A method and apparatus for determining recommended content, and an electronic device

Abstract

The application discloses a method and device for determining recommended content and an electronic device. The method comprises: obtaining a plurality of first recommended content; determining first semantic similarity between the first recommended content; determining second recommended content using a large language model according to the first semantic similarity and first features of the first recommended content, wherein the second semantic similarity between the second recommended content and any first recommended content is less than the first semantic similarity; and displaying the second recommended content. The method eliminates the condition that the plurality of first recommended content has high homogeneity, makes the second recommended content more consistent with the content expected to be seen by the user, and greatly improves the accuracy and recommendation efficiency of the recommended content.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus and electronic device for determining recommended content. Background Technology

[0002] Currently, recommendation algorithms are increasingly used for content recommendation, such as in streaming media and product recommendations. However, existing recommendation algorithms suffer from problems such as overly singular optimization objectives (e.g., focusing solely on user clicks) and overly limited consideration of factors (e.g., having few attribute tags for recommended content). This leads to local optima, resulting in highly homogeneous recommendations received by users, and consequently, low accuracy in determining the recommended content. Summary of the Invention

[0003] The purpose of this application is to provide a method, apparatus, and electronic device for determining recommended content, which can solve the problem of high homogeneity of recommended content received by users.

[0004] In a first aspect, embodiments of this application provide a method for determining recommended content, including:

[0005] Get multiple top-recommended content;

[0006] Determine the first semantic similarity between the first recommended content;

[0007] Based on the first semantic similarity and the first feature of the first recommended content, the second recommended content is determined using a large language model. The second semantic similarity between the second recommended content and any of the first recommended content is less than the first semantic similarity.

[0008] The second recommended content will be displayed.

[0009] In one possible implementation, determining the second recommended content using a large language model based on the first semantic similarity and the first feature of the first recommended content includes:

[0010] Determine the comparison result between the first semantic similarity and the first preset similarity;

[0011] If the comparison result indicates that the first semantic similarity is greater than the first preset similarity, the first feature of the first recommended content is determined, and the second recommended content is determined using a large language model.

[0012] In one possible implementation, the determining method further includes:

[0013] The first recommended content is feature extracted using a large language model to obtain the first feature, which includes at least one of topic, role, and opinion.

[0014] In one possible implementation, determining the second recommended content using a large language model includes:

[0015] For each of the first features, a third semantic similarity is determined among the first features of multiple first recommended contents;

[0016] Based on the first feature whose third semantic similarity is greater than the second preset similarity, the second feature is determined using the large language model, and the fourth semantic similarity between the second feature and any of the first features is less than the second preset similarity.

[0017] Based on the second feature, the second recommended content is generated using the large language model, or based on the second feature, the second recommended content is obtained by filtering from the database.

[0018] In one possible implementation, the determining method further includes:

[0019] Collect user feedback information regarding the second recommended content, wherein the feedback information includes at least the user's browsing time and number of clicks on the second recommended content;

[0020] Based on the feedback information, a third recommended content is determined and displayed.

[0021] In one possible implementation, obtaining multiple first recommended contents includes:

[0022] If one of the first recommended content items is generated at the same time, then the first recommended content items within the cumulative time period are retrieved; or

[0023] If multiple first recommended content items are generated at the same time, then the preset number of first recommended content items at the current time are obtained.

[0024] In one possible implementation, determining the first semantic similarity between the first recommended content includes:

[0025] Based on different time periods, determine the top recommended content belonging to the same time period;

[0026] For the first recommended content belonging to the same time period, determine the first semantic similarity.

[0027] In one possible implementation, the determining method further includes:

[0028] Receive a first operation, the first operation including an operation performed on a preset area;

[0029] Based on the first operation, at least one first feature within the first recommendation is displayed, and each first feature is located in a different region;

[0030] Receive a second operation, the second operation including an operation performed on the region where the first feature is located;

[0031] Based on the second operation, the second recommended content is displayed.

[0032] Secondly, embodiments of this application also provide a device for determining recommended content, comprising:

[0033] The acquisition module is configured to acquire multiple first-recommended content items.

[0034] The first determining module is configured to determine the first semantic similarity between the first recommended content;

[0035] The second determining module is configured to determine the second recommended content using a large language model based on the first semantic similarity and the first feature of the first recommended content, wherein the second semantic similarity between the second recommended content and any of the first recommended content is less than the first semantic similarity.

[0036] The first display module is configured to display the second recommended content.

[0037] Thirdly, embodiments of this application also provide an electronic device, including: a processor and a memory, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via a bus, and when the machine-readable instructions are executed by the processor, the steps of the method for determining the recommended content described in any one of the above are performed.

[0038] Fourthly, embodiments of this application also provide a storage medium, wherein a computer-readable storage medium stores a computer program, which, when executed by a processor, performs the steps of the method for determining the recommended content described in any one of the above-mentioned embodiments.

[0039] In this embodiment, a large language model is used to determine the second recommended content based on the first semantic similarity between multiple first recommended content items and the first features of the first recommended content items. The second semantic similarity between the second recommended content item and any first recommended content item is less than the first semantic similarity, thereby eliminating the situation where multiple first recommended content items have high homogeneity. This makes the second recommended content item more in line with the content that users expect to see, greatly improving the accuracy and efficiency of the recommended content. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating a method for determining recommended content provided in this application is shown;

[0042] Figure 2 The flowchart illustrates a method for determining recommended content using a large language model to determine the second recommended content provided in this application.

[0043] Figure 3 A flowchart illustrating the content display using an electronic device in a method for determining recommended content provided in this application is shown.

[0044] Figure 4 This application provides a schematic diagram of the interface for displaying content in an electronic device.

[0045] Figure 5 A schematic diagram of the structure of a device for determining recommended content provided in this application is shown;

[0046] Figure 6 A schematic diagram of the structure of an electronic device provided in this application is shown. Detailed Implementation

[0047] Various embodiments and features of this application are described herein with reference to the accompanying drawings.

[0048] It should be understood that various modifications can be made to the embodiments described herein. Therefore, the above description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this application will be apparent to those skilled in the art.

[0049] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0050] These and other features of this application will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.

[0051] It should also be understood that although this application has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of this application, which have the features described in the claims and are therefore all within the scope of protection defined herein.

[0052] The above and other aspects, features and advantages of this application will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.

[0053] Specific embodiments of this application are described thereafter with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of this application, which can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that could obscure the application. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but merely serve as the basis and representative basis for the claims to teach those skilled in the art to use this application in a variety of substantially any suitable detailed structures.

[0054] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in other embodiments,” all of which may refer to one or more of the same or different embodiments according to this application.

[0055] To facilitate understanding of this application, a method for determining recommended content provided in this application will be described in detail first. Here, the execution subject of the embodiments of this application can be a processor or controller of an electronic device, or a server, etc.

[0056] Figure 1 A flowchart of the method for determining recommended content provided in the embodiments of this application is shown, wherein the specific steps include S101-S104.

[0057] S101, obtain multiple top recommended content.

[0058] In practice, a recommendation algorithm is pre-built to determine and display recommended content to the user. The recommended content determined by the algorithm represents content that the user might expect to see. For example, the recommendation algorithm is embedded in the Xiaohongshu application. When a user opens the Xiaohongshu application, the recommendation algorithm filters recommended content from the database according to its own calculation rules and displays that recommended content.

[0059] Among these methods, when multiple first-recommended content items are obtained, it can be connected to the recommendation algorithm to directly obtain the content output by the recommendation algorithm and use the content output by the recommendation algorithm as the first-recommended content; it can also directly collect the content in the currently displayed page as the first-recommended content, etc.

[0060] Here, the primary recommended content includes text, images, and either text or images.

[0061] S102, determine the first semantic similarity between the first recommended content.

[0062] After obtaining multiple primary recommended content items, semantic analysis is performed on these items to determine the primary semantic similarity among them. Optionally, for every two primary recommended content items, the sub-semantic similarity between them is calculated, resulting in multiple sub-semantic similarities. The average of these multiple sub-semantic similarities is then calculated to obtain the primary semantic similarity.

[0063] S103, based on the first semantic similarity and the first feature of the first recommended content, the second recommended content is determined using a large language model. The second semantic similarity between the second recommended content and any first recommended content is less than the first semantic similarity.

[0064] After obtaining the initial semantic similarity, it is further compared with a first preset similarity to determine the comparison result. If the comparison result shows that the initial semantic similarity is greater than the first preset similarity, it indicates that the multiple first recommended content items exhibit a high degree of homogeneity, meaning they represent an information cocoon phenomenon. If the comparison result shows that the initial semantic similarity is less than or equal to the first preset similarity, it indicates that the multiple first recommended content items have high accuracy and match the content expected by the user.

[0065] Furthermore, when it is determined that the first semantic similarity of the comparison result is greater than the first preset similarity, the first feature of the first recommended content is determined, and the second recommended content is determined using a large language model.

[0066] As one example, a Large Language Model (LLM) can be used to extract features from the first recommended content, resulting in a first feature. This first feature includes at least one of the following: topic, roles included in the first recommended content, and the viewpoint corresponding to the first recommended content. Of course, other semantic analysis algorithms can also be used to extract features from the first recommended content, as long as the first feature can be obtained.

[0067] After obtaining the first feature, a large language model is used to determine the second recommended content based on the first feature. Specifically, the second semantic similarity between the second recommended content determined by the large language model and any first recommended content is less than the first semantic similarity, meaning the second semantic similarity is less than the first preset similarity. This indicates that there is no high degree of homogeneity between the second recommended content and the first recommended content.

[0068] S104 displays the second recommended content.

[0069] After determining the second recommended content, the second recommended content is displayed so that users can see it. This updates the first recommended content and eliminates the situation where multiple first recommended content items are highly homogeneous.

[0070] As one example, Figure 2 A flowchart of a method for determining the second recommended content using a large language model is shown, wherein the specific steps include S201-S203.

[0071] S201, for each first feature, determine the third semantic similarity between the first features of multiple first recommended contents.

[0072] S202, based on the first feature whose third semantic similarity is greater than the second preset similarity, the second feature is determined using a large language model, and the fourth semantic similarity between the second feature and any first feature is less than the second preset similarity.

[0073] S203. Based on the second feature, generate second recommended content using a large language model, or based on the second feature, filter second recommended content from the database.

[0074] When determining the second recommended content, firstly, for each first feature, determine the third semantic similarity between multiple first recommended content items for that first feature. For example, for the first feature "topic," perform semantic analysis on the topics of any two first recommended content items to obtain the third semantic similarity of the topics of the first recommended content; for the first feature "opinion," perform semantic analysis on the opinions of any two first recommended content items to obtain the third semantic similarity of the opinions of the first recommended content, and so on.

[0075] After obtaining the third semantic similarity of each first feature, the third semantic similarity is compared with the second preset similarity. If the third semantic similarity is greater than the second preset similarity, it indicates that there is a high degree of homogeneity among the first features corresponding to the third semantic similarity. At this time, based on the first feature whose third semantic similarity is greater than the second preset similarity, the second feature is determined using a large language model. Optionally, the large language model can be used to generate features with lower semantic similarity to the first feature, or the large language model can be used to search and filter features with lower semantic similarity to the first feature from the database, etc. As long as the fourth semantic similarity between the determined second feature and any first feature is less than the second preset similarity, the embodiments of this application do not impose specific limitations on this.

[0076] After obtaining the second feature, a large language model is used to generate the second recommended content. For example, after determining that the second feature includes text such as topic, role, and opinion, the second recommended content can be obtained by processing the topic, role, and opinion using generation rules. This second recommended content is also text.

[0077] In another example, after obtaining the second feature, the second recommended content can be obtained by filtering from the database based on the second feature. For example, the second feature is the total number of pixels and the element "leaves". The database is searched for second recommended content that matches the total number of pixels and contains the element "leaves". In this case, the second recommended content is an image.

[0078] The second recommended content determined through the above steps is at least one, and the second semantic similarity between each second recommended content and any first recommended content is less than the first semantic similarity.

[0079] In this embodiment of the application, when obtaining multiple first recommended content items, it is first determined whether the recommendation algorithm outputs one or multiple items at a time. If the recommendation algorithm outputs one item at a time, it indicates that one first recommended content item is generated at the same time. In this case, the first recommended content items within the cumulative time period are obtained to obtain multiple first recommended content items. When the recommendation algorithm outputs multiple items, it indicates that multiple first recommended content items are generated at the same time. In this case, a preset number of first recommended content items at the current time are obtained.

[0080] Similarly, when retrieving multiple primary recommended content items, the number of items on the currently displayed page is identified. If the current displayed page contains only one item, such as an image, then the content from multiple displayed pages within the accumulated time period is retrieved to obtain multiple primary recommended content items. If the current displayed page contains multiple items, such as multiple text items, then a preset number of items are retrieved as multiple primary recommended content items.

[0081] The cumulative duration and preset quantity can be adjusted according to the application scenario of the recommendation algorithm, the attributes of the content, and the recommendation volume of the recommendation algorithm.

[0082] Considering that users may use different applications at different times, or browse different content within the same application at different times, when determining the first semantic similarity between recommended content, we can first identify the first recommended content belonging to the same time period based on different time periods. Then, we can determine the first semantic similarity for the first recommended content belonging to the same time period. This allows for a more accurate determination of whether the recommended content is highly homogeneous, avoiding misjudgments that waste resources (e.g., determining second recommended content when the first recommended content is already accurate), while also not affecting user experience and providing a higher level of satisfaction.

[0083] Furthermore, after displaying the second recommended content to the user, feedback information generated by the user regarding the second recommended content can be collected in real time. The feedback information includes at least the browsing time and the number of clicks on the second recommended content. When the browsing time of the second recommended content reaches a preset time, it indicates that the second recommended content is the content that the user expects to see; when the number of clicks on the second recommended content reaches a preset number, it indicates that the second recommended content is the content that the user expects to see.

[0084] The system uses feedback information to filter out the second set of recommended content that the user expects to see. This second set of recommended content is then used to update the recommendation algorithm, such as by adjusting its parameters, resulting in an updated algorithm. Finally, the updated algorithm is used to determine and display the third set of recommended content. This ensures that the third set of recommended content is more accurate than the first set and avoids highly homogenized content.

[0085] Regarding the above determination method, when displaying content using the display device of an electronic device, it can be referred to... Figure 3 The flowchart shown includes specific steps S301-S304.

[0086] S301, Receive a first operation, the first operation including an operation performed on a preset area.

[0087] S302, based on the first operation, display at least one first feature within the first recommendation, each first feature being located in a different region.

[0088] S303, receive a second operation, the second operation including an operation performed on the region where the first feature is located.

[0089] S304, based on the second operation, displays the second recommended content.

[0090] Combination Figure 4 The diagram illustrates the interface of the display device for content display. Optionally, this interface can be a webpage display interface, an application display interface, or a streaming media / social media display interface, etc. The display area corresponding to bubble A in the interface is a pre-defined preset area used to display the first recommended content. For example, when a user runs a video playback application on their electronic device, multiple first recommended contents are determined based on a recommendation algorithm when the user opens the application; these are the contents the user expects to see. This first recommended content is an identifier image for a video, and it is displayed in the display area corresponding to bubble A.

[0091] In practice, the user performs the first operation on the preset area, namely the display area corresponding to bubble A, such as moving the mouse to the area or clicking the touch screen corresponding to the area.

[0092] After the input device of the electronic device acquires the first operation, it sends the first operation to the processor or controller of the electronic device for subsequent processing.

[0093] After receiving a first operation, the processor or controller of the electronic device extracts features from the first recommended content using a large language model based on the first operation to obtain a first feature. This first feature can be one or more. Then, at least one first feature from the first recommendation is displayed via a display device, with each first feature located in a different area. Figure 4 The display area corresponding to bubble B shown is the area where the first feature is located. Of course, Figure 4 There are 3 bubbles B in the middle, but Figure 4 This is just one example; the number of bubbles B can be adjusted according to actual needs. For instance, when a user performs the first action, i.e., moves the mouse to the display area corresponding to bubble A, the display areas corresponding to the three bubbles B will respectively display the theme, role, and update date of the first recommended content. The theme, role, and update date are the first features of the first recommended content. Of course, the first features of the first recommended content can be different in different scenarios.

[0094] Furthermore, a second operation is received in real time. The second operation includes an operation performed on the area where the first feature is located. The second operation can be the same as the first operation, and the second operation is also collected and sent by the input device.

[0095] In response to the second operation, that is, based on the first feature corresponding to the second operation, a second feature is determined using a large language model. The fourth semantic similarity between the second feature and the first feature is less than the second preset similarity. Then, based on the second feature, a second recommended content is generated using the large language model, or based on the second feature, the second recommended content is obtained by filtering from the database. After obtaining the second recommended content, it is displayed in the display area corresponding to bubble C, so that users can clearly, accurately and efficiently obtain the second recommended content.

[0096] It is worth noting that the layout in the interface is not limited to Figure 4Alternatively, it can be in list format. For example, the entire display area of ​​the interface is divided into a left display area, a middle display area, and a right display area. The first recommended content is displayed sequentially from top to bottom in the left display area. After receiving a first operation, based on the first operation, the first feature of the first recommended content specified by the user is displayed in the middle display area. Then, a second operation is received, and based on the second operation, the second recommended content corresponding to the first feature specified by the user is displayed, and so on. Furthermore, this interface layout can be set by the electronic device manufacturer or by the user; this application embodiment does not specifically limit this.

[0097] Based on the same inventive concept, the second aspect of this application also provides a method for determining recommended content and a device for determining recommended content. Since the principle of the device for determining recommended content in this application is similar to the method for determining recommended content mentioned above, the implementation of the electronic device can refer to the implementation of the method, and the repeated parts will not be described again.

[0098] Figure 5 A schematic diagram of the device for determining recommended content provided in an embodiment of this application is shown, specifically including:

[0099] Module 501 is configured to retrieve multiple first-recommended content items;

[0100] The first determining module 502 is configured to determine the first semantic similarity between the first recommended content;

[0101] The second determining module 503 is configured to determine the second recommended content using a large language model based on the first semantic similarity and the first feature of the first recommended content, wherein the second semantic similarity between the second recommended content and any of the first recommended content is less than the first semantic similarity.

[0102] The first display module 504 is configured to display the second recommended content.

[0103] In yet another embodiment, the second determining module 503 is specifically configured as follows:

[0104] Determine the comparison result between the first semantic similarity and the first preset similarity;

[0105] If the comparison result indicates that the first semantic similarity is greater than the first preset similarity, the first feature of the first recommended content is determined, and the second recommended content is determined using a large language model.

[0106] In yet another embodiment, the apparatus for determining recommended content further includes an extraction module 505, configured as follows:

[0107] The first recommended content is feature extracted using a large language model to obtain the first feature, which includes at least one of topic, role, and opinion.

[0108] In yet another embodiment, the second determining module 503 is further configured to:

[0109] For each of the first features, a third semantic similarity is determined among the first features of multiple first recommended contents;

[0110] Based on the first feature whose third semantic similarity is greater than the second preset similarity, the second feature is determined using the large language model, and the fourth semantic similarity between the second feature and any of the first features is less than the second preset similarity.

[0111] Based on the second feature, the second recommended content is generated using the large language model, or based on the second feature, the second recommended content is obtained by filtering from the database.

[0112] In yet another embodiment, the device for determining recommended content further includes a third determining module 506, configured as follows:

[0113] Collect user feedback information regarding the second recommended content, wherein the feedback information includes at least the user's browsing time and number of clicks on the second recommended content;

[0114] Based on the feedback information, a third recommended content is determined and displayed.

[0115] In yet another embodiment, the acquisition module 501 is specifically configured as follows:

[0116] If one of the first recommended content items is generated at the same time, then the first recommended content items within the cumulative time period are retrieved; or

[0117] If multiple first recommended content items are generated at the same time, then the preset number of first recommended content items at the current time are obtained.

[0118] In yet another embodiment, the first determining module 502 is specifically configured as follows:

[0119] Based on different time periods, determine the top recommended content belonging to the same time period;

[0120] For the first recommended content belonging to the same time period, determine the first semantic similarity.

[0121] In yet another embodiment, the device for determining recommended content further includes a second display module 507, configured as follows:

[0122] Receive a first operation, the first operation including an operation performed on a preset area;

[0123] Based on the first operation, at least one first feature within the first recommendation is displayed, and each first feature is located in a different region;

[0124] Receive a second operation, the second operation including an operation performed on the region where the first feature is located;

[0125] Based on the second operation, the second recommended content is displayed.

[0126] In this embodiment, a large language model is used to determine the second recommended content based on the first semantic similarity between multiple first recommended content items and the first features of the first recommended content items. The second semantic similarity between the second recommended content item and any first recommended content item is less than the first semantic similarity, thereby eliminating the situation where multiple first recommended content items have high homogeneity. This makes the second recommended content item more in line with the content that users expect to see, greatly improving the accuracy and efficiency of the recommended content.

[0127] This application provides a storage medium, which is a computer-readable medium, storing a computer program. When executed by a processor, the computer program implements the method provided in any embodiment of this application, including the following steps S11 to S14:

[0128] S11, obtain multiple top-recommended content;

[0129] S12, determine the first semantic similarity between the first recommended content;

[0130] S13, based on the first semantic similarity and the first feature of the first recommended content, the second recommended content is determined using a large language model, wherein the second semantic similarity between the second recommended content and any of the first recommended content is less than the first semantic similarity;

[0131] S14, display the second recommended content.

[0132] In this embodiment, a large language model is used to determine the second recommended content based on the first semantic similarity between multiple first recommended content items and the first features of the first recommended content items. The second semantic similarity between the second recommended content item and any first recommended content item is less than the first semantic similarity, thereby eliminating the situation where multiple first recommended content items have high homogeneity. This makes the second recommended content item more in line with the content that users expect to see, greatly improving the accuracy and efficiency of the recommended content.

[0133] This application provides an electronic device, the structural schematic diagram of which is shown below. Figure 6As shown, it includes at least a memory 601 and a processor 602. The memory 601 stores a computer program, and the processor 602 implements the method provided in any embodiment of this application when executing the computer program in the memory 601. Exemplarily, the steps of the electronic device computer program are as follows: S21 to S24:

[0134] S21, retrieve multiple top-recommended content;

[0135] S22, determine the first semantic similarity between the first recommended content;

[0136] S23, based on the first semantic similarity and the first feature of the first recommended content, the second recommended content is determined using a large language model, wherein the second semantic similarity between the second recommended content and any of the first recommended content is less than the first semantic similarity;

[0137] S24, display the second recommended content.

[0138] In this embodiment, a large language model is used to determine the second recommended content based on the first semantic similarity between multiple first recommended content items and the first features of the first recommended content items. The second semantic similarity between the second recommended content item and any first recommended content item is less than the first semantic similarity, thereby eliminating the situation where multiple first recommended content items have high homogeneity. This makes the second recommended content item more in line with the content that users expect to see, greatly improving the accuracy and efficiency of the recommended content.

[0139] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk. Optionally, in this embodiment, the processor executes the method steps described in the above embodiments according to the program code stored in the storage medium. Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, which will not be repeated here. Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed on a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be executed in a different order than those described here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any specific hardware and software combination.

[0140] Furthermore, although exemplary embodiments have been described herein, their scope includes any and all embodiments based on this application that have equivalent elements, modifications, omissions, combinations (e.g., schemes involving intersections of various embodiments), adaptations, or alterations. Elements in the claims will be interpreted broadly based on the language used in the claims and are not limited to the examples described in this specification or during the implementation of this application, which will be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered illustrative only, and the true scope and spirit are indicated by the full scope of the following claims and their equivalents.

[0141] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. Other embodiments can be used by those skilled in the art when reading the above description. Furthermore, in the above detailed description, various features may be grouped together to simplify the application. This should not be construed as an intention that a disclosed feature not claimed is necessary for any claim. Rather, the subject matter of the application may be less than all the features of a particular disclosed embodiment. Thus, the following claims are incorporated herein by reference as examples or embodiments, wherein each claim is an independent, separate embodiment, and these embodiments are contemplated to be combined with each other in various combinations or arrangements. The scope of this application should be determined by reference to the appended claims and the full scope of their equivalents.

[0142] The foregoing has described in detail several embodiments of this application, but this application is not limited to these specific embodiments. Those skilled in the art can make various variations and modifications based on the concept of this application, and all such variations and modifications should fall within the scope of protection claimed in this application.

Claims

1. A method for determining recommended content, comprising: obtaining a plurality of first recommended content; determining first semantic similarities between the first recommended content; obtaining first features of the first recommended content according to the first semantic similarities, the first features being obtained when the first semantic similarities are greater than first preset similarities; determining second features by using a large language model based on the first features, to obtain second recommended content based on the second features, the second recommended content having second semantic similarities with any of the first recommended content being less than the first semantic similarities; displaying the second recommended content.

2. The method of claim 1, wherein the obtaining first features of the first recommended content according to the first semantic similarities comprises: determining a comparison result between the first semantic similarities and the first preset similarities; and determining the first features of the first recommended content if the comparison result indicates that the first semantic similarities are greater than the first preset similarities.

3. The method of claim 1 or 2, further comprising: extracting the first features by using the large language model on the first recommended content, the first features including at least one of a theme, a role, and a viewpoint.

4. The method of claim 1, wherein the determining second features by using the large language model based on the first features, to obtain second recommended content based on the second features comprises: determining third semantic similarities between the first features of the plurality of first recommended content for each of the first features; and determining second features by using the large language model based on the first features having third semantic similarities greater than second preset similarities, the second features having fourth semantic similarities with any of the first features being less than the second preset similarities; and generating the second recommended content by using the large language model based on the second features, or screening the second recommended content from a database based on the second features.

5. The method of claim 1, further comprising: collecting feedback information generated by a user for the second recommended content, the feedback information including at least a browsing duration and a click count of the second recommended content by the user; and determining and displaying third recommended content based on the feedback information.

6. The method of claim 1, wherein the obtaining a plurality of first recommended content comprises: obtaining the first recommended content within a cumulative duration if one of the first recommended content is generated at the same time; or obtaining a preset number of first recommended content at a current time if a plurality of the first recommended content are generated at the same time.

7. The method of claim 1, wherein the determining first semantic similarities between the first recommended content comprises: determining first recommended content belonging to a same time period based on different time periods; and determining the first semantic similarities for the first recommended content belonging to the same time period.

8. The method of claim 1, further comprising: receiving a first operation, the first operation including an operation performed on a preset area. ​ Based on the first operation, at least one first feature of the first recommended content is displayed, each of the first features being in a different area; A second operation is received, the second operation including an operation performed on the area where the first feature is located; Based on the second operation, the second recommended content is displayed.

9. A recommended content determination apparatus, comprising: an acquisition module configured to acquire a plurality of first recommended contents; a first determination module configured to determine first semantic similarities between the first recommended contents; a second determination module configured to obtain, according to the first semantic similarities, first features of the first recommended contents, the first features being obtained when the first semantic similarities are greater than a first preset similarity; determine, based on the first features, second features by using a large language model, and obtain, based on the second features, second recommended contents, the second semantic similarities between the second recommended contents and any of the first recommended contents being less than the first semantic similarities; a first display module configured to display the second recommended contents.

10. An electronic device comprising: a processor and a memory, the memory storing machine readable instructions executable by the processor, when the electronic device is running, the processor and the memory communicate through the bus, the machine readable instructions executed by the processor perform the steps of the recommended content determination method according to any one of claims 1 to 8.

Images