, accommodation booking status, document processing status, and user geographical location; wherein, the status triggering condition is a logical condition defined based on the combination of values ​​of multiple status items in the user scenario state.

[0100] Optionally, as an embodiment, the external events include at least one of the following: weather warning events, flight change events, traffic control events, and safety notice events provided in real time by external data sources; wherein, the value of the event region in the event triggering conditions is set based on the destination parameter in the key parameters.

[0101] Optionally, as an embodiment, the push module 403 is specifically used to determine a target event from multiple external events when there are multiple different types of external events, all of which meet their respective event triggering conditions, based on the urgency of each external event and the degree of matching between each external event and the user's scenario state; and push first recommended content based on the target event.

[0102] The content recommendation device in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, Ultra-Mobile Personal Computers (UMPCs), netbooks, or Personal Digital Assistants (PDAs), etc., while non-mobile electronic devices can be servers, Network Attached Storage (NAS), Personal Computers (PCs), Televisions (TVs), ATMs, or self-service machines, etc. This application embodiment does not impose specific limitations.

[0103] The content recommendation device in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0104] The content recommendation device provided in this application embodiment can achieve... Figure 1 or Figure 2 To avoid repetition, the various processes implemented in the method embodiment shown will not be described again here.

[0105] Optionally, such as Figure 5 As shown, this application embodiment also provides an electronic device 500, including a processor 501, a memory 502, and a program or instructions stored in the memory 502 and executable on the processor 501. When the program or instructions are executed by the processor 501, they implement the various processes of the above-mentioned recommended method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0106] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0107] Figure 6 This is a schematic diagram of the hardware structure of an electronic device that implements an embodiment of this application.

[0108] The electronic device 600 includes, but is not limited to, components such as: radio frequency unit 601, network module 602, audio output unit 603, input unit 604, sensor 605, display unit 606, user input unit 607, interface unit 608, memory 609, and processor 610.

[0109] Those skilled in the art will understand that the electronic device 600 may also include a power supply (such as a battery) for supplying power to various components. The power supply may be logically connected to the processor 610 through a power management system, thereby enabling functions such as managing charging, discharging, and power consumption through the power management system. Figure 6 The electronic device structure shown does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown, or combine certain components, or have different component arrangements, which will not be elaborated here.

[0110] The processor 610 is configured to determine the user's intent based on the user's input; acquire the user scenario state and external events corresponding to the user's intent; and push first recommended content based on the external events when the user scenario state meets preset state triggering conditions and the external events meet preset event triggering conditions.

[0111] As can be seen, in this embodiment, by parsing user intent and continuously tracking their user scenario state, the recommended content is ensured to be closely related to the current scenario, thus improving the accuracy of the recommended content. Simultaneously, proactively monitoring external events associated with intent can trigger warnings or suggestions immediately when changes in the objective environment affect decision-making, achieving a shift from passive response to proactive care. Furthermore, the recommendation action is triggered only when the monitored user scenario state meets preset state triggering conditions and the external event meets preset event triggering conditions, ensuring both accuracy and timeliness.

[0112] Optionally, as an embodiment, the processor 610 is specifically configured to determine the atomic intent corresponding to the user input content, and extract key parameters associated with the atomic intent; determine the user scenario state to be monitored based on the atomic intent; and determine the external events to be monitored based on the key parameters.

[0113] Optionally, as an embodiment, the user scenario state includes at least one of the following: transportation ticket booking status, accommodation booking status, document processing status, and user geographical location; wherein, the status triggering condition is a logical condition defined based on the combination of values ​​of multiple status items in the user scenario state.

[0114] Optionally, as an embodiment, the external events include at least one of the following: weather warning events, flight change events, traffic control events, and safety notice events provided in real time by external data sources; wherein, the value of the event region in the event triggering conditions is set based on the destination parameter in the key parameters.

[0115] Optionally, as an embodiment, the processor 610 is specifically configured to, when there are multiple different types of external events, and all of them satisfy their respective event triggering conditions, determine a target event from the multiple external events according to the urgency of each external event and the degree of matching between each external event and the user scenario state; and push first recommended content based on the target event.

[0116] It should be understood that, in this embodiment, the input unit 604 may include a graphics processing unit (GPU) 6041 and a microphone 6042. The GPU 6041 processes image data of still images or videos obtained by an image capture device (such as a camera) in video capture mode or image capture mode. The display unit 606 may include a display panel 6061, which may be configured in the form of a liquid crystal display, an organic light-emitting diode, etc. The user input unit 607 includes a touch panel 6071 and other input devices 6072. The touch panel 6071 is also called a touch screen. The touch panel 6071 may include a touch detection device and a touch controller. Other input devices 6072 may include, but are not limited to, a physical keyboard, function keys (such as volume control buttons, power buttons, etc.), a trackball, a mouse, and a joystick, which will not be described in detail here. The memory 609 can be used to store software programs and various data, including but not limited to applications and operating systems. Processor 610 can integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into processor 610.

[0117] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the recommended method embodiments described above and achieve the same technical effects. To avoid repetition, these will not be repeated here.

[0118] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0119] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-mentioned recommended method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0120] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0121] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0122] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0123] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A content recommendation method, characterized in that, The method includes: Determine the user's intent based on the user's input; Obtain the user scenario state and external events corresponding to the user intent; If the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions, the first recommended content is pushed based on the external event.

2. The method according to claim 1, characterized in that, The step of determining the user's intent based on the user's input includes: Determine the atomic intent corresponding to the user input content, and extract the key parameters associated with the atomic intent; The acquisition of the user scenario state and external events corresponding to the user intent includes: Based on the atomic intent, determine the user scenario state that needs to be monitored; and based on the key parameters, determine the external events that need to be monitored.

3. The method according to claim 1, characterized in that, The user scenario status includes at least one of the following: transportation ticket booking status, accommodation booking status, document processing status, and user geographical location; The state triggering condition is a logical condition defined based on the combination of values ​​of multiple state items in the user scenario state.

4. The method according to claim 2, characterized in that, The external events include at least one of the following: weather warning events, flight change events, traffic control events, and safety notice events provided in real time by external data sources; The value of the event region in the event triggering condition is set based on the destination parameter in the key parameters.

5. The method according to any one of claims 1-4, characterized in that, The method of pushing the first recommended content based on the external event includes: When there are multiple different types of external events, and all of them meet their respective event triggering conditions, a target event is determined from the multiple external events based on the urgency of each external event and the degree of matching between each external event and the user scenario state. The first recommended content is pushed based on the target event.

6. A content recommendation device, characterized in that, The device includes: The determination module is used to determine the user's intent based on the user's input. The acquisition module is used to acquire the user scenario state and external events corresponding to the user intent; The push module is used to push first recommended content based on the external event when the user scenario state meets the preset state triggering conditions and the external event meets the preset event triggering conditions.

7. The apparatus according to claim 6, characterized in that, The determining module is specifically used to determine the atomic intent corresponding to the user input content, and to extract key parameters associated with the atomic intent; The acquisition module is specifically used to determine the user scenario state to be monitored based on the atomic intent; and to determine the external events to be monitored based on the key parameters.

8. The apparatus according to claim 6, characterized in that, The user scenario status includes at least one of the following: transportation ticket booking status, accommodation booking status, document processing status, and user geographical location; wherein, the status triggering condition is a logical condition defined based on the combination of values ​​of multiple status items in the user scenario status.

9. The apparatus according to claim 7, characterized in that, The external events include at least one of the following: weather warning events, flight change events, traffic control events, and safety notice events provided in real time by external data sources; wherein, the value of the event region in the event triggering conditions is set based on the destination parameter in the key parameters.

10. The apparatus according to any one of claims 6-9, characterized in that, The push module is specifically used to determine a target event from multiple external events when there are multiple different types of external events, all of which meet their respective event triggering conditions, based on the urgency of each external event and the degree of matching between each external event and the user's scenario state. The first recommended content is pushed based on the target event.

11. An electronic device, characterized in that, The electronic device includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the content recommendation method as described in any one of claims 1-5.

Images