Network audio-visual content recommendation method and system based on multiple roles

Abstract

The invention discloses a network audio-visual content recommendation method and system based on multiple roles. The method comprises the following steps: receiving a role selected by a user and sent by a front end; querying a content preference tag set corresponding to the role from a pre-constructed tag set, establishing a content tag index, and searching all matched program contents according to the content tag index; selecting a plurality of program contents from all the program contents according to a preset time priority or popularity priority, and forming a candidate recommendation set; de-duplicating the program contents in the candidate recommendation set, and sorting the de-duplicated program contents according to the content popularity to obtain a recommendation list; and sending the recommendation list to a front end for playing in sequence. According to the method, a simple rule matching algorithm is adopted, programs can be directly recommended according to predefined role content preferences, dependence on a complex machine learning model is avoided, content recommendation can be immediately generated after role selection, waiting for model training or data synchronization is not needed, and response delay is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of audio-visual content recommendation, and in particular to a method and system for recommending network audio-visual content based on multiple roles. Background Art

[0002] The rapid development of internet technology and the widespread adoption of smart devices have revolutionized the dissemination and consumption of online audiovisual content. Users access a wealth of audiovisual content through various online platforms, such as video streaming services, social media, and music platforms. To adapt to the flood of information, online audiovisual content recommendation systems have emerged as a crucial technical tool for improving user experience, increasing platform engagement, and optimizing content distribution efficiency. Recommendation algorithms based on user interests, behavioral data, and content features intelligently analyze user preferences and recommend personalized content tailored to their needs, significantly changing how users consume content.

[0003] Patent document No. CN115695897A discloses a multi-screen interactive content recommendation method, storage medium and device based on multiple roles. The method is applied to an electronic device, comprising the following steps: when connected to another electronic device for screen interaction, recording the interaction history under the current role, saving it to a database, and forming a training sample for a first recommendation model; recording the connection and usage habits of the current role with different electronic devices, saving it to a database, and forming a training sample for a second recommendation model; when establishing a connection with another electronic device, generating a first recommended content based on the first recommendation model, generating a second recommended content based on the second recommendation model, and fusing the first recommended content and the second recommended content to generate a third recommended content for final presentation, such as Figure 1 shown.

[0004] It can be seen that the existing technology mainly involves a multi-screen interactive content recommendation method based on multiple roles, including recording the interaction history and device connection habits of user roles to generate personalized recommended content. This method analyzes user behavior data through multiple models (first recommendation model and second recommendation model) to generate the final recommended content. However, this patent document has the following shortcomings when implementing multi-screen interactive content recommendation based on multiple roles:

[0005] (1) Over-reliance on device connection habits: This method requires recording device connection usage habits, which may bring privacy issues. Users may not want their device connection and usage habits to be recorded and analyzed.

[0006] (2) High complexity of recommendation model: This method needs to generate recommendation content based on the interaction history and device connection habits of multiple roles, and train and integrate multiple recommendation models, which leads to high complexity of recommendation algorithm, increases system burden and maintenance difficulty.

[0007] (3) Lack of adaptability: If users’ viewing habits change frequently in a multi-device scenario, the model may not be able to adjust in time, and the recommended content may lag behind the changes in users’ interests.

[0008] (4) Data synchronization delay problem: The database and model training process is carried out in the back-end server. Data transmission and synchronization may cause delays, affecting real-time performance and user experience.

[0009] (5) User privacy risk: This involves recording and analyzing a large amount of user behavior data and device usage data. If data security measures are not in place, there may be a risk of user privacy leakage. Summary of the Invention

[0010] To this end, the present application provides a method and system for recommending online audio-visual content based on multiple roles to solve the problems of high complexity and response delay of existing recommendation models.

[0011] In order to achieve the above objectives, this application provides the following technical solutions:

[0012] In a first aspect, a method for recommending online audiovisual content based on multiple roles is provided. The method is applied to a backend that maintains a role-label mapping table and a label-content category matching rule table, including:

[0013] Step 1: Receive the user-selected role sent by the front-end;

[0014] Step 2: Querying a content preference tag set corresponding to the role from a pre-built tag set; the tag set is a set of recommended content category tags preset based on the role's attribute information;

[0015] Step 3: Create a content tag index based on the content preference tag set, and search for all matching program contents based on the content tag index;

[0016] Step 4: Select multiple pieces of program content from all program content according to a preset time priority or popularity priority, and form a candidate recommendation set;

[0017] Step 5: Deduplication of program contents in the candidate recommendation set, and sorting the deduplication of program contents according to content popularity to obtain a recommendation list;

[0018] Step 6: Send the recommendation list to the front end for playback in sequence.

[0019] Optionally, the method further includes: collecting user behavior data, calculating content preference tag set weights based on the user behavior data, and adjusting role preferences based on the content preference tag set weights.

[0020] Optionally, the behavior data includes the role selection timestamp, the tag of the content clicked in each recommendation list, the length of time the user stays on the content page, the classification of the viewed content, and the number of times the user quickly jumps out of the content.

[0021] Optionally, in step 1, the roles include children, parents and elderly people.

[0022] Optionally, in step 3, an Elasticsearch or Solr method is used when establishing a content tag index according to the content preference tag set.

[0023] Optionally, in step 5, when deduplicating the program contents in the candidate recommendation set, filtering is performed based on region, time period, holidays, expired content, and offline content.

[0024] Optionally, in step 5, the content popularity includes the number of views and the number of likes.

[0025] In a second aspect, a multi-role-based online audiovisual content recommendation system is provided, wherein the system is provided at a backend, and the backend maintains a role-label mapping table and a label-content category matching rule table, including:

[0026] The front-end data receiving module is used to receive the user-selected role sent by the front-end;

[0027] A content preference tag set query module is used to query a content preference tag set corresponding to the role from a pre-built tag set; the tag set is a set of recommended content category tags preset based on the attribute information of the role;

[0028] a program content matching module, configured to establish a content tag index based on the content preference tag set, and search for all matching program contents based on the content tag index;

[0029] A candidate recommendation set selection module is used to select multiple program contents from all program contents according to a preset time priority or popularity priority, and form a candidate recommendation set;

[0030] A recommendation list generating module is used to remove duplicate program contents from the candidate recommendation set and sort the removed duplicate program contents according to content popularity to obtain a recommendation list;

[0031] The data sending module is used to send the recommendation list to the front end for playing in sequence.

[0032] In a third aspect, a computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for recommending network audio-visual content based on multiple roles when executing the computer program.

[0033] In a fourth aspect, a computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for recommending network audio-visual content based on multiple roles.

[0034] Compared with the prior art, this application has at least the following beneficial effects:

[0035] 1. The present application provides a method for recommending online audio-visual content based on multiple roles, which receives the user-selected role sent by the front end; queries the content preference tag set corresponding to the role from the pre-built tag set; establishes a content tag index based on the content preference tag set, and searches for all matching program content based on the content tag index; selects multiple program contents from all program contents according to a pre-set time priority or popularity priority, and forms a candidate recommendation set; deduplicates the program content in the candidate recommendation set, and sorts the deduplicated program content according to content popularity to obtain a recommendation list; and sends the recommendation list to the front end for playback in sequence. The method for recommending online audio-visual content based on multiple roles provided by the present application adopts a simple rule matching algorithm, which can directly recommend programs based on predefined role content preferences, does not rely on complex machine learning models, and can generate content recommendations immediately after role selection, without waiting for model training or data synchronization, thereby reducing response delays.

[0036] 2. This application collects user behavior data, calculates the content preference tag set weight based on the user behavior data, and adjusts the role preference based on the content preference tag set weight, thereby dynamically optimizing the recommendation effect and having strong adaptability. In addition, this application only collects user behavior data and does not store it, which fundamentally reduces the risk of user privacy leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] To more intuitively illustrate the prior art and the present application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be considered as limiting conditions for implementing the present application; for example, based on the technical concepts disclosed in this application and the exemplary drawings, those skilled in the art are capable of easily making routine adjustments or further optimizations to the addition / reduction / attribution division, specific shapes, positional relationships, connection methods, dimensional ratios, etc. of certain units (components).

[0038] Figure 1 A schematic diagram of an existing multi-role-based multi-screen interactive content recommendation method;

[0039] Figure 2 A flowchart of a method for recommending online audio-visual content based on multiple roles provided in Example 1 of the present application;

[0040] Figure 3This is a structural diagram of a multi-role based online audio-visual content recommendation method provided in Example 1 of the present application. DETAILED DESCRIPTION

[0041] The present application will be further described below in detail through specific embodiments in conjunction with the accompanying drawings.

[0042] In the description of this application: unless otherwise specified, the meaning of "plurality" is two or more. The terms "first", "second", "third", etc. in this application are intended to distinguish the objects referred to and do not have any special meaning in terms of technical connotation (for example, they should not be understood as emphasizing the importance or order, etc.). Expressions such as "including", "comprising", "having", etc. also mean "not limited to" (certain units, components, materials, steps, etc.).

[0043] The terms such as "upper", "lower", "left", "right", "middle", etc. cited in this application are usually used to indicate the general relative position relationship for the convenience of intuitive understanding by referring to the drawings, and are not absolute limitations on the position relationship in the actual product.

[0044] Example 1

[0045] This embodiment provides a method for recommending online audiovisual content based on multiple roles. The method is applied to a backend that maintains a role-label mapping table and a label-content category matching rule table. For example:

[0046] The role-tag mapping table (Role→Tags) includes:

[0047] Children → ["Cartoon", "Animation", "Puzzle"];

[0048] Parents → ["News","TV Series","Variety Show"];

[0049] Elderly → ["opera","health","nostalgic movies"].

[0050] The tag-content category rule table (Tag→ProgramCategory) includes:

[0051] "Cartoon" → ["Boonie Bears", "Pleasant Goat", "Super Wings"];

[0052] "Opera" → ["Peking Opera", "Yue Opera", "Huangmei Opera"].

[0053] See also Figure 2 and Figure 3 This embodiment provides a method for recommending online audio-visual content based on multiple roles, including:

[0054] S1: Receives the user-selected role sent by the front-end;

[0055] Specifically, this embodiment defines several roles (e.g., parents, children, and elderly people) during initialization, and presets content preferences and viewing preferences for each role. Users can directly select roles on the front-end device without logging in or other operations.

[0056] That is, in this embodiment, an independent content preference profile (i.e., tag set) is established for each character during the initialization phase. This profile is a set of recommended content category tags (ContentTags) preset based on the character's attribute information (e.g., age group, interest preferences, content acceptance threshold, etc.), for example:

[0057] Child role: ["animation", "cartoon", "children's educational", "parent-child interaction"];

[0058] Parent role: ["news","life services","variety shows","TV series"];

[0059] Elderly roles: ["opera","classic old film","health and wellness","documentary"].

[0060] During runtime, the user selects a character on the front-end device (for example, by selecting an avatar or icon via a remote control). The front-end immediately sends the selected character ID to the back-end recommendation service module, which then loads the recommended content.

[0061] S2: Query the content preference tag set corresponding to the role from the pre-built tag set; the tag set is a set of recommended content category tags preset based on the role's attribute information;

[0062] Specifically, this step is a role preference mapping, and the corresponding content preference tag set can be found through the role ID.

[0063] S3: Create a content tag index based on the content preference tag set, and search for all matching program contents based on the content tag index;

[0064] Specifically, when establishing a content tag index based on a content preference tag set, an Elasticsearch or Solr method is used to quickly find program content that matches the tag.

[0065] S4: selecting multiple pieces of program content from all program content according to a preset time priority or popularity priority, and forming a candidate recommendation set;

[0066] Specifically, in this step, N latest and most popular contents (configurable, for example, time priority or popularity priority) can be taken out from all program contents to form a candidate recommendation set.

[0067] S5: Deduplication of program contents in the candidate recommendation set is performed, and the deduplication of program contents is sorted according to content popularity to obtain a recommendation list;

[0068] Specifically, this step deduplicates the program content in the candidate recommendation set based on parameters such as region, time period, holidays, expired and offline content (for example, cartoons are recommended in the morning and bedtime stories are recommended in the evening). After deduplicating the candidate content, it is sorted according to parameters such as the number of recent plays, clicks, viewing time, playback volume, likes or update time to obtain a recommendation list.

[0069] S6: Send the recommendation list to the front end for playback in sequence.

[0070] Specifically, the first M items in the recommendation list are sent to the front-end display module (such as TV, Pad or mobile phone) for users to choose to play.

[0071] This embodiment provides a method for recommending online audio-visual content based on multiple roles, further comprising:

[0072] S7: Adaptive content update, which means regularly analyzing the viewing behavior after the character is selected (without recording the specific viewing content), and automatically adjusting the character's content preferences based on the overall trend.

[0073] Adaptive content updating refers to automatically adjusting the preference weights of different roles based on the user's usage behavior in different roles over the long term, thereby dynamically optimizing the recommendation effect. Specifically, it includes:

[0074] S701: Collecting user behavior data;

[0075] Specifically, this embodiment collects but does not store the specific content viewed by the user. Instead, it collects data in the following general dimensions:

[0076] (1) Role selection timestamp;

[0077] (2) The label of the content clicked in each recommendation list;

[0078] (3) The length of time users stay on the content page (e.g., more than 30 seconds is considered effective viewing);

[0079] (4) categorization of viewing content (mapped through a content tagging system);

[0080] (5) The number of times users quickly jump out of the content (reverse preference indicator).

[0081] S702: Calculate the content preference tag set weight according to the user's behavior data, and adjust the role preference according to the content preference tag set weight.

[0082] Specifically, each tag is assigned an adjustable weight, for example, an initial value of 1.0. The following algorithm is executed periodically:

[0083] W_new = W_old + α * number of positive behaviors - β * number of negative behaviors

[0084] Among them, α is the positive feedback gain coefficient (for example: click, complete viewing); β is the negative feedback penalty coefficient (for example: quick exit, short stay); W_new controls the sorting priority of the tag in the recommendation set; W_old also refers to the sorting priority of the tag in the recommendation set. W_old and W_new only differ in the corresponding "number of iterations", and the initial value is 1.0 (that is, W_1 = 1.0).

[0085] In this embodiment, the update cycle and strategy include:

[0086] (1) Statistics are collected every 24 hours;

[0087] (2) persisting a new role preference configuration once a week;

[0088] (3) A sliding window (e.g., the last 7 days) can be set to avoid overfitting;

[0089] (4) Supports aggregate analysis of the behavior of the same role on different devices (clustered by role ID).

[0090] This embodiment can ensure that the recommended content is more in line with the character's interests by regularly updating the character's content preferences.

[0091] This embodiment provides a multi-role-based online audiovisual content recommendation method that uses a simple rule-matching algorithm (a non-machine learning recommendation method based on tags and a rule table) to directly recommend program content based on preset content preferences. For example, the child role is matched with cartoons and animation content; the parent role is matched with news and variety shows. The entire process typically completes in milliseconds, ensuring that recommended content is quickly presented after the user selects a role, enhancing the smoothness of the experience. This method eliminates the need for complex model training, making it easy to deploy, debug, and update in real time. It is particularly suitable for application scenarios with limited terminal performance or high requirements for recommendation interpretability.

[0092] The multi-role based online audio-visual content recommendation method provided in this embodiment has the following advantages:

[0093] (1) Improved efficiency and response speed: Because this embodiment uses a rule-matching algorithm, recommended content is generated faster without the need for complex model training and data synchronization. As a result, users can instantly obtain personalized recommendations, significantly improving response speed.

[0094] (2) Simplified user operations: The existing technology requires frequent logins and account switching to achieve multi-role recommendations, while this embodiment avoids these tedious operations through a simple role selection process, making the user experience more convenient and intuitive.

[0095] (3) Improved privacy protection: It avoids the collection and storage of user behavior and device usage data, fundamentally reducing the risk of user privacy leakage. This improvement is particularly important in the context of users' increasing concern about data privacy.

[0096] (4) Reduced algorithm complexity and resource consumption: The existing multi-model fusion strategy is abandoned and a simple rule matching method is adopted. This not only reduces the algorithm complexity, but also reduces the consumption of computing resources, making it easier to expand and maintain.

[0097] (5) Improved content accuracy in multi-user scenarios: By presetting the content preferences of roles, content recommendations can be made more accurately. For different family members, their interests can be directly matched without relying on a long data accumulation and learning process.

[0098] Example 2

[0099] This embodiment provides a multi-role-based online audiovisual content recommendation system. The system is set at a backend, and the backend maintains a role-label mapping table and a label-content category matching rule table, including:

[0100] The front-end data receiving module is used to receive the user-selected role sent by the front-end;

[0101] A content preference tag set query module is used to query a content preference tag set corresponding to the role from a pre-built tag set; the tag set is a set of recommended content category tags preset based on the attribute information of the role;

[0102] a program content matching module, configured to establish a content tag index based on the content preference tag set, and search for all matching program contents based on the content tag index;

[0103] A candidate recommendation set selection module is used to select multiple program contents from all program contents according to a preset time priority or popularity priority, and form a candidate recommendation set;

[0104] A recommendation list generating module is used to remove duplicate program contents from the candidate recommendation set and sort the removed duplicate program contents according to content popularity to obtain a recommendation list;

[0105] The data sending module is used to send the recommendation list to the front end for playing in sequence.

[0106] Regarding the specific implementation content of each module in a multi-role-based network audio-visual content recommendation system, please refer to the above definition of a multi-role-based network audio-visual content recommendation method, which will not be repeated here.

[0107] Example 3

[0108] This embodiment provides a computer device including a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps of a method for recommending online audio-visual content based on multiple roles.

[0109] Example 4

[0110] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of a method for recommending network audio-visual content based on multiple roles are implemented.

[0111] The technical features of the above embodiments can be combined arbitrarily (as long as there is no contradiction in the combination of these technical features). In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described; these embodiments that are not explicitly written should also be considered to be within the scope of this specification.

Claims

1. A method for recommending online audio-visual content based on multiple roles, characterized in that: The method is applied to a backend, which maintains a role-label mapping table and a label-content category matching rule table, including: Step 1: Receive the user-selected role sent by the front-end; Step 2: Querying a content preference tag set corresponding to the role from a pre-built tag set; the tag set is a set of recommended content category tags preset based on the role's attribute information; Step 3: Create a content tag index based on the content preference tag set, and search for all matching program contents based on the content tag index; Step 4: Select multiple pieces of program content from all program content according to a preset time priority or popularity priority, and form a candidate recommendation set; Step 5: Deduplication of program contents in the candidate recommendation set, and sorting the deduplication of program contents according to content popularity to obtain a recommendation list; Step 6: Send the recommendation list to the front end for playback in sequence.

2. The method for recommending online audio-visual content based on multiple roles according to claim 1, characterized in that: Also includes: Collect user behavior data, calculate content preference tag set weights based on the user behavior data, and adjust role preferences based on the content preference tag set weights.

3. The method for recommending online audio-visual content based on multiple roles according to claim 2, characterized in that: The behavioral data includes the timestamp of character selection, the label of the content clicked in each recommendation list, the length of time the user stays on the content page, the category of the content viewed, and the number of times the user quickly jumps out of the content.

4. The method for recommending online audio-visual content based on multiple roles according to claim 1, characterized in that: In step 1, the roles include children, parents and elderly people.

5. The method for recommending online audio-visual content based on multiple roles according to claim 1, characterized in that: In step 3, the Elasticsearch or Solr method is used to establish a content tag index based on the content preference tag set.

6. The method for recommending online audio-visual content based on multiple roles according to claim 1, characterized in that: In step 5, when deduplicating the program contents in the candidate recommendation set, the program contents are filtered according to region, time period, holidays, expired content, and offline content.

7. The method for recommending online audio-visual content based on multiple roles according to claim 1, characterized in that: In step 5, the content popularity includes the number of views and the number of likes.

8. A multi-role based online audio-visual content recommendation system, characterized in that: The system is set up at the back end, and the back end maintains a role-label mapping table and a label-content category matching rule table, including: The front-end data receiving module is used to receive the user-selected role sent by the front-end; A content preference tag set query module is used to query a content preference tag set corresponding to the role from a pre-built tag set; the tag set is a set of recommended content category tags preset based on the attribute information of the role; a program content matching module, configured to establish a content tag index based on the content preference tag set, and search for all matching program contents based on the content tag index; A candidate recommendation set selection module is used to select multiple program contents from all program contents according to a preset time priority or popularity priority, and form a candidate recommendation set; A recommendation list generating module is used to remove duplicate program contents from the candidate recommendation set and sort the removed duplicate program contents according to content popularity to obtain a recommendation list; The data sending module is used to send the recommendation list to the front end for playing in sequence.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images