Content recommendation method and device based on user interest, equipment and storage medium

Abstract

The present disclosure provides a content recommendation method and device based on user interest, equipment and storage medium, relates to the field of artificial intelligence, in particular to the field of large model and intelligent recommendation. The specific implementation scheme is: obtaining current scene information and a pre-stored interest model; wherein the current scene information represents the space-time scene in which the user is currently located, and the interest model is a user portrait model representing the field of content that the user is interested in; updating the interest model according to the current scene information to obtain an updated interest model; determining the content to be recommended according to the updated interest model and the device type of the terminal device currently used by the user, and pushing the content to be recommended to the user. The accuracy of content recommendation is improved, and the user experience is improved.

Description

Technical Field

[0001] This disclosure relates to large models and intelligent recommendation in the field of artificial intelligence, and in particular to a method, apparatus, device and storage medium for content recommendation based on user interests. Background Technology

[0002] Users' interests change over time and in different contexts; depending on their interests, users will want to watch different content.

[0003] Traditional content recommendation models analyze user needs based on static historical behavior, which often cannot flexibly respond to shifts in user interests, resulting in low accuracy of content recommendations and impacting user experience. Summary of the Invention

[0004] This disclosure provides a method, apparatus, device, and storage medium for content recommendation based on user interests.

[0005] According to a first aspect of this disclosure, a content recommendation method based on user interests is provided, comprising:

[0006] Acquire current scene information and pre-stored interest models; wherein, the current scene information represents the spatiotemporal scene in which the user is currently located, and the interest model is a user profile model representing the domain of content that the user is interested in;

[0007] Based on the current scene information, the interest model is updated to obtain the updated interest model;

[0008] Based on the updated interest model and the device type of the user's current terminal device, the content to be recommended is determined and pushed to the user.

[0009] According to a second aspect of this disclosure, a content recommendation device based on user interests is provided, comprising:

[0010] The acquisition unit is used to acquire current scene information and pre-stored interest models; wherein, the current scene information represents the spatiotemporal scene in which the user is currently located, and the interest model is a user profile model that represents the domain of content that the user is interested in;

[0011] An update unit is used to update the interest model based on the current scene information to obtain an updated interest model.

[0012] The recommendation unit is used to determine the content to be recommended based on the updated interest model and the device type of the terminal device currently used by the user, and to push the content to be recommended to the user.

[0013] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0014] At least one processor; and

[0015] A memory that is communicatively connected to the at least one processor;

[0016] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect of this disclosure.

[0017] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the method described in accordance with a first aspect of this disclosure.

[0018] According to a fifth aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect of this disclosure.

[0019] The technology disclosed herein improves the accuracy of content recommendation and enhances user experience.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0021] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0022] Figure 1 This is a flowchart illustrating a content recommendation method based on user interests provided according to an embodiment of this disclosure;

[0023] Figure 2 This is a schematic diagram illustrating the optimization process of the interest model provided according to embodiments of this disclosure;

[0024] Figure 3 This is a flowchart illustrating a content recommendation method based on user interests provided according to an embodiment of this disclosure;

[0025] Figure 4 This is a schematic diagram illustrating the updating process of the interest model according to embodiments of this disclosure;

[0026] Figure 5 This is a flowchart illustrating a content recommendation method based on user interests provided according to an embodiment of this disclosure;

[0027] Figure 6This is a schematic diagram of a device type-based content recommendation process provided according to an embodiment of this disclosure;

[0028] Figure 7 This is a flowchart illustrating a content recommendation method based on user interests provided according to an embodiment of this disclosure;

[0029] Figure 8 This is a structural block diagram of a content recommendation device based on user interests provided according to an embodiment of the present disclosure;

[0030] Figure 9 This is a structural block diagram of a content recommendation device based on user interests provided according to an embodiment of the present disclosure;

[0031] Figure 10 This is a block diagram of an electronic device used to implement the user interest-based content recommendation method of the embodiments of this disclosure;

[0032] Figure 11 This is a block diagram of an electronic device used to implement the user interest-based content recommendation method of the embodiments of this disclosure. Detailed Implementation

[0033] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0034] With the development of artificial intelligence technology, when users use terminal devices, it is possible to predict the content that users may be interested in and automatically recommend content to them to improve the user experience. However, user interests change with time and space, and traditional recommendation methods cannot flexibly cope with the instantaneous shifts in user interests.

[0035] Currently, many recommendation systems analyze user needs based on historical user behavior, often failing to accurately reflect real-time shifts in user interests. If a user's interests change rapidly within a short period, traditional methods cannot capture this quickly. For example, current recommendation systems struggle to handle changes in interests caused by frequent switching between different times, locations, and devices, reducing the accuracy of content recommendations and impacting user experience.

[0036] This disclosure provides a content recommendation method, apparatus, device, and storage medium based on user interests, applicable to large models and intelligent recommendation fields in the field of artificial intelligence, to improve the accuracy and flexibility of content recommendation and enhance user experience.

[0037] It should be noted that the data in this embodiment is not specific to any particular user and does not reflect the personal information of any particular user. It should also be noted that the data in this embodiment comes from a publicly available dataset.

[0038] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0039] To help readers gain a deeper understanding of the implementation principles of this disclosure, the following will be discussed in conjunction with... Figures 1-11 The embodiments are further refined.

[0040] Figure 1 This is a flowchart illustrating a content recommendation method based on user interests according to an embodiment of this disclosure. This method can be executed by a content recommendation device based on user interests. Figure 1 As shown, the method includes the following steps:

[0041] S101. Obtain current scene information and pre-stored interest model; wherein, current scene information represents the spatiotemporal scene in which the user is currently located, and interest model is a user profile model representing the domain of content that the user is interested in.

[0042] For example, users can browse various content of interest using a terminal device, such as articles, images, and videos. Each user has their own interest model, which is a user profile model that represents the areas of content the user is interested in. User profiles can be built based on information such as the user's location, occupation, and browsing history. For example, if a user is a software engineer, their interest model could include technical areas related to software development. Storing user interest models allows for easy retrieval and targeted content recommendations.

[0043] It can acquire the user's current scene information in real time. This current scene information represents the user's current spatiotemporal context, that is, it determines the user's current time and space context. For example, the current time context might be a weekday, and the space context might be an office. In other words, it can acquire the user's current geographical location information and time information, and determine the user's current geographical location information and time information as the current scene information.

[0044] S102. Update the interest model based on the current scene information to obtain the updated interest model.

[0045] For example, the current scene information may change at any time. Based on the current scene information, the currently acquired interest model can be updated in real time to obtain an updated interest model. The updated interest model can be stored, meaning that each time an interest model is acquired and stored, it is the latest interest model.

[0046] Based on current scene information, the system can predict the areas a user might be interested in, and update the existing areas in the interest model accordingly. For example, if the predicted area is not already in the interest model, it can be added to obtain an updated model. Specifically, if the current scene information indicates that the user is off-get off work and at home, it can be predicted that the user's current area of ​​interest is leisure and entertainment, rather than technology. Therefore, a leisure and entertainment area can be added to the interest model to facilitate the recommendation of leisure and entertainment content.

[0047] Different scenarios can be pre-associated with different domains. Based on the current scenario information, the system can find domains related to the current scenario and identify them as areas the user might be interested in. Alternatively, it can predict the user's current areas of interest based on the domains of content the user browsed within the same historical timeframe within the current scenario. For example, the domains of content the user browsed in the same scenario can be used as the user's areas of interest in that scenario.

[0048] In this embodiment, the method further includes: acquiring the user's historical behavior data; wherein, the historical behavior data includes the user's behavior data within a preset first time period and the user's behavior data within a preset second time period, the preset first time period being longer than the preset second time period, and the behavior data representing the user's operation behavior and operation content; and updating the interest model based on the historical behavior data to obtain the updated interest model.

[0049] Specifically, in addition to updating the interest model based on the current scene information, it can also obtain the user's historical behavior data and update the interest model based on the user's historical behavior data.

[0050] The system pre-sets a first time period and a second time period. The first time period can be longer than the second time period; for example, the first time period might cover the past year, and the second time period might cover the past month. In other words, the first time period can span a longer period than the second time period, and the first time period can include the second time period. The historical behavioral data corresponding to the first time period represents the user's long-term behavioral data, while the historical behavioral data corresponding to the second time period represents the user's short-term behavioral data. For example, if the first time period is the past year and the second time period is the past month, then all behavioral data from the user within the past year and the past month will be retrieved.

[0051] Behavioral data can characterize user actions and the content of those actions. Actions can include browsing web pages, clicking videos, searching, saving, liking, etc. The content of actions refers to the objects on which the actions are performed, such as the web pages the user browses or the posts the user saves.

[0052] Based on acquired historical behavioral data, the system predicts areas of potential user interest, updates the current interest model, and stores the updated model. For example, if a user frequently views fitness videos within a month, short-term behavioral data is sufficient to determine their interest in fitness, eliminating the need to process long-term behavioral data and reducing computational load. Conversely, if a user views fitness videos three times a month, short-term behavioral data alone is insufficient to determine their interest; combining long-term behavioral data accurately identifies browsing patterns, allowing the fitness category to be added to the interest model.

[0053] The advantage of this setup is that by acquiring behavioral data over both short and long periods, it is possible to efficiently and accurately predict the areas of interest that users are interested in, thereby improving the accuracy of interest models, which in turn improves the accuracy of content recommendations and enhances the user experience.

[0054] In this embodiment, the interest model includes multiple keywords and the weights corresponding to each keyword. Keywords represent domains, and weights represent the degree of user interest. The interest model is updated based on historical behavior data to obtain an updated interest model, including: extracting keywords from historical behavior data to obtain keywords corresponding to the historical behavior data, which are historical words; and updating the interest model based on the historical words to obtain an updated interest model.

[0055] Specifically, users can be interested in multiple fields; that is, the interest model can represent multiple fields. Different fields can correspond to different keywords, and the interest model can include multiple keywords to represent the fields. For example, the keyword for the fitness field is "fitness." Each keyword in the interest model can also have its own weight, which represents the degree of user interest in the field; the higher the weight, the greater the degree of interest.

[0056] Once historical behavioral data is obtained, whether it's long-term or short-term, keywords can be extracted. For example, a user's web browsing history can be retrieved, keywords extracted from the content of the web pages read, and reading time can be captured to infer user interests and preferences. User click and interaction records can also be retrieved; by combining user likes, shares, and page exits, the degree of interest in specific content can be statistically analyzed, and keywords from that specific content can be extracted. Furthermore, a user's search history can be retrieved, and keywords searched by the user at specific times or locations can be analyzed as extracted keywords. Keywords extracted from historical behavioral data are then designated as historical terms.

[0057] Large models can be used to continuously track users' historical behavioral data, such as web browsing, video clicks, keyword searches, collections and likes. By deeply analyzing this data, historical words can be extracted through the embedding vectors of text and video.

[0058] Specifically, a deep learning model in NLP (Natural Language Processing) can be pre-built for keyword extraction; for example, the BERT (Bidirectional Encoder Representations from Transformers) model can be used. Historical behavioral data can be input into the BERT model; for example, articles visited or content clicked by the user can be input into the model to extract keywords from this content. In this embodiment, the update of the interest model can use a hybrid architecture, specifically, an architecture combining deep learning and temporal tracking. For example, keywords can be extracted from textual or multimodal data using a large language model based on BERT; LSTM (Long Short-Term Memory) can be used to dynamically monitor short-term behavioral shifts; and TF-IDF (Term Frequency-Inverse Document Frequency) and collaborative filtering algorithms, combined with long-term user behavioral data, can achieve full-cycle coverage of the user profile. In this embodiment, the specific model architecture of the large model is not limited.

[0059] When the large model learns over different time periods and discovers that users' areas of interest are changing, the weights of the corresponding keywords can be gradually reduced, and newly generated keywords can be added to update the model in a timely manner. In other words, after obtaining historical keywords, the interest model can be updated based on the existing keywords. For example, historical keywords can be added to the interest model, or the weights of existing keywords in the interest model can be updated.

[0060] The beneficial effects of this setup are that it dynamically tracks users' short-term interest shifts using LSTM, allowing for flexible adjustments to the keyword weights in the interest model; it uses TF-IDF and collaborative filtering algorithms, combined with browsing history, search behavior, and favorites behavior, to update the interest model and uncover users' potential long-term interests. This achieves comprehensive confirmation of both short- and long-term interests through large models and natural language processing techniques, improving the accuracy of the interest model and consequently enhancing the accuracy of content recommendations.

[0061] In this embodiment, the interest model is updated based on historical words to obtain an updated interest model. This includes: if historical words exist in the interest model, the weights corresponding to the historical words in the interest model are adjusted to obtain an updated interest model.

[0062] Specifically, the system determines whether historical terms exist in the interest model. If so, it recommends content related to those terms based on the interest model. To enable users to access content in that area more quickly, the weights of historical terms in the interest model can be adjusted, for example, by increasing their weight. When recommending content, content from areas with higher weights is prioritized.

[0063] The advantage of this setup is that adjusting the weights in the interest model does not require rebuilding the interest model, thus improving adjustment efficiency and the accuracy of content recommendation.

[0064] In this embodiment, the interest model is updated based on historical words to obtain an updated interest model. This includes: if a historical word does not exist in the interest model, the historical word is added to the interest model, and a preset initial weight is assigned to the historical word to obtain an updated interest model.

[0065] Specifically, if it is determined that no historical words have been extracted from the interest model, then the interest model is considered unable to recommend content related to the historical words to the user. Historical words can then be added as new keywords to the interest model.

[0066] Each keyword in the interest model has its own weight. Therefore, it's necessary to assign weights to historical words added to the interest model. A preset weight can be assigned as the initial weight, which can be adjusted during the interest model's updates. Alternatively, historical words can be weighted based on their frequency of occurrence in historical behavioral data; for example, higher frequency results in a higher weight, thus prioritizing content from the areas associated with those historical words when recommending content to users.

[0067] The benefit of this setup is that adding keywords to the interest model enables it to accurately represent the areas of interest to users, thereby improving the accuracy of content recommendations.

[0068] S103. Based on the updated interest model and the device type of the user's current terminal device, determine the content to be recommended and push the content to be recommended to the user.

[0069] For example, the device type of the user's current terminal device is determined. For instance, the device type could be a mobile device, a PC (Personal Computer), etc. Different device types may be adapted to different content. For example, mobile devices are more suitable for displaying short videos, while PC devices are more suitable for displaying long articles.

[0070] Based on the updated interest model, the system determines the user's current areas of interest and searches for content within those areas using online or offline methods. Depending on the user's current device type, it selects device-compatible content from this pool as recommendations. This recommended content is then pushed to the user's device for viewing. For example, if the interest model indicates the user's current interest is in entertainment news, the system searches for current entertainment news or trending topics. If the user's device is a mobile device, it searches for short videos from current entertainment news and trending topics and pushes them to the user.

[0071] In this embodiment, the method further includes: acquiring the user's operation behavior on the content to be recommended and determining the keywords corresponding to the content to be recommended; adjusting the weights of the keywords corresponding to the content to be recommended in the updated interest model based on the user's operation behavior on the content to be recommended, to obtain the adjusted model; wherein, the adjusted model represents the updated interest model after the weight adjustment; determining new content to be recommended based on the adjusted model and the device type of the terminal device currently used by the user, and pushing the new content to be recommended to the user.

[0072] Specifically, after pushing content to be recommended to users, the system can monitor users' actions related to that content to determine whether the content is of interest to the user. If so, the content recommendation is considered accurate, and no changes to the interest model are needed; otherwise, the content recommendation is considered inaccurate, and the interest model needs to be updated to recommend content to the user again.

[0073] By capturing real-time user interactions with the content to be recommended, we can identify user behavior, such as clicking, dwell time, liking, sharing, and exiting. Dwell time refers to how long a user stays on the content after clicking on it, thus determining whether the user is satisfied with the content.

[0074] First, identify keywords for the content to be recommended. Based on user behavior, determine the user's level of interest in the domain associated with those keywords, and adjust the weight of each keyword in the interest model accordingly. For example, longer dwell time indicates higher interest, so the weight can be increased; shorter dwell time indicates lower interest, so the weight can be decreased. The adjusted interest model is then designated as the revised model. Based on the revised model and the device type, new content to be recommended is identified and continued to be presented to the user, with ongoing feedback on user behavior.

[0075] For example, feedback on click behavior can determine that users frequently click on certain content, indicating high user interest in that content, and dynamically increase the keyword weight of that content's domain. Feedback on prolonged browsing behavior suggests that users spend a significant amount of time on a particular piece of content, indicating strong user interest and a desire for more related recommendations, thus increasing the weight of keywords in that domain. Feedback on skipping or swiping indicates user disinterest in that type of content, and a decreasing algorithm can be used to gradually eliminate these uninteresting domains. In this embodiment, time-series deep learning models, such as RNNs (Recurrent Neural Networks) and LSTMs, can be used to adjust weights in real time to adapt to rapid changes in short-term demand. For instance, a large language model can dynamically capture and analyze user feedback based on time, location, and device, and iterate in real time by acquiring user feedback data to progressively optimize the content recommendation strategy in a closed-loop manner, thereby improving recommendation accuracy.

[0076] The benefit of this setup is that it continuously optimizes and adjusts the interest model based on user behavior feedback, generating increasingly accurate personalized user models. Each change in interests triggers fine-tuning and learning within the interest model, ensuring that subsequent content recommendations align more closely with the user's current interests, thus improving the accuracy of content recommendations and enhancing the user experience.

[0077] Figure 2 This is a schematic diagram illustrating the optimization process of the interest model. Figure 2In this process, for content pushed to users, the system tracks user behavior. For clicks, the weight of keywords corresponding to that content can be increased to continuously recommend relevant content. For prolonged engagement, the weight of keywords can be increased again to continue recommending related content. For closing or swiping, the weight of keywords can be decreased to reduce the amount of related content recommended. Based on the adjusted interest model, the recommendation list is updated. This list can include multiple pieces of content that the user might be interested in, and new content is selected from the list to recommend to the user. For example, the content ranked first in the recommendation list can be recommended to the user.

[0078] In this embodiment, the user's current spatiotemporal scene can be determined, thereby identifying the user's current scene information. The user's current interest model is obtained, which represents the domains of content the user is interested in. Based on the current scene information, the interest model is updated in real time, ensuring that the updated model reflects the user's current areas of interest. Based on the updated interest model and the device type of the user's current terminal device, content to be recommended is determined and pushed to the user. This allows users to view content that is interesting and compatible with their terminal device, improving the accuracy and flexibility of content recommendation and enhancing the user experience.

[0079] Figure 3 This is a flowchart illustrating a content recommendation method based on user interests, provided in an embodiment of this disclosure.

[0080] In this embodiment, the interest model is updated based on the current scene information to obtain an updated interest model, including: predicting the user's preferred domain based on the current scene information; wherein the preferred domain represents the predicted domain that the user is interested in; and updating the interest model based on the preferred domain to obtain an updated interest model.

[0081] This embodiment is based on the above embodiment, such as Figure 3 As shown, the method includes the following steps:

[0082] S301. Obtain current scene information and pre-stored interest model; wherein, current scene information represents the spatiotemporal scene in which the user is currently located, and interest model is a user profile model representing the domain of content that the user is interested in.

[0083] For example, this step can refer to step S101 above, and will not be repeated here.

[0084] S302. Based on the current scenario information, predict the user's preferred domain; whereby the preferred domain represents the predicted domain that the user is interested in.

[0085] For example, based on current scene information, the user's current areas of interest can be predicted and designated as their preferred areas. For instance, if it is determined that the user is currently in a shopping mall, then the user's areas of interest can be predicted to be promotional activities, shopping, etc., and these areas can be designated as their current preferred areas.

[0086] In this embodiment, the current scene information includes time information and spatial information; predicting the user's preference domain based on the current scene information includes: determining the current scene category of the user based on the time information and spatial information; wherein, the scene category represents the user's schedule in the current scene; and predicting the user's preference domain based on the current scene category.

[0087] Specifically, the current scenario information can include current time and spatial information. Time information can represent the current date and time, while spatial information can represent the current location and place. Based on the current time and spatial information, the user's current scenario category is determined. The scenario category can represent the user's schedule in the current scenario. That is, in different scenario categories, the user may have different schedules. For example, in a travel scenario, the user's schedule may be visiting tourist attractions; in a work scenario, the user's schedule may be researching information.

[0088] Different time and spatial information can correspond to different scene categories. Multiple scene categories are preset. For example, if the time information is 8:00 AM on a weekday and the spatial information is the entrance of the residential area, the scene category can be "commuting scenario".

[0089] Users may have different preference areas in different scenarios. Based on the current scenario category, we can determine the areas the user might be interested in, which can be considered as preference areas. For example, in a commuting scenario, a user might be interested in traffic congestion, and the predicted preference area could be the area of ​​road traffic.

[0090] The advantage of this setup is that, based on the current scene information, it can determine what category of scene the user is currently in, and thus update the user's interests and preferences in real time according to the actual scene, flexibly responding to scene changes and improving the accuracy of content recommendations.

[0091] In this embodiment, determining the user's current scene category based on time and spatial information includes: determining the user's current time category based on time information, wherein the time category represents the user's state at the current time; determining the user's current spatial category based on spatial information, wherein the spatial category represents the user's state at the current location; and determining the user's current scene category based on the time category and spatial category.

[0092] Specifically, multiple time categories are pre-set, allowing users to have different states within each category. For example, time categories can be divided into weekdays, weekends, mornings, evenings, etc. The system determines the user's current time category based on the current time information. For instance, a pre-set time synchronization component can continuously detect the current time and determine its category. In other words, it can pre-set the relationships between different dates and time categories, and / or between different points in time and time categories, determining the corresponding time category based on the current time information.

[0093] Multiple space categories are pre-set, and users can have different states corresponding to different space categories. For example, space categories can be divided into office areas, tourist areas, and homes. Based on spatial information, the user's current space category is determined. For example, a preset geographic data matching tool, such as a high-precision map, can be used to map the current location coordinates to different space categories. That is, the association between different location coordinates and space categories can be preset, and the corresponding space category can be determined based on the current spatial information.

[0094] By combining time and space categories, the current scene category of the user can be determined. Different time and space categories can be preset, along with corresponding relationships between scene categories. Based on these preset relationships, the scene category corresponding to the time and space categories can be determined. For example, if the time category is weekday and the space category is office area, the scene category could be "going to work." Or, if the time category is weekend and the space category is outdoors, the scene category could be "outing."

[0095] The benefits of this setup are that by combining temporal and spatial information, scene information can be determined, improving the accuracy of scene category identification and adapting to changing recommendation needs across different time periods and locations. This also improves the accuracy of interest model updates, ensuring accurate content recommendations for users.

[0096] In this embodiment, predicting the user's preferred domain based on the current scene category includes: determining the domain corresponding to the current scene category of the user as the preferred domain based on a preset first association relationship; wherein, the preset first association relationship represents the association relationship between the scene category and the domain.

[0097] Specifically, a first association is pre-set and stored, which represents the relationship between scene categories and domains. After determining the user's current scene category, the corresponding domain can be found based on the pre-set first association, serving as the user's current preferred domain. For example, if the current scene category is "going to work," the preferred domain could be a technology-related domain; if the current scene category is "outing," the preferred domain could be a travel-related domain.

[0098] The advantage of this setup is that, based on the preset associations, the preferred domains can be quickly determined, improving the efficiency of determining preferred domains and thus improving the efficiency of content recommendation.

[0099] S303. Update the interest model according to the preference domain to obtain the updated interest model.

[0100] For example, it is determined whether the preference domain is the domain represented in the interest model. If so, it is considered that the interest model can recommend content that the user is interested in; if not, it is considered that the interest model cannot recommend content that the user is interested in. The preference domain can be added to the interest model so as to recommend relevant content in the preference domain to the user.

[0101] In this embodiment, the system predicts the user's current interests based on their spatiotemporal context, and updates the interest model accordingly. This approach considers that user interests change with the spatiotemporal context, allowing for flexible adjustments to the interest model and ensuring that recommended content closely follows these changes, thus enhancing the user experience.

[0102] In this embodiment, the interest model includes multiple keywords and the weights corresponding to each keyword. Keywords represent domains, and weights represent the degree of user interest. The interest model is updated according to the preferred domains to obtain the updated interest model, including: determining the keywords corresponding to the preferred domains as target words according to a preset second association relationship; wherein the preset second association relationship represents the association relationship between the domains and the keywords; and updating the interest model according to the target words to obtain the updated interest model.

[0103] Specifically, an interest model can include multiple keywords, each representing a different domain. Each keyword has its own weight, which indicates the degree of user interest in the domain associated with that keyword. A higher weight indicates a higher level of interest, while a lower weight indicates a lower level of interest.

[0104] A second association is pre-defined, representing the relationship between different domains and keywords. After determining the preferred domain, keywords corresponding to that domain can be identified as target words based on the second association. The current interest is then updated based on the target words; for example, the target words can be added to the interest model, or the weights of keywords in the interest model can be adjusted based on the target words.

[0105] The advantage of this setup is that it allows for the rapid identification of keywords to be updated as target words, thereby improving the efficiency of updating the interest model and enabling fast content recommendations for users.

[0106] In this embodiment, the interest model is updated based on the target word to obtain the updated interest model. This includes: if the target word exists in the interest model, the weights corresponding to the target word in the interest model are adjusted to obtain the updated interest model.

[0107] Specifically, the system determines whether the target word exists in the interest model. If so, it recommends content related to the target word's domain to the user based on the interest model. To enable users to access content in that domain more quickly, the weight of the target word in the interest model can be adjusted. For example, the weight of the target word can be increased according to a preset incremental algorithm. When recommending content, content from domains with higher weights is prioritized.

[0108] The advantage of this setup is that adjusting the weights in the interest model does not require rebuilding the interest model, thus improving adjustment efficiency and the accuracy of content recommendation.

[0109] In this embodiment, the interest model is updated according to the target word to obtain the updated interest model. This includes: if the target word does not exist in the interest model, the target word is added to the interest model and a preset initial weight is assigned to the target word to obtain the updated interest model.

[0110] Specifically, if no target word is identified in the interest model, it is considered that the interest model cannot recommend content in the domain of the target word to the user. The target word can then be added as a new keyword to the interest model.

[0111] Each keyword in the interest model has its own weight. Therefore, it is necessary to assign weights to the target words added to the interest model. A preset weight can be assigned to the target words as the initial weight, which can be adjusted during the update process of the interest model. In this embodiment, the target words represent the areas of interest that the user is currently interested in, that is, the content that the user is most likely to be recommended in that area. Therefore, a larger initial weight can be preset to prioritize recommending content in the areas of the target words to the user.

[0112] The benefit of this setup is that adding keywords to the interest model enables it to accurately represent the areas of interest to users, thereby improving the accuracy of content recommendations.

[0113] Figure 4 This is a schematic diagram illustrating the update process of the interest model. Figure 4 In this process, the system obtains the user's current spatial and time information via GPS (Global Positioning System). This spatial and time information is then used to define the current scene information. Based on the current scene information, the system matches the user's current spatial and time categories. Combining the spatial and time categories, the system adjusts the keywords and weights in the interest model, outputting an updated interest model.

[0114] S304. Based on the updated interest model and the device type of the user's current terminal device, determine the content to be recommended and push the content to be recommended to the user.

[0115] For example, this step can refer to step S103 above, and will not be repeated here.

[0116] In this embodiment, the user's current spatiotemporal scene can be determined, thereby identifying the user's current scene information. The user's current interest model is obtained, which represents the domains of content the user is interested in. Based on the current scene information, the interest model is updated in real time, ensuring that the updated model reflects the user's current areas of interest. Based on the updated interest model and the device type of the user's current terminal device, content to be recommended is determined and pushed to the user. This allows users to view content that is interesting and compatible with their terminal device, improving the accuracy and flexibility of content recommendation and enhancing the user experience.

[0117] Figure 5 This is a flowchart illustrating a content recommendation method based on user interests, provided in an embodiment of this disclosure.

[0118] In this embodiment, determining the content to be recommended based on the updated interest model and the device type of the user's current terminal device includes: determining candidate content from a preset content database based on the updated interest model; wherein the candidate content is content that the user is interested in; determining the device type of the user's current terminal device, and determining the content to be recommended from the candidate content based on the device type of the terminal device.

[0119] This embodiment is based on the above embodiment, such as Figure 5 As shown, the method includes the following steps:

[0120] S501. Obtain current scene information and pre-stored interest model; wherein, current scene information represents the spatiotemporal scene in which the user is currently located, and interest model is a user profile model representing the domain of content that the user is interested in.

[0121] For example, this step can refer to step S101 above, and will not be repeated here.

[0122] S502. Update the interest model based on the current scene information to obtain the updated interest model.

[0123] For example, this step can refer to step S102 above, and will not be repeated here.

[0124] S503. Based on the updated interest model, determine candidate content from the preset content database; wherein, the candidate content is content that the user is interested in.

[0125] For example, the updated interest model can represent the domain of content that a user is currently interested in. Based on the updated interest model, content corresponding to that domain can be searched online or offline. In this embodiment, a content database can be pre-set, storing content from various domains. Based on the updated interest model, the keywords contained in the model and the weight of each keyword are determined. Keywords can be sorted according to their weights. Based on the keyword's ranking position, content in the domain corresponding to that keyword is searched from the pre-set content database. For example, content in the domain corresponding to the top two keywords can be searched.

[0126] The number of content items can be preset to avoid finding too much content. The content database can be filtered based on metrics such as view count, likes, and shares to identify potentially recommended content. The found content is stored in a preset content list, for example, 10 items ranked by view count. Content retrieved from the database is then designated as candidate content, representing items the user might be interested in. This embodiment does not specifically limit the method for finding candidate content. For example, popular content can be prioritized based on its publication time. Alternatively, content related to the user's long-term reading or work can be selected as candidate content, allowing for dynamic adjustments based on short-term fluctuations and long-term interest.

[0127] S504. Determine the device type of the terminal device currently being used by the user, and determine the content to be recommended from the candidate content based on the device type of the terminal device.

[0128] For example, when a user views content using a terminal device, the device type of the user's current terminal device is determined; for example, the device type could be a mobile device or a PC. Users exhibit different usage habits and behavioral patterns when using different types of devices; therefore, different content needs to be recommended to users based on their device type.

[0129] It can record the content viewed by users on different types of devices within a historical time period. For example, when using a PC, users tend to read professional reports and detailed articles; when using a mobile device, users tend to view short videos or instant content. Based on the content viewed by users on the current type of device within a historical time period, it searches for highly similar content from the candidate content and uses it as the current recommended content.

[0130] In this embodiment, the type of device the user is using is detected. For example, the device type can be determined by directly obtaining the device resolution and device ID through an API (Application Program Interface). Different recommended content is determined for different device types, making the recommendations more contextualized and in line with the device's usage needs, greatly improving the consistency of the recommendation experience across devices.

[0131] In this embodiment, determining the content to be recommended from the candidate content based on the device type of the terminal device includes: determining the content format corresponding to the device type of the terminal device as the target format according to a preset third association relationship; wherein, the preset third association relationship represents the association relationship between the device type and the content format; and determining the candidate content of the target format as the content to be recommended.

[0132] Specifically, different types of devices are suited to displaying different content formats. A third-party relationship can be pre-set and stored, representing the connection between device type and content format. For example, PC devices are best suited to displaying long articles, while mobile devices are best suited to displaying short videos.

[0133] After determining the device type of the user's current terminal device, a content format corresponding to the device type is determined based on a preset third association relationship, and this format is used as the target format. Content in the target format is then retrieved from the candidate content and used as the current content to be recommended. For example, the target format content can be searched sequentially from top to bottom in the candidate content list, and the first piece of content in the target format can be used as the content to be recommended. If no content in the target format is found among the candidate content, the first piece of content in the list can be used as the content to be recommended.

[0134] Figure 6This is a schematic diagram of the content recommendation process based on device type. Figure 6 First, the device's attribute information is obtained, which may include resolution, device ID, and user activity over a historical period. Based on this attribute information, the device type is determined. For different device types, targeted content to be recommended is identified and output to the device.

[0135] The advantage of this setup is that by automatically detecting the devices used by users, the type of content to be recommended can be adjusted accordingly, enabling content switching and optimization across devices, ensuring that matching personalized services can be provided on different devices, and improving the user experience.

[0136] In this embodiment, the user's current spatiotemporal scene can be determined, thereby identifying the user's current scene information. The user's current interest model is obtained, which represents the domains of content the user is interested in. Based on the current scene information, the interest model is updated in real time, ensuring that the updated model reflects the user's current areas of interest. Based on the updated interest model and the device type of the user's current terminal device, content to be recommended is determined and pushed to the user. This allows users to view content that is interesting and compatible with their terminal device, improving the accuracy and flexibility of content recommendation and enhancing the user experience.

[0137] Figure 7 This is a flowchart illustrating a content recommendation method based on user interests, provided in an embodiment of this disclosure.

[0138] This embodiment is based on the above embodiment, such as Figure 7 As shown, the method includes the following steps:

[0139] S701: Obtain the current geographical location, current time, device type, and historical behavior data.

[0140] For example, the current geographic location can be location coordinates obtained via GPS, the current time can be the current date and time, the device type refers to the type of terminal device currently being used by the user, and historical behavior data can include the user's short-term and long-term behavior data. The short-term and long-term behavior data can include the user's historical behavior data within a first time period and the user's historical behavior data within a second time period. The first time period can be longer than the second time period, and the second time period can be included within the first time period. For example, the first time period might be the past year, and the second time period might be the past month.

[0141] S702. Generate current scene information based on the current geographical location and current time.

[0142] For example, spatial information can be determined based on the current geographical location, and temporal information can be determined based on the current time. Based on the current spatial and temporal information, the current scene information can be determined. That is, the current scene information can include both spatial and temporal information.

[0143] S703. Update the keywords and weights of the interest model based on the current scene information.

[0144] For example, based on the current scene information, the user's current scene category is determined. Different scene categories can correspond to different preference domains, indicating that the user's areas of interest differ in different scenes. The current interest model is then updated based on the current preference domain.

[0145] The interest model includes multiple keywords, each with its own weight. Keywords and their weights can be updated within the model. This improves the accuracy of recommendations based on user time and location, and allows for the recommendation of appropriate content according to the environment.

[0146] S704. Update the keywords and weights of the interest model based on historical behavior data.

[0147] For example, keywords and / or weights in the interest model can also be updated based on users' short- and long-term historical behavioral data. This enables the detection of popular short-term content and its mixing with long-term content, ensuring that recommended content satisfies current interests while also addressing long-term needs.

[0148] S705. Determine the updated interest model.

[0149] For example, an updated interest model is obtained based on current scene information and historical behavior data.

[0150] S706. Determine the content to be recommended based on the updated interest model and device type.

[0151] For example, based on the updated interest model, a large amount of content can be filtered, such as by rating each piece of content and selecting the highest-rated content as the content to be recommended.

[0152] By detecting the type of device a user is using, targeted content recommendation strategies can be triggered accordingly. For example, device resolution, device ID, and user behavior can be directly obtained through system APIs. User behavior could include the number of times a user views short videos or the time spent on long texts on different devices. The format of the recommended content can then be adjusted and optimized based on the device type.

[0153] S707: Display the content to be recommended on the terminal device.

[0154] For example, the display format of the content to be recommended is adapted to the terminal device, so that the content to be recommended can be displayed on the terminal device.

[0155] S708. Obtain user actions related to recommended content.

[0156] For example, once a user clicks on any content to be recommended, the user's actions regarding that content can be obtained.

[0157] S709. Update the interest model based on the user's actions towards the recommended content.

[0158] For example, the interest model can be further optimized based on user behavior. User behavior is captured in real time as feedback on the recommended content. User behavior can include clicking, swiping, skipping, exiting, liking, commenting, sharing, etc. After the content to be recommended is pushed to the front-end interface, user behavior is captured and recorded in real time. These behaviors reflect the user's level of attention to the recommended content and influence the adjustment of the interest model.

[0159] These user actions can be used to rate recommended content. For example, if a user shares or likes recommended content, the content can receive a higher rating. A higher rating indicates greater user interest in the content's relevant domain, increasing the weight of keywords related to that domain in the interest model and thus adjusting the model itself. This closed-loop feedback optimization approach allows for continuous improvement of the interest model, maintaining its interactive "learning and adaptation" characteristics, preventing outdated or biased recommendations, and enhancing the user experience.

[0160] For example, if a user browses some technical articles on a PC on Monday morning, and then goes camping over the weekend and uses their phone to search for camping-related content, the system can intelligently capture these changes in user behavior and push personalized content to the user.

[0161] Specifically, the GPS positioning module detects that the user has arrived at a scenic spot outside the city over the weekend, automatically entering the "Weekend Scene, Outdoor Scene." It captures changes in the user's device, switching from a PC during the weekday to a mobile device. Based on specific device characteristics, it automatically reduces lengthy recommendations, prioritizing short and practical videos or articles. Combining this with articles about similar travel or equipment viewed by the user in previous weeks, it recommends relevant content. The user clicked on multiple camping gear guides, stayed on these pages for a long time, searched for related camping gear, and added some outdoor products to their favorites. Based on this user feedback, the system uses LSTM algorithms and specific user behavior analysis to update the weekend's interest model, adjusting the priority of the next batch of camping-related content in real time, ultimately ensuring that recommended content matches the real-time scene and individual needs. The next time the user goes to another campsite, it can prioritize pushing different types of equipment reviews, related experience-sharing videos, and other content based on time and historical records.

[0162] Figure 8 This is a structural block diagram of a content recommendation device based on user interests, provided in an embodiment of this disclosure. For ease of explanation, only the parts relevant to the embodiments of this disclosure are shown. (Refer to...) Figure 8 The content recommendation device 800 based on user interests includes: an acquisition unit 801, an update unit 802, and a recommendation unit 803.

[0163] The acquisition unit 801 is used to acquire current scene information and pre-stored interest models; wherein, the current scene information represents the spatiotemporal scene in which the user is currently located, and the interest model is a user profile model that represents the domain of content that the user is interested in;

[0164] The update unit 802 is used to update the interest model according to the current scene information to obtain the updated interest model;

[0165] The recommendation unit 803 is used to determine the content to be recommended based on the updated interest model and the device type of the terminal device currently used by the user, and to push the content to be recommended to the user.

[0166] Figure 9 A structural block diagram of a content recommendation device based on user interests provided in this disclosure embodiment is shown below. Figure 9 As shown, the content recommendation device 900 based on user interests includes an acquisition unit 901, an update unit 902, and a recommendation unit 903. The update unit 902 includes a domain prediction module 9021 and a model update module 9022.

[0167] Domain prediction module 9021 is used to predict the user's preferred domain based on the current scene information; wherein, the preferred domain represents the predicted domain that the user is interested in;

[0168] The model update module 9022 is used to update the interest model according to the preference domain to obtain the updated interest model.

[0169] In one example, the current scene information includes time information and spatial information; the domain prediction module 9021 includes:

[0170] The category determination submodule is used to determine the current scene category of the user based on the time information and the spatial information; wherein, the scene category represents the user's schedule in the current scene;

[0171] The domain prediction submodule is used to predict the user's preferred domain based on the current scene category.

[0172] In one example, the category determines the submodule, specifically used for:

[0173] Based on the time information, the user's current time category is determined; wherein, the time category represents the user's state at the current time;

[0174] Based on the spatial information, the current spatial category of the user is determined; wherein, the spatial category represents the user's state in the current location;

[0175] The current scene category of the user is determined based on the time category and the space category.

[0176] In one example, the domain prediction submodule is specifically used for:

[0177] Based on a preset first association relationship, the domain corresponding to the scene category in which the user is currently located is determined as the preferred domain; wherein, the preset first association relationship represents the association relationship between scene category and domain.

[0178] In one example, the interest model includes multiple keywords and weights corresponding to each keyword, where the keywords represent a domain and the weights represent the user's level of interest; the model update module 9022 includes:

[0179] The target determination submodule is used to determine the keywords corresponding to the preference domain as target words based on a preset second association relationship; wherein, the preset second association relationship represents the association relationship between the domain and the keywords;

[0180] The model update submodule is used to update the interest model based on the target word to obtain the updated interest model.

[0181] In one example, the model update submodule is specifically used for:

[0182] If the target word exists in the interest model, the weights corresponding to the target word in the interest model are adjusted to obtain the updated interest model.

[0183] In one example, the model update submodule is specifically used for:

[0184] If the target word does not exist in the interest model, then the target word is added to the interest model and a preset initial weight is assigned to the target word to obtain the updated interest model.

[0185] In one example, recommended unit 903 includes:

[0186] The candidate determination module is used to determine candidate content from a preset content database based on the updated interest model; wherein the candidate content is content that the user is interested in.

[0187] The content recommendation module is used to determine the device type of the terminal device currently used by the user, and to determine the content to be recommended from the candidate content based on the device type of the terminal device.

[0188] In one example, the content recommendation module includes:

[0189] The format determination submodule is used to determine the content format corresponding to the device type of the terminal device as the target format based on a preset third association relationship; wherein, the preset third association relationship represents the association relationship between the device type and the content format;

[0190] The content determination submodule is used to determine candidate content in the target format, which is the content to be recommended.

[0191] One example also includes:

[0192] A data acquisition unit is used to acquire the user's historical behavior data; wherein, the historical behavior data includes the user's behavior data within a preset first time period and the user's behavior data within a preset second time period, the preset first time period being longer than the preset second time period, and the behavior data characterizing the user's operation behavior and operation content;

[0193] The interest model update unit is used to update the interest model based on the historical behavior data to obtain the updated interest model.

[0194] In one example, the interest model includes multiple keywords and weights corresponding to each keyword, where the keywords represent a domain and the weights represent the user's level of interest; the interest model update unit includes:

[0195] The historical word extraction module is used to extract keywords from the historical behavior data to obtain the keywords corresponding to the historical behavior data, which are historical words.

[0196] The historical word update module is used to update the interest model based on the historical words to obtain the updated interest model.

[0197] In one example, the history word update module includes:

[0198] The weight adjustment submodule is used to adjust the weights corresponding to the historical words in the interest model if the historical words exist in the interest model, so as to obtain the updated interest model.

[0199] In one example, the history word update module includes:

[0200] The historical word addition submodule is used to add the historical word to the interest model if the historical word does not exist in the interest model, and assign a preset initial weight to the historical word to obtain the updated interest model.

[0201] One example also includes:

[0202] An operation acquisition unit is used to acquire the user's operation behavior on the content to be recommended, and to determine the keywords corresponding to the content to be recommended;

[0203] The model adjustment unit is used to adjust the weights of the keywords corresponding to the content to be recommended in the updated interest model according to the user's operation behavior on the content to be recommended, so as to obtain the adjusted model; wherein, the adjusted model represents the updated interest model after the weight adjustment.

[0204] The re-recommendation unit is used to determine new content to be recommended based on the adjusted model and the device type of the terminal device currently used by the user, and to push the new content to be recommended to the user.

[0205] According to embodiments of this disclosure, this disclosure also provides an electronic device.

[0206] Figure 10 A structural block diagram of an electronic device provided in this disclosure embodiment, such as... Figure 10As shown, the electronic device 1000 includes: at least one processor 1002; and a memory 1001 communicatively connected to the at least one processor 1002; wherein the memory stores instructions executable by the at least one processor 1002, the instructions being executed by the at least one processor 1002 to enable the at least one processor 1002 to execute the user interest-based content recommendation method of this disclosure.

[0207] The electronic device 1000 also includes a receiver 1003 and a transmitter 1004. The receiver 1003 is used to receive instructions and data sent by other devices, and the transmitter 1004 is used to send instructions and data to external devices.

[0208] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0209] According to embodiments of this disclosure, this disclosure also provides a computer program product comprising: a computer program stored in a readable storage medium, at least one processor of an electronic device being able to read the computer program from the readable storage medium, and the at least one processor executing the computer program causing the electronic device to perform the scheme provided in any of the above embodiments.

[0210] Figure 11 A schematic block diagram of an example electronic device 1100 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0211] like Figure 11 As shown, device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in read-only memory (ROM) 1102 or a computer program loaded from storage unit 1108 into random access memory (RAM) 1103. The RAM 1103 may also store various programs and data required for the operation of device 1100. The computing unit 1101, ROM 1102, and RAM 1103 are interconnected via bus 1104. Input / output (I / O) interface 1105 is also connected to bus 1104.

[0212] Multiple components in device 1100 are connected to I / O interface 1105, including: input unit 1106, such as keyboard, mouse, etc.; output unit 1107, such as various types of monitors, speakers, etc.; storage unit 1108, such as disk, optical disk, etc.; and communication unit 1109, such as network card, modem, wireless transceiver, etc. Communication unit 1109 allows device 1100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0213] The computing unit 1101 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1101 performs the various methods and processes described above, such as a content recommendation method based on user interests. For example, in some embodiments, the content recommendation method based on user interests can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed on device 1100 via ROM 1102 and / or communication unit 1109. When the computer program is loaded into RAM 1103 and executed by the computing unit 1101, one or more steps of the content recommendation method based on user interests described above can be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to perform a content recommendation method based on user interests by any other suitable means (e.g., by means of firmware).

[0214] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0215] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0216] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on 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 of the foregoing.

[0217] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0218] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0219] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or simply "VPS") services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0220] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0221] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A content recommendation method based on user interests, comprising: The system acquires current scene information and pre-stored interest models; wherein, the current scene information represents the spatiotemporal scene in which the user is currently located, and the interest model is a user profile model representing the domain of content that the user is interested in; the current scene information includes time information and spatial information. Based on the time information, the user's current time category is determined; wherein, the time category represents the user's state at the current time; the time category includes one or more of weekday, weekend, morning, or evening; Based on the spatial information, the current spatial category of the user is determined; wherein, the spatial category represents the user's state in the current location; the spatial category includes office area, tourist area, or home. Based on the time category and the space category, the current scene category of the user is determined; wherein, the scene category represents the user's schedule in the current scene; Based on the user's current scenario category, predict the user's preferred domain; wherein, the preferred domain represents the predicted domain that the user is interested in; the domain includes at least one of the following: technology domain, travel domain, road travel domain, shopping domain, and fitness domain; The interest model is updated based on the preference domain to obtain the updated interest model; Based on the updated interest model and the device type of the user's current terminal device, determine the content to be recommended and push the content to be recommended to the user; The step of predicting the user's preferred domain based on the current scene category includes: Based on a preset first association relationship, the domain corresponding to the scene category in which the user is currently located is determined as the preferred domain; wherein, the preset first association relationship represents the association relationship between scene category and domain; The interest model includes multiple keywords and weights corresponding to each keyword. The keywords represent a domain, and the weights represent the user's level of interest. Updating the interest model based on the preference domain to obtain an updated interest model includes: Based on a preset second association relationship, keywords corresponding to the preference domain are determined as target words; wherein, the preset second association relationship represents the association relationship between the domain and the keywords; If the target word exists in the interest model, the weights corresponding to the target word in the interest model are adjusted to obtain the updated interest model. If the target word does not exist in the interest model, then the target word is added to the interest model and a preset initial weight is assigned to the target word to obtain the updated interest model; The method further includes: Obtain the user's historical behavior data; wherein, the historical behavior data includes the user's behavior data within a preset first time period and the user's behavior data within a preset second time period, the preset first time period being longer than the preset second time period, and the behavior data characterizing the user's operational behavior and operational content; The historical behavior data is processed by keyword extraction to obtain the keywords corresponding to the historical behavior data, which are historical words. If the historical word exists in the interest model, the weights corresponding to the historical word in the interest model are adjusted to obtain the updated interest model. If the historical word does not exist in the interest model, then the historical word is added to the interest model and a preset initial weight is assigned to the historical word to obtain the updated interest model.

2. The method according to claim 1, wherein, The step of determining the content to be recommended based on the updated interest model and the device type of the user's current terminal device includes: Based on the updated interest model, candidate content is determined from a preset content database; wherein, the candidate content is content that the user is interested in. Determine the device type of the terminal device currently being used by the user, and determine the content to be recommended from the candidate content based on the device type of the terminal device.

3. The method according to claim 2, wherein, The step of determining the content to be recommended from the candidate content based on the device type of the terminal device includes: Based on a preset third association relationship, the content format corresponding to the device type of the terminal device is determined as the target format; wherein, the preset third association relationship represents the association relationship between the device type and the content format; Candidate content in the target format is determined as the content to be recommended.

4. The method according to any one of claims 1-3, further comprising: The user's actions on the content to be recommended are obtained, and the keywords corresponding to the content to be recommended are determined. Based on the user's interaction with the content to be recommended, the weights of the keywords corresponding to the content to be recommended in the updated interest model are adjusted to obtain the adjusted model; wherein, the adjusted model represents the updated interest model after the weights are adjusted. Based on the adjusted model and the device type of the user's current terminal device, new content to be recommended is determined and pushed to the user.

5. A content recommendation device based on user interests, comprising: The acquisition unit is used to acquire current scene information and pre-stored interest models; wherein, the current scene information represents the spatiotemporal scene in which the user is currently located, and the interest model is a user profile model that represents the domain of content that the user is interested in; An update unit is used to update the interest model based on the current scene information to obtain an updated interest model. The recommendation unit is used to determine the content to be recommended based on the updated interest model and the device type of the terminal device currently used by the user, and to push the content to be recommended to the user. The current scene information includes time information and spatial information; the update unit is specifically used to determine the user's current time category based on the time information, wherein the time category represents the user's state at the current time; determine the user's current spatial category based on the spatial information; determine the user's current scene category based on the time category and the spatial category; predict the user's preference domain based on the user's current scene category; and update the interest model based on the preference domain to obtain an updated interest model; wherein the spatial category represents the user's state in the current location, and the spatial category includes office area, tourist area, or home; the time category includes one or more of weekdays, weekends, mornings, or evenings; the scene category represents the user's schedule in the current scene; and the preference domain represents the predicted areas of interest for the user, and the areas include at least one of the following: technology, travel, road travel, shopping, and fitness. Specifically, when the update unit predicts the user's preference domain based on the user's current scene category, it is used for: Based on a preset first association relationship, the domain corresponding to the scene category in which the user is currently located is determined as the preferred domain; wherein, the preset first association relationship represents the association relationship between scene category and domain; The interest model includes multiple keywords and weights for each keyword, whereby the keywords represent a domain and the weights represent the degree of user interest. The updating unit is used to update the interest model according to the preference domain, and when obtaining the updated interest model, it is specifically used for: Based on a preset second association relationship, keywords corresponding to the preference domain are determined as target words; wherein, the preset second association relationship represents the association relationship between the domain and the keywords; If the target word exists in the interest model, the weights corresponding to the target word in the interest model are adjusted to obtain the updated interest model. If the target word does not exist in the interest model, then the target word is added to the interest model and a preset initial weight is assigned to the target word to obtain the updated interest model; The updating unit is used to update the interest model according to the preference domain, and when obtaining the updated interest model, it is further used to: Obtain the user's historical behavior data; wherein, the historical behavior data includes the user's behavior data within a preset first time period and the user's behavior data within a preset second time period, the preset first time period being longer than the preset second time period, and the behavior data characterizing the user's operational behavior and operational content; The historical behavior data is processed by keyword extraction to obtain the keywords corresponding to the historical behavior data, which are historical words. If the historical word exists in the interest model, the weights corresponding to the historical word in the interest model are adjusted to obtain the updated interest model. If the historical word does not exist in the interest model, then the historical word is added to the interest model and a preset initial weight is assigned to the historical word to obtain the updated interest model.

6. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-4.

7. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-4.

8. A computer program product, wherein, Includes a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1-4.

Images