Information recommendation method and device, electronic equipment and medium

Abstract

The invention discloses an information recommendation method and device, electronic equipment and a medium, and is applied to a recommendation system.The method comprises the steps that input data of a user is obtained, then candidate preference information is determined, competition preference information corresponding to the candidate preference information is searched for in a stable preference set, challenge intensity parameters are determined according to the candidate preference information, and the challenge intensity parameters are selected according to the challenge intensity parameters; determining a stable strength parameter according to the competition preference information, comparing the challenge strength parameter with the stable strength parameter, taking the candidate preference information and / or the competition preference information as stable preference information according to a comparison result, and updating a stable preference set, and determining stable preference information corresponding to the real-time interaction instruction from the updated stable preference set as target preference information, and determining an information recommendation result according to the target preference information and the real-time interaction instruction and recommending the information recommendation result to the user. According to the method, the stable preference can be accurately determined, the recommendation result fits the real intention and long-term demand of the user, and the accuracy of the personalized recommendation service is improved.

Description

Technical Field

[0001] This application belongs to the field of information recommendation, specifically relating to an information recommendation method, apparatus, electronic device, and medium. Background Technology

[0002] With the rapid development of intelligent services, people's demand for personalized services is growing. Personalized services need to provide content that matches users' interests and needs. In this context, accurately grasping user preferences has become a crucial foundation for personalized and intelligent recommendations or services. User preferences not only include their long-term stable interests and habits, but may also involve short-term exploratory behavior or temporary changes in interests. Therefore, effectively distinguishing and handling these different types of preferences has become key to improving the quality of personalized services.

[0003] However, existing technologies often employ a simple preference update overwrite method when processing user preferences. This approach directly overwrites old preferences with new ones when there is a conflict between the user's new and historical preferences. While this can reflect the user's latest interests to some extent, it struggles to distinguish between stable long-term preferences and short-term exploratory behavior. This can lead to updated preferences that fail to meet user needs in terms of stability and accuracy, resulting in recommendations that do not match the user's actual requirements and impacting the accuracy of personalized recommendation services. Summary of the Invention

[0004] The purpose of this application is to provide an information recommendation method, apparatus, electronic device, and medium.

[0005] In a first aspect, embodiments of this application provide an information recommendation method applied to a recommendation system, the recommendation system including a stable set of preferences corresponding to users, the method comprising: Obtain user input data and determine candidate preference information based on the input data; Search for competing preference information that corresponds to the candidate preference information in the stable preference set corresponding to the user; Based on the candidate preference information, the challenge intensity parameter corresponding to the candidate preference information is determined, and based on the competition preference information, the stability intensity parameter corresponding to the competition preference information is determined; The challenge strength parameter and the stability strength parameter are compared to obtain the comparison result. Based on the comparison results, the candidate preference information and / or the competing preference information are used as stable preference information, and the stable preference set is updated using the stable preference information. Obtain the user's real-time interaction command, and determine the stable preference information corresponding to the real-time interaction command from the updated stable preference set as the target preference information; The information recommendation result is determined based on the target preference information and the real-time interaction command, and then recommended to the user.

[0006] Optionally, the recommendation system further includes a temporary preference set and a historical preference set corresponding to the user, and the method further includes: Use any one of the temporary preference set, the stable preference set, and the historical preference set as reference preference information; The number of times the reference preference information is mentioned in the input data is obtained, as well as the emotional intensity parameter and mention time of each mention of the reference preference information in the input data; wherein, the input data is the user's input data from a preset historical time up to the current time; Obtain the time difference between the mentioned time and the current time; The reference confidence level of the reference preference information is determined based on the number of mentions, the emotional intensity parameter, the time difference, and a preset time decay coefficient; wherein the reference confidence level increases with the number of mentions and the emotional intensity parameter, and decays with the time difference.

[0007] Optionally, the temporary preference set includes several temporary preference information items; after the steps of obtaining user input data and determining candidate preference information based on the input data, the method further includes: The candidate preference information is used as the temporary preference information, and the temporary preference set is updated using the temporary preference information; In the set of temporary preferences, obtain several temporary confidence levels corresponding to several temporary preference information, and add the several temporary confidence levels to obtain the total value of temporary confidence. Search the temporary competitive preference information that corresponds to the candidate preference information in the temporary preference set; The provisional confidence level corresponding to the candidate preference information is used as the candidate confidence level; The ratio of the candidate confidence level to the total value of the provisional confidence level is used as the first decay weight; The temporary confidence level corresponding to the temporary competitive preference information is taken as the temporary competitive confidence level; The temporary competitive confidence is attenuated according to the first attenuation weight to obtain the attenuated temporary competitive confidence. Based on the candidate confidence and the decayed temporary competitive confidence, the candidate preference information and / or the temporary competitive preference information are retained as the temporary preference in the temporary preference set.

[0008] Optionally, determining the challenge intensity parameter corresponding to the candidate preference information based on the candidate preference information, and determining the stability intensity parameter corresponding to the competition preference information based on the competition preference information, includes: Within a preset time window, obtain several historical confidence levels corresponding to the candidate preference information and the historical mention times corresponding to each of the several historical confidence levels; The average rate of change of confidence corresponding to the candidate preference within the time window is determined based on several historical confidence levels and the historical mention times corresponding to those historical confidence levels. The average rate of change of confidence level is used as a challenge intensity parameter. Obtain the duration of the competitive preference information's persistence in the stable preference set; Obtain the first mention time of the last mention of the competitive preference information in the input data; Obtain the first time difference between the first mentioned time and the current time; The stability strength parameter corresponding to the competitive preference information is determined based on the duration of existence, the preset baseline stability duration, the first time difference, and the preset stability decay coefficient; wherein the stability strength parameter increases with the duration of existence and decays with the first time difference.

[0009] Optionally, the step of using the candidate preference information and / or the competing preference information as stable preference information based on the comparison result, and updating the stable preference set using the stable preference information, includes: If the challenge intensity parameter is less than the stability intensity parameter, then in the stable preference set, obtain several stable confidence levels corresponding to several stable preference information respectively, and add the several stable confidence levels to obtain the total stable confidence value; The ratio of the candidate confidence level to the total stable confidence level is used as the second decay weight; The stable confidence level corresponding to the competitive preference information is used as the competitive confidence level; The competitive confidence score is attenuated according to the second attenuation weight to obtain the attenuated competitive confidence score; Based on the candidate confidence and the decayed competing confidence, the candidate preference information and / or the competing preference information are retained as stable preference information in the stable preference set.

[0010] Optionally, the step of using the candidate preference information and / or the competing preference information as stable preference information based on the comparison result, and updating the stable preference set using the stable preference information, further includes: If the challenge intensity parameter is greater than or equal to the stability intensity parameter, then a first enhancement weight is determined based on the challenge intensity parameter and the stability intensity parameter; The candidate confidence level is enhanced based on the first enhancement weight to obtain the enhanced candidate confidence level. The third attenuation weight is determined based on the first enhancement weight; The competitive confidence is attenuated according to the second attenuation weight and the third attenuation weight to obtain the attenuated competitive confidence. Based on the enhanced candidate confidence and the weakened competitive confidence, the candidate preference information and / or the competitive preference information are retained as stable preference information in the stable preference set.

[0011] Optionally, the recommendation system further includes an archived preference set corresponding to the user, and the method further includes: If the candidate confidence level is greater than or equal to a preset promotion threshold, search for competing preference information corresponding to the candidate preference information in the stable preference set corresponding to the user. If there is no competing preference information corresponding to the candidate preference information in the stable preference set, the candidate confidence is enhanced according to the preset promotion reward weight, and the candidate preference information is stored as stable preference information in the stable preference set according to the enhanced candidate confidence. In the stable preference set, if the stability confidence level corresponding to any stable preference information is less than a preset degradation threshold, then the stable preference information is stored as historical preference information in the historical preference set. If, in the temporary preference set, the temporary confidence level corresponding to any temporary preference information is less than the downgrade threshold, then the temporary preference information is stored as archived preference information in the archived preference set. In the historical preference set, obtain the duration of persistence of any historical preference information in the stable preference set and the duration of persistence in the historical preference set; If the duration of the historical preference information in the historical preference set is greater than the duration of the historical preference information in the stable preference set, then the historical preference information is stored as archived preference information in the archived preference set.

[0012] Optionally, the method further includes: If the candidate preference information is the same as any of the historical preference information in the historical preference set, then the confidence peak value corresponding to the historical preference information is obtained; Obtain the first sentiment intensity parameter of the last mention of the candidate preference information in the input data; The second enhancement weight is determined based on the confidence peak value, the first emotional intensity parameter, the duration of the historical preference information in the stable preference set, and the duration of the historical preference information in the historical preference set. Obtain the historical confidence level corresponding to the historical preference information; The historical confidence level is enhanced according to the second enhancement weight to obtain the enhanced historical confidence level; Based on the enhanced historical confidence level, the historical preference information is stored in the temporary preference set.

[0013] Optionally, the method further includes: The preset statistical time period prior to the current moment is taken as the current period; Within the current period, first preference flow information is obtained among the temporary preference set, the stable preference set, the historical preference set, and the archived preference set; The weight parameters of the recommendation system are adjusted according to the first preference flow information; wherein the weight parameters include at least one of the following: promotion threshold, demotion threshold, promotion reward weight, time decay coefficient, and baseline stability duration. In the next cycle of the current cycle, obtain the second preference flow information between the temporary preference set, the stable preference set, the historical preference set and the archived preference set, and obtain the feedback from the user on the information recommendation results; If the second preference flow information does not meet the preset conditions, or if the feedback indicates that the information recommendation result does not meet the user's expectations, then the weight parameters of the recommendation system are readjusted according to the second preference flow information.

[0014] Secondly, embodiments of this application provide an information recommendation device applied to a recommendation system, the recommendation system including a stable set of preferences corresponding to a user, the device comprising: The candidate preference acquisition module is used to acquire user input data and determine candidate preference information based on the input data; The competitive preference acquisition module is used to search for competitive preference information corresponding to the candidate preference information in the stable preference set corresponding to the user. The strength parameter acquisition module is used to determine the challenge strength parameter corresponding to the candidate preference information based on the candidate preference information, and to determine the stability strength parameter corresponding to the competition preference information based on the competition preference information. The comparison result acquisition module is used to compare the challenge intensity parameter and the stability intensity parameter to obtain the comparison result. A stable preference update module is used to take the candidate preference information and / or the competing preference information as stable preference information based on the comparison results, and update the stable preference set using the stable preference information; The target preference acquisition module is used to acquire the user's real-time interaction command and determine the stable preference information corresponding to the real-time interaction command from the updated stable preference set as the target preference information; The recommendation result acquisition module is used to determine the information recommendation result based on the target preference information and the real-time interaction command, and recommend it to the user.

[0015] Thirdly, embodiments of this application provide an electronic device, including a processor, a memory, and a program or instructions stored in the memory and capable of running on the processor, wherein the program or instructions, when executed by the processor, implement the method described above.

[0016] Fourthly, a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the method described above.

[0017] The embodiments of this application have the following advantages: In this embodiment, user input data is acquired, candidate preference information is determined based on the input data, competing preference information corresponding to the candidate preference information is searched in the user's corresponding stable preference set, challenge intensity parameters corresponding to the candidate preference information are determined based on the candidate preference information, and stable intensity parameters corresponding to the competing preference information are determined based on the competing preference information. This application, by establishing a stable preference set and introducing a comparison between challenge intensity parameters and stable intensity parameters, can distinguish between users' long-term stable preferences and short-term exploratory behavior. The challenge intensity parameters and stable intensity parameters are compared to obtain a comparison result. Based on the comparison result, candidate preference information and / or competing preference information are used as stable preference information, and the stable preference set is updated using the stable preference information. This application no longer simply performs an overwrite update, but determines stable preferences based on the comparison result of challenge intensity parameters and stable intensity parameters, thereby significantly improving the accuracy of preference judgment. Real-time user interaction commands are acquired, and stable preference information corresponding to the real-time interaction commands is determined from the updated stable preference set as target preference information. Information recommendation results are determined based on the target preference information and the real-time interaction commands and recommended to the user. The information recommendation results in this application are determined based on target preference information in a stable preference set, which better aligns with the user's true intentions and long-term needs, thereby improving the accuracy of personalized recommendation services. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0019] Figure 1 This is a flowchart illustrating the steps of an information recommendation method provided in an embodiment of this application; Figure 2 This is a schematic diagram of a preference management process provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an information recommendation device provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been presented in the various embodiments of this application to enable readers to better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and updates based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0021] With the rapid development of artificial intelligence, people's demand for personalization is growing, and user preference modeling is an important foundation for personalized and intelligent recommendations and services. Existing technologies mainly rely on static user tags and simple preference update coverage, lacking explicit management of the entire preference lifecycle. They also struggle to distinguish between stable long-term preferences, short-term exploratory behavior, and regression of old preferences, failing to adequately balance the stability and flexibility of preferences.

[0022] For example, existing technologies lack the ability to manage the entire lifecycle of preferences, and cannot flexibly handle issues such as users' short-term trial preferences, long-term stable preferences, and regression of old preferences. User profiles are chaotic, and personalized recommendation results are also chaotic, presenting users with recommendation results that are sometimes good and sometimes bad.

[0023] Furthermore, the existing technology's mechanism for dealing with preference conflicts is rather mechanical. When users' new and old interests conflict, the existing technology usually simply "favors the new and rejects the old," replacing the old preferences with the new ones. It lacks a "competitive" or "game-like" decision-making process, which is prone to misjudgment and leads to valuable long-term preferences being easily covered up.

[0024] Secondly, existing technologies lack or are inefficient in their historical preference restart (reflow) mechanisms. When a user develops a renewed liking for an old preference, existing technologies typically start accumulating all preferences again, resulting in a slow response and an inability to "quickly awaken" old preferences.

[0025] Furthermore, most of the key parameters used to manage preferences in recommendation systems rely on human experience and are set in advance, making it impossible to make appropriate adaptive adjustments based on the unique behavioral patterns of different users.

[0026] In existing technologies, parameters are static and lack adaptability. Key parameters used in recommendation systems to manage preferences, such as decay factors and confidence thresholds, are usually globally fixed and cannot be adjusted according to the differences in individual user "variability" (such as some users having specific interests while others have varied interests) and the overall operational health, thus failing to achieve personalized optimal configuration.

[0027] Therefore, this application provides an information recommendation method, apparatus, electronic device, and medium. By establishing a stable preference set and introducing a comparison between a challenge intensity parameter and a stability intensity parameter, it can distinguish between a user's long-term stable preferences and short-term exploratory behavior. This application no longer simply performs overwrite updates, but determines stable preferences based on the comparison results of the challenge intensity parameter and the stability intensity parameter, thereby significantly improving the accuracy of preference judgment. The information recommendation results of this application are determined based on target preference information in the stable preference set, which better aligns with the user's true intentions and long-term needs, and can improve the accuracy of personalized recommendation services.

[0028] Reference Figure 1 The diagram illustrates a flowchart of the steps of an information recommendation method provided in an embodiment of this application.

[0029] In the embodiments of this application, the information recommendation method can be applied to a recommendation system.

[0030] A recommender system can refer to a computer system based on software, hardware, or a combination of both, specifically configured to perform user preference modeling and personalized information services. By acquiring and analyzing user input data, a recommender system dynamically constructs, updates, and manages a structured, multi-state set of preferences, and generates and recommends personalized information results based on the stable set of preferences for the user's real-time interaction commands.

[0031] A recommender system includes a stable set of preferences corresponding to each user. This stable preference set refers to a specific set of data used in the recommender system to store verified preference information that represents a particular user's long-term core interests. The preferences in the stable preference set are a subset of preferences with high stability and confidence, extracted from a large amount of user behavior through screening mechanisms such as confidence assessment and competitive game theory. During recommendation decisions, the recommender system will prioritize and primarily rely on the preference information in the stable preference set to generate the final recommendation results, ensuring the accuracy and stability of the recommendations.

[0032] The method may specifically include the following steps: Step 101: Obtain user input data and determine candidate preference information based on the input data.

[0033] In this embodiment of the application, user input data can be obtained, which is various types of raw information generated by the user and input into the recommendation system. The form of input data may include, but is not limited to, text, voice, click behavior, browsing history, search query, etc.

[0034] Then, the input text information can be extracted from the input data. If the input data is in speech form, it can be converted into text form.

[0035] Input text information can be fed into a preference recognition model for preference identification to obtain candidate preference information. Candidate preference information refers to the user interests and preferences to be processed, identified from the user input data.

[0036] In this context, a preference recognition model refers to a computational model based on artificial intelligence algorithms. Its function is to automatically identify, extract, and structure the preference information expressed by users from their raw input data (such as text, voice, and behavior logs). Preference recognition models can be built based on technologies such as natural language processing and deep learning. They can understand the semantics of user input, determine its sentiment tendency, and extract preference content, preference categories, and sentiment intensity parameters representing user interests. Ultimately, they output structured preference data for subsequent preference management processes.

[0037] In practical implementation, all preference information in the recommender system can be stored as a preference entity. Each preference entity can include the following: Preference coding: A globally unique identifier for each preference entity within the system. Preference coding is used to accurately locate, retrieve, and track the entire lifecycle of a specific preference, ensuring the uniqueness and traceability of the data.

[0038] Preference content: Describes the specific object or theme that the preference refers to. Examples include "Movies: Sci-Fi," "Music: Classical," and "Restaurants: Sichuan Cuisine." Preference content defines the object of a user's interest.

[0039] Preference Categories: High-level abstract groupings that categorize preferred content. Examples include multiple major categories such as "movies," "music," and "food."

[0040] Confidence score: A dynamically changing quantitative value. The confidence score comprehensively represents the current importance, stability, and activity level of a preference to a user. This score is the core decision-making basis driving the flow of preference states. It can determine whether a preference is promoted or downgraded.

[0041] Historical Track: A log sequence or data structure. A historical track is used to record, in chronological order, all transitions of a preference entity across different states of preference sets, changes in confidence scores, and timestamps of key operations.

[0042] Step 102: Search for competing preference information corresponding to the candidate preference information in the stable preference set corresponding to the user.

[0043] In this embodiment, competing preference information corresponding to candidate preference information can be found in the user's stable preference set. Competing preference information refers to existing preferences in the stable preference set that have the same preference category as candidate preference information but different preference content. For example, if the preference category and content of candidate preference information are "Movies: Science Fiction," the preference category and content of competing preference information could be "Movies: Comedy," thus constituting a competitive relationship within the same preference category.

[0044] Step 103: Determine the challenge intensity parameter corresponding to the candidate preference information based on the candidate preference information, and determine the stability intensity parameter corresponding to the competition preference information based on the competition preference information.

[0045] In this embodiment, the challenge intensity parameter corresponding to the candidate preference information can be determined based on the candidate preference information. The challenge intensity parameter is a dynamic indicator used to quantify the recent activity and development trend of the candidate preference information.

[0046] In this embodiment, a stability strength parameter corresponding to the competitive preference information can be determined based on the competitive preference information. The stability strength parameter is a comprehensive index used to quantify the steady-state maintenance capability of competitive preference information. It reflects the robustness of the competitive preference information against new preference shocks.

[0047] Step 104: Compare the challenge strength parameter and the stability strength parameter to obtain the comparison result.

[0048] In this embodiment, the challenge strength parameter and the stability strength parameter can be compared to obtain a comparison result. The comparison result represents the state indicator characterizing the dynamic game outcome relationship between competitive preference information and candidate preference information, that is, between new and old preferences.

[0049] In the embodiments of this application, the comparison result can be generated directly by comparing the numerical values ​​of the challenge intensity parameter and the stability intensity parameter. The comparison result can also include two explicit conclusions: maintaining stability or initiating a challenge.

[0050] Step 105: Based on the comparison results, the candidate preference information and / or the competing preference information are used as stable preference information, and the stable preference information is used to update the stable preference set.

[0051] In this embodiment, candidate preference information and / or competing preference information can be used as stable preference information based on the comparison results, and the stable preference set can be updated using the stable preference information. Stable preference information can refer to preference data that represents a user's long-term interests.

[0052] In practice, the confidence levels of candidate preference information and / or competing preference information can be adjusted based on the specific comparison results, namely, based on the two scenarios of maintaining stability or challenging the initiation. Then, based on the adjusted confidence levels, the candidate preference information and / or competing preference information can be categorized into the stable preference set.

[0053] It is understandable that a stable preference set can simultaneously contain one or more preference information of the same preference type. In other words, as long as the confidence levels corresponding to candidate preference information and / or competing preference information each meet the threshold requirements, both candidate and competing preference information can be retained in the stable preference set. For example, "Movies: Sci-Fi" and "Movies: Comedy" may both be long-term interests of a user.

[0054] Step 106: Obtain the user's real-time interaction command, and determine the stable preference information corresponding to the real-time interaction command from the updated stable preference set as the target preference information.

[0055] In this embodiment, the user's real-time interaction command can be obtained, and the stable preference information corresponding to the real-time interaction command can be determined from the updated stable preference set as the target preference information. The target preference information can refer to the stable preference information that best matches the user's current real-time interaction command, selected in real-time from the updated stable preference set.

[0056] Real-time interactive commands refer to service requests that users send to the recommendation system in real time through interactive methods such as voice, touch, or gestures.

[0057] In its implementation, when the recommendation system receives a user's real-time interaction command, it immediately searches and matches it within the latest set of stable preferences, identifying the stable preference with the highest confidence level that is semantically relevant to the current interaction command as the target preference. This ensures that the recommendation system uses the latest and most stable user interest features that have been dynamically verified each time it provides a recommendation to the user.

[0058] Step 107: Determine the information recommendation result based on the target preference information and the real-time interaction command, and recommend it to the user.

[0059] In this embodiment, information recommendation results can be determined and recommended to the user based on target preference information and real-time interaction commands. The information recommendation results can refer to personalized recommendation content generated by the recommendation system after parsing the real-time interaction commands based on target preferences.

[0060] In practical implementation, the recommendation system can perform multi-dimensional matching of target preference information and real-time interaction commands to filter out candidate options that best match the user's long-term interests and meet immediate needs. After sorting and optimization, a final recommendation list is formed and presented to the user through screens such as the central control screen and entertainment screen. Alternatively, it can directly use candidate options from the recommendation list to make recommendations to the user in the form of voice or other methods.

[0061] In this embodiment, user input data is acquired, candidate preference information is determined based on the input data, competing preference information corresponding to the candidate preference information is searched in the user's corresponding stable preference set, challenge intensity parameters corresponding to the candidate preference information are determined based on the candidate preference information, and stable intensity parameters corresponding to the competing preference information are determined based on the competing preference information. This application, by establishing a stable preference set and introducing a comparison between challenge intensity parameters and stable intensity parameters, can distinguish between users' long-term stable preferences and short-term exploratory behavior. The challenge intensity parameters and stable intensity parameters are compared to obtain a comparison result. Based on the comparison result, candidate preference information and / or competing preference information are used as stable preference information, and the stable preference set is updated using the stable preference information. This application no longer simply performs an overwrite update, but determines stable preferences based on the comparison result of challenge intensity parameters and stable intensity parameters, thereby significantly improving the accuracy of preference judgment. Real-time user interaction commands are acquired, and stable preference information corresponding to the real-time interaction commands is determined from the updated stable preference set as target preference information. Information recommendation results are determined based on the target preference information and the real-time interaction commands and recommended to the user. The information recommendation results in this application are determined based on target preference information in a stable preference set, which better aligns with the user's true intentions and long-term needs, thereby improving the accuracy of personalized recommendation services.

[0062] In one optional embodiment of this application, the recommendation system further includes a temporary preference set and a historical preference set corresponding to the user.

[0063] The temporary preference set is a dynamic storage area used to store newly identified or unverified preferences. All newly added preference information that has not been fully verified can be observed and evaluated in this set. The historical preference set is a dynamic storage area used to store historical preference information that has been downgraded from the stable preference set but has not yet been archived. This historical preference information was once a user's stable preference, but its recent activity level is low.

[0064] The method further includes the following steps: S1001, take any one of the preference information in the temporary preference set, the stable preference set, and the historical preference set as reference preference information; S1002, obtain the number of times the reference preference information is mentioned in the input data, and the emotional intensity parameter and mention time of each mention of the reference preference information in the input data; wherein, the input data is the user's input data from a preset historical time up to the current time; S1003, Obtain the time difference between the mentioned time and the current time; S1004, determine the reference confidence level of the reference preference information based on the number of mentions, the emotional intensity parameter, the time difference, and a preset time decay coefficient; wherein the reference confidence level increases with the number of mentions and the emotional intensity parameter, and decays with the time difference.

[0065] In the embodiments of this application, any one of the preference information from the temporary preference set, the stable preference set, and the historical preference set can be used as the reference preference information. It is understood that the reference preference information can be any one of the temporary preference information, stable preference information, and historical preference information.

[0066] In this embodiment, the number of times reference preference information is mentioned in the input data, as well as the emotional intensity parameter and mention time of each mention of reference preference information in the input data, can be obtained. The input data is user input data from a preset historical time up to the current time.

[0067] The number of mentions can refer to the total number of times that reference preference information is explicitly expressed or implicitly reflected in user input data within a time range from a preset historical time to the current time.

[0068] The sentiment intensity parameter refers to the quantitative value obtained after performing sentiment analysis on user input each time reference preference information is mentioned. It is used to characterize the intensity of a user's subjective emotion when expressing preferences. The sentiment intensity parameter can be identified by the aforementioned preference recognition model.

[0069] The mention time can refer to the precise timestamp on the timeline corresponding to each mention of reference preference information.

[0070] Understandably, since the input data is the user's input data from a preset historical time up to the current time, the current time will be continuously updated as time changes, and therefore the input data will also be continuously updated.

[0071] In this embodiment, the time difference between the mention time and the current time can be obtained, and the reference confidence level of the reference preference information can be determined based on the number of mentions, the sentiment intensity parameter, the time difference, and a preset time decay coefficient. The reference confidence level increases with the number of mentions and the sentiment intensity parameter, and decays with the time difference.

[0072] The time decay coefficient can refer to a preset parameter used to control the rate at which confidence decays over time. The time decay coefficient can determine the magnitude of the decay of confidence over time.

[0073] In practical implementation, the reference confidence level can be determined using the following formula:

[0074] Here, Score represents the reference confidence level of the reference preference information. In other words, in the recommender system, the reference confidence level of any preference information in the temporary preference set, stable preference set, and historical preference set is calculated based on this formula.

[0075] This represents the emotional intensity parameter when the reference preference information is mentioned for the i-th time. The emotional intensity parameter can be set from 0 to 5 points. 0 points corresponds to neutral or no emotional inclination, 1-2 points corresponds to a slight preference, 3-4 points corresponds to a clear preference, and 5 points corresponds to a strong emotional expression.

[0076] This represents the confidence score decay value between the mention time and the current time when the reference preference information is mentioned for the i-th time. Wherein, This represents the time difference between the mention time and the current time when the reference preference information is mentioned for the i-th time. This value can be set according to the actual situation, for example, in days. This represents the time decay coefficient, which can control the decay rate. It can also be set according to the actual situation, such as 0.05, 0.1, etc.

[0077] This represents the frequency accumulation, which is calculated from when i=1. The scores calculated separately for each subsequent mention of preference are then cumulatively added together.

[0078] Because the confidence score calculation formula includes frequency accumulation, and the sentiment intensity parameter when the reference preference information is mentioned for the i-th time is also used as a parameter in the frequency accumulation, when the current reference preference information is mentioned, the sentiment intensity parameter when the reference preference information is mentioned for the i-th time will be added to the original confidence score. This confidence score will become larger and larger, that is, the confidence score will increase with the number of mentions and the sentiment intensity parameter.

[0079] Furthermore, when the current reference preference information is mentioned for the i-th time, the confidence score decay value between the mention time and the current time is... In other words, after the current reference preference information is mentioned for the i-th time, and if it is not mentioned again, time progresses, and the current moment also moves forward. The confidence score at the time of the i-th mention of the current reference preference information decays. The time difference between the mentioned time and the current time The value will increase over time, and consequently the confidence score decay value will decrease, resulting in a smaller and smaller confidence score. In other words, the confidence score will decay over time.

[0080] This application introduces temporary preference sets, stable preference sets, and historical preference sets, and combines mention frequency, sentiment intensity parameters, time difference, and time decay coefficients to dynamically calculate the confidence level of reference preference information. This effectively distinguishes between users' long-term stable preferences and short-term exploratory behavior. Furthermore, the introduction of sentiment intensity parameters more accurately reflects users' subjective emotions, improving the accuracy of preference management. The time decay coefficient ensures the timeliness of preferences, avoiding interference from outdated preferences in recommendation results, thereby improving the stability of personalized recommendations and user satisfaction.

[0081] In one optional embodiment of this application, the temporary preference set includes several temporary preference information.

[0082] After step 101, the method further includes: S1011, the candidate preference information is used as the temporary preference information, and the temporary preference set is updated using the temporary preference information; S1012, in the temporary preference set, obtain several temporary confidence levels corresponding to several temporary preference information respectively, and add the several temporary confidence levels to obtain the total value of temporary confidence; S1013, Search for temporary competing preference information corresponding to the candidate preference information in the temporary preference set; S1014, the temporary confidence level corresponding to the candidate preference information is used as the candidate confidence level; S1015, the ratio of the candidate confidence level to the total value of the temporary confidence level is used as the first decay weight; S1016, the temporary confidence level corresponding to the temporary competitive preference information is taken as the temporary competitive confidence level; S1017, The temporary competitive confidence is attenuated according to the first attenuation weight to obtain the attenuated temporary competitive confidence. S1018, based on the candidate confidence and the decayed temporary competitive confidence, the candidate preference information and / or the temporary competitive preference information are retained as the temporary preference in the temporary preference set.

[0083] In this embodiment of the application, candidate preference information can be used as temporary preference information, and the temporary preference set can be updated using the temporary preference information. In other words, candidate preference information can first enter the temporary preference set.

[0084] In this embodiment, several temporary confidence levels corresponding to several temporary preference information can be obtained from the temporary preference set, and the temporary confidence levels are added together to obtain a total temporary confidence value. Then, temporary competing preference information corresponding to the candidate preference information can be searched in the temporary preference set. Here, the temporary confidence level is the confidence level corresponding to the temporary preference information.

[0085] Temporary competing preference information refers to existing preferences in the temporary preference set that have the same preference category as the candidate preference information but different preference content. For example, if the preference category and preference content of the candidate preference information are "movies: science fiction", the preference category and preference content of the temporary competing preference information could be "movies: comedy", and the two constitute a competitive relationship under the same preference category.

[0086] In this embodiment, the temporary confidence level corresponding to the candidate preference information can be used as the candidate confidence level, and the ratio of the candidate confidence level to the total value of the temporary confidence level can be used as the first decay weight.

[0087] The temporary confidence level corresponding to the temporary competitive preference information is used as the temporary competitive confidence level. The temporary competitive confidence level is then attenuated according to the first attenuation weight to obtain the attenuated temporary competitive confidence level.

[0088] Then, based on the candidate confidence and the decayed temporary competitive confidence, the candidate preference information and / or temporary competitive preference information can be retained as temporary preferences in the temporary preference set.

[0089] It is understandable that a temporary preference set can contain one or more preference information of the same preference type simultaneously. In other words, as long as the confidence levels corresponding to the candidate preference information and / or the temporary competing preference information each meet the threshold requirements, both the candidate preference information and the temporary competing preference information can be retained in the stable preference set. For example, "Movie: Sci-Fi" and "Movie: Comedy" may both be short-term interests of a user.

[0090] In practice, if temporary competing preference information corresponding to candidate preference information is found in the temporary preference set, competition within the pool can be carried out, which is competition between preferences in the temporary preference set.

[0091] The decayed temporary competitive confidence level can be determined using the following formula: Score_B1 = Score_old_B1 * (1 - w1) Where Score_B1 represents the decayed temporary competitive confidence, Score_old_B1 represents the temporary competitive confidence, and w1 represents the first decay weight.

[0092] In practical implementation, the first attenuation weight can be determined using the following formula: w1 = Score_A / Σ(Score1_all) Where Score_A represents the candidate confidence level, and Σ(Score1_all) represents the total temporary confidence level, which is the sum of the temporary confidence levels corresponding to all temporary preference information in the temporary preference set.

[0093] The setting of the first decay weight indicates that the higher the confidence level of the new preference, i.e., the candidate preference information, the greater the impact on the old preference, i.e., the competing preference information. The setting of the temporary confidence total value indicates that the more preferences there are in the temporary preference set, the more easily the user's preferences are changed or the user expects to explore new preferences. In this case, the first decay weight should be reduced to reduce the impact of the new preference, i.e., the candidate preference information, on the old preference, i.e., the competing preference information.

[0094] This application effectively manages new user preference information by introducing a temporary preference set and a temporary confidence level mechanism, and dynamically adjusts the confidence level of temporary preference information through pool competition. The competitive relationship between candidate preference information and temporary competing preference information is quantified through a first decay weight, ensuring that the impact of new preferences on old preferences is proportional to their confidence level, while also considering the overall size of the temporary preference set to avoid excessive impact. This enhances the flexibility of preference management, better balances users' different short-term interests, and thus improves the accuracy of personalized recommendations and user satisfaction.

[0095] In one optional embodiment of this application, step 103 includes the following steps: S1021, within a preset time window, obtain several historical confidence levels corresponding to the candidate preference information and the historical mention times corresponding to the several historical confidence levels respectively; S1022, determine the average rate of change of confidence corresponding to the candidate preference within the time window based on the aforementioned historical confidence levels and the historical mention times corresponding to the aforementioned historical confidence levels respectively; S1023, the average rate of change of confidence level is used as the challenge intensity parameter; S1024, Obtain the duration of the competitive preference information in the stable preference set; S1025, Obtain the first mention time of the last mention of the competitive preference information in the input data; S1026, Obtain the first time difference between the first mentioned time and the current time; S1027, determine the stability strength parameter corresponding to the competition preference information based on the duration of existence, the preset baseline stability duration, the first time difference, and the preset stability decay coefficient; wherein, the stability strength parameter increases with the duration of existence and decays with the first time difference.

[0096] In this embodiment of the application, within a preset time window, several historical confidence levels corresponding to candidate preference information and the historical mention times corresponding to the several historical confidence levels can be obtained.

[0097] Historical confidence level refers to the confidence level of candidate preference information at historical mention times, where historical mention times are the time points when the input data corresponding to each historical confidence level was collected.

[0098] In this embodiment of the application, the average rate of change of confidence corresponding to candidate preferences within a time window can be determined based on several historical confidence levels and the historical mention times corresponding to several historical confidence levels, and the average rate of change of confidence can be used as a challenge intensity parameter.

[0099] In practical implementation, the challenge intensity parameter can be determined using the following formula: Trend_A=mean( ) Where Trend_A represents the challenge intensity parameter corresponding to the candidate preference information. This represents the historical confidence score calculated when candidate preference information is mentioned for the i-th time. This represents the historical confidence level calculated when candidate preference information is mentioned for the (i-1)th time. This represents the historical moment when candidate preference information is mentioned for the i-th time. This represents the historical mention time when candidate preference information is mentioned for the (i-1)th time. `mean()` represents the average of all rates of change within a certain time window. The time window can be set according to the actual situation, such as one week, half a month, etc.

[0100] This application introduces a challenge intensity parameter to dynamically assess the confidence trend of candidate preference information within a preset time window. The challenge intensity parameter calculates the average rate of change of confidence based on historical confidence levels and historical mention times, thereby quantifying the activity and stability of candidate preference information. This effectively identifies dynamic changes in user preferences, avoiding misjudging highly volatile short-term interests as long-term stable preferences, and provides accurate data for subsequent preference competition and updates, thus improving the accuracy and flexibility of personalized recommendations.

[0101] In this embodiment of the application, the duration of the competitive preference information in the stable preference set can be obtained, the first mention time of the last mention of the competitive preference information in the input data can be obtained, and the first time difference between the first mention time and the current time can be obtained.

[0102] In this embodiment, the stability strength parameter corresponding to the competition preference information can be determined based on the duration of existence, a preset baseline stability duration, a first time difference, and a preset stability decay coefficient. The stability strength parameter increases with the duration of existence and decays with the first time difference.

[0103] In practical implementation, the stability strength parameter can be determined using the following formula: Stability_B=

[0104] Where Stability_B represents the stability strength parameter corresponding to the competitive preference information, and T represents the duration of the competitive preference information in the stable preference set. Indicates the baseline stability time. This represents the first time difference, which is the time difference between the first mention of the last mention of competitive preference information in the input data and the current time. This represents the stable attenuation coefficient.

[0105] As time progresses, the current moment also moves forward. The duration T of the competitive preference information in the stable preference set will become larger and larger, and consequently, the stability strength parameter will become larger and larger. In other words, the stability strength parameter will increase with the duration of existence.

[0106] If the current reference preference information is not mentioned, time progresses, and the current moment also moves forward. The first time difference between the first mention moment of the last mention of competing preference information in the input data and the current moment is... It will become larger and larger, and then the stability strength parameter will become smaller and smaller, that is, the stability strength parameter decays with the first time difference.

[0107] This application introduces a stability strength parameter, which comprehensively considers the duration of competitive preference information, the last mention time, and the time difference to dynamically assess its stability and activity. The stability strength parameter increases with the duration of existence, indicating that the longer a preference exists in the stable preference set, the more stable it is. Simultaneously, it decays with the time difference, ensuring the timeliness of the preference. This effectively distinguishes between long-term stable preferences and short-term interests, avoiding misjudging the stability of a preference due to its long period of inactivity. Therefore, it provides a more accurate basis for preference competition and updates, improving the accuracy and flexibility of personalized recommendations.

[0108] In one optional embodiment of this application, step 104 includes the following steps: S1041, compare the challenge strength parameter and the stability strength parameter; S1042, if the challenge intensity parameter is less than the stability intensity parameter, then maintaining stability will be taken as the comparison result; S1043, if the challenge intensity parameter is greater than or equal to the stability intensity parameter, then the challenge initiation is taken as the comparison result.

[0109] In the embodiments of this application, the challenge strength parameter and the stability strength parameter can be compared.

[0110] In the specific implementation, if the challenge strength parameter Trend_A is less than the stability strength parameter Stability_B, then maintaining stability can be used as the comparison result. Here, maintaining stability can refer to the game result generated when the challenge strength parameter is less than the stability strength parameter, indicating that the current impact of the candidate preference information is not enough to shake the solid position of the existing competitive preference information, and the stable pattern with the original competitive preference information as the core will be maintained. However, the candidate preference information still has an impact on the competitive preference information, so only the competitive confidence of the competitive preference information is attenuated.

[0111] In the specific implementation, if the challenge strength parameter Trend_A is greater than or equal to the stability strength parameter Stability_B, the challenge initiation can be used as the comparison result. Here, challenge initiation can refer to the game result generated when the challenge strength parameter is greater than or equal to the stability strength parameter, indicating that the candidate preference information has sufficient impact to launch an effective challenge to the existing competitive preference information. At this time, the candidate confidence of the candidate preference information will be enhanced, and the competitive confidence of the competitive preference information will be weakened.

[0112] This application introduces a comparison mechanism between challenge strength parameters and stability strength parameters to dynamically assess the impact of candidate preference information on competing preference information. When the challenge strength parameter is less than the stability strength parameter, the existing competitive preference information remains stable, with only attenuation processing applied. When the challenge strength parameter is greater than or equal to the stability strength parameter, candidate preference information is allowed to initiate a challenge, increasing its confidence and attenuating the competing preference information. This effectively balances the competitive relationship between new and old preferences, avoids preference loss due to simple overwriting, and ensures timely response to changes in user interests, improving the flexibility and accuracy of personalized recommendations.

[0113] In one optional embodiment of this application, step 105 includes the following steps: S1051, if the challenge intensity parameter is less than the stability intensity parameter, then in the stable preference set, obtain several stable confidence levels corresponding to several stable preference information respectively, and add the several stable confidence levels to obtain the total stable confidence value; S1052, the ratio of the candidate confidence level to the total value of the stable confidence level is used as the second decay weight; S1053, the stable confidence level corresponding to the competitive preference information is taken as the competitive confidence level; S1054, The competitive confidence is attenuated according to the second attenuation weight to obtain the attenuated competitive confidence; S1055, based on the candidate confidence and the decayed competitive confidence, the candidate preference information and / or the competitive preference information are retained as the stable preference information in the stable preference set.

[0114] In this embodiment, the challenge strength parameter is less than the stability strength parameter, meaning the comparison result is stability. In this case, several stable confidence levels can be obtained from the stable preference set, corresponding to different stable preference information, and these stable confidence levels are summed to obtain the total stable confidence value. Here, stable confidence level refers to the current confidence value of each preference information in the stable preference set. The total stable confidence value can be a summary value obtained by accumulating the confidence levels of all stable preference information within the stable preference set.

[0115] In this embodiment, the ratio of candidate confidence to total stable confidence is used as the second attenuation weight, and the stable confidence corresponding to the competitive preference information is used as the competitive confidence. The competitive confidence is attenuated according to the second attenuation weight to obtain the attenuated competitive confidence.

[0116] Candidate preference information and / or competing preference information can be retained as stable preference information in the stable preference set based on the candidate confidence and the decayed competing confidence.

[0117] In practical implementation, assuming the comparison result remains stable, the decayed competition confidence can be obtained using the following formula: Score_B2 = Score_old_B2 * (1 - w2) Wherein, Score_B2 represents the decayed competitive confidence score when the comparison result remains stable, Score_old_B2 represents the competitive confidence score, and w2 represents the second decay weight.

[0118] In practical implementation, the second attenuation weight can be determined using the following formula: w2 = Score_A / Σ(Score2_all) Where Score_A represents the candidate confidence level, and Σ(Score2_all) represents the total stable confidence level, which is the sum of the stable confidence levels corresponding to all stable preference information in the stable preference set.

[0119] This application, while maintaining a stable comparison result, dynamically adjusts the confidence level of competing preference information by calculating the ratio of candidate confidence level to the total stable confidence level as a second decay weight. This effectively balances the competitive relationship between candidate and competing preference information, preventing the competing preference information from being excessively weakened due to the impact of candidate preference information, while ensuring a more reasonable confidence distribution of the stable preference set, thereby improving the stability and accuracy of personalized recommendations.

[0120] In one optional embodiment of this application, step 105 further includes the following steps: S1061, if the challenge intensity parameter is greater than or equal to the stability intensity parameter, then a first enhancement weight is determined based on the challenge intensity parameter and the stability intensity parameter; S1062, The candidate confidence is enhanced according to the first enhancement weight to obtain the enhanced candidate confidence; S1063, determine the third attenuation weight based on the first enhancement weight; S1064, The competitive confidence is attenuated according to the second attenuation weight and the third attenuation weight to obtain the attenuated competitive confidence. S1065, based on the enhanced candidate confidence and the weakened competitive confidence, the candidate preference information and / or the competitive preference information are retained as the stable preference information in the stable preference set.

[0121] In this embodiment, the challenge strength parameter is greater than or equal to the stability strength parameter; that is, the comparison result indicates a challenge is initiated. At this point, a first enhancement weight can be determined based on the challenge strength parameter and the stability strength parameter. The candidate confidence level is then enhanced based on the first enhancement weight to obtain the enhanced candidate confidence level. In a specific implementation, the ratio of the challenge strength parameter to the stability strength parameter can be used as the first enhancement weight.

[0122] In practical implementation, when the comparison result indicates a challenge has been initiated, the enhanced candidate confidence can be obtained using the following formula: Score_A1= *Score_A Wherein, Score_A1 represents the enhanced candidate confidence when the comparison result is a challenge initiated, Score_A represents the candidate confidence, Trend_A represents the challenge strength parameter corresponding to the candidate preference information, and Stability_B represents the stability strength parameter corresponding to the competition preference information. This indicates the first enhanced weight.

[0123] In this embodiment, the third attenuation weight can be determined based on the first enhancement weight. In a specific implementation, the reciprocal of the third attenuation weight can be used as the third attenuation weight. That is, the ratio of the stability strength parameter to the challenge strength parameter is used as the third attenuation weight.

[0124] In this embodiment of the application, the competitive confidence can be attenuated according to the second attenuation weight and the third attenuation weight to obtain the attenuated competitive confidence.

[0125] In practical implementation, when the comparison result is a challenge initiated, the decayed competition confidence can be obtained using the following formula: Score_B3= *Score_old_B2*(1-w2) Wherein, Score_B3 represents the decayed competition confidence when the comparison result is a challenge initiated, Score_old_B2 represents the competition confidence, w2 represents the second decay weight, Trend_A represents the challenge strength parameter corresponding to the candidate preference information, and Stability_B represents the stability strength parameter corresponding to the competition preference information. This represents the third decay weight.

[0126] In this embodiment of the application, candidate preference information and / or competitive preference information can be retained as stable preference information in the stable preference set based on the enhanced candidate confidence and the decayed competitive confidence.

[0127] This application, when the comparison result indicates a challenge has been initiated, dynamically adjusts the confidence levels of candidate and competing preference information by introducing a first enhancement weight and a third decay weight. The enhanced candidate confidence is amplified by the ratio of the challenge strength parameter to the stable strength parameter, ensuring that new preferences can effectively challenge old preferences. Simultaneously, the decayed competing confidence is reduced using a second and third decay weight, preventing old preferences from being excessively weakened. This approach scientifically balances the competitive relationship between new and old preferences, ensuring more reasonable updates to the stable preference set, thereby improving the flexibility and accuracy of personalized recommendations.

[0128] In one optional embodiment of this application, the method further includes the following steps: S1071, if the candidate confidence level is greater than or equal to a preset promotion threshold, search for competitive preference information corresponding to the candidate preference information in the stable preference set corresponding to the user; S1072, if there is no competing preference information corresponding to the candidate preference information in the stable preference set, then the candidate confidence is enhanced according to the preset promotion reward weight, and the candidate preference information is stored as stable preference information in the stable preference set according to the enhanced candidate confidence.

[0129] In this embodiment, when the candidate confidence level is greater than or equal to a preset promotion threshold, competing preference information corresponding to the candidate preference information can be searched in the user's corresponding stable preference set. In other words, the step of searching for competing preference information corresponding to the candidate preference information in the user's corresponding stable preference set is triggered when the candidate confidence level is greater than or equal to the preset promotion threshold.

[0130] In this embodiment of the application, if there is no competing preference information corresponding to the candidate preference information in the stable preference set, the candidate confidence is enhanced according to the preset promotion reward weight, and the candidate preference information is stored as stable preference information in the stable preference set according to the enhanced candidate confidence.

[0131] In the specific implementation, the step of searching for competing preference information corresponding to candidate preference information in the user's stable preference set is triggered when the candidate confidence is greater than or equal to the preset promotion threshold. That is to say, the score of the candidate preference information has exceeded the preset promotion threshold. If there is no competing preference information corresponding to the candidate preference information in the stable preference set, the candidate preference information can be used as stable preference information and directly stored in the stable preference set.

[0132] In practical implementation, to protect the stability of the stable preference set and prevent newly promoted stable preference information from unexpectedly degrading, a promotion reward weight can be set to enhance the newly promoted stable preference information. The candidate confidence score enhanced by the preset promotion reward weight can be obtained using the following formula: Score_A2=α*Score_A Where Score_A2 represents the candidate confidence score enhanced by the preset promotion reward weight, and Score_A represents the candidate confidence score. This indicates the weight of promotion rewards.

[0133] Furthermore, the time decay coefficient in the calculation formula can be modified when calculating the stability confidence of stable preference information in a stable preference set. This allows for control over the attenuation rate, and can also be set according to actual conditions, such as adjusting the time attenuation coefficient. The values ​​were changed from 0.05 and 0.1 to smaller values ​​such as 0.02 and 0.01.

[0134] In addition, a protection period can be granted. For example, for a specific period of time, newly promoted stable preference information is not subject to decay or downgrading. After the specific period, the time difference between the mention time and the current time will begin to accumulate. The confidence level is attenuated.

[0135] This application, by setting a promotion threshold and promotion reward weight, can dynamically determine whether to promote candidate preference information to stable preference information when the confidence level of candidate preference information reaches or exceeds a preset threshold. If there is no corresponding competing preference information in the stable preference set, the confidence level of the candidate is directly enhanced according to the promotion reward weight, and it is stored in the stable preference set. This can promptly identify and introduce new preferences with high confidence, avoid the neglect of high-quality candidate preferences due to preference competition mechanisms, and ensure the diversity and timeliness of the stable preference set, thereby improving the accuracy and flexibility of personalized recommendations.

[0136] In one optional embodiment of this application, the recommendation system further includes an archive preference set corresponding to the user.

[0137] Archived preference sets can be long-term storage areas used to store completely deactivated preference data. Archived preference sets can store the complete trajectory of archived preference information. Within the archived preference set, last-place items can also be evicted based on factors such as the number of archived preference information entries, the time of the archived preference information, and the trajectory length (information richness) of the archived preference information.

[0138] The method further includes the following steps: S1073, in the stable preference set, if the stability confidence level corresponding to any stable preference information is less than a preset downgrade threshold, then the stable preference information is stored as historical preference information in the historical preference set. S1074, in the temporary preference set, if the temporary confidence level corresponding to any of the temporary preference information is less than the downgrade threshold, then the temporary preference information is stored as archived preference information in the archived preference set.

[0139] In this embodiment, if the stability confidence level corresponding to any stable preference information in the stable preference set is less than a preset downgrade threshold, the stable preference information is stored as historical preference information in the historical preference set. The downgrade threshold can refer to a confidence threshold; when the confidence level of stable preference information is lower than the downgrade threshold, it indicates that the stable preference information no longer qualifies as a user's core interest.

[0140] In this embodiment, if the temporary confidence level of any temporary preference information in the temporary preference set is less than the downgrade threshold, the temporary preference information is stored as archived preference information in the archived preference set. When the confidence level of temporary preference information is lower than the downgrade threshold, it indicates that the temporary preference information no longer qualifies as a user's temporary interest. The downgrade threshold can also be set to different values ​​in different preference sets.

[0141] This application introduces a downgrade threshold to dynamically monitor the confidence level of preference information in both stable and temporary preference sets, enabling timely identification and processing of low-confidence preference information. If the confidence level of stable preference information falls below the downgrade threshold, it is downgraded to historical preference information and stored in the historical preference set, ensuring the conciseness and timeliness of the stable preference set. If the confidence level of temporary preference information falls below the downgrade threshold, it is downgraded to archived preference information, avoiding redundancy in the temporary preference set. This effectively cleans up inefficient preferences and optimizes the structure of the preference set.

[0142] In one optional embodiment of this application, the method further includes the following steps: S1075, in the historical preference set, obtain the duration of continuous existence of any historical preference information in the stable preference set and the duration of continuous existence in the historical preference set; S1076, if the duration of the historical preference information in the historical preference set is longer than the duration of the historical preference information in the stable preference set, then the historical preference information is stored as archived preference information in the archived preference set.

[0143] In this embodiment of the application, in the historical preference set, the duration of any historical preference information in the stable preference set and the duration of historical preference information in the historical preference set can be obtained. If the duration of historical preference information in the historical preference set is longer than the duration of historical preference information in the stable preference set, then the historical preference information is stored as archived preference information in the archived preference set.

[0144] In practice, when the retention time of historical preference information in the historical set exceeds its longest retention time in the stable set, or when the confidence level of historical preference information continuously decays to the archiving threshold, the historical preference information will be moved to the archived preference set. Archived preferences no longer participate in daily recommendation calculations, but their complete historical trajectory data is still retained for user interest evolution analysis or model optimization. The archiving threshold can be set according to actual needs.

[0145] This application's embodiments introduce an archived preference set, enabling further classification and storage of historical preference information, ensuring the long-term preservation of data on completely inactive preferences. By comparing the duration of historical preference information's persistence in the stable preference set and the historical preference set, it is possible to dynamically determine whether it should be archived. Archived preference information no longer participates in recommendation calculations, but its complete trajectory data is still retained, providing support for subsequent user interest evolution analysis and model optimization. This effectively cleans up inefficient data, optimizes system resource allocation, provides a data foundation for preference evolution research, and enhances the intelligence level of the recommendation system.

[0146] In one optional embodiment of this application, the method further includes the following steps: S1081, if the candidate preference information is the same as any of the historical preference information in the historical preference set, then obtain the confidence peak value corresponding to the historical preference information; S1082, Obtain the first sentiment intensity parameter of the last mention of the candidate preference information in the input data; S1083, determine the second enhancement weight based on the confidence peak value, the first emotional intensity parameter, the duration of the historical preference information in the stable preference set, and the duration of the historical preference information in the historical preference set; S1084, Obtain the historical confidence level corresponding to the historical preference information; S1085, The historical confidence level is enhanced according to the second enhancement weight to obtain the enhanced historical confidence level; S1086, Based on the enhanced historical confidence level, store the historical preference information into the temporary preference set.

[0147] In this embodiment, if the candidate preference information is the same as any historical preference information in the historical preference set, then the peak confidence level corresponding to the historical preference information is obtained. The peak confidence level refers to the maximum confidence level of the historical preference information in the stable preference set before it falls into the historical preference set.

[0148] In this embodiment of the application, a first sentiment intensity parameter can be obtained from the last mention of candidate preference information in the input data, and a second enhancement weight can be determined based on the confidence peak, the first sentiment intensity parameter, the duration of historical preference information in the stable preference set, and the duration of historical preference information in the historical preference set.

[0149] The second enhancement weight can be obtained using the following formula:

[0150] in, This represents the second enhancement weight, I_current represents the first sentiment intensity parameter, and Score_peak represents the peak confidence level corresponding to historical preference information. This indicates the duration of historical preference information within a stable preference set. This indicates the duration for which historical preference information persists in the historical preference set. `max()` represents the maximum value, and `min()` represents the minimum value. In other words... The value ranges from 1.2 to 1.5.

[0151] In this embodiment of the application, the historical confidence level corresponding to the historical preference information can be obtained, and the historical confidence level can be enhanced according to the second enhancement weight to obtain the enhanced historical confidence level. Based on the enhanced historical confidence level, the historical preference information can be stored in the temporary preference set.

[0152] In practical implementation, the enhanced historical confidence level can be obtained using the following formula: Score_A3=Score_A*γ Here, Score_A3 represents the enhanced historical confidence level, and Score_A represents the candidate confidence level, which is the candidate confidence level of the candidate preference information mentioned this time. This indicates the second enhanced weight.

[0153] In practical implementation, this process can represent the repatriation of historical preference information, much like an experienced employee re-entering a "probationary period," but with a higher starting point. A fast track can also be set up simultaneously. For historical preference information repatriated to the temporary preference set, the promotion threshold for historical preference information can be set lower than for entirely new preference information, because historical preference information was previously either temporary or stable preference information for the user.

[0154] This application's embodiments, by introducing a second enhancement weight, can dynamically assess the potential value of historical preference information and enhance it based on its peak confidence level, sentiment intensity parameter, and duration of persistence in different sets. The enhanced historical confidence level is used to re-store the historical preference information into a temporary preference set, giving it the opportunity to participate in recommendation calculations again. This effectively mines the potential value of historical preference information, preventing it from being completely forgotten due to preference downgrades, while improving the flexibility of preference updates and the accuracy of personalized recommendations, providing users with recommended content that better aligns with their long-term interests.

[0155] In one optional embodiment of this application, the method further includes the following steps: S1091, take the preset statistical time period before the current time as the current period; S1092, within the current period, obtain the first preference flow information among the temporary preference set, the stable preference set, the historical preference set, and the archived preference set; S1093, adjust the weight parameters of the recommendation system according to the first preference flow information; wherein, the weight parameters include at least one of promotion threshold, demotion threshold, promotion reward weight, time decay coefficient, and baseline stability duration; S1094, in the next cycle of the current cycle, obtain the second preference flow information between the temporary preference set, the stable preference set, the historical preference set and the archived preference set, and obtain the feedback from the user on the information recommendation results; S1095, if the second preference flow information does not meet the preset conditions, or if the feedback indicates that the information recommendation result does not meet the user's expectations, then the weight parameters of the recommendation system are readjusted according to the second preference flow information.

[0156] In this embodiment of the application, a preset statistical time period prior to the current time can be used as the current period; Within the current period, obtain the first preference flow information among the temporary preference set, stable preference set, historical preference set, and archived preference set.

[0157] Among them, the first preference flow information can refer to the quantitative statistical data formed by monitoring the state migration of preference entities between temporary sets, stable sets, historical sets and archived sets within the current period.

[0158] The first preference flow information can record the following indicators: The number and proportion of preferences that flow from the temporary preference set to the stable preference set.

[0159] The number and proportion of preferences that flow from the temporary or stable preference set to the historical preference set. The number and proportion of preferences that flow from the historical preference set to the temporary preference set.

[0160] The number and proportion of preferences that flow from the historical preference set to the archived preference set.

[0161] The weight parameters of the recommendation system are adjusted based on the first preference flow information; wherein, the weight parameters include at least one of the following: promotion threshold, demotion threshold, promotion reward weight, time decay coefficient, and baseline stability duration.

[0162] In practical implementation, for example, when the number and proportion of preferences flowing from the historical preference set to the archived preference set are large, it indicates that the stable preference information is more likely to be erroneously downgraded. In this case, the protection of stable preference information should be strengthened, and the difficulty of inter-pool games should be increased. This can be achieved by adding a penalty coefficient to the challenge intensity parameter, adding a protection coefficient to the stability intensity parameter, or lowering the downgrade threshold corresponding to the stable preference information.

[0163] In this embodiment, within the next cycle of the current cycle, second preference flow information is obtained among the temporary preference set, stable preference set, historical preference set, and archived preference set, and user feedback on the information recommendation results is also obtained. The data type recorded in the second preference flow information is the same as that in the first preference flow information; however, the data content recorded in the second preference flow information is collected within the next cycle of the current cycle. Feedback may include score feedback directly obtained from user-initiated ratings, and may also include evaluation data on the recommendation results indirectly obtained from user interaction behaviors such as dwell time.

[0164] In this embodiment of the application, if the second preference flow information does not meet the preset conditions, or if the feedback indication information recommendation result does not meet the user's expectations, the weight parameters of the recommendation system can be readjusted according to the second preference flow information.

[0165] In practice, when the second preference flow index is detected to be outside the normal fluctuation range of the data, or when user feedback data indicates that the recommendation effect is not as expected, the abnormal changes in the preference migration rate between each preference set can be analyzed, and key parameters such as the promotion threshold and time decay coefficient can be recalibrated in combination with user feedback data.

[0166] This application dynamically monitors preference flow information and user feedback, enabling real-time adjustment of the recommendation system's weight parameters to ensure continuous optimization of recommendation performance. By analyzing the flow data between temporary preference sets, stable preference sets, historical preference sets, and archived preference sets, it can identify abnormal patterns in preference migration and calibrate key parameters such as promotion and demotion thresholds based on user feedback. This allows for flexible responses to changes in user interests and uncertainties in system operation, improving the recommendation system's adaptability and the accuracy of recommendation results, and providing users with a more personalized recommendation experience that better meets their current needs.

[0167] In the specific implementation, refer to Figure 2 The diagram shows a schematic representation of a preference management process provided in an embodiment of this application.

[0168] The recommender system includes a temporary preference set, a stable preference set, a historical preference set, and an archived preference set.

[0169] It can obtain user input data and determine candidate preference information based on the input data.

[0170] If the candidate preference information already exists in the temporary preference set, stable preference set, or historical preference set, then the candidate confidence of the candidate preference information can be updated.

[0171] If the candidate preference information is not found in the temporary preference set, stable preference set, or historical preference set, then the candidate preference information can be placed into the temporary preference set first.

[0172] Then, the temporary competing preference information corresponding to the candidate preference information is searched in the temporary preference set. If no temporary competing preference information exists, the candidate preference information will be directly retained in the temporary preference set. If temporary competing preference information exists, pool competition can be performed according to the steps S1012-S1018 of the embodiments of this application.

[0173] If, during the retention of candidate preference information in the temporary preference set, the confidence level of the candidate preference information exceeds the promotion threshold, then the competing preference information corresponding to the candidate preference information is searched in the user's corresponding stable preference set. If no competing preference information exists, the candidate preference information is directly retained in the stable preference set. If competing preference information exists, then inter-pool game can be performed according to the steps S103-S105 of the embodiments of this application.

[0174] In the stable preference set, if the stability confidence level corresponding to any stable preference information is less than the preset downgrade threshold, then the stable preference information is stored as historical preference information in the historical preference set. If the temporary confidence level of any temporary preference information in the temporary preference set is less than the downgrade threshold, then the temporary preference information will be stored as archived preference information in the archived preference set.

[0175] In the historical preference set, obtain the duration of any historical preference information in the stable preference set and the duration of any historical preference information in the historical preference set. If the duration of historical preference information in the historical preference set is greater than the duration of historical preference information in the stable preference set, then the historical preference information is stored as archived preference information in the archived preference set.

[0176] If the candidate preference information is the same as any historical preference information in the historical preference set, the historical preference information can be returned to the temporary preference set according to the steps of S1082-S1086 in the embodiments of this application.

[0177] Meanwhile, the recommendation system includes an adaptive agent, which can refer to a software entity with autonomous decision-making and learning capabilities, capable of autonomously adjusting its operating parameters. Following the steps S1091-S1095 of the embodiments of this application, the adaptive agent can observe the temporary preference set, stable preference set, historical preference set, and archived preference set, receive user feedback data, and adjust the weight parameters of the recommendation system.

[0178] Reference Figure 3 The diagram illustrates a structural schematic of an information recommendation device according to an embodiment of this application, applied to a recommendation system. The device includes: The candidate preference acquisition module 301 is used to acquire user input data and determine candidate preference information based on the input data. The competitive preference acquisition module 302 is used to search for competitive preference information corresponding to the candidate preference information in the stable preference set corresponding to the user. The intensity parameter acquisition module 303 is used to determine the challenge intensity parameter corresponding to the candidate preference information based on the candidate preference information, and to determine the stability intensity parameter corresponding to the competition preference information based on the competition preference information. The comparison result acquisition module 304 is used to compare the challenge intensity parameter and the stability intensity parameter to obtain a comparison result. The stable preference update module 305 is used to take the candidate preference information and / or the competing preference information as stable preference information according to the comparison result, and update the stable preference set using the stable preference information; The target preference acquisition module 306 is used to acquire the user's real-time interaction command and determine the stable preference information corresponding to the real-time interaction command from the updated stable preference set as the target preference information; The recommendation result acquisition module 307 is used to determine the information recommendation result based on the target preference information and the real-time interaction command and recommend it to the user.

[0179] In one optional embodiment of this application, the recommendation system further includes a temporary preference set and a historical preference set corresponding to the user, and the apparatus further includes: The first confidence calculation module is used to use any one of the preference information from the temporary preference set, the stable preference set, and the historical preference set as reference preference information; The second confidence calculation module is used to obtain the number of times the reference preference information is mentioned in the input data, as well as the emotional intensity parameter and mention time of each mention of the reference preference information in the input data; wherein, the input data is the user's input data from a preset historical time up to the current time; The third confidence calculation module is used to obtain the time difference between the mentioned time and the current time. The fourth confidence calculation module is used to determine the reference confidence of the reference preference information based on the number of mentions, the emotional intensity parameter, the time difference, and a preset time decay coefficient; wherein the reference confidence increases with the number of mentions and the emotional intensity parameter, and decays with the time difference.

[0180] In one optional embodiment of this application, the temporary preference set includes several temporary preference information items; after the steps of obtaining user input data and determining candidate preference information based on the input data, the apparatus further includes: The first internal competition module is used to use the candidate preference information as the temporary preference information and update the temporary preference set using the temporary preference information; The second internal competition module is used to obtain several temporary confidence levels corresponding to several temporary preference information in the temporary preference set, and add the several temporary confidence levels to obtain the total value of temporary confidence. The third internal competition module is used to search for temporary competitive preference information corresponding to the candidate preference information in the temporary preference set; The fourth internal competition module is used to take the temporary confidence level corresponding to the candidate preference information as the candidate confidence level; The fifth internal competition module is used to use the ratio of the candidate confidence level to the total value of the temporary confidence level as the first decay weight; The sixth internal competition module is used to take the temporary confidence level corresponding to the temporary competition preference information as the temporary competition confidence level; The seventh internal competition module is used to perform attenuation processing on the temporary competition confidence based on the first attenuation weight to obtain the attenuated temporary competition confidence. The eighth internal competition module is used to retain the candidate preference information and / or the temporary competition preference information as the temporary preference in the temporary preference set based on the candidate confidence and the decayed temporary competition confidence.

[0181] In one optional embodiment of this application, the intensity parameter acquisition module 303 includes: The first challenge intensity acquisition submodule is used to acquire, within a preset time window, several historical confidence levels corresponding to the candidate preference information and the historical mention times corresponding to the several historical confidence levels respectively; The second challenge intensity acquisition submodule is used to determine the average rate of change of confidence corresponding to the candidate preference within the time window based on several historical confidence levels and the historical mention times corresponding to several historical confidence levels respectively. The third challenge intensity acquisition submodule is used to use the average rate of change of confidence as the challenge intensity parameter. The first stability intensity acquisition submodule is used to acquire the duration of the competitive preference information in the stable preference set; The second stable strength acquisition submodule is used to acquire the first mention time of the last mention of the competitive preference information in the input data; The third stable intensity acquisition submodule is used to acquire the first time difference between the first mentioned time and the current time; The fourth stability intensity acquisition submodule is used to determine the stability intensity parameter corresponding to the competition preference information based on the duration of existence, the preset baseline stability duration, the first time difference, and the preset stability decay coefficient; wherein the stability intensity parameter increases with the duration of existence and decays with the first time difference.

[0182] In one optional embodiment of this application, the stable preference update module 305 includes: The first stability maintenance submodule is used to obtain several stability confidence levels corresponding to several stability preference information from the stability preference set if the challenge intensity parameter is less than the stability intensity parameter, and add the several stability confidence levels to obtain the total stability confidence value. The second stability maintenance submodule is used to use the ratio of the candidate confidence level to the total stable confidence level as the second decay weight; The third stability maintenance submodule is used to take the stability confidence level corresponding to the competitive preference information as the competitive confidence level; The fourth stability maintenance submodule is used to perform attenuation processing on the competitive confidence based on the second attenuation weight to obtain the attenuated competitive confidence. The fifth stability maintenance submodule is used to retain the candidate preference information and / or the competitive preference information as stable preference information in the stable preference set based on the candidate confidence and the decayed competitive confidence.

[0183] In one optional embodiment of this application, the stable preference update module 305 further includes: The first challenge initiation submodule is used to determine a first enhancement weight based on the challenge intensity parameter and the stable intensity parameter if the challenge intensity parameter is greater than or equal to the stable intensity parameter. The second challenge initiation submodule is used to enhance the candidate confidence based on the first enhancement weight to obtain the enhanced candidate confidence. The third challenge initiation submodule is used to determine the third attenuation weight based on the first enhancement weight; The fourth challenge initiation submodule is used to perform attenuation processing on the competition confidence based on the second attenuation weight and the third attenuation weight to obtain the attenuated competition confidence. The fifth challenge initiation submodule is used to retain the candidate preference information and / or the competitive preference information as stable preference information in the stable preference set based on the enhanced candidate confidence and the decayed competitive confidence.

[0184] In one optional embodiment of this application, the recommendation system further includes an archive preference set corresponding to the user, and the apparatus further includes: The competitive preference lookup module is used to search for competitive preference information corresponding to the candidate preference information in the stable preference set corresponding to the user when the candidate confidence level is greater than or equal to a preset promotion threshold. The promotion module is used to enhance the candidate confidence based on a preset promotion reward weight if there is no competing preference information corresponding to the candidate preference information in the stable preference set, and store the candidate preference information as stable preference information in the stable preference set based on the enhanced candidate confidence. The first downgrade module is used to store the stable preference information as historical preference information in the historical preference set if the stability confidence level corresponding to any stable preference information is less than a preset downgrade threshold. The second downgrade module is used to store the temporary preference information as archived preference information in the archived preference set if the temporary confidence level corresponding to any temporary preference information is less than the downgrade threshold. The first archiving module is used to obtain, from the historical preference set, the duration of continuous existence of any historical preference information in the stable preference set and the duration of continuous existence in the historical preference set; The second archiving module is used to store the historical preference information as archived preference information in the archived preference set if the duration of the historical preference information in the historical preference set is greater than the duration of the historical preference information in the stable preference set.

[0185] In one optional embodiment of this application, the apparatus further includes: The first feedback module is used to obtain the confidence peak value corresponding to the historical preference information if the candidate preference information is the same as any of the historical preference information in the historical preference set. The second feedback module is used to obtain the first sentiment intensity parameter of the last mention of the candidate preference information in the input data; The third feedback module is used to determine the second enhancement weight based on the confidence peak value, the first emotional intensity parameter, the duration of the historical preference information in the stable preference set, and the duration of the historical preference information in the historical preference set. The fourth feedback module is used to obtain the historical confidence level corresponding to the historical preference information; The fifth backflow module is used to enhance the historical confidence score according to the second enhancement weight to obtain the enhanced historical confidence score; The sixth reflow module is used to store the historical preference information into the temporary preference set based on the enhanced historical confidence level.

[0186] In one optional embodiment of this application, the apparatus further includes: The first parameter update module is used to take the preset statistical time period before the current time as the current period; The second parameter update module is used to obtain first preference flow information among the temporary preference set, the stable preference set, the historical preference set, and the archived preference set within the current period; The third parameter update module is used to adjust the weight parameters of the recommendation system according to the first preference flow information; wherein, the weight parameters include at least one of the following: promotion threshold, demotion threshold, promotion reward weight, time decay coefficient, and baseline stability duration; The fourth parameter update module is used to obtain second preference flow information between the temporary preference set, the stable preference set, the historical preference set and the archived preference set in the next cycle of the current cycle, and to obtain the feedback from the user on the information recommendation results. The fifth parameter update module is used to readjust the weight parameters of the recommendation system based on the second preference flow information if the second preference flow information does not meet the preset conditions, or if the feedback indicates that the information recommendation result does not meet the user's expectations.

[0187] As the apparatus embodiment is basically similar to the method embodiment, it is described in a relatively simple manner. For relevant details, please refer to the description of the method embodiment.

[0188] An embodiment of this application also provides an electronic device, which may include a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the method described above.

[0189] An embodiment of this application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the method described above.

[0190] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0191] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0192] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0193] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0194] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0195] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0196] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other modifications and updates to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all modifications and updates falling within the scope of the embodiments of the present application.

[0197] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device 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 terminal device. 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 terminal device that includes the aforementioned element.

[0198] The above provides a detailed description of the information recommendation method, apparatus, electronic device, and medium. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. An information recommendation method, characterized in that, Applied to a recommender system, the recommender system comprising a stable set of preferences corresponding to a user, the method includes: Obtain user input data and determine candidate preference information based on the input data; Search for competing preference information that corresponds to the candidate preference information in the stable preference set corresponding to the user; Based on the candidate preference information, the challenge intensity parameter corresponding to the candidate preference information is determined, and based on the competition preference information, the stability intensity parameter corresponding to the competition preference information is determined; The challenge strength parameter and the stability strength parameter are compared to obtain the comparison result. Based on the comparison results, the candidate preference information and / or the competing preference information are used as stable preference information, and the stable preference set is updated using the stable preference information. Obtain the user's real-time interaction command, and determine the stable preference information corresponding to the real-time interaction command from the updated stable preference set as the target preference information; The information recommendation result is determined based on the target preference information and the real-time interaction command, and then recommended to the user.

2. The method according to claim 1, characterized in that, The recommendation system further includes a temporary preference set and a historical preference set corresponding to the user, and the method further includes: Use any one of the temporary preference set, the stable preference set, and the historical preference set as reference preference information; The number of times the reference preference information is mentioned in the input data is obtained, as well as the emotional intensity parameter and mention time of each mention of the reference preference information in the input data; wherein, the input data is the user's input data from a preset historical time up to the current time; Obtain the time difference between the mentioned time and the current time; The reference confidence level of the reference preference information is determined based on the number of mentions, the emotional intensity parameter, the time difference, and a preset time decay coefficient; wherein the reference confidence level increases with the number of mentions and the emotional intensity parameter, and decays with the time difference.

3. The method according to claim 2, characterized in that, The temporary preference set includes several temporary preference information; After obtaining user input data and determining candidate preference information based on the input data, the method further includes: The candidate preference information is used as the temporary preference information, and the temporary preference set is updated using the temporary preference information; In the set of temporary preferences, obtain several temporary confidence levels corresponding to several temporary preference information, and add the several temporary confidence levels to obtain the total value of temporary confidence. Search the temporary competitive preference information that corresponds to the candidate preference information in the temporary preference set; The provisional confidence level corresponding to the candidate preference information is used as the candidate confidence level; The ratio of the candidate confidence level to the total value of the provisional confidence level is used as the first decay weight; The temporary confidence level corresponding to the temporary competitive preference information is taken as the temporary competitive confidence level; The temporary competitive confidence is attenuated according to the first attenuation weight to obtain the attenuated temporary competitive confidence. Based on the candidate confidence and the decayed temporary competitive confidence, the candidate preference information and / or the temporary competitive preference information are retained as the temporary preference in the temporary preference set.

4. The method according to claim 2, characterized in that, The steps of determining the challenge intensity parameter corresponding to the candidate preference information based on the candidate preference information, and determining the stability intensity parameter corresponding to the competition preference information based on the competition preference information, include: Within a preset time window, obtain several historical confidence levels corresponding to the candidate preference information and the historical mention times corresponding to each of the several historical confidence levels; The average rate of change of confidence corresponding to the candidate preference within the time window is determined based on several historical confidence levels and the historical mention times corresponding to those historical confidence levels. The average rate of change of confidence level is used as a challenge intensity parameter. Obtain the duration of the competitive preference information's persistence in the stable preference set; Obtain the first mention time of the last mention of the competitive preference information in the input data; Obtain the first time difference between the first mentioned time and the current time; The stability strength parameter corresponding to the competitive preference information is determined based on the duration of existence, the preset baseline stability duration, the first time difference, and the preset stability decay coefficient; wherein the stability strength parameter increases with the duration of existence and decays with the first time difference.

5. The method according to claim 3, characterized in that, The step of using the candidate preference information and / or the competing preference information as stable preference information based on the comparison results, and updating the stable preference set using the stable preference information, includes: If the challenge intensity parameter is less than the stability intensity parameter, then in the stable preference set, obtain several stable confidence levels corresponding to several stable preference information respectively, and add the several stable confidence levels to obtain the total stable confidence value; The ratio of the candidate confidence level to the total stable confidence level is used as the second decay weight; The stable confidence level corresponding to the competitive preference information is used as the competitive confidence level; The competitive confidence score is attenuated according to the second attenuation weight to obtain the attenuated competitive confidence score; Based on the candidate confidence and the decayed competing confidence, the candidate preference information and / or the competing preference information are retained as stable preference information in the stable preference set.

6. The method according to claim 5, characterized in that, The step of using the candidate preference information and / or the competing preference information as stable preference information based on the comparison results, and updating the stable preference set using the stable preference information, further includes: If the challenge intensity parameter is greater than or equal to the stability intensity parameter, then a first enhancement weight is determined based on the challenge intensity parameter and the stability intensity parameter; The candidate confidence level is enhanced based on the first enhancement weight to obtain the enhanced candidate confidence level. The third attenuation weight is determined based on the first enhancement weight; The competitive confidence is attenuated according to the second attenuation weight and the third attenuation weight to obtain the attenuated competitive confidence. Based on the enhanced candidate confidence and the weakened competitive confidence, the candidate preference information and / or the competitive preference information are retained as stable preference information in the stable preference set.

7. The method according to claim 3, characterized in that, The recommendation system further includes an archive preference set corresponding to the user, and the method further includes: If the candidate confidence level is greater than or equal to a preset promotion threshold, search for competing preference information corresponding to the candidate preference information in the stable preference set corresponding to the user. If there is no competing preference information corresponding to the candidate preference information in the stable preference set, the candidate confidence is enhanced according to the preset promotion reward weight, and the candidate preference information is stored as stable preference information in the stable preference set according to the enhanced candidate confidence. In the stable preference set, if the stability confidence level corresponding to any stable preference information is less than a preset degradation threshold, then the stable preference information is stored as historical preference information in the historical preference set. If, in the temporary preference set, the temporary confidence level corresponding to any temporary preference information is less than the downgrade threshold, then the temporary preference information is stored as archived preference information in the archived preference set. In the historical preference set, obtain the duration of persistence of any historical preference information in the stable preference set and the duration of persistence in the historical preference set; If the duration of the historical preference information in the historical preference set is greater than the duration of the historical preference information in the stable preference set, then the historical preference information is stored as archived preference information in the archived preference set.

8. The method according to claim 7, characterized in that, The method further includes: If the candidate preference information is the same as any of the historical preference information in the historical preference set, then the confidence peak value corresponding to the historical preference information is obtained; Obtain the first sentiment intensity parameter of the last mention of the candidate preference information in the input data; The second enhancement weight is determined based on the confidence peak value, the first emotional intensity parameter, the duration of the historical preference information in the stable preference set, and the duration of the historical preference information in the historical preference set. Obtain the historical confidence level corresponding to the historical preference information; The historical confidence level is enhanced according to the second enhancement weight to obtain the enhanced historical confidence level; Based on the enhanced historical confidence level, the historical preference information is stored in the temporary preference set.

9. The method according to claim 8, characterized in that, The method further includes: The preset statistical time period prior to the current moment is taken as the current period; Within the current period, first preference flow information is obtained among the temporary preference set, the stable preference set, the historical preference set, and the archived preference set; The weight parameters of the recommendation system are adjusted according to the first preference flow information; wherein the weight parameters include at least one of the following: promotion threshold, demotion threshold, promotion reward weight, time decay coefficient, and baseline stability duration. In the next cycle of the current cycle, obtain the second preference flow information between the temporary preference set, the stable preference set, the historical preference set and the archived preference set, and obtain the feedback from the user on the information recommendation results; If the second preference flow information does not meet the preset conditions, or if the feedback indicates that the information recommendation result does not meet the user's expectations, then the weight parameters of the recommendation system are readjusted according to the second preference flow information.

10. An information recommendation device, characterized in that, Applied to a recommendation system, the recommendation system comprising a stable set of preferences corresponding to a user, the device includes: The candidate preference acquisition module is used to acquire user input data and determine candidate preference information based on the input data; The competitive preference acquisition module is used to search for competitive preference information corresponding to the candidate preference information in the stable preference set corresponding to the user. The strength parameter acquisition module is used to determine the challenge strength parameter corresponding to the candidate preference information based on the candidate preference information, and to determine the stability strength parameter corresponding to the competition preference information based on the competition preference information. The comparison result acquisition module is used to compare the challenge intensity parameter and the stability intensity parameter to obtain the comparison result. A stable preference update module is used to take the candidate preference information and / or the competing preference information as stable preference information based on the comparison results, and update the stable preference set using the stable preference information; The target preference acquisition module is used to acquire the user's real-time interaction command and determine the stable preference information corresponding to the real-time interaction command from the updated stable preference set as the target preference information; The recommendation result acquisition module is used to determine the information recommendation result based on the target preference information and the real-time interaction command, and recommend it to the user.

11. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored on the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the method as described in any one of claims 1-9.

12. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the method as described in any one of claims 1-9.

Images