Information response method and apparatus

Abstract

The application discloses an information response method and device, and belongs to the technical field of artificial intelligence. The method comprises the following steps: receiving first information input by a user; determining the correlation degree between the first information and a first scene, wherein the first scene is constructed according to pre-collected scene information; and in the case that the correlation degree meets a preset condition, generating and displaying second information for responding to the first information based on the first scene.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, specifically relating to an information response method and apparatus. Background Technology

[0002] With the rapid development of artificial intelligence technology, artificial intelligence (AI) assistants have been widely used in smart terminal devices, providing users with services such as information inquiry, travel planning, life services, and work assistance.

[0003] Currently, AI assistants typically perform semantic understanding based on user-input text queries and combine historical dialogues as context to generate responses. This generation method results in highly generalized responses, leading to poor accuracy. Summary of the Invention

[0004] The purpose of this application is to provide an information response method and apparatus that can solve the problem of poor accuracy in responding to user input.

[0005] In a first aspect, embodiments of this application provide an information response method, the method comprising:

[0006] Receive the first information input by the user;

[0007] Determine the correlation between the first information and the first scenario, which is constructed based on pre-collected scenario information;

[0008] If the correlation degree meets the preset conditions, second information is generated and displayed based on the first scenario to respond to the first information.

[0009] Secondly, embodiments of this application provide an information response device, the device comprising:

[0010] The receiving module is used to receive the initial information input by the user.

[0011] The determination module is used to determine the correlation between the first information and the first scene, wherein the first scene is constructed based on pre-collected scene information;

[0012] The generation module is used to generate and display second information in response to the first information based on the first scenario, provided that the correlation degree meets the preset conditions.

[0013] Thirdly, embodiments of this application provide an electronic device including a processor and a memory, wherein the memory stores a program or instructions executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the information response method as described in the first aspect.

[0014] Fourthly, embodiments of this application provide a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps in the information response method as described in the first aspect.

[0015] Fifthly, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0016] In a sixth aspect, embodiments of this application provide a computer program product stored in a storage medium, which is executed by at least one processor to implement the method described in the first aspect.

[0017] In this embodiment of the application, when the user inputs first information, the system can obtain the user's current scenario and dynamically determine the relevance of the first information to the scenario, thereby selectively introducing relevant scenarios into the conversation generation process and generating more personalized answers that better meet the user's actual needs. Attached Figure Description

[0018] Figure 1 This is a flowchart of an information response method provided in an embodiment of this application;

[0019] Figure 2 This is a flowchart of a scene information processing method provided in an embodiment of this application;

[0020] Figure 3 This is a structural diagram of a context-aware AI assistant personalized response device provided in an embodiment of this application;

[0021] Figure 4 This is a structural diagram of an information response device provided in an embodiment of this application;

[0022] Figure 5 This is one of the structural diagrams of an electronic device provided in the embodiments of this application;

[0023] Figure 6 This is the second structural diagram of an electronic device provided in the embodiments of this application. Detailed Implementation

[0024] 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.

[0025] 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 use of data can be interchanged 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.

[0026] Currently, AI assistants primarily rely on user input to respond, and even when contextual information is incorporated, it's often limited to past conversations. However, in real-world applications, users often ask questions in specific, real-world scenarios, such as while traveling, in inclement weather, on business trips, or with low battery. Although user input is expressed in natural language, their true needs often implicitly depend on the current physical environment, device status, and individual behavior.

[0027] To improve response accuracy, contextual information such as location, weather, or schedule information can be introduced during the dialogue to enhance the practicality of the dialogue results. A unified contextual understanding and management mechanism can be adopted to enable it to function flexibly and accurately in general conversational scenarios.

[0028] The information response method provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0029] See Figure 1 , Figure 1 This is a flowchart of an information response method provided in this application, such as... Figure 1 As shown, information response methods include:

[0030] Step 101: Receive the first information input by the user;

[0031] The first piece of information can be input by the user through text, voice, image, or other means. This information can be query information or everyday conversation information. For example, a user might input, "How do I get to the subway station?".

[0032] Step 102: Determine the correlation between the first information and the first scene, wherein the first scene is constructed based on pre-collected scene information;

[0033] A scenario can be constructed based on pre-collected scenario information. The user's terminal device can collect the user's current scenario information in real time or at regular intervals, and update it periodically or under certain conditions. The scenario information can include user behavior data, the operating status data of the user's device, and environmental data of the user's current environment.

[0034] In some implementations, multi-source scene information can be collected through different means, thereby extracting or generating scene information based on the multi-source scene information to improve the accuracy of the scene information.

[0035] Based on the collected scene information, scene semantic units can be constructed as the current first scene. One or more first scenes can be constructed, and each first scene can be described by scene type, effective time, spatial applicable scope, etc.

[0036] For example, based on the user's current location "subway station", the time is Friday "9:10", the user's commuting habit "subway", the time of leaving home "9:00", the meeting time "9:30", the travel scenario "on the subway", the behavior scenario "listening to music", and the daily scenario "before the work meeting starts".

[0037] The aforementioned scenario information can be collected before receiving the first piece of information input by the user, or it can be collected in real time after receiving the first piece of information.

[0038] A first scene is constructed based on the collected scene information, and the correlation between the first information and the first scene is determined. In some implementations, the semantic similarity between the first information and the first scene can be determined based on the semantics of the first information and the semantics of the first scene. It can also be determined whether the input time of the first information is within the valid time range of the first scene, or whether the first information is within a pre-defined first scene. The correlation between the first information and the first scene is determined based on the magnitude of semantic similarity, whether it is within a valid time range, or whether it is within the first scene.

[0039] Step 103: If the correlation degree meets the preset conditions, generate and display the second information based on the first scenario to respond to the first information.

[0040] If the correlation between the first piece of information and the first scene meets a preset condition, it indicates a high correlation between the two, and the corresponding first scene is used as context to generate the second piece of information. The preset condition can be that the correlation between the first piece of information and the first scene is greater than a preset threshold, or that the correlation falls within a preset ranking range.

[0041] The second information is generated based on the first information input by the user and the first scenario in which the user is currently located. Taking the user's current scenario into account can improve the accuracy of the second information generation.

[0042] For example, when a user enters "How do I get to the subway station?", the system constructs a first scenario based on pre-collected scenario information, determines that the user's input is related to the weather scenario, and outputs prompts that, in addition to the route to the subway station, may also include "It is currently raining, remember to bring an umbrella".

[0043] When the first piece of information has a high degree of relevance to multiple first scenarios, the second piece of information can be output based on the first piece of information input by the user, while also considering multiple first scenarios. In the example above, the second piece of information can be output based on the user's current travel scenario "morning rush hour" and the weather scenario "rainy day," to provide more accurate prompts.

[0044] After generating the second piece of information, the system outputs a personalized conversational response, or second information, to the user, based on the first scenario. In most cases, the system directly displays the final generated response in natural language, making the user perceive the response as more relevant to their current situation. For example, even with the same "travel advice," the response (second information) will show significant differences depending on the scenario (such as time, location, or mode of transportation). The system can also provide brief prompts explaining which scenario factors, such as weather, mode of transportation, or current time of day, were considered in the response, thereby enhancing the user's trust in the AI ​​assistant's ability to "understand the current context."

[0045] When outputting the second information, it can be displayed on the screen, for example, by text or images. In addition, the second information can also be output via audio.

[0046] By using the above method, response information is generated based on the correlation between the first information and the first scenario, combined with scenarios with high correlation, thereby reducing interference from irrelevant information and improving the accuracy of the response information.

[0047] In this embodiment of the application, when the user inputs first information, the system can obtain the user's current scenario and dynamically determine the relevance of the first information to the scenario, thereby selectively introducing relevant scenarios into the conversation generation process and generating more personalized answers that better meet the user's actual needs.

[0048] Optionally, determining the correlation between the first information and the first scenario includes:

[0049] Obtain the correlation factors between the first information and the first scenario, wherein the correlation factors include at least one of the following: semantic similarity factor, time validity factor, scenario confidence factor, and scenario prior factor;

[0050] The correlation between the first information and the first scenario is determined based on each of the at least one related factors and the weight corresponding to each related factor.

[0051] Upon receiving initial user input, semantic parsing can be performed to obtain a semantic representation. This semantic representation can include extracted keywords, intent type, and semantic feature vectors, which can be generated using a pre-trained semantic encoding model.

[0052] The scene semantic unit is treated as an independent searchable semantic item. For each first scene, a relevance score is calculated between it and the current first information. This relevance score is used to characterize the degree of association between the first information and the first scene.

[0053] The semantic similarity factor can be expressed as Based on the similarity between the query semantic vector and the semantic vector of the scene semantic unit, the degree of closeness between the two in the semantic space is measured, as expressed in Formula 1:

[0054] (1)

[0055] in, It is a query semantic vector. It is a scene semantic vector. It refers to semantic similarity.

[0056] The time-efficiency factor can be expressed as: If the current time falls within the time validity period of the scene semantic unit, the weight of the time validity factor is higher; if the current time deviates from the time validity period of the scene semantic unit, the weight of the time validity factor decreases exponentially.

[0057] Confidence weighting factor It is determined based on the confidence field of the scene semantic unit.

[0058] Scenario Prior Factors The weighting can be determined based on the relationship between the query intent of the first piece of information and the semantic dimension of the scene semantic unit. Furthermore, different prior weights (pre-set) can be assigned to different semantic dimensions. For example, in a travel-related query, the weights for travel status, location, and weather scene type would be higher.

[0059] Based on one or more of the above-mentioned relevant factors and the weight of each relevant factor, the comprehensive relevance score can be calculated using Formula 2.

[0060] (2)

[0061] in, It is a configurable weight parameter.

[0062] The relevance score can be used to characterize the degree of association between the first scene and the first piece of information.

[0063] For example, if the user's first input is "What should I pay attention to when taking the subway this morning?", relevance calculations are performed based on multiple pre-constructed first scenarios (i.e., scenario semantic units). The scenario semantic units are shown in Table 1.

[0064] Table 1

[0065] Scene semantic unit Sim Time Confidence Prior Score Subway high high 0.95 high 0.9 Go to work (go_to_work) mid high 0.90 mid 0.7 Listen to music (listening_music) low high 0.80 low 0.6 Work meeting low mid 0.95 high 0.8

[0066] Based on the aforementioned multiple relevance factors and their corresponding weights, the degree of association between the first information and each scene semantic unit can be obtained, thereby determining whether the corresponding scene needs to be introduced to improve the accuracy of the second information generation.

[0067] Optionally, the number of the first scenarios is N, where N is an integer greater than 1; the step of generating and displaying second information based on the first scenarios to respond to the first information when the correlation degree meets a preset condition includes:

[0068] The N first scenarios are sorted according to their degree of relevance.

[0069] Among the N first scenarios, M first scenarios are selected whose sorting is within a preset range or whose relevance is greater than a preset value, where M is an integer greater than 1 and M≤N;

[0070] Based on the M first scenarios, second information is generated and displayed to respond to the first information.

[0071] Based on the constructed scene semantic units (i.e., the first scene), a relevance score can be calculated for each scene semantic unit relative to the first information input by the current user. The conversation assistant responds to the relevance score, classifies the scene semantic units for use, and introduces them into the conversation generation process in a controlled manner.

[0072] Specifically, the N first scenarios are sorted according to their relevance to the first information, and only the M first scenarios with relevance scores higher than a preset value or ranked at the top are retained as the information set that can participate in generating this session. Semantic units of scenarios with low relevance scores can be ignored.

[0073] For example, the three acquired scenarios are ranked according to their relevance as Scenario 1, Scenario 3, and Scenario 2. In some implementations, the first two scenarios with the highest relevance (i.e., Scenario 1 and Scenario 3) can be acquired. In some implementations, scenarios with relevance scores higher than a preset value (including Scenario 1, Scenario 3, and Scenario 2) can be acquired.

[0074] In some implementations, a scenario that simultaneously satisfies both of the above conditions can be selected, namely, a scenario where the ranking is within a preset range and the correlation is greater than a preset value, in order to improve response accuracy.

[0075] During the generation process, relevance scores can be used to characterize the importance of different scene semantic units to the primary information, thereby influencing their participation and influence in the generation process. For example, scene semantic units with higher relevance scores will be prioritized for guiding the focus and detail supplementation of the answer content, while scene semantic units with lower relevance scores but still selected will only participate in the generation as background constraints.

[0076] Before or during the generation of a session, scene semantic units need to perform privacy protection processing on scene information, anonymize sensitive fields such as user identity and location, encrypt information during the transmission of input to the large model, and ensure that scene information is not stored or reused for a long time after the session ends.

[0077] In practical applications, not all user queries require contextual information to aid understanding. Directly concatenating all contextual information into the large model inputting the user query, without control and filtering of the injected content, can easily lead to generated results that deviate from the user's original intent. It also increases computational and power consumption overhead on the device side, which is not conducive to real-time operation on smart terminal devices.

[0078] By employing the above method, the system dynamically determines whether contextual information needs to be introduced and, if so, which contextual information to introduce, based on the user's query content during the session, thereby reducing interference from irrelevant contextual information in the model's generated results. Furthermore, this method can identify N scenarios highly relevant to the initial information, allowing for the generation of response information based on these N scenarios, thus improving response accuracy.

[0079] Optionally, before receiving the first information input by the user, the method further includes:

[0080] Collect scene information associated with the user, the scene information including at least one of the following:

[0081] The environmental data of the user's environment;

[0082] The user's behavioral data;

[0083] The user's schedule data;

[0084] The status data of the terminal device used by the user;

[0085] The user's preference data.

[0086] Among them, the environmental data of the user's environment may include location information, time information, weather information, etc.

[0087] User behavior data can include behavioral data related to the user's current behavioral state, such as application (APP) clicks, movement behavior, and physiological behaviors such as sleep.

[0088] User schedule data includes schedule-related data related to the user's schedule, such as meetings, business trips, and travel tickets;

[0089] The status data of the terminal device includes device status data related to the operating status of the terminal device, such as device power and charging status;

[0090] User preference data includes preference-related data related to users' long-term or short-term usage habits, such as consumption information, entertainment (music, video, reading) preferences, etc.

[0091] The aforementioned scene information (also known as scene-aware data) is acquired according to a preset collection frequency and a dynamically adjusted collection strategy. Different scene information can correspond to different collection frequencies. Environmental and behavioral data with high real-time requirements can be collected at a frequency of seconds; schedule and preference data with low change frequency can be collected at a frequency of days or triggered when the status changes; while device status data is collected triggered when the status changes.

[0092] Simultaneously, the data collection frequency can be dynamically adjusted based on the terminal device's battery level, system load, user activity level, and the magnitude of scene changes. The collection frequency is reduced when the terminal device is detected to be stationary or in a stable scene, and increased when location changes, behavioral changes, or the activation of an AI assistant are detected. This ensures real-time scene awareness while reducing computational resource consumption and device power consumption.

[0093] The aforementioned raw scene perception data is only used as state input during the acquisition phase and does not directly participate in the generation of session responses.

[0094] By collecting the aforementioned multi-source data, various scenario information can be obtained to construct different scenarios, thereby improving the comprehensiveness of scenario information and generating response information based on the user's scenario.

[0095] Optionally, the scene information includes multi-source data collected through different means, and before determining the correlation between the first information and the first scene, the method further includes:

[0096] The multi-source data shall be subjected to at least one of the following alignment processes: time alignment, event-driven alignment, semantic alignment, and feature space alignment.

[0097] The multi-source data after alignment is standardized.

[0098] Among them, multi-source data can be raw data from different sensors and applications (such as Global Positioning System (GPS) location, weather application programming interface (API), user behavior, etc.).

[0099] After collecting the aforementioned multi-source data, the multi-source data can be uniformly parsed, aligned, and standardized to construct a unified scene semantic unit representation that can be understood and utilized by downstream modules.

[0100] like Figure 2 As shown, data alignment is performed on multi-source data to address alignment issues at the temporal, spatial, and semantic levels, thereby constructing a highly consistent and highly fusionable temporal sequence representation. Specific methods may include, but are not limited to, the following:

[0101] Time alignment: By unifying the time axis, time window, and time decay mechanism, the sensing data with different sampling frequencies and reports are aligned. For example, multiple data collected by sensors of terminal devices such as Wireless Fidelity (WiFi) connection, GPS, and pedometer are aligned according to the usage timestamp.

[0102] For example, during a user's walk from the office to the subway station between 10:00 and 10:05, GPS location data, pedometer data, and Wi-Fi connection data are collected simultaneously. Aligning all data points to the 10:00-10:05 time window allows the system to accurately identify the complete scenario of "walking from the office to the subway station between 10:00 and 10:05," avoiding misidentification due to time discrepancies.

[0103] Event-driven alignment: Using the key events of a particular event as trigger points, related multi-source data is reorganized and aligned. For example, location changes or schedule events are used to perform context alignment analysis on the data before and after them. Event priority and processing mechanisms are introduced. Different priorities are assigned to different events based on event type and event importance. When event conflicts are detected, the event with higher priority is selected as the main alignment event, and the remaining events are processed later. Event importance can be determined based on preset rules or model training.

[0104] For example, a user's schedule might include "company meeting at 10:00 AM." By collecting the user's context information, including time 9:50 AM and location "company," the system can identify the scenario where the user needs to attend a meeting at the company.

[0105] Semantic alignment: Mapping scene information from different data sources but pointing to the same semantic meaning to a unified semantic dimension. For example, using knowledge graphs to classify the tags "running" and "high heart rate" into the "exercising" scene.

[0106] For example, a user might simultaneously use a motion sensor and a health app to monitor their vital signs. Mapping "running" to the semantic tag "exercise" (similarity 0.92) and "high exercise intensity" to the semantic tag "exercise" (similarity 0.95) allows "running" and "high exercise intensity" to be integrated into a single exercise scenario, avoiding redundancy and errors in scenario recognition.

[0107] Feature space alignment: Through feature normalization, dimensionality mapping, and weight adjustment, the comparability and fusion of features from different types of scenes can be achieved. For example, contrastive learning can be used to encode screen frames, text, and sensor data into the same semantic space. Specifically, data from different modalities are first processed by their respective feature encoders to learn intra-modal representations, obtaining intermediate feature representations with unified dimensional constraints. Then, cross-modal functions project the features of each modality into the same semantic space. Cross-modal mapping can be implemented using contrastive learning or shared projection layers. Furthermore, different fusion weights can be assigned based on the stability, confidence, and noise level of the features from different modalities.

[0108] For example, a user views an image of "rain" on their phone screen (visual data), while the weather API returns a text description of "light rain" (semantic data). By extracting visual and text features, different weights are assigned to the visual and semantic data based on confidence and stability, and the data is then fused to obtain the scenario corresponding to the two sets of data as "rain".

[0109] Based on the aligned data, standardization is performed to construct a unified data representation system, and standardized representation is carried out at the structural and semantic levels.

[0110] Structure standardization: Using a unified data structure format to standardize fields, constrain types, and organize hierarchy in the scenario data. Structured data representation methods can include, but are not limited to, text structures, key-value pair structures, binary structures, or in-memory object structures.

[0111] Semantic standardization: Mapping scene semantics into vectorized representations to support subsequent similarity calculations, relevance assessments, and model input. Vectorization is generated through a pre-trained semantic encoding model: First, the structured standardized data is converted into semantic description text that the model can process. Then, the pre-trained model encodes the semantic description text, generating fixed-dimensional scene vectors through pooling operations. The vector dimension is fixed at 256 dimensions.

[0112] Contextual information from different sources (such as location, weather, device status, etc.) usually exists in raw data or unstructured form, lacking a unified semantic abstraction and structured representation, making it difficult to effectively integrate, filter and reuse, and easily forming redundant or conflicting information in the dialogue generation process.

[0113] By using the above methods, the collected multi-source scene data can be transformed into unified standard data, which can improve the accuracy and comprehensiveness of the data.

[0114] Optionally, before determining the correlation between the first information and the first scenario, the method further includes:

[0115] Based on the scene type to which the scene information belongs, a first scene corresponding to the scene information is constructed. The first scene is described by at least one of the following: scene type, scene semantics, scene confidence, time validity, and spatial usage scope.

[0116] After obtaining standardized scene information, the scene information is mapped into a set of structured scene semantic units (i.e., the first scene) based on the scene type to which the scene information belongs. This enables the scene semantic units to have semantic interpretability, searchability, and computability, which is used to support subsequent scene relevance matching and decision processing, and realize the transformation from low-level data to high-level semantic concepts.

[0117] Scene semantic units can be described by scene type, scene semantic value, confidence level, temporal validity, spatial applicability, and semantic vector representation. The scene type can be a predefined category, such as travel scenario or work scenario. One or more corresponding scene types can be identified for the same scene information. The confidence level reflects the confidence level of the scene information after aligning or fusing the collected raw multi-source data.

[0118] Since scene information changes over time, the corresponding scene will also change. Time validity can be the proximity of the current scene's time range; when the time outside this range is long, the time validity is low. Spatial applicability can be the spatial scope of the scene, such as a region.

[0119] As shown in Table 2, Table 2 is a scene semantic unit constructed based on scene information.

[0120] Table 2

[0121] Location time Front-end application Commuting habits Time away from home Meeting time Scene semantic unit metro station Friday 9:10 music subway 9:00 9:30 { "params": [{ "dimension": "travel scenario", "value": "on the subway","confidence":0.95, "time_validity":"2026-01-30T09:00-09:30","spatial_scope": "Shenzhen Metro (Station A)","embedding":[...]},{ "dimension": "behavior scenario", "value": "listening to music", "confidence":0.80, "embedding":[...]},{ "dimension": "schedule scenario", "value": "before the start of a work meeting","confidence":0.95,"time_validity":"2026-01-30T09:10~09:30","spatial_scope": "company location", "embedding":[...]} ]}

[0122] When a user initiates a query (enters the first piece of information), based on the correlation calculation results between the query semantics and the scene semantic units, it is determined whether the corresponding scene information needs to be introduced, and a subset of scene semantic units related to the current query is selected as the set of injectable context information.

[0123] Using the above method, the correlation between the scene description information and the first information can be determined, thus improving the accuracy of the correlation calculation.

[0124] As a specific example, such as Figure 3 As shown, a context-aware AI assistant personalized response device includes:

[0125] The scene perception module is used to continuously collect multi-source raw scene perception data in the background.

[0126] The scene parsing module is used to parse, align, standardize, and semantically abstract scene-aware data, constructing a unified scene semantic unit representation that can be understood and utilized by downstream modules. Specifically, after processing multi-source data, scene semantic units are constructed and represented using scene type, scene semantic value, confidence level, temporal validity, spatial applicability, and semantic vector.

[0127] The scenario triggering module is used to calculate the relevance based on query semantics and scenario semantic units when a user initiates a query. Based on the relevance, it determines whether scenario information needs to be introduced and filters out scenario semantic units relevant to the current query. A small model running on the client side can be used for relevance determination to improve accuracy.

[0128] The conversation assistant response module is used to generate the current conversation based on the selected scene semantic units.

[0129] The presentation module is used to output personalized conversational response content to the user, generated based on the scene's semantic units.

[0130] Current AI assistants struggle to perceive the user's current physical environment and device operating status. The methods described above enable a strong correlation between conversational responses and the user's real-world context, improving response relevance; avoiding interference from irrelevant scene information in the model generation process; reducing edge computing power and consumption, improving system real-time performance; and enhancing the user's perception of the intelligent assistant's "scene understanding ability."

[0131] The embodiments of this application can achieve the following beneficial effects:

[0132] 1. Unify and abstract multi-source device sensing data into structured scene semantic units to achieve a unified semantic expression of the user's current scene;

[0133] 2. Through the context (i.e. scenario) semantic fusion mechanism, ensure the consistency, real-time performance and reliability of context information from different data sources;

[0134] 3. When a session occurs, a context-relevance judgment mechanism based on user queries is introduced to dynamically filter context information related to the current query, rather than injecting all or fixed amounts of information.

[0135] 4. Adopt a controlled contextual information injection method to enable contextual information to participate in the generation process of large model answers as generation constraints or auxiliary context, thereby improving the relevance and practicality of the answers without changing the user's original intent.

[0136] The information response method provided in this application can be executed by an information response device. This application uses an information response device executing the information response method as an example to illustrate the information response device provided in this application.

[0137] See Figure 4 , Figure 4 This application provides an information response device, such as... Figure 4 As shown, the information response device 400 includes:

[0138] The receiving module 401 is used to receive the first information input by the user;

[0139] The determining module 402 is used to determine the correlation between the first information and the first scene, wherein the first scene is constructed based on pre-collected scene information;

[0140] The generation module 403 is used to generate and display second information in response to the first information based on the first scenario when the correlation degree meets the preset conditions.

[0141] Optionally, the determining module includes:

[0142] The first acquisition submodule is used to acquire the correlation factors between the first information and the first scenario, wherein the correlation factors include at least one of the following: semantic similarity factor, time validity factor, scenario confidence factor, and scenario prior factor;

[0143] The determination submodule is used to determine the correlation between the first information and the corresponding scenario based on each of the at least one related factors and the weight corresponding to each of the related factors.

[0144] Optionally, the number of the first scenes is N, where N is an integer greater than 1; the generation module includes:

[0145] The sorting submodule is used to sort the N first scenarios according to the degree of relevance.

[0146] The second acquisition submodule is used to acquire M first scenarios from the N first scenarios, where the sorting is within a preset range or the correlation is greater than a preset value, where M is an integer greater than 1 and M≤N;

[0147] A generation submodule is used to generate and display second information in response to the first information based on the M first scenarios.

[0148] Optionally, the device further includes:

[0149] The data acquisition module is used to acquire scene information associated with the user, the scene information including at least one of the following:

[0150] The environmental data of the user's environment;

[0151] The user's behavioral data;

[0152] The user's schedule data;

[0153] The status data of the terminal device used by the user;

[0154] The user's preference data.

[0155] Optionally, the scene information includes multi-source data collected through different methods, and the device further includes:

[0156] The first processing module is used to perform at least one of the following alignment processes on the multi-source data: time alignment, event-driven alignment, semantic alignment, and feature space alignment.

[0157] The second processing module is used to standardize the multi-source data after the alignment process.

[0158] Optionally, the device further includes:

[0159] The construction module is used to construct a first scene corresponding to the scene information based on the scene type to which the scene information belongs. The first scene is described by at least one of scene type, scene semantics, scene confidence, time validity, and spatial usage scope.

[0160] In this embodiment of the application, when the user inputs first information, the system can obtain the user's current scenario and dynamically determine the relevance of the first information to the scenario, thereby selectively introducing relevant scenarios into the conversation generation process and generating more personalized answers that better meet the user's actual needs.

[0161] The information response device in this application embodiment can be an electronic device or a component within an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices besides a terminal. For example, the electronic device can be a mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc. It can also be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM, or self-service machine, etc. This application embodiment does not specifically limit the device.

[0162] The information response device in this application embodiment can be a device with an operating system. The operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit the specific operating system.

[0163] The information response device provided in this application embodiment can achieve... Figure 1 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0164] Optionally, such as Figure 5As shown, this application embodiment also provides an electronic device 500, including a processor 501 and a memory 502. The memory 502 stores a program or instructions that can run on the processor 501. When the program or instructions are executed by the processor 501, they implement the various steps of the above-described information response method embodiment and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0165] 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.

[0166] Figure 6 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of this application.

[0167] 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.

[0168] 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.

[0169] The user input unit 607 is used to: receive first information input by the user;

[0170] The processor 610 is used to determine the correlation between the first information and the first scene, which is constructed based on pre-collected scene information; when the correlation meets a preset condition, it generates and displays second information based on the first scene to respond to the first information.

[0171] In this embodiment of the application, when the user inputs first information, the system can obtain the user's current scenario and dynamically determine the relevance of the first information to the scenario, thereby selectively introducing relevant scenarios into the conversation generation process and generating more personalized answers that better meet the user's actual needs.

[0172] Optionally, the processor 610 performs the determination of the correlation between the first information and the first scene, including:

[0173] Obtain the correlation factors between the first information and the first scenario, wherein the correlation factors include at least one of the following: semantic similarity factor, time validity factor, scenario confidence factor, and scenario prior factor;

[0174] The correlation between the first information and the first scenario is determined based on each of the at least one related factors and the weight corresponding to each related factor.

[0175] Optionally, the number of the first scenarios is N, where N is an integer greater than 1; the processor 610 executes the step of generating and displaying second information based on the first scenarios to respond to the first information when the correlation degree meets the preset conditions, including:

[0176] The N first scenarios are sorted according to their degree of relevance.

[0177] Among the N first scenarios, M first scenarios are selected whose sorting is within a preset range or whose relevance is greater than a preset value, where M is an integer greater than 1 and M≤N;

[0178] Based on the M first scenarios, second information is generated and displayed to respond to the first information.

[0179] Optionally, the input unit 604 is used for:

[0180] Collect scene information associated with the user, the scene information including at least one of the following:

[0181] The environmental data of the user's environment;

[0182] The user's behavioral data;

[0183] The user's schedule data;

[0184] The status data of the terminal device used by the user;

[0185] The user's preference data.

[0186] Optionally, the scene information includes multi-source data collected through different methods, and the processor 610 is further used for:

[0187] The multi-source data shall be subjected to at least one of the following alignment processes: time alignment, event-driven alignment, semantic alignment, and feature space alignment.

[0188] The multi-source data after alignment is standardized.

[0189] Optionally, the processor 610 is further configured to:

[0190] Based on the scene type to which the scene information belongs, a first scene corresponding to the scene information is constructed. The first scene is described by at least one of the following: scene type, scene semantics, scene confidence, time validity, and spatial usage scope.

[0191] 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, or the like. The user input unit 607 includes at least one of 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, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be described in detail here.

[0192] The memory 609 can be used to store software programs and various data. The memory 609 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 609 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory 609 in this embodiment includes, but is not limited to, these and any other suitable types of memory.

[0193] Processor 610 may include one or more processing units; optionally, processor 610 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 610.

[0194] 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 above-described information response method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0195] 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.

[0196] 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-described information response method embodiments and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0197] 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.

[0198] This application provides a computer program product, which is stored in a storage medium and executed by at least one processor to implement the various processes of the information response method embodiments described above, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0199] 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.

[0200] 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) and includes 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.

[0201] 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. An information response method, characterized in that, include: Receive the first information input by the user; Determine the correlation between the first information and the first scenario, which is constructed based on pre-collected scenario information; If the correlation degree meets the preset conditions, second information is generated and displayed based on the first scenario to respond to the first information.

2. The method according to claim 1, characterized in that, Determining the correlation between the first information and the first scenario includes: Obtain the correlation factors between the first information and the first scenario, wherein the correlation factors include at least one of the following: semantic similarity factor, time validity factor, scenario confidence factor, and scenario prior factor; The correlation between the first information and the first scenario is determined based on each of the at least one related factors and the weight corresponding to each related factor.

3. The method according to claim 1 or 2, characterized in that, The number of the first scenarios is N, where N is an integer greater than 1; the step of generating and displaying second information based on the first scenarios to respond to the first information, when the correlation degree meets a preset condition, includes: The N first scenarios are sorted according to their degree of relevance. Among the N first scenarios, M first scenarios are selected whose sorting is within a preset range or whose relevance is greater than a preset value, where M is an integer greater than 1 and M≤N; Based on the M first scenarios, second information is generated and displayed to respond to the first information.

4. The method according to claim 1 or 2, characterized in that, Before receiving the first information input by the user, the method further includes: Collect scene information associated with the user, the scene information including at least one of the following: The environmental data of the user's environment; The user's behavioral data; The user's schedule data; The status data of the terminal device used by the user; The user's preference data.

5. The method according to claim 4, characterized in that, The scene information includes multi-source data collected through different means. Before determining the correlation between the first information and the first scene, the method further includes: The multi-source data shall be subjected to at least one of the following alignment processes: time alignment, event-driven alignment, semantic alignment, and feature space alignment. The multi-source data after alignment is standardized.

6. An information response device, characterized in that, include: The receiving module is used to receive the initial information input by the user. The determination module is used to determine the correlation between the first information and the first scene, wherein the first scene is constructed based on pre-collected scene information; The generation module is used to generate and display second information in response to the first information based on the first scenario, provided that the correlation degree meets the preset conditions.

7. The apparatus according to claim 6, characterized in that, The determining module includes: The first acquisition submodule is used to acquire the correlation factors between the first information and the first scenario, wherein the correlation factors include at least one of the following: semantic similarity factor, time validity factor, scenario confidence factor, and scenario prior factor; The determination submodule is used to determine the correlation between the first information and the first scenario based on each of the at least one related factors and the weight corresponding to each of the related factors.

8. The apparatus according to claim 6 or 7, characterized in that, The number of the first scenes is N, where N is an integer greater than 1; the generation module includes: The sorting submodule is used to sort the N first scenarios according to the degree of relevance. The second acquisition submodule is used to acquire M first scenarios from the N first scenarios, where the sorting is within a preset range or the correlation is greater than a preset value, where M is an integer greater than 1 and M≤N; A generation submodule is used to generate and display second information in response to the first information based on the M first scenarios.

9. The apparatus according to claim 6 or 7, characterized in that, The device further includes: The data acquisition module is used to acquire scene information associated with the user, the scene information including at least one of the following: The environmental data of the user's environment; The user's behavioral data; The user's schedule data; The status data of the terminal device used by the user; The user's preference data.

10. The apparatus according to claim 9, characterized in that, The scene information includes multi-source data collected through different methods, and the device further includes: The first processing module is used to perform at least one of the following alignment processes on the multi-source data: time alignment, event-driven alignment, semantic alignment, and feature space alignment. The second processing module is used to standardize the multi-source data after the alignment process.

Images