Text creation customized material personalized recommendation method and system

Abstract

The application discloses a creative customization material individualized recommendation method and system, and relates to the technical field of data processing. The method comprises the following steps: when a target user selects a recommended material in a material recommendation display area to browse, based on historical browsing materials and interaction data of the historical browsing materials, a core interest label and a core interest material of the target user are determined; based on a matching degree between a label of a recommended non-browsing material and the core interest label, a first candidate recommended material is determined; based on a heat value and a scene adaptation degree of a non-recommended material, a second candidate recommended material is determined; based on a similarity between the non-recommended material and the core interest material, a third candidate recommended material is determined; and based on the first candidate recommended material, the second candidate recommended material and the third candidate recommended material, a material recommendation display area is updated. In the above manner, the known interest and the potential interest of the user are fully considered, and suitable materials can be more comprehensively and accurately recommended to the user.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method and system for personalized recommendation of customized cultural and creative materials. Background Technology

[0002] With the booming development of the cultural and creative industries, online cultural and creative customization platforms (such as personalized gifts, customized T-shirts, and exclusive cultural and creative product designs) are becoming increasingly popular. Most of these platforms rely on periodic model training based on users' historical behavior (such as past clicks and purchase records) and recommend materials to users according to the model's output. This approach over-depends on users' historical behavior and ignores users' potential interests and needs, resulting in simplistic recommendation results.

[0003] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide a personalized recommendation method and system for customized cultural and creative materials, which aims to solve the technical problem that traditional recommendation schemes in the prior art rely solely on users' historical behavior, over-rely on known interests, ignore users' potential interest needs, and result in monotonous recommendation results.

[0005] To achieve the above objectives, this application provides a method for personalized recommendation of customized cultural and creative materials, the method comprising: When a target user selects and browses recommended materials in the material recommendation display area, the system updates the recommended unviewed materials, unrecommended materials, historically viewed materials, and the interaction data of the historically viewed materials in real time. Based on the historical browsing materials and the interaction data of the historical browsing materials, the core interest tags and core interest materials of the target user are determined; Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, the first candidate recommended materials are determined; Based on the popularity and scene suitability of the unrecommended materials, a second candidate recommended material is determined; Based on the similarity between the unrecommended materials and the core interest materials, a third candidate recommended material is determined; The material recommendation display area is updated based on the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material.

[0006] In one embodiment, the step of determining the target user's core interest tags and core interest materials based on the historical browsing materials and the interaction data of the historical browsing materials includes: Based on the interaction data of the historical browsing materials, valid sample materials are determined from the historical browsing materials. The interaction data includes the number of times it is viewed, the duration of stay, the collection status, and the number of times it is shared. Based on the interaction data of the effective sample materials, calculate the interest score of the effective sample materials; The interest score of the effective sample material is used as the interest score corresponding to the tag of the effective sample material. Based on the interest score corresponding to the tag of the effective sample material, the core interest tags of the target user are selected. Based on the interest scores of the valid sample materials, the core interest materials of the target user are determined from the valid sample materials.

[0007] In one embodiment, the step of calculating the interest score of the effective sample material based on the interaction data of the effective sample material includes: Obtain the first correspondence between the number of views, dwell time, collection status, number of shares and the initial interest score, and determine the initial interest score of the effective sample material based on the interaction data of the effective sample material and the first correspondence. Based on the most recent interaction time corresponding to the interaction data of the effective sample material, determine the time decay factor of the effective sample material; The initial interest score of the effective sample material is corrected based on the time decay factor of the effective sample material to obtain the interest score of the effective sample material.

[0008] In one embodiment, the step of determining the first candidate recommended material based on the matching degree between the tags of the recommended unviewed material and the core interest tags includes: Based on the tags of the recommended unviewed materials and the core interest tags, the overlapping tags of the recommended unviewed materials are determined; Obtain the second correspondence between the number of overlapping tags, the matching weight of overlapping tags, and the initial matching degree. Based on the second correspondence and the number and matching weight of overlapping tags corresponding to the recommended unviewed material, obtain the initial matching degree between the tags of the recommended unviewed material and the core interest tags. Obtain the recommended materials in the same batch corresponding to the recommended unviewed materials in the historical browsing materials, and determine the bias correction ratio based on the browsing ratio of the same type of materials in the same batch corresponding to the recommended unviewed materials; Based on the bias correction ratio corresponding to the recommended unviewed material, the initial matching degree between the tags of the recommended unviewed material and the core interest tags is corrected to obtain the matching degree between the tags of the recommended unviewed material and the core interest tags; Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, the first candidate recommended materials with a matching degree greater than a preset matching threshold are selected from the recommended unviewed materials.

[0009] In one embodiment, the step of determining the second candidate recommended material based on the popularity value and scene suitability of the unrecommended material includes: Based on the degree of thematic repetition between the unrecommended materials and the historically viewed materials, the first materials to be removed are determined; Based on the tag overlap between the unrecommended material and the first candidate recommended material, a second material to be removed is determined; Remove the first and second rejected materials from the unrecommended materials to obtain the target unrecommended materials; Based on the popularity value and scene suitability of the target unrecommended material, as well as the weights corresponding to the popularity value and scene suitability, the recommendation score of the target unrecommended material is calculated; Based on the recommendation score of the target unrecommended material, a second candidate recommended material with a recommendation score greater than a preset score threshold is selected from the target unrecommended material.

[0010] In one embodiment, the step of determining the second candidate recommended material based on the popularity value and scene suitability of the unrecommended material further includes: Extract the popularity metrics of the unrecommended materials, including the number of clicks, complete views rate, percentage of favorites, and positive review rate within a preset time period; Based on the popularity index of the unrecommended material and the popularity weight corresponding to the popularity index, the initial popularity value of the unrecommended material is determined; Based on the release time of the unrecommended material, the initial popularity value of the unrecommended material is adjusted to obtain the popularity value of the unrecommended material; Based on the real-time customized scenario of the target user, the adaptation score of the unrecommended material in the real-time customized scenario is determined. The real-time customized scenario includes time-dimensional scenario, location-dimensional scenario, and device-dimensional scenario. The scene adaptation degree of the unrecommended material is determined based on the adaptation score of the unrecommended material in the real-time customized scene and the adaptation weight of the real-time customized scene.

[0011] In one embodiment, the step of determining a third candidate recommended material based on the similarity between the unrecommended material and the core interest material includes: Based on the target feature extraction model, similar feature vectors of the unrecommended materials and similar feature vectors of the core interest materials are extracted. The similar feature vectors include semantic feature vectors, visual feature vectors, and attribute feature vectors. The similar feature vectors of the unrecommended materials are fused to obtain the fused feature vector of the unrecommended materials, and the similar feature vectors of the core interest materials are fused to obtain the fused feature vector of the core interest materials. Based on the fusion feature vector of the unrecommended material and the fusion feature vector of the core interest material, the similarity between the unrecommended material and the core interest material is determined; Based on the similarity between the unrecommended materials and the core interest materials, a third candidate recommended material with a similarity greater than a preset similarity threshold is selected from the unrecommended materials.

[0012] In one embodiment, the step of updating the material recommendation display area based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material includes: Based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material, the target recommendation material is determined; Based on the recommendation type of the target recommended material, a personalized display strategy for the target recommended material is determined. The personalized display strategy includes display position, visual tags, and annotation colors. Based on the target recommended materials and the personalized display strategy of the target recommended materials, the material recommendation display area is updated.

[0013] In one embodiment, the step of determining the target recommendation material based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material includes: The target user's historical interaction behavior in the material recommendation display area is obtained. Based on the historical interaction behavior, a corresponding recommendation ratio is assigned to the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material. The historical interaction behavior includes browsing type, number of consecutive selections, and number of valid browsings. Based on the preset recommendation quantity and the recommendation ratio, the recommendation quantities of the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material are determined respectively; Based on the number of recommendations for the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material, a target recommendation material is selected from the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material.

[0014] Furthermore, to achieve the above objectives, this application also proposes a personalized recommendation system for customized cultural and creative materials, which includes: The data update module is used to update the recommended materials, unviewed materials, unrecommended materials, historically viewed materials, and interactive data of the historically viewed materials in real time when the target user selects and browses the recommended materials in the material recommendation display area. The material filtering module is used to determine the core interest tags and core interest materials of the target user based on the historical browsing materials and the interaction data of the historical browsing materials; The material filtering module is also used to determine the first candidate recommended material based on the matching degree between the tags of the recommended unviewed materials and the core interest tags; The material filtering module is also used to determine a second candidate recommended material based on the popularity value and scene adaptability of the unrecommended material; The material filtering module is also used to determine a third candidate recommended material based on the similarity between the unrecommended material and the core interest material; The material recommendation module is used to update the material recommendation display area based on the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material.

[0015] In addition, to achieve the above objectives, this application also proposes a personalized recommendation device for customized cultural and creative materials. The personalized recommendation device for customized cultural and creative materials includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. The computer program is configured to implement the steps of the personalized recommendation method for customized cultural and creative materials as described above.

[0016] In addition, to achieve the above objectives, the present invention also proposes a storage medium, which is a computer-readable storage medium, and stores a computer program on the storage medium. When the computer program is executed by a processor, it implements the steps of the personalized recommendation method for customized cultural and creative materials as described above.

[0017] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the personalized recommendation method for customized cultural and creative materials as described above.

[0018] This application provides a personalized recommendation method for cultural and creative customized materials. When a target user selects recommended materials in the material recommendation display area to browse, the method updates the recommended unbrowsed materials, unrecommended materials, historically browsed materials, and interaction data of historically browsed materials in real time. Based on historically browsed materials and their interaction data, the method determines the target user's core interest tags and core interest materials. Based on the matching degree between the tags of the recommended unbrowsed materials and the core interest tags, the method determines the first candidate recommended materials. Based on the popularity value and scene adaptability of the unrecommended materials, the method determines the second candidate recommended materials. Based on the similarity between the unrecommended materials and the core interest materials, the method determines the third candidate recommended materials. Based on the first, second, and third candidate recommended materials, the material recommendation display area is updated. This application predicts two types of potential interest content. First, it predicts content the user might browse in the future from previously recommended content that the user hasn't currently viewed, using this as the first type of recommendation candidate. Second, it predicts content the user might browse in the future from content that hasn't been recommended to the user, using this as the second type of recommendation candidate. Finally, it combines this prediction with a third type of recommendation candidate based on the user's historical preferences to generate a personalized recommendation list. This approach fully considers both the user's known and potential interests, enabling more comprehensive and accurate recommendations of suitable content. It increases the probability of content discovery and adoption, improves the overall utilization rate of content on the customized platform, and solves the technical problem of traditional recommendation schemes that rely solely on user history, over-rely on known interests, ignore potential user needs, and result in limited recommendation outcomes. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating an embodiment of the method for personalized recommendation of cultural and creative customized materials for this application; Figure 2 A flowchart illustrating Example 2 of the method for personalized recommendation of cultural and creative customized materials for this application; Figure 3 This is a schematic diagram of the module structure of the personalized recommendation system for customized cultural and creative materials in an embodiment of this application; Figure 4 This is a schematic diagram of the device structure of the hardware operating environment involved in the personalized recommendation method for customized cultural and creative materials in this application embodiment.

[0022] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0024] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0025] The main solution of this application embodiment is as follows: when a target user selects recommended materials in the material recommendation display area for browsing, the system updates the interactive data of unbrowsed materials, unrecommended materials, historically browsed materials, and historically browsed materials in real time; based on the historically browsed materials and their interactive data, the system determines the target user's core interest tags and core interest materials; based on the matching degree between the tags of unbrowsed materials and the core interest tags, the system determines the first candidate recommended materials; based on the popularity value and scene adaptability of unrecommended materials, the system determines the second candidate recommended materials; based on the similarity between unrecommended materials and the core interest materials, the system determines the third candidate recommended materials; and based on the first, second, and third candidate recommended materials, the material recommendation display area is updated.

[0026] This application provides a solution that predicts two types of potential interest materials. First, it predicts materials the user might browse in the future from those previously recommended but not currently viewed, serving as the first type of recommendation candidate material. Second, it predicts materials the user might browse in the future from those not previously recommended to the user, serving as the second type of recommendation candidate material. Finally, it combines these with a third type of recommendation candidate material corresponding to the user's historical preferences to generate a personalized recommendation list. This approach fully considers both the user's known and potential interests, enabling more comprehensive and accurate recommendations of suitable materials. It increases the probability of material discovery and adoption, improves the overall utilization rate of materials on the customized platform, and solves the technical problem of traditional recommendation schemes that rely solely on user history, over-rely on known interests, ignore potential user needs, and result in limited recommendation outcomes.

[0027] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a personalized recommendation device for customized cultural and creative materials. This embodiment does not specifically limit this. The following uses a personalized recommendation device for customized cultural and creative materials as an example to describe this embodiment and the following embodiments.

[0028] This application provides a method for personalized recommendation of customized cultural and creative materials, referring to... Figure 1 , Figure 1 A flowchart illustrating the first embodiment of the personalized recommendation method for customized cultural and creative materials in this application.

[0029] In this embodiment, the personalized recommendation method for cultural and creative customized materials includes steps S10 to S60: Step S10: When the target user selects recommended materials in the material recommendation display area to browse, update the recommended unbrowsed materials, unrecommended materials, historically browsed materials, and the interaction data of the historically browsed materials in real time; It should be noted that the target user is the user currently selecting materials for cultural and creative product customization on the platform. The target user can choose the category of cultural and creative products they customize according to their own needs, without specific limitations. The customization platform has a personalized recommendation system for cultural and creative product customization materials, which can apply the steps of the personalized recommendation method for cultural and creative product customization materials described in this embodiment.

[0030] Additionally, it should be noted that the material recommendation display area is the area used to display recommended materials. It can be placed anywhere suitable on the current webpage, such as at the bottom or right side of the page; there are no specific limitations. Whether the material recommendation display area appears on the current page can be set by the target user. The recommended materials displayed in the material recommendation display area are the recommended materials. All materials in the customization platform are stored in the material database.

[0031] Understandably, if a target user clicks on a recommended material in the recommended material display area, they will be redirected to the details page of that recommended material, indicating that the target user is selecting and browsing that material. If a target user selects / clicks on a recommended material in the recommended material display area, key data such as recommended unviewed materials, unrecommended materials, historically viewed materials, and interaction data of historically viewed materials will be updated immediately based on the user's latest material browsing activity, in order to conduct a new round of material recommendations.

[0032] It's important to clarify that "recommended but unviewed content" refers to content that has been recommended to the target user but not viewed by them, while "unrecommended content" refers to content that has never been recommended to the target user. "Historically viewed content" refers to content that the target user has viewed. Interaction data for historically viewed content refers to the target user's interactions with that content. This data includes at least the number of views, dwell time, favorite status, and number of shares. The number of views refers to the number of times the target user clicked to enter / viewed the content details page. The dwell time refers to the duration of viewing the content after entering the details page. The number of shares refers to the number of times the target user shared the corresponding content with other users / friends. Favorit status includes two states: favorited and not favorited. Favorited means the corresponding content has been favorited by the target user, while not favorited means the corresponding content has not yet been favorited by the target user.

[0033] In this embodiment, each time a target user selects a recommended material to browse, the system will update the recommended unbrowsed materials, unrecommended materials, historically browsed materials, and the interaction data of each historically browsed material in real time.

[0034] Step S20: Based on the historical browsing materials and the interaction data of the historical browsing materials, determine the core interest tags and core interest materials of the target user; It should be noted that each material will have corresponding tags, such as category tags, style tags, and theme tags. Category tags can be descriptions of the material category, such as: people, plants, and text. Style tags can be descriptions of the style applicable to the material, such as: forest style, fresh and simple, traditional Chinese style, and anime. Theme tags can be descriptions of the theme applicable to the material, such as: birthday, workplace, and anniversary. This embodiment does not make specific limitations on these.

[0035] As can be understood, core interest tags refer to tags that the target users are particularly interested in, i.e., tags that the target users prefer, while core interest content refers to content that the target users are particularly interested in, i.e., content that the target users prefer. Multiple core interest tags and core interest content can typically be set.

[0036] In one feasible implementation, step S20 may include steps S201 to S204: Step S201: Based on the interaction data of the historical browsing materials, determine the valid sample materials in the historical browsing materials; It's important to note that valid sample content refers to content that the target user has actually viewed. When browsing recommended content, target users may accidentally click or click randomly / blindly, resulting in a short dwell time. Therefore, the dwell time of historical viewed content can be used to filter out valid sample content.

[0037] Understandably, a duration threshold is set, for example, 3 seconds. If the dwell time on historical browsing content is greater than or equal to this threshold, it indicates that the target user has effectively browsed the historical browsing content, and the historical browsing content can be used as a valid sample content. If the dwell time on historical browsing content is less than this threshold, it indicates that the target user has not effectively browsed the historical browsing content, and the historical browsing content cannot be used as a valid sample content.

[0038] Step S202: Calculate the interest score of the effective sample material based on the interaction data of the effective sample material; It should be noted that the interest score is an indicator used to measure the degree of interest / preference of the target user. The higher the interest score, the more interested the target user is.

[0039] In one feasible implementation, step S202 may include: obtaining a first correspondence between the number of views, dwell time, collection status, number of shares and initial interest score; determining the initial interest score of the effective sample material based on the interaction data of the effective sample material and the first correspondence; determining the time decay factor of the effective sample material based on the most recent interaction time corresponding to the interaction data of the effective sample material; and correcting the initial interest score of the effective sample material based on the time decay factor of the effective sample material to obtain the interest score of the effective sample material.

[0040] It should be noted that the initial interest score refers to a preliminary calculated interest score, which needs further refinement to obtain an accurate score. The initial interest score can be calculated using the number of views, dwell time, collection status, and number of shares. During calculation, the number of views, dwell time, collection status, and number of shares need to be expressed as corresponding scores. For example, for the number of views, 0 times equals 0 points, 1-3 times equals 2 points, 4-5 times equals 4 points, 6-8 times equals 6 points, 9-10 times equals 8 points, and more than 10 times equals 10 points; for dwell time, less than 3 seconds equals 0 points, greater than or equal to 3 seconds and less than 10 seconds equals 2 points, greater than or equal to 10 seconds and less than 30 seconds equals 6 points, and greater than or equal to 30 seconds equals 10 points; for collection status, collected equals 10 points, not collected equals 0 points; for the number of shares, 0 times equals 0 points, 1-3 times equals 2 points, 4-5 times equals 4 points, 6-8 times equals 6 points, 9-10 times equals 8 points, and more than 10 times equals 10 points.

[0041] Understandably, the primary correspondence between pageview counts, dwell time, favorite status, share counts, and the initial interest score—that is, the formula for calculating the initial interest score—is as follows:

[0042] In the formula, Indicates the initial interest score. , , , These represent the scores corresponding to the number of views, dwell time, collection status, and number of shares, respectively. , , , These represent the weights corresponding to the number of views, dwell time, collection status, and number of shares, respectively. The specific values ​​can be flexibly adjusted according to actual needs. By substituting the number of views, dwell time, collection status, number of shares, and the corresponding scores and weights of the initial interest scores into the first correspondence mentioned above, the initial interest score of each valid sample material can be calculated.

[0043] It should be understood that the most recent interaction time is the time of the most recent interaction with the material. For example, if the most recent interaction was browsing, then the time of the most recent click would be considered the most recent interaction time. This embodiment introduces a time decay factor. The further the most recent interaction time is from the current time, the smaller the value of the time decay factor; the closer the most recent interaction time is to the current time, the larger the value of the time decay factor. The time decay factor can be calculated based on the most recent interaction time and the decay coefficient, as shown below:

[0044] In the formula, Indicates the time decay factor. This represents the attenuation coefficient, which can be set to 0.002. This represents the number of days since the most recent interaction. A time decay factor is used to correct the initial interest score of valid sample materials, avoiding interference from outdated data in the score calculation, to obtain the final interest score. The calculation formula is:

[0045] In the formula, Indicates interest score, Indicates the initial interest score. This represents the time decay factor.

[0046] Step S203: Use the interest score of the effective sample material as the interest score corresponding to the tag of the effective sample material, and filter out the core interest tags of the target user based on the interest score corresponding to the tag of the effective sample material; It is understandable that the interest score of valid sample material can also be considered as the interest score of the tags of that valid sample material. In specific implementation, the interest scores of the same tags can be added together to obtain the total interest score of each tag. Then, all tags of valid sample material can be sorted in descending order according to the total interest score or the average value of the interest scores. According to a set number (e.g., 10), the corresponding tags are selected from high to low as core interest tags.

[0047] Step S204: Based on the interest scores of the valid sample materials, determine the core interest materials of the target user from the valid sample materials.

[0048] Understandably, the valid sample materials are sorted in descending order according to their interest scores, and the corresponding valid sample materials are selected as core interest materials in descending order of the set number (e.g., 15).

[0049] Step S30: Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, determine the first candidate recommended materials; It's important to note that the first-choice recommended content refers to content among the unviewed recommended content that has a high probability of being viewed by the target user. The matching degree between the tags of the unviewed recommended content and the core interest tags, i.e., the degree of matching between the tags of the unviewed content and the core interest tags, is typically used as the first-choice recommended content.

[0050] In one feasible implementation, step S30 may include steps S301 to S305: Step S301: Based on the tags of the recommended unviewed materials and the core interest tags, determine the overlapping tags of the recommended unviewed materials; It should be noted that overlapping tags refer to tags that overlap with the core interest tags among all the tags of each recommended unviewed material.

[0051] Step S302: Obtain the second correspondence between the number of overlapping tags, the matching weight of overlapping tags and the initial matching degree. Based on the second correspondence and the number and matching weight of overlapping tags corresponding to the recommended unviewed material, obtain the initial matching degree between the tags of the recommended unviewed material and the core interest tags. It is understandable that the initial matching degree is the preliminary calculated matching degree, which needs to be further corrected to obtain an accurate matching degree. When calculating the initial matching degree, each core interest tag is assigned a corresponding weight, i.e., a matching weight. In this embodiment, the matching weight of a core interest tag can be set based on its interest score; the higher the interest score, the greater the matching weight. The matching weight of overlapping tags is determined according to the matching weight of the core interest tags. The second correspondence between the number of overlapping tags, the matching weight of overlapping tags, and the initial matching degree, i.e., the calculation formula for the initial matching degree, is as follows:

[0052] In the formula, Indicates the initial matching degree. Indicates the number of overlapping tags. Indicates the number of core interest tags. Indicates the first The matching weight of each overlapping label Indicates the first The matching weights of each core interest tag are calculated. Substituting the number of overlapping tags and their matching weights, along with the number of core interest tags and their matching weights, into the second correspondence mentioned above, the initial matching degree between the tags of the recommended unviewed materials and the core interest tags is calculated.

[0053] Step S303: Obtain the recommended materials in the same batch corresponding to the recommended unviewed materials in the historical browsing materials, and determine the bias correction ratio based on the browsing ratio of the same type of materials in the same batch of recommended materials corresponding to the recommended unviewed materials; It should be noted that "recommended materials in the same batch" refers to other materials pushed along with the unviewed recommended materials when recommending them to the target user. "Materials of the same type in the same batch" refers to materials of the same type as the unviewed recommended materials in the same batch. "Same type" means that at least one tag is the same.

[0054] It is understood that the browsing ratio refers to the proportion of similar content that the target user has viewed. This embodiment introduces a bias correction ratio based on the browsing ratio of similar content, which is the proportion by which the initial matching degree needs to be corrected. For example, if the browsing ratio is greater than a set threshold (e.g., 60%), it indicates that the content pushed this time generally matches user preferences, and the bias correction ratio is set to 10%; otherwise, the bias correction ratio is set to 0%. This can be flexibly adjusted according to the actual situation, and no specific limitation is made.

[0055] Step S304: Based on the bias correction ratio corresponding to the recommended unviewed material, correct the initial matching degree between the tags of the recommended unviewed material and the core interest tags to obtain the matching degree between the tags of the recommended unviewed material and the core interest tags; It should be understood that the initial match rate is adjusted according to the bias correction ratio. For example, if the bias correction ratio is 10%, the initial match rate is increased by 10%, which means multiplying the initial match rate by 110%. The value obtained after adjusting the initial match rate is the final match rate. The range of the match rate is usually [0,1].

[0056] Step S305: Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, select the first candidate recommended materials with a matching degree greater than a preset matching threshold from the recommended unviewed materials.

[0057] It is understood that the preset matching threshold is a pre-set threshold for the matching degree, for example, 0.65, which can be flexibly adjusted according to the actual situation, and there is no specific limitation on it. In this embodiment, recommended unviewed materials with a matching degree greater than the preset matching threshold are used as the first candidate recommended materials.

[0058] Step S40: Based on the popularity value and scene adaptability of the unrecommended materials, determine the second candidate recommended materials; It's important to note that popularity is an indicator of how popular the content is, typically ranging from [0,1]. A higher value indicates greater acceptance of the content among all users on the customization platform. Scene adaptability is an indicator of whether the content accurately matches the target user's current environment / scene, also typically ranging from [0,1]. A higher value indicates a better fit for the target user's current time, location, and device status.

[0059] It is understandable that the second candidate recommended material refers to the material with a high probability of being viewed by the target user, determined from the unrecommended materials based on its popularity and scene suitability. In this embodiment, suitable unrecommended materials are selected as the second candidate recommended materials according to their popularity and scene suitability.

[0060] In one possible implementation, steps A11-A15 may be included before step S40: Step A11: Extract the popularity metrics of the unrecommended materials. The popularity metrics include the number of clicks, the complete view rate, the percentage of favorites, and the positive review rate within a preset time period. It should be noted that the popularity index refers to the core indicators that can be used to calculate the popularity value, that is, the indicators that can reflect the popularity value of the material, including the number of clicks within a preset time period, the complete view rate within a preset time period, the percentage of favorites within a preset time period, and the positive review rate within a preset time period.

[0061] Additionally, it should be noted that the preset time period is a pre-defined cycle for extracting popularity metrics, for example, 7 days. Assuming the preset time period is set to 7 days, the popularity metrics could be the number of clicks in the past 7 days, the complete pageview rate in the past 7 days, the percentage of items saved in the past 7 days, and the positive review rate in the past 7 days.

[0062] It's understandable that click-through rate (CTR) refers to the total number of times all users on the platform clicked on the unrecommended content. Full view rate is the ratio of the number of users who stayed on the details page for at least 3 seconds after clicking on the unrecommended content to the CTR. Collection rate is the ratio of the number of users who collected the unrecommended content to the total number of users who clicked on it. Positive review rate is the proportion of positive reviews among all reviews of the unrecommended content. CTR needs to be uniformly normalized to the value range of [0,1].

[0063] Step A12: Determine the initial popularity value of the unrecommended material based on the popularity index of the unrecommended material and the popularity weight corresponding to the popularity index; It should be understood that this embodiment assigns a corresponding weight to each popularity indicator when calculating the popularity value, i.e., popularity weight. The specific value can be flexibly adjusted according to the actual situation. For example, the popularity weight of click volume is 0.3, the popularity weight of complete view rate is 0.4, the popularity weight of collection ratio is 0.2, and the popularity weight of positive evaluation rate is 0.1. No specific limitation is made on this. Based on the popularity weight of each popularity indicator, the popularity indicators of unrecommended materials are weighted and summed to initially calculate the popularity value, i.e., the initial popularity value. Further adjustments are needed to ensure accuracy. The calculation formula for the initial popularity value is as follows:

[0064] In the formula, Indicates the initial heat value. , , , These represent click-through rate, complete page view rate, percentage of saved items, and positive review rate, respectively. , , , These represent the popularity weights corresponding to clicks, complete pageview rate, percentage of favorites, and positive review rate, respectively. The specific values ​​can be flexibly adjusted according to actual needs. We use multi-dimensional metrics for weighted calculations to avoid the bias of relying solely on click-through rate statistics.

[0065] Step A13: Based on the release time of the unrecommended material, adjust the initial popularity value of the unrecommended material to obtain the popularity value of the unrecommended material; It is understandable that the content database is constantly updated. New content, due to its more recent release date, has limited accuracy in terms of its popularity metrics, failing to precisely represent its popularity. Therefore, this embodiment adjusts the initial popularity value of unrecommended content based on its release date to obtain the final popularity value. For example, new content released within 3 days receives an additional 10% increase in popularity (initial popularity value increases by 10%), while older content released more than 30 days ago automatically experiences a 15% decrease in popularity (initial popularity value decreases by 15%). No specific limitations are imposed on this.

[0066] Step A14: Based on the real-time customized scenario of the target user, determine the adaptation score of the unrecommended material in the real-time customized scenario. The real-time customized scenario includes time-dimensional scenario, location-dimensional scenario, and device-dimensional scenario. It should be noted that real-time customized scenarios refer to the environment / scene in which the target user is currently located. Real-time customized scenarios are characterized from the dimensions of time, location, and device. Therefore, real-time customized scenarios can be considered to include time-dimensional scenarios, location-dimensional scenarios, and device-dimensional scenarios. Time-dimensional scenarios can be represented by weekdays / rest days and morning / noon / evening time periods. Location-dimensional scenarios can be represented by the attributes of the current location, such as: scenic spot, home, office building. Device-dimensional scenarios can be represented by the type of device currently being used, such as: mobile phone, computer, tablet.

[0067] Understandably, the adaptation score is also calculated from the time, location, and device dimensions. In practice, a fixed initial adaptation score (e.g., 1 point) is typically set for all unrecommended content. Then, the initial adaptation score is increased or decreased according to the time, location, and device dimensions to obtain the adaptation score of the unrecommended content in each of these dimensions. For example, if the time dimension is a weekday evening, the initial adaptation score of quiet content will increase by 10%; if the location dimension is a scenic area, the initial adaptation score of landscape content will increase by 10%; and if the device dimension is a computer, the initial adaptation score of high-definition content will increase by 10%.

[0068] Step A15: Determine the scene adaptation degree of the unrecommended material based on the adaptation score of the unrecommended material in the real-time customized scene and the adaptation weight of the real-time customized scene.

[0069] It should be understood that when calculating scene adaptation, corresponding weights, i.e., adaptation weights, are set for the time dimension scene, location dimension scene, and device dimension scene. For example, the adaptation weight for the time dimension scene is set to 0.35, the adaptation weight for the location dimension scene is set to 0.35, and the adaptation weight for the device dimension scene is set to 0.3. Based on the adaptation weights for the time dimension scene, location dimension scene, and device dimension scene, the adaptation scores of the unrecommended materials in these three dimensions are weighted and summed to calculate the scene adaptation degree of the unrecommended materials.

[0070] In one feasible implementation, step S40 may include steps S401 to S405: Step S401: Based on the degree of theme repetition between the unrecommended materials and the historical browsing materials, determine the first materials to be removed; It's important to note that theme duplication refers to the degree of repetition of the themes applied to the content. Generally, it's determined by comparing unrecommended content with content viewed within the last 30 days. Theme duplication can be represented as the ratio of the number of duplicate themes to the total number of themes applied to unrecommended content.

[0071] Understandably, unrecommended content with a topic repetition rate exceeding a set threshold (e.g., 80%) is considered the first batch of content to be removed.

[0072] Step S402: Based on the tag overlap between the unrecommended material and the first candidate recommended material, determine the second material to be removed; It's important to note that tag overlap refers to the degree of overlap between tags used in content. Tag overlap is determined based on the tags of unrecommended content and the tags of the first-choice recommended content. Tag overlap can be represented as the ratio of the number of overlapping tags to the total number of tags for unrecommended content.

[0073] Understandably, unrecommended content with a tag overlap exceeding a set threshold (e.g., 70%) is considered as the second batch of content to be removed.

[0074] Step S403: Remove the first and second rejected materials from the unrecommended materials to obtain the target unrecommended materials; Understandably, the first and second removal materials are removed from the unrecommended materials to filter out homogeneous materials. The remaining materials are the target unrecommended materials.

[0075] Step S404: Calculate the recommendation score of the target unrecommended material based on its popularity value and scene adaptability, as well as the weights corresponding to the popularity value and scene adaptability. Understandably, after determining the popularity and scene suitability of the unrecommended content, a weighted sum of the popularity and scene suitability of the target unrecommended content can be performed according to the weights corresponding to the popularity and scene suitability, to comprehensively calculate the recommendation score for each target unrecommended content. The weight of the popularity value can be set to 0.4, and the weight of the scene suitability can be set to 0.6. When calculating the recommendation score, the scene suitability is usually normalized to the range [0,1].

[0076] Step S405: Based on the recommendation score of the target unrecommended material, select a second candidate recommended material whose recommendation score is greater than a preset score threshold from the target unrecommended material.

[0077] It should be understood that the preset score threshold is a pre-defined threshold for the recommended score, for example, 0.8, without specific limitations. Target unrecommended content with a recommended score greater than the preset threshold is used as the second candidate recommended content, ensuring both content freshness and relevance to the target user's potential needs in the current scenario. If any of the selected second candidate recommended content is identical to the first candidate recommended content, that identical content is removed from the second candidate recommended content. If the final number of selected second candidate recommended content is small (below the minimum selection number, for example, 10), the preset score threshold can be appropriately lowered, and the selection process can be repeated. If the final number of selected second candidate recommended content is large (exceeding the maximum selection number, for example, 20), the second candidate recommended content is sorted in descending order according to the recommended score, and the content with the highest recommended score is selected according to the maximum selection number.

[0078] Step S50: Based on the similarity between the unrecommended material and the core interest material, determine the third candidate recommended material; It should be noted that the third candidate recommended material refers to material identified from the unrecommended materials that is similar to the core interest material. In this embodiment, suitable unrecommended materials are selected as the third candidate recommended materials based on the similarity between the unrecommended materials and the core interest material.

[0079] In one feasible implementation, step S50 may include steps S501 to S504: Step S501: Based on the target feature extraction model, extract the similar feature vectors of the unrecommended materials and the similar feature vectors of the core interest materials. The similar feature vectors include semantic feature vectors, visual feature vectors, and attribute feature vectors. It should be noted that the target feature extraction model, which is used to extract similar feature vectors, has been pre-trained based on similar feature vectors from the target user's historical browsing content. The target feature extraction model can be trained based on a neural network model.

[0080] Understandably, in practice, invalid samples with a dwell time of less than 3 seconds can be removed before feature extraction begins to avoid low-quality samples interfering with similarity determination.

[0081] Additionally, it should be noted that similarity feature vectors, which are features used to measure similarity, include semantic feature vectors, visual feature vectors, and attribute feature vectors. Semantic tag vectors are uniquely encoded based on the introductory keywords of the material (keywords in the description / introduction of the material on the details page), style tags, and theme tags. Visual feature vectors are extracted main color tones and textures. Attribute feature vectors are structured vectors that integrate the material's pricing, carrier type (applicable to cultural and creative product categories), and creator's style.

[0082] Step S502: The similar feature vectors of the unrecommended materials are fused to obtain the fused feature vector of the unrecommended materials, and the similar feature vectors of the core interest materials are fused to obtain the fused feature vector of the core interest materials. Understandably, the fusion of similar feature vectors typically involves assigning corresponding weights to the sub-vectors (semantic feature vector, visual feature vector, and attribute feature vector). For example, the semantic feature vector might have a weight of 1.2, the visual feature vector 1.1, and the attribute feature vector 1.0. Each element in the sub-vector is multiplied by its corresponding weight, and then these elements are concatenated horizontally in sequence to obtain an initial fused vector. This initial fused vector is then normalized so that the magnitude of the entire vector equals 1, resulting in the final fused feature vector. Subsequent similarity calculations will then have a stable similarity range within [0,1], preventing the interference of different dimensional magnitudes on calculation accuracy.

[0083] It should be understood that similar feature vectors of unrecommended materials are fused to obtain fused feature vectors of unrecommended materials, and similar feature vectors of core interest materials are fused to obtain fused feature vectors of core interest materials.

[0084] Step S503: Based on the fusion feature vector of the unrecommended material and the fusion feature vector of the core interest material, determine the similarity between the unrecommended material and the core interest material; It is understood that in this embodiment, the cosine similarity between the unrecommended material and the core interest material is used as the similarity between the unrecommended material and the core interest material.

[0085] Step S504: Based on the similarity between the unrecommended materials and the core interest materials, select a third candidate recommended material from the unrecommended materials whose similarity is greater than a preset similarity threshold.

[0086] It should be understood that the preset similarity threshold is a pre-defined threshold for similarity, such as 0.72, and no specific limitation is made. Unrecommended creatives with a similarity greater than the preset threshold are selected as third-choice recommended creatives. These creatives perfectly match the clearly defined preferences of the target users, which can significantly improve click-through rates. If any of the selected third-choice recommended creatives is identical to the first / second-choice recommended creative, then that identical creative is removed from the third-choice recommended creatives. If the number of third-choice recommended creatives selected is small (below the minimum selection number, e.g., 10), the preset similarity threshold can be appropriately lowered, and the selection process can be repeated. If the number of third-choice recommended creatives selected is large (exceeding the maximum selection number, e.g., 20), then the third-choice recommended creatives are sorted in descending order of similarity, and the creatives with the highest similarity ranking are selected according to the maximum selection number.

[0087] Step S60: Update the material recommendation display area based on the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material.

[0088] Understandably, based on the first, second, and third candidate recommended materials, the materials that need to be recommended now are determined, and the material recommendation display area is updated.

[0089] This embodiment provides a personalized recommendation method for customized cultural and creative materials. When a target user selects recommended materials in the material recommendation display area to browse, the method updates the interactive data of unbrowsed materials, unrecommended materials, historically browsed materials, and historically browsed materials in real time. Based on the historically browsed materials and their interactive data, the method determines the target user's core interest tags and core interest materials. Based on the matching degree between the tags of the unbrowsed materials and the core interest tags, the method determines the first candidate recommended materials. Based on the popularity value and scene adaptability of the unrecommended materials, the method determines the second candidate recommended materials. Based on the similarity between the unrecommended materials and the core interest materials, the method determines the third candidate recommended materials. Based on the first, second, and third candidate recommended materials, the material recommendation display area is updated. This embodiment can predict two types of potential interest materials. It predicts materials that the user may browse in the future from those previously recommended but not currently viewed by the user, and uses them as the first type of recommendation candidate materials. It also predicts materials that the user may browse in the future from those that have not been recommended to the user, and uses them as the second type of recommendation candidate materials. Combined with the third type of recommendation candidate materials corresponding to the user's historical preferences, a personalized recommendation list is generated. By fully considering the user's known interests and potential interests, it can recommend suitable materials to the user more comprehensively and accurately, improve the probability of material discovery and adoption, and enhance the overall utilization rate of materials on the customized platform.

[0090] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 Step S60 may include steps S601 to S603: Step S601: Determine the target recommendation material based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material; It should be noted that the target recommended material is the material that is finally determined to be recommended at this time.

[0091] In one feasible implementation, step S601 may include steps S6011 to S6013: Step S6011: Obtain the target user's historical interaction behavior in the material recommendation display area. Based on the historical interaction behavior, assign corresponding recommendation ratios to the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material. The historical interaction behavior includes browsing type, number of consecutive selections, and number of valid browsings. Understandably, historical interaction behavior refers to the target user's recent actions, including browsing type, number of consecutive selections, and number of effective browsings. Browsing type refers to which category (first, second, or third candidate recommended material) the target user clicked on belongs to. Number of consecutive selections refers to the number of consecutive clicks the target user made on the first, second, or third candidate recommended material. Number of effective browsings refers to the number of times the target user effectively browsed the first, second, or third candidate recommended material (dwell time greater than or equal to 3 seconds).

[0092] It should be understood that the recommendation ratio refers to the ratio of the number of first-candidate recommended materials, second-candidate recommended materials, and third-candidate recommended materials that can be displayed in the material recommendation display area to the total number of displayed materials. In this embodiment, the recommendation ratios of the first-candidate recommended materials, second-candidate recommended materials, and third-candidate recommended materials are determined based on the target user's historical interaction behavior.

[0093] For example, the recommendation ratio of the first and second candidate recommended materials is first initialized to 30%, and the recommendation ratio of the third candidate recommended material is initialized to 40%. If the target user clicks on the first candidate recommended material three times in a row, the recommendation ratio of the first candidate recommended material is increased by 10%, while the recommendation ratios of the second and third candidate recommended materials are decreased by 5%. If the target user skips the second type of candidate material twice in a row, the recommendation ratio of the second candidate recommended material is decreased by 10%, while the recommendation ratios of the first and third candidate recommended materials are increased by 5%. If the target user clicks on the third type of candidate material twice in a row, the recommendation ratio of the third candidate recommended material is increased by 10%, while the recommendation ratios of the first and second candidate recommended materials are decreased by 5%.

[0094] In addition, the proportion of materials from the same creator and on the same theme in each category of candidate materials can be limited to no more than 20%.

[0095] Step S6012: Based on the preset recommendation quantity and the recommendation ratio, determine the recommendation quantity of the first candidate recommendation material, the second candidate recommendation material and the third candidate recommendation material respectively; It should be noted that the preset recommendation quantity is the total number of materials that can be displayed in the material recommendation display area, while the recommendation quantity is the number of materials that can be displayed in the material recommendation display area. Multiplying the preset recommendation quantity by the recommendation ratio allows you to calculate the recommendation quantity of the first, second, and third candidate recommended materials.

[0096] Step S6013: Based on the number of recommendations of the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material, select the target recommendation material from the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material.

[0097] Understandably, based on the number of recommendations for the first, second, and third candidate recommended materials, the top-ranked materials from each category are selected as the target recommended materials.

[0098] Step S602: Based on the recommendation type of the target recommendation material, determine the personalized display strategy of the target recommendation material. The personalized display strategy includes display position, visual tags, and annotation colors. It should be noted that if the target recommended material is the first candidate recommended material, the recommendation type is type one; if the target recommended material is the second candidate recommended material, the recommendation type is type two; and if the target recommended material is the third candidate recommended material, the recommendation type is type three.

[0099] As can be understood, a personalized display strategy is a personalized display method. A personalized display strategy includes display location, visual labels, and annotation colors. Display location refers to the position of the content within the recommended display area; visual labels refer to prominent category labels; and annotation colors refer to the colors used for annotation.

[0100] Step S603: Update the material recommendation display area based on the target recommended material and the personalized display strategy of the target recommended material.

[0101] For example, the material recommendation display area is divided into three zones. The first zone is the "You Might Have Missed" zone, displaying the first type of target recommendation materials. A clock icon or a "Take Another Look" prompt can be used as a visual label, and the area is marked in blue. The second zone is the "New Try" or "Discover and Explore" zone, displaying the second type of target recommendation materials. A rocket icon or a "New" icon can be used as a visual label, and the area is marked in yellow. The third zone is the "Based on Your Preferences" zone, displaying the third type of target recommendation materials. A heart icon or a crown icon can be used as a visual label, and the area is marked in green. The material recommendation display area can be divided into top, center, and bottom sections, or left, center, and right sections. In practice, the display position, visual labels, and marking colors for each type of target recommendation material can be adjusted accordingly. After determining the personalized display strategy for the target recommendation materials, the material recommendation display area is updated.

[0102] This embodiment provides a personalized recommendation method for customized cultural and creative materials. Based on first, second, and third candidate recommended materials, a target recommended material is determined. Based on the recommendation type of the target recommended material, a personalized display strategy is determined, including display position, visual tags, and highlighting colors. Based on the target recommended material and its personalized display strategy, the material recommendation display area is updated. Through special visual highlighting methods, users can easily identify different types of recommended materials, effectively guiding user browsing behavior, increasing user browsing willingness, and extending the time users spend on the customization platform.

[0103] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the personalized recommendation method for cultural and creative customized materials in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0104] This application also provides a personalized recommendation system for customized cultural and creative materials. Please refer to [link / reference]. Figure 3 The personalized recommendation system for customized cultural and creative materials includes: The data update module 10 is used to update the recommended unviewed materials, unrecommended materials, historically viewed materials, and the interaction data of the historically viewed materials in real time when the target user selects recommended materials in the material recommendation display area for browsing. The material filtering module 20 is used to determine the core interest tags and core interest materials of the target user based on the historical browsing materials and the interaction data of the historical browsing materials; The material filtering module 20 is also used to determine the first candidate recommended material based on the matching degree between the tags of the recommended unviewed materials and the core interest tags; The material filtering module 20 is also used to determine a second candidate recommended material based on the popularity value and scene adaptability of the unrecommended material; The material filtering module 20 is also used to determine a third candidate recommended material based on the similarity between the unrecommended material and the core interest material; The material recommendation module 30 is used to update the material recommendation display area based on the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material.

[0105] In one feasible implementation, the material filtering module 20 is further configured to determine valid sample materials from the historically viewed materials based on the interaction data of the historically viewed materials, wherein the interaction data includes the number of views, dwell time, collection status, and number of shares; Based on the interaction data of the effective sample materials, calculate the interest score of the effective sample materials; The interest score of the effective sample material is used as the interest score corresponding to the tag of the effective sample material. Based on the interest score corresponding to the tag of the effective sample material, the core interest tags of the target user are selected. Based on the interest scores of the valid sample materials, the core interest materials of the target user are determined from the valid sample materials.

[0106] In one feasible implementation, the material filtering module 20 is further configured to obtain a first correspondence between the number of views, dwell time, collection status, number of shares and initial interest score, and determine the initial interest score of the effective sample material based on the interaction data of the effective sample material and the first correspondence. Based on the most recent interaction time corresponding to the interaction data of the effective sample material, determine the time decay factor of the effective sample material; The initial interest score of the effective sample material is corrected based on the time decay factor of the effective sample material to obtain the interest score of the effective sample material.

[0107] In one feasible implementation, the material filtering module 20 is further configured to determine the overlapping tags of the recommended unviewed materials based on the tags of the recommended unviewed materials and the core interest tags; Obtain the second correspondence between the number of overlapping tags, the matching weight of overlapping tags, and the initial matching degree. Based on the second correspondence and the number and matching weight of overlapping tags corresponding to the recommended unviewed material, obtain the initial matching degree between the tags of the recommended unviewed material and the core interest tags. Obtain the recommended materials in the same batch corresponding to the recommended unviewed materials in the historical browsing materials, and determine the bias correction ratio based on the browsing ratio of the same type of materials in the same batch corresponding to the recommended unviewed materials; Based on the bias correction ratio corresponding to the recommended unviewed material, the initial matching degree between the tags of the recommended unviewed material and the core interest tags is corrected to obtain the matching degree between the tags of the recommended unviewed material and the core interest tags; Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, the first candidate recommended materials with a matching degree greater than a preset matching threshold are selected from the recommended unviewed materials.

[0108] In one feasible implementation, the material filtering module 20 is further configured to determine the first material to be removed based on the degree of theme repetition between the unrecommended material and the historically viewed material; Based on the tag overlap between the unrecommended material and the first candidate recommended material, a second material to be removed is determined; Remove the first and second rejected materials from the unrecommended materials to obtain the target unrecommended materials; Based on the popularity value and scene suitability of the target unrecommended material, as well as the weights corresponding to the popularity value and scene suitability, the recommendation score of the target unrecommended material is calculated; Based on the recommendation score of the target unrecommended material, a second candidate recommended material with a recommendation score greater than a preset score threshold is selected from the target unrecommended material.

[0109] In one feasible implementation, the material filtering module 20 is further used to extract the popularity index of the unrecommended materials, the popularity index including the number of clicks, the complete view rate, the percentage of favorites, and the positive evaluation rate within a preset time period; Based on the popularity index of the unrecommended material and the popularity weight corresponding to the popularity index, the initial popularity value of the unrecommended material is determined; Based on the release time of the unrecommended material, the initial popularity value of the unrecommended material is adjusted to obtain the popularity value of the unrecommended material; Based on the real-time customized scenario of the target user, the adaptation score of the unrecommended material in the real-time customized scenario is determined. The real-time customized scenario includes time-dimensional scenario, location-dimensional scenario, and device-dimensional scenario. The scene adaptation degree of the unrecommended material is determined based on the adaptation score of the unrecommended material in the real-time customized scene and the adaptation weight of the real-time customized scene.

[0110] In one feasible implementation, the material screening module 20 is further configured to extract similar feature vectors of the unrecommended materials and similar feature vectors of the core interest materials based on the target feature extraction model. The similar feature vectors include semantic feature vectors, visual feature vectors, and attribute feature vectors. The similar feature vectors of the unrecommended materials are fused to obtain the fused feature vector of the unrecommended materials, and the similar feature vectors of the core interest materials are fused to obtain the fused feature vector of the core interest materials. Based on the fusion feature vector of the unrecommended material and the fusion feature vector of the core interest material, the similarity between the unrecommended material and the core interest material is determined; Based on the similarity between the unrecommended materials and the core interest materials, a third candidate recommended material with a similarity greater than a preset similarity threshold is selected from the unrecommended materials.

[0111] In one feasible implementation, the material recommendation module 30 is further configured to determine a target recommendation material based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material; Based on the recommendation type of the target recommended material, a personalized display strategy for the target recommended material is determined. The personalized display strategy includes display position, visual tags, and annotation colors. Based on the target recommended materials and the personalized display strategy of the target recommended materials, the material recommendation display area is updated.

[0112] In one feasible implementation, the material recommendation module 30 is further configured to acquire the target user's historical interaction behavior in the material recommendation display area, and based on the historical interaction behavior, assign corresponding recommendation ratios to the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material, respectively. The historical interaction behavior includes browsing type, number of consecutive selections, and number of valid browsings. Based on the preset recommendation quantity and the recommendation ratio, the recommendation quantities of the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material are determined respectively; Based on the number of recommendations for the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material, a target recommendation material is selected from the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material.

[0113] The personalized recommendation system for customized cultural and creative materials provided in this application adopts the personalized recommendation method for customized cultural and creative materials in the above embodiments. It can solve the technical problem that traditional recommendation schemes rely solely on users' historical behavior, excessively depend on known interests, ignore users' potential interest needs, and result in monotonous recommendation results. Compared with the prior art, the beneficial effects of the personalized recommendation system for customized cultural and creative materials provided in this application are the same as those of the personalized recommendation method for customized cultural and creative materials provided in the above embodiments, and other technical features of the personalized recommendation system for customized cultural and creative materials are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0114] This application provides a personalized recommendation device for customized cultural and creative materials. The device includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the personalized recommendation method for customized cultural and creative materials in Embodiment 1 above.

[0115] The following is for reference. Figure 4The diagram illustrates a structural schematic suitable for implementing a personalized recommendation device for customized cultural and creative materials in the embodiments of this application. The personalized recommendation device for customized cultural and creative materials in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 4 The illustrated personalized recommendation device for customized cultural and creative materials is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0116] like Figure 4 As shown, the personalized recommendation device for customized cultural and creative materials may include a processing system 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in ROM (Read Only Memory) 1002 or the program loaded from storage system 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the personalized recommendation device for customized cultural and creative materials. The processing system 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input systems 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output systems 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage systems 1003 including, for example, magnetic tapes, hard disks, etc.; and communication systems 1009. Communication system 1009 allows the personalized recommendation device for creative and cultural materials to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a personalized recommendation device for creative and cultural materials with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0117] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication system, or installed from storage system 1003, or installed from ROM 1002. When the computer program is executed by processing system 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0118] The personalized recommendation device for customized cultural and creative materials provided in this application adopts the personalized recommendation method for customized cultural and creative materials in the above embodiments. It can solve the technical problem that traditional recommendation schemes rely solely on users' historical behavior, excessively depend on known interests, ignore users' potential interest needs, and result in monotonous recommendation results. Compared with the prior art, the beneficial effects of the personalized recommendation device for customized cultural and creative materials provided in this application are the same as those of the personalized recommendation method for customized cultural and creative materials provided in the above embodiments, and other technical features in this personalized recommendation device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0119] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0120] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0121] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the personalized recommendation method for cultural and creative customized materials in the above embodiments.

[0122] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0123] The aforementioned computer-readable storage medium may be included in the personalized recommendation device for customized cultural and creative materials; or it may exist independently and not be assembled into the personalized recommendation device for customized cultural and creative materials.

[0124] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the personalized recommendation device for cultural and creative customized materials, the personalized recommendation device for cultural and creative customized materials: when a target user selects and browses recommended materials in the material recommendation display area, it updates in real time the recommended unviewed materials, unrecommended materials, historically viewed materials, and interactive data of historically viewed materials; based on historically viewed materials and their interactive data, it determines the target user's core interest tags and core interest materials; based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, it determines the first candidate recommended materials; based on the popularity value and scene adaptability of the unrecommended materials, it determines the second candidate recommended materials; based on the similarity between the unrecommended materials and the core interest materials, it determines the third candidate recommended materials; and based on the first, second, and third candidate recommended materials, it updates the material recommendation display area.

[0125] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0128] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described personalized recommendation method for customized cultural and creative materials. This solves the technical problem that traditional recommendation schemes rely solely on users' historical behavior, excessively depend on known interests, ignore potential user interests, and result in monotonous recommendation results. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the personalized recommendation method for customized cultural and creative materials provided in the above embodiments, and will not be elaborated upon here.

[0129] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the personalized recommendation method for customized cultural and creative materials as described above.

[0130] The computer program product provided in this application can solve the technical problem that traditional recommendation schemes rely solely on users' historical behavior, excessively depend on known interests, ignore users' potential interest needs, and result in monotonous recommendation results. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the personalized recommendation method for customized cultural and creative materials provided in the above embodiments, and will not be repeated here.

[0131] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

[0132] Furthermore, all user-related data involved in this application (e.g., historical browsing materials, interaction data, historical interaction behavior, real-time customized scenarios, etc.) were obtained with the user's permission or consent. In other words, when this application is applied to specific products or technologies, user permission is required to acquire and process the relevant data, and the processing of the data must comply with the relevant laws, regulations, and regulatory standards of the relevant countries and regions. For example, when it is necessary to obtain the user's current geographical location, a location acquisition prompt can be displayed on the user's terminal. After receiving confirmation from the user regarding the location acquisition prompt, the terminal can obtain the user's current geographical location.

Claims

1. A method for personalized recommendation of cultural and creative customized materials, characterized in that, The method includes: When a target user selects and browses recommended materials in the material recommendation display area, the system updates the recommended unviewed materials, unrecommended materials, historically viewed materials, and the interaction data of the historically viewed materials in real time. Based on the historical browsing materials and the interaction data of the historical browsing materials, the core interest tags and core interest materials of the target user are determined; Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, a first candidate recommended material is determined, specifically including: based on the tags of the recommended unviewed materials and the core interest tags, determining the overlapping tags of the recommended unviewed materials; obtaining a second correspondence between the number of overlapping tags, the matching weight of overlapping tags, and the initial matching degree; based on the second correspondence and the number and matching weight of overlapping tags corresponding to the recommended unviewed materials, obtaining the initial matching degree between the tags of the recommended unviewed materials and the core interest tags; obtaining the same batch of recommended materials corresponding to the recommended unviewed materials in the historical viewed materials; determining a bias correction ratio based on the viewing ratio of similar materials in the same batch of recommended materials corresponding to the recommended unviewed materials; correcting the initial matching degree between the tags of the recommended unviewed materials and the core interest tags based on the bias correction ratio corresponding to the recommended unviewed materials, obtaining the matching degree between the tags of the recommended unviewed materials and the core interest tags; and filtering out the first candidate recommended materials with a matching degree greater than a preset matching threshold from the recommended unviewed materials based on the matching degree between the tags of the recommended unviewed materials and the core interest tags. Based on the popularity value and scene adaptability of the unrecommended material, a second candidate recommended material is determined. Specifically, based on the real-time customized scene of the target user, the adaptation score of the unrecommended material in the real-time customized scene is determined. The real-time customized scene includes time-dimensional scene, location-dimensional scene, and device-dimensional scene. Based on the adaptation score of the unrecommended material in the real-time customized scene and the adaptation weight of the real-time customized scene, the scene adaptability of the unrecommended material is determined. Based on the similarity between the unrecommended materials and the core interest materials, a third candidate recommended material is determined; The material recommendation display area is updated based on the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material.

2. The method as described in claim 1, characterized in that, The step of determining the target user's core interest tags and core interest materials based on the historical browsing materials and the interaction data of the historical browsing materials includes: Based on the interaction data of the historical browsing materials, valid sample materials are determined from the historical browsing materials. The interaction data includes the number of times it is viewed, the duration of stay, the collection status, and the number of times it is shared. Based on the interaction data of the effective sample materials, calculate the interest score of the effective sample materials; The interest score of the effective sample material is used as the interest score corresponding to the tag of the effective sample material. Based on the interest score corresponding to the tag of the effective sample material, the core interest tags of the target user are selected. Based on the interest scores of the valid sample materials, the core interest materials of the target user are determined from the valid sample materials.

3. The method as described in claim 2, characterized in that, The step of calculating the interest score of the effective sample material based on the interaction data of the effective sample material includes: Obtain the first correspondence between the number of views, dwell time, collection status, number of shares and the initial interest score, and determine the initial interest score of the effective sample material based on the interaction data of the effective sample material and the first correspondence. Based on the most recent interaction time corresponding to the interaction data of the effective sample material, determine the time decay factor of the effective sample material; The initial interest score of the effective sample material is corrected based on the time decay factor of the effective sample material to obtain the interest score of the effective sample material.

4. The method as described in claim 1, characterized in that, The step of determining the second candidate recommended material based on the popularity value and scene suitability of the unrecommended material includes: Based on the degree of thematic repetition between the unrecommended materials and the historically viewed materials, the first materials to be removed are determined; Based on the tag overlap between the unrecommended material and the first candidate recommended material, a second material to be removed is determined; Remove the first and second rejected materials from the unrecommended materials to obtain the target unrecommended materials; Based on the popularity value and scene suitability of the target unrecommended material, as well as the weights corresponding to the popularity value and scene suitability, the recommendation score of the target unrecommended material is calculated; Based on the recommendation score of the target unrecommended material, a second candidate recommended material with a recommendation score greater than a preset score threshold is selected from the target unrecommended material.

5. The method as described in claim 1, characterized in that, Before the step of determining the second candidate recommended material based on the popularity value and scene suitability of the unrecommended material, the method further includes: Extract the popularity metrics of the unrecommended materials, including the number of clicks, complete views rate, percentage of favorites, and positive review rate within a preset time period; Based on the popularity index of the unrecommended material and the popularity weight corresponding to the popularity index, the initial popularity value of the unrecommended material is determined; Based on the release time of the unrecommended material, the initial popularity value of the unrecommended material is adjusted to obtain the popularity value of the unrecommended material.

6. The method as described in claim 1, characterized in that, The step of determining the third candidate recommended material based on the similarity between the unrecommended material and the core interest material includes: Based on the target feature extraction model, similar feature vectors of the unrecommended materials and similar feature vectors of the core interest materials are extracted. The similar feature vectors include semantic feature vectors, visual feature vectors, and attribute feature vectors. The similar feature vectors of the unrecommended materials are fused to obtain the fused feature vector of the unrecommended materials, and the similar feature vectors of the core interest materials are fused to obtain the fused feature vector of the core interest materials. Based on the fusion feature vector of the unrecommended material and the fusion feature vector of the core interest material, the similarity between the unrecommended material and the core interest material is determined; Based on the similarity between the unrecommended materials and the core interest materials, a third candidate recommended material with a similarity greater than a preset similarity threshold is selected from the unrecommended materials.

7. The method as described in claim 1, characterized in that, The step of updating the material recommendation display area based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material includes: Based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material, the target recommendation material is determined; Based on the recommendation type of the target recommended material, a personalized display strategy for the target recommended material is determined. The personalized display strategy includes display position, visual tags, and annotation colors. Based on the target recommended materials and the personalized display strategy of the target recommended materials, the material recommendation display area is updated.

8. The method as described in claim 7, characterized in that, The step of determining the target recommendation material based on the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material includes: The target user's historical interaction behavior in the material recommendation display area is obtained. Based on the historical interaction behavior, a corresponding recommendation ratio is assigned to the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material. The historical interaction behavior includes browsing type, number of consecutive selections, and number of valid browsings. Based on the preset recommendation quantity and the recommendation ratio, the recommendation quantities of the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material are determined respectively; Based on the number of recommendations for the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material, a target recommendation material is selected from the first candidate recommendation material, the second candidate recommendation material, and the third candidate recommendation material.

9. A personalized recommendation system for customized cultural and creative materials, characterized in that, The system includes: The data update module is used to update the recommended materials, unviewed materials, unrecommended materials, historically viewed materials, and interactive data of the historically viewed materials in real time when the target user selects and browses the recommended materials in the material recommendation display area. The material filtering module is used to determine the core interest tags and core interest materials of the target user based on the historical browsing materials and the interaction data of the historical browsing materials; The material filtering module is also used to determine the first candidate recommended material based on the matching degree between the tags of the recommended unviewed materials and the core interest tags; The material filtering module is also used to determine a second candidate recommended material based on the popularity value and scene adaptability of the unrecommended material; The material filtering module is also used to determine a third candidate recommended material based on the similarity between the unrecommended material and the core interest material; The material recommendation module is used to update the material recommendation display area based on the first candidate recommended material, the second candidate recommended material, and the third candidate recommended material; The material filtering module is also used to determine the overlapping tags of the recommended unviewed materials based on the tags of the recommended unviewed materials and the core interest tags; Obtain the second correspondence between the number of overlapping tags, the matching weight of overlapping tags, and the initial matching degree. Based on the second correspondence and the number and matching weight of overlapping tags corresponding to the recommended unviewed material, obtain the initial matching degree between the tags of the recommended unviewed material and the core interest tags. Obtain the recommended materials in the same batch corresponding to the recommended unviewed materials in the historical browsing materials, and determine the bias correction ratio based on the browsing ratio of the same type of materials in the same batch corresponding to the recommended unviewed materials; Based on the bias correction ratio corresponding to the recommended unviewed material, the initial matching degree between the tags of the recommended unviewed material and the core interest tags is corrected to obtain the matching degree between the tags of the recommended unviewed material and the core interest tags; Based on the matching degree between the tags of the recommended unviewed materials and the core interest tags, the first candidate recommended materials with a matching degree greater than a preset matching threshold are selected from the recommended unviewed materials; The material filtering module is also used to determine the adaptation score of the unrecommended material in the real-time customized scenario based on the real-time customized scenario of the target user. The real-time customized scenario includes time-dimensional scenario, location-dimensional scenario and device-dimensional scenario. The scene adaptation degree of the unrecommended material is determined based on the adaptation score of the unrecommended material in the real-time customized scene and the adaptation weight of the real-time customized scene.

Images