Method, device and equipment for determining user preference based on multiple rounds of probing, and medium

Abstract

The application discloses a user preference determination method and device based on multi-round detection, equipment and medium, relates to the technical field of user preference detection, and comprises the following steps: selecting a first label from a representative label set of a preset first-level detection space to perform user preference detection, and determining a first target label; selecting a second label associated with the first target label from a representative label set of a preset second-level detection space to perform user preference detection, and determining a second target label; selecting a third label associated with the second target label from a representative label set of a preset final-level detection space to perform user preference detection, and determining a third target label; and determining a feature label set based on the first target label, the second target label and the third target label. Through a three-level detection progressive architecture, hierarchical and accurate mining of user preferences is realized, the breadth of user interest is covered, the depth of subdivided demand is captured, the one-sidedness of traditional single detection is overcome, and the content matching accuracy is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of user preference detection technology, and in particular to a method, apparatus, device and medium for determining user preferences based on multi-round detection. Background Technology

[0002] With the rapid development of internet technology and the explosive growth of digital content, users are facing increasingly significant problems of information overload. Against this backdrop, content recommendation systems have gradually become one of the core technological means to improve user experience and enhance platform stickiness. Their core objective is to analyze user behavior and preferences to achieve personalized content distribution, thereby helping users efficiently obtain information of interest.

[0003] Traditional user preference detection methods primarily rely on surface analysis of explicit behavioral data, such as users' browsing history, click behavior, and page dwell time. These methods infer users' interests by statistically analyzing high-frequency behavioral data and establishing simple association rules or weighting models.

[0004] However, the effectiveness of recommender systems heavily relies on the accuracy and comprehensiveness of user preference detection. How to accurately capture users' complex and diverse content preferences and generate highly relevant recommendations remains a key challenge that urgently needs to be overcome in the current technological field. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and medium for determining user preferences based on multi-round detection, aiming to solve the problem of low accuracy of recommendation results in existing recommendation systems.

[0006] In a first aspect, embodiments of the present invention provide a user preference determination method based on multi-round probing, comprising:

[0007] Select the first label from the representative label set of the preset primary detection space to detect user preferences and determine the first target label;

[0008] Select a second label associated with the first target label from a representative label set in a preset secondary detection space to detect user preferences and determine the second target label;

[0009] Select a third label associated with the second target label from the representative label set of the preset final-level detection space to detect user preferences and determine the third target label;

[0010] A set of feature labels is determined based on the first target label, the second target label, and the third target label.

[0011] Secondly, embodiments of the present invention also provide a user preference determination apparatus based on multi-round detection, which includes a unit for performing the above-described method.

[0012] Thirdly, embodiments of the present invention also provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0013] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described method.

[0014] This invention provides a method, apparatus, device, and medium for determining user preferences based on multi-round detection. The method includes: selecting a first tag from a representative tag set of a preset first-level detection space to detect user preferences and determine a first target tag; selecting a second tag associated with the first target tag from a representative tag set of a preset second-level detection space to detect user preferences and determine a second target tag; selecting a third tag associated with the second target tag from a representative tag set of a preset final-level detection space to detect user preferences and determine a third target tag; and determining a set of feature tags based on the first target tag, the second target tag, and the third target tag. This invention achieves hierarchical and precise mining of user preferences through a three-level progressive detection architecture: first-level detection locates basic interest directions (e.g., "movies"), second-level detection associates and refines fields (e.g., "science fiction movies"), and final-level detection deepens specific preferences (e.g., "hard science fiction"). The synergistic integration of the three levels of tags forms a structured set of feature tags, covering both the breadth of user interests and the depth of subdivided needs, overcoming the one-sidedness of traditional single-round detection, providing high-granularity data support for personalized recommendations, and significantly improving content matching accuracy. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the user preference determination method based on multi-round detection provided in an embodiment of the present invention;

[0017] Figure 2 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0020] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0021] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0022] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0023] Please see Figure 1 This invention proposes a method for determining user preferences based on multi-round detection, the method comprising the following steps:

[0024] S1, select the first label from the representative label set of the preset primary detection space to detect user preferences and determine the first target label.

[0025] S2, select a second label associated with the first target label from the representative label set of the preset secondary detection space to detect user preferences and determine the second target label.

[0026] S3, select a third label associated with the second target label from the representative label set of the preset final detection space to detect user preferences and determine the third target label.

[0027] S4, determine a set of feature labels based on the first target label, the second target label, and the third target label.

[0028] In practice, a primary detection space, a secondary detection space, and a final detection space are pre-constructed, each containing a representative label set. The labels in these representative label sets are hierarchically layered, progressing progressively deeper. The label set in the secondary detection space contains more granular preference dimensions than that in the primary detection space; similarly, the label set in the final detection space contains even more granular preference dimensions than that in the secondary detection space.

[0029] This invention employs a three-level detection method to determine the feature tag set. Specifically, it sequentially performs user preference detection by selecting a first tag from a representative tag set in a preset first-level detection space to determine a first target tag; by selecting a second tag associated with the first target tag from a representative tag set in a preset second-level detection space to determine a second target tag; and by selecting a third tag associated with the second target tag from a representative tag set in a preset final-level detection space to determine a third target tag. Finally, the first target tag, the second target tag, and the third target tag are integrated to obtain the feature tag set.

[0030] User preference detection involves providing users with content corresponding to selected tags, collecting user behavior data, determining user preferences based on the user behavior data, and determining tags (the first target tag, the second target tag, and the third target tag) that match the user based on the user preferences.

[0031] In this embodiment of the invention, a three-tiered progressive detection architecture—a primary detection space, a secondary detection space, and a final detection space—is constructed to achieve a layer-by-layer mining of user preferences, from broad to granular. First, the primary detection space selects a first tag from a representative tag set for initial preference localization (e.g., the "movie" category), providing a basic direction for subsequent detection. Second, the secondary detection space selects a second tag (e.g., "science fiction movie") based on the correlation with the first target tag, achieving a more refined focus on the area of ​​interest. Finally, the final detection space further correlates the second target tag with a third tag (e.g., "hard science fiction"), completing a deep characterization of user preferences. This hierarchical detection mechanism effectively avoids the one-sidedness of traditional single-detection methods, ensuring that the feature tag set covers both the main areas of user interest and accurately captures specific preference details. Through the collaborative integration of the three levels of tags, the system can output a structured user preference profile (e.g., "movie → science fiction movie → hard science fiction"), providing high-granularity data support for personalized recommendations and significantly improving the matching accuracy between recommended content and users' actual needs.

[0032] In some preferred embodiments, the above step "selecting a first label from a representative label set of a preset primary detection space to detect user preferences and determine a first target label" specifically includes the following steps: selecting a portion of the labels from the representative label set of the primary detection space as the first label for user preference detection; determining whether the confidence level of the first label meets a preset confidence level requirement; if the confidence level of the first label does not meet the preset confidence level requirement, proceeding to the step of selecting a portion of the labels from the representative label set of the primary detection space as the first label for user preference detection; if the confidence level of the first label meets the preset confidence level requirement, using the first label as the first target label.

[0033] In practice, confidence level is a key indicator for assessing the degree of match between detected labels and users' true preferences. The logic for determining confidence level is as follows:

[0034] 1. Data Source

[0035] Explicit feedback: Obtain explicit feedback from users on tags directly through the I / O module (such as "like / dislike" selection, rating).

[0036] Implicit behavior: Analyze user interaction behavior with tagged content (such as click-through rate, dwell time, frequency of repeat visits, sharing / favoriting actions).

[0037] Confidence score dynamic calculation rules:

[0038] Explicit feedback score: User-initiated feedback (such as ratings) directly affects confidence level, and different ratings correspond to different scores.

[0039] Implicit behavior quantification score: The implicit behavior quantification score is obtained by calculating weights based on behavioral data (e.g., click-through rate × 30% + dwell time × 50% + sharing × 20%). Furthermore, the confidence level is the sum of the explicit feedback score and the implicit behavior quantification score.

[0040] Furthermore, confidence can also be predicted based on user behavior data using a pre-trained neural network model. Specifically, user behavior data (explicit feedback and implicit behavior) is input into a pre-trained neural network model, and the neural network model outputs its confidence.

[0041] In this embodiment, by determining whether the confidence level of the first tag meets the preset requirements, the system can proactively identify the reliability of the detection results: if the confidence level is insufficient (e.g., the user's feedback on the "movie" tag is ambiguous), a new tag is selected for re-detection to avoid misjudgments caused by accidental behavior or data noise; if the confidence level meets the requirements, the first target tag is locked as a valid output. This mechanism forms a closed-loop process of "detection-verification-optimization," significantly improving the accuracy of the first-level detection.

[0042] For example, when a user shows little interest in the initial tag "children's content," the system automatically switches to related tags such as "documentary" for re-verification until a high-confidence interest direction is found. This adaptive capability based on real-time feedback fundamentally solves the problem of preference misjudgment caused by static analysis in traditional methods, ensuring that subsequent secondary and final-level detections are based on a reliable foundation.

[0043] In some preferred embodiments, the above step "selecting a portion of the labels from the representative label set of the primary detection space as the first label for user preference detection" includes: if it is the first selection, randomly selecting a portion of the labels from the representative label set of the primary detection space as the first label for user preference detection; if it is not the first selection, selecting a portion of the labels from the representative label set of the primary detection space as the first label based on user behavior data.

[0044] In this embodiment, the label selection strategy of the primary detection space is optimized: the initial random selection avoids cold start bias (e.g., defaulting to cover major categories such as "movies" and "music"), and subsequent user behavior data based on user preferences focuses on high-probability directions (e.g., prioritizing detection of "documentaries" based on click records).

[0045] This strategy significantly improves detection efficiency—the initial broad-based approach ensures no interest dimensions are overlooked, while subsequent behavior-driven approaches quickly converge to true preferences. For example, when a user frequently browses food-related content, the system automatically lowers the detection priority of the "sports" tag and concentrates resources on verifying the "cooking" related tag. This dynamic resource allocation mechanism reduces invalid computation by approximately 40% while ensuring high confidence in the first-level detection results, providing an efficient starting point for the hierarchical detection chain.

[0046] In some preferred embodiments, the above step "selecting a second label associated with the first target label from a representative label set in a preset secondary detection space to perform user preference detection and determine the second target label" specifically includes the following steps: selecting a portion of the labels associated with the first target label from the representative label set in the secondary detection space as the second label for user preference detection; determining whether the coverage of the second label meets a preset coverage requirement; if the coverage of the second label does not meet the preset coverage requirement, proceeding to the step of selecting a portion of the labels associated with the first target label from the representative label set in the secondary detection space as the second label for user preference detection; if the coverage of the second label meets the preset coverage requirement, using the second label as the second target label.

[0047] In practical implementation, tag coverage refers to the extent to which the system can effectively cover the user's actual content preferences using the current tag set during the user preference detection process. In other words, it's the proportion of user behavior data that the selected tags can cover, specifically the proportion of user behavior data associated with the tags.

[0048] In this embodiment, by determining whether the coverage of the second tag meets the requirements (e.g., whether the "science fiction movie" tag covers enough user behavior data), the system can evaluate the comprehensiveness of the subdivided interest areas: if the coverage is insufficient (e.g., only covering 30% of behavior), tag reselection and re-probing are triggered (e.g., supplementing with the "action movie" tag); if the requirements are met, the second target tag is output.

[0049] This mechanism primarily addresses the issue of missing interest dimensions. For example, when a user simultaneously prefers "science fiction movies" and "comedy movies," traditional methods might overlook the latter due to a single detection, while this scheme ensures that all significant interest directions are identified through coverage iteration. This dynamic expansion mechanism gives the secondary detection results both depth and breadth, providing a complete and detailed interest framework for the final detection.

[0050] Furthermore, the coverage of the labels can also be predicted based on user behavior data using a pre-trained neural network model. Specifically, the user behavior data corresponding to the labels is input into the pre-trained neural network model, and the neural network model outputs the coverage of the labels.

[0051] In some preferred embodiments, the above step "selecting a third label associated with the second target label from a representative label set in a preset final-level detection space to perform user preference detection and determine the third target label" specifically includes the following steps: selecting a portion of the labels associated with the second target label from the representative label set in the final-level detection space as the third label for user preference detection; determining whether the coverage of the third label meets a preset coverage requirement; if the coverage of the third label does not meet the preset coverage requirement, proceeding to the step of selecting a portion of the labels associated with the second target label from the representative label set in the final-level detection space as the third label for user preference detection; if the coverage of the third label meets the preset coverage requirement, using the third label as the third target label.

[0052] In practical implementation, tag coverage refers to the extent to which the system can effectively cover the user's actual content preferences using the current tag set during the user preference detection process. In other words, it's the proportion of user behavior data that the selected tags can cover, specifically the proportion of user behavior data associated with those tags.

[0053] In this embodiment, by verifying the coverage of the third label (e.g., whether "hard science fiction" fully reflects the user's specific preferences), the system ensures the integrity of the finest-grained preferences: if the target is not met (e.g., the user also prefers the "space opera" subcategory), the labels are expanded and re-detected; if the target is met, the third target label is output. This mechanism focuses on optimizing the fine-grained representation of preferences.

[0054] For example, when a user has a strong preference for the "time travel" theme within "hard science fiction," coverage checks can drive the system to add relevant tags, avoiding mistaking broad tags (such as "science fiction movies") for the final result. This deep coverage capability enables end-level probing to accurately capture the differentiated details of user interests, providing a high-value data source for the feature tag set.

[0055] Furthermore, the coverage of the labels can also be predicted based on user behavior data using a pre-trained neural network model. Specifically, the user behavior data corresponding to the labels is input into the pre-trained neural network model, and the neural network model outputs the coverage of the labels.

[0056] In some preferred embodiments, the above step "determine the feature label set based on the first target label, the second target label, and the third target label" specifically includes the following steps: performing feature integration processing on the first target label, the second target label, and the third target label to obtain the feature label set.

[0057] In practice, a mutually exclusive label pool is set up in the primary detection space. The first target label, the second target label, and the third target label are all stored in the mutually exclusive label pool. The mutually exclusive label pool is responsible for integrating the detection results of multiple levels to ensure the comprehensiveness, accuracy, and non-conflictibility of the final user label set. Its core task is to integrate the detection results of the primary detection space (broad domain), the secondary detection space (sub-domain), and the final detection space (specific content) into a unified user preference label through "integrated feature label operation".

[0058] Feature integration processing specifically includes:

[0059] Deduplication: If there is semantic overlap between the tags transmitted in different probe spaces (e.g., the second-level tag "science fiction movie" and the final-level tag "hard science fiction"), they are merged into a unified tag with higher confidence (e.g., "hard science fiction movie").

[0060] Conflict resolution: If there are contradictions between tags (e.g., a user simultaneously tags "likes comedy" and "dislikes funny clips"), the choice is made based on confidence level or user feedback priority.

[0061] The final result is a set of feature labels that are free of redundancy and conflict.

[0062] In this embodiment, feature integration processing merges the three-level target labels into a unified label set. This operation first addresses the redundancy problem of multi-level labels (such as merging "movie" and "science fiction movie"), then constructs a hierarchical structure (such as "movie → science fiction movie → hard science fiction"), and finally generates a priority ranking based on confidence weights. For example, when the second-level label "science fiction movie" (confidence 85%) and the last-level label "hard science fiction" (confidence 90%) have a semantic relationship, the system integrates them into a high-weight path "science fiction movie - hard science fiction".

[0063] This structured output enables the recommendation system to accurately match content levels (such as prioritizing hard science fiction movies), while avoiding the decision ambiguity of traditional flat tag sets, significantly improving recommendation efficiency and user satisfaction.

[0064] In some preferred embodiments, the method further includes the following steps: determining recommended content based on the feature tag set; and pushing the recommended content to the user.

[0065] In this embodiment, the feature tag set is directly linked to the recommendation application. Based on high-precision tags generated by multi-level detection (such as "hard science fiction"), the system can filter highly matching content (such as "Interstellar") and achieve closed-loop verification through push notifications. User feedback on recommended content (such as click-through rates) can inversely optimize tag confidence, forming a reinforced "detection-recommendation-feedback" loop.

[0066] This mechanism completely solves the recommendation bias problem caused by coarse preference data in traditional recommendation systems, such as mistakenly recommending romantic comedies to users with the general tag "movie". Through real-time data iteration, the system continuously adapts to changes in user interests (such as shifting from "science fiction" to "documentary"), achieving dynamic and accurate recommendations.

[0067] This invention proposes a method for determining user preferences based on multi-round detection. The method includes: selecting a first tag from a representative tag set of a preset first-level detection space to detect user preferences and determine a first target tag; selecting a second tag associated with the first target tag from a representative tag set of a preset second-level detection space to detect user preferences and determine a second target tag; selecting a third tag associated with the second target tag from a representative tag set of a preset final-level detection space to detect user preferences and determine a third target tag; and determining a set of feature tags based on the first target tag, the second target tag, and the third target tag. This invention achieves hierarchical and precise mining of user preferences through a three-level progressive detection architecture: first-level detection locates basic interest directions (e.g., "movies"), second-level detection associates and refines fields (e.g., "science fiction movies"), and final-level detection deepens specific preferences (e.g., "hard science fiction"). The synergistic integration of the three levels of tags forms a structured set of feature tags, covering both the breadth of user interests and the depth of subdivided needs, overcoming the one-sidedness of traditional single-round detection, providing high-granularity data support for personalized recommendations, and significantly improving content matching accuracy.

[0068] Corresponding to the above-described user preference determination method based on multi-round probing, the present invention also provides a user preference determination apparatus based on multi-round probing. This apparatus includes a unit for executing the aforementioned user preference determination method based on multi-round probing, and can be configured in a desktop computer, tablet computer, laptop computer, or other terminal. Specifically, the user preference determination apparatus based on multi-round probing includes:

[0069] The first detection unit is used to select a first label from a representative label set of a preset primary detection space to detect user preferences and determine the first target label;

[0070] The second detection unit is used to select a second label associated with the first target label from a representative label set in a preset secondary detection space to detect user preferences and determine the second target label.

[0071] The third detection unit is used to select a third label associated with the second target label from a representative label set in a preset final-level detection space to detect user preferences and determine the third target label.

[0072] An integration unit is used to determine a set of feature labels based on the first target label, the second target label, and the third target label.

[0073] In some preferred embodiments, the step of selecting a first label from a representative label set of a preset primary detection space to detect user preferences and determine a first target label includes:

[0074] Select a portion of the labels from the representative label set of the primary detection space as the first label for user preference detection;

[0075] Determine whether the confidence level of the first tag meets the preset confidence level requirement;

[0076] If the confidence level of the first tag does not meet the preset confidence level requirement, proceed to the step of selecting a portion of the tags from the representative tag set of the first-level detection space as the first tag for user preference detection;

[0077] If the confidence level of the first label meets the preset confidence level requirement, the first label is used as the first target label.

[0078] In some preferred embodiments, the step of selecting a second label associated with the first target label from a representative label set in a preset secondary detection space to perform user preference detection and determine the second target label includes:

[0079] Select a portion of the labels associated with the first target label from the representative label set of the secondary detection space as the second label for user preference detection;

[0080] Determine whether the coverage of the second tag meets the preset coverage requirement;

[0081] If the coverage of the second tag does not meet the preset coverage requirement, proceed to the step of selecting a portion of the tags associated with the first target tag from the representative tag set of the secondary detection space as the second tag for user preference detection;

[0082] If the coverage of the second tag meets the preset coverage requirement, the second tag will be used as the second target tag.

[0083] In some preferred embodiments, the step of selecting a third label associated with the second target label from a representative label set in a preset final-level detection space to perform user preference detection and determine the third target label includes:

[0084] Select a portion of the labels associated with the second target label from the representative label set of the final-level detection space as the third label for user preference detection;

[0085] Determine whether the coverage of the third tag meets the preset coverage requirement;

[0086] If the coverage of the third tag does not meet the preset coverage requirement, proceed to the step of selecting a portion of the tags associated with the second target tag from the representative tag set of the final level detection space as the third tag for user preference detection;

[0087] If the coverage of the third tag meets the preset coverage requirement, the third tag will be used as the third target tag.

[0088] In some preferred embodiments, determining the feature label set based on the first target label, the second target label, and the third target label includes:

[0089] The first target label, the second target label, and the third target label are combined to obtain the feature label set.

[0090] In some preferred embodiments, the user preference determination device based on multi-round detection further includes:

[0091] A determining unit is used to determine recommended content based on the set of feature tags;

[0092] The push unit is used to push the recommended content to the user.

[0093] In some preferred embodiments, selecting a subset of labels from the representative label set of the primary detection space as the first label for user preference detection includes:

[0094] If this is the first selection, a portion of the labels are randomly selected from the representative label set of the primary detection space as the first label for user preference detection;

[0095] If this is not the first selection, a subset of labels will be selected from the representative label set of the primary detection space based on user behavior data as the first label.

[0096] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned user preference determination device and its units based on multi-round detection can be referred to the corresponding descriptions in the foregoing method embodiments. For the sake of convenience and brevity, these will not be repeated here.

[0097] The aforementioned user preference determination device based on multi-round detection can be implemented as a computer program, which can, for example... Figure 2 It runs on the computer device shown.

[0098] Please see Figure 2 , Figure 2This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be an electronic device with communication functions, such as a smartphone, tablet, laptop, desktop computer, personal digital assistant, or wearable device. The server can be a standalone server or a server cluster composed of multiple servers.

[0099] The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0100] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, it causes the processor 502 to execute a user preference determination method based on multi-round probing.

[0101] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0102] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a user preference determination method based on multi-round probing.

[0103] The network interface 505 is used for network communication with other devices. Those skilled in the art will understand that the above structure is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. A specific computer device 500 may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements.

[0104] The processor 502 is used to run a computer program 5032 stored in a memory to implement the steps of a user preference determination method based on multi-round detection provided in any embodiment of the present invention.

[0105] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0106] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0107] Therefore, the present invention also provides a storage medium. This storage medium may be a computer-readable storage medium. The storage medium stores a computer program. When executed by a processor, the computer program causes the processor to perform the steps of a user preference determination method based on multi-round probing provided in any embodiment of the present invention.

[0108] The storage medium is a physical, non-transient storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk, or any other physical storage medium capable of storing program code. The computer-readable storage medium can be non-volatile or volatile.

[0109] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0110] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0111] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0112] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0113] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0114] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0115] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for determining user preferences based on multi-round probing, characterized in that, The method comprises the following steps: selecting a first label from a preset representative label set of a first-level detection space for user preference detection, and determining a first target label; selecting a second label associated with the first target label from a preset representative label set of a second-level detection space for user preference detection, and determining a second target label; selecting a third label associated with the second target label from a preset representative label set of a final-level detection space for user preference detection, and determining a third target label; determining a feature label set based on the first target label, the second target label, and the third target label; wherein the step of selecting a first label from a preset representative label set of a first-level detection space for user preference detection, and determining a first target label, comprises the following steps: selecting part of the labels in the representative label set of the first-level detection space as first labels for user preference detection; determining whether the confidence of the first label meets a preset confidence requirement; if the confidence of the first label does not meet the preset confidence requirement, returning to the step of selecting part of the labels in the representative label set of the first-level detection space as first labels for user preference detection; if the confidence of the first label meets the preset confidence requirement, taking the first label as the first target label; the step of selecting a second label associated with the first target label from a preset representative label set of a second-level detection space for user preference detection, and determining a second target label, comprises the following steps: selecting part of the labels in the representative label set of the second-level detection space associated with the first target label as second labels for user preference detection; determining whether the coverage of the second label meets a preset coverage requirement; if the coverage of the second label does not meet the preset coverage requirement, returning to the step of selecting part of the labels in the representative label set of the second-level detection space associated with the first target label as second labels for user preference detection; if the coverage of the second label meets the preset coverage requirement, taking the second label as the second target label; the step of selecting a third label associated with the second target label from a preset representative label set of a final-level detection space for user preference detection, and determining a third target label, comprises the following steps: selecting part of the labels in the representative label set of the final-level detection space associated with the second target label as third labels for user preference detection; determining whether the coverage of the third label meets a preset coverage requirement; if the coverage of the third label does not meet the preset coverage requirement, returning to the step of selecting part of the labels in the representative label set of the final-level detection space associated with the second target label as third labels for user preference detection; if the coverage of the third label meets the preset coverage requirement, taking the third label as the third target label. 2.The method of claim 1, wherein, the step of determining a feature label set based on the first target label, the second target label, and the third target label, comprises the following step: performing feature integration processing on the first target label, the second target label, and the third target label to obtain the feature label set. 3.The method of claim 1, wherein, The method further comprises: determining recommended content based on the feature tag set; pushing the recommended content to a user. 4.The method of claim 1, wherein, The selecting part of the representative tag set of the first-level detection space as the first tag for user preference detection comprises: if it is the first selection, randomly selecting part of the representative tag set of the first-level detection space as the first tag for user preference detection; if it is not the first selection, selecting part of the representative tag set of the first-level detection space as the first tag based on user behavior data.

5. A user preference determination apparatus based on multi-round probing, characterized by, The computer device comprises a memory and a processor, the memory has a computer program stored thereon, and the processor implements the method of any one of claims 1-4 when executing the computer program.

6. A computer device, comprising: The computer device comprises a memory and a processor, the memory has a computer program stored thereon, and the processor implements the method of any one of claims 1-4 when executing the computer program.

7. A computer-readable storage medium, characterized in that, The storage medium has a computer program stored thereon, and the computer program can implement the method of any one of claims 1-4 when executed by a processor.

Images