Information providing method, display method, apparatus, device, and storage medium

Abstract

The application discloses an information providing method, a display method, apparatuses, devices and a storage medium, and relates to the technical field of computers. The method comprises the following steps: after receiving an object recommendation request sent by a client, obtaining recommendation information of a recommended object, wherein the object recommendation request is a request sent by the client after receiving an operation of viewing comment information of first display content; determining a first emoticon based on the recommendation information; adding the first emoticon to the recommendation information to obtain imitated comment information, wherein the imitated comment information is used for synchronous display with at least one piece of comment information of the first display content; and sending the imitated comment information to the client. When the object recommendation is performed in the comment interface through the above method, the recommendation information will be added with the matched emoticon (the first emoticon), so that the display content is more abundant.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an information provision method, display method, apparatus, device, and storage medium. Background Technology

[0002] Recommendation information is information used to recommend objects (such as advertisements). The objects recommended by recommendation information include products, events, applications, stores, accounts, etc. Therefore, recommendation information has a wide range of applications in various scenarios.

[0003] For example, in the comment interface of displayed content (such as short videos), the recommendation information (such as advertising copy, images of recommended objects, etc.) provided by the supplier (such as advertisers) is often directly displayed.

[0004] However, when performing object recommendations in the comments interface, the content displayed is rather limited as it only shows recommendations provided by the aforementioned suppliers. Summary of the Invention

[0005] This application provides an information providing method, display method, apparatus, device, and storage medium. The technical solutions provided by this application are as follows:

[0006] According to one aspect of the embodiments of this application, an information display method is provided, the method comprising:

[0007] In response to an operation to view comment information of the first displayed content, at least one comment of the first displayed content is obtained, as well as simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes recommendation information of the recommended object and a first emoticon added to the recommendation information, the first emoticon being determined based on the recommendation information;

[0008] A comment interface is displayed that shows the first displayed content, and the comment interface displays the at least one comment and the simulated comment.

[0009] In response to the operation on the simulated comment information, the associated interface of the recommended object is displayed.

[0010] According to one aspect of the embodiments of this application, an information providing method is provided, the method comprising:

[0011] After receiving an object recommendation request from the client, the recommendation information of the recommended object is obtained. The object recommendation request is a request sent by the client after receiving the operation of viewing the comment information of the first displayed content.

[0012] The first emoji is determined based on the recommended information;

[0013] The first emoji is added to the recommendation information to obtain pseudo-comment information, wherein the pseudo-comment information is used to be displayed synchronously with at least one comment on the first displayed content;

[0014] The simulated comment information is sent to the client.

[0015] According to one aspect of the embodiments of this application, an information display device is provided, the device comprising:

[0016] The acquisition module is used to acquire at least one comment of the first displayed content in response to the operation of viewing the comment information of the first displayed content, as well as simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes the recommendation information of the recommended object and a first emoticon added to the recommendation information, the first emoticon being determined based on the recommendation information;

[0017] A comment display module is used to display a comment interface for the first displayed content, and to display the at least one comment and the simulated comment information in the comment interface;

[0018] The associated display module is used to display the associated interface of the recommended object in response to the operation on the simulated comment information.

[0019] According to one aspect of the embodiments of this application, an information providing apparatus is provided, the apparatus comprising:

[0020] The acquisition module is used to acquire the recommendation information of the recommended object after receiving the object recommendation request sent by the client. The object recommendation request is a request sent by the client after receiving the operation of viewing the comment information of the first displayed content.

[0021] The determining module is used to determine the first emoji based on the recommendation information;

[0022] An add module is used to add the first emoji to the recommendation information to obtain simulated comment information, wherein the simulated comment information is used to be displayed synchronously with at least one comment on the first displayed content;

[0023] The sending module is used to send the simulated comment information to the client.

[0024] According to one aspect of the embodiments of this application, a computer device is provided, the computer device including a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the above-described information display method, or to implement the above-described information provision method.

[0025] According to one aspect of the embodiments of this application, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the above-described information display method or the above-described information provision method.

[0026] According to one aspect of the embodiments of this application, a computer program product is provided, the computer program product including a computer program stored in a computer-readable storage medium, wherein a processor reads from the computer-readable storage medium and executes the computer program to implement the above-described information display method, or to implement the above-described information provision method.

[0027] The technical solutions provided in this application have at least the following beneficial effects:

[0028] Upon receiving an object recommendation request from the client (triggered by viewing the comment information of the first displayed content), the system retrieves the recommendation information for the recommended object and adds a first emoticon determined based on this information to the recommendation information, thus generating pseudo-comment information. This pseudo-comment information is displayed synchronously with the comment information of the first displayed content. Therefore, when object recommendations are performed in the comment interface (i.e., the interface used to present comment information) using the above method, the recommendation information will include a matching emoticon (the first emoticon), resulting in richer displayed content. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of a computer system provided in one embodiment of this application;

[0030] Figure 2 This is a flowchart of an information display method provided in one embodiment of this application;

[0031] Figure 3 This is a schematic diagram illustrating the addition of a first emoji to text recommendation information according to one embodiment of this application;

[0032] Figure 4 This is a flowchart of an information display method provided in another embodiment of this application;

[0033] Figure 5 This is a schematic diagram of a comment interface provided in one embodiment of this application;

[0034] Figure 6 This is a schematic diagram of a comment interface provided in another embodiment of this application;

[0035] Figure 7 This is a flowchart of an information provision method provided in one embodiment of this application;

[0036] Figure 8This is a schematic diagram of an information provision method provided in one embodiment of this application;

[0037] Figure 9 This is a flowchart of an information provision method provided in another embodiment of this application;

[0038] Figure 10 This is a schematic diagram illustrating the correspondence between emojis and their codes according to an embodiment of this application;

[0039] Figure 11 This is a flowchart of an information provision method provided in another embodiment of this application;

[0040] Figure 12 This is a flowchart of a language model training method provided in one embodiment of this application;

[0041] Figure 13 This is a schematic diagram illustrating the construction of a large language model provided in one embodiment of this application;

[0042] Figure 14 This is a schematic diagram of sample engineering processing provided in one embodiment of this application;

[0043] Figure 15 This is a schematic diagram of an information provision scheme provided in one embodiment of this application;

[0044] Figure 16 This is a flowchart of an information display method in a short video scenario provided by an embodiment of this application;

[0045] Figure 17 This is a flowchart of an embodiment of the information provision method in a short video scenario provided by this application;

[0046] Figure 18 This is a block diagram of an information display device provided in one embodiment of this application;

[0047] Figure 19 This is a block diagram of an information providing device according to an embodiment of this application;

[0048] Figure 20 This is a structural block diagram of a computer device provided in one embodiment of this application. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0050] Before introducing the technical solutions of this application, some terms involved in this application will be explained. The following related explanations are optional solutions and can be arbitrarily combined with the technical solutions of the embodiments of this application, all of which fall within the protection scope of the embodiments of this application. The embodiments of this application include at least some of the following contents.

[0051] Comments interface: An interface used to display and allow users to post comments on a specific piece of content.

[0052] Comment section ads: A type of advertising that displays ad content in the comment section, usually in the form of a comment.

[0053] Native recommendation information: Integrate recommendation information (such as advertisements) with content on the platform that displays the recommendation information (such as video comments), making the recommendation information resemble part of the platform's content.

[0054] Emoji: A small digital image used to express emotions or thoughts in electronic messaging and social media, such as emojis.

[0055] Transformer: The Transformer is a network structure in deep learning based on the attention mechanism. This mechanism allows AI (Artificial Intelligence) models to learn the weights between different positions in an input sequence, and it is widely used in natural language processing and computer vision.

[0056] Pre-trained language model: an artificial intelligence model, also known as a large language model (LLM), which is trained on a large dataset to learn general knowledge and feature representations, and is therefore widely used in various natural language processing tasks.

[0057] Native recommendation information is an important means to improve the effectiveness of recommendation information, and it has a wide range of applications in some advertising scenarios, such as:

[0058] Native splash screen ads: Design splash screen ads to resemble the application interface, making them less noticeable to users.

[0059] Native video ad integration: seamlessly integrates video ad content with video content, ensuring the ad content is relevant to the video content and providing a better user experience.

[0060] Currently, many platforms offer comment functions to users, allowing them to post comments on content displayed on the platform (such as videos, products, stores, and social media content posted by other users). Consequently, there is an increasing demand for object recommendations within the comment interface.

[0061] Based on this, this application provides a native recommendation information solution for use in comment interfaces. This solution enhances the nativeness of recommendation information by adding emojis, making the information more attractive and interactive. By leveraging the visual appeal and emotional expressiveness of emojis, it strengthens the emotional connection between the recommendation information and the user, increasing click-through rates and user engagement. Simultaneously, by reducing interference with user attention, it enhances the natural dissemination of recommendation information. The following embodiments will provide a more detailed description of the above solution.

[0062] Please refer to Figure 1 This illustration shows a schematic diagram of a computer system provided in one embodiment of this application. The computer system may include: a terminal device 10 and a server 20.

[0063] Terminal device 10 can be an electronic device such as a mobile phone, tablet computer, multimedia playback device, PC (Personal Computer), wearable device, in-vehicle terminal device, VR (Virtual Reality) device, AR (Augmented Reality) device, MR (Mixed Reality) device, etc. Terminal device 10 can install and run a client application for the target application, which can be any application that supports displaying a comment interface, such as a social application, shopping application, video application (such as a short video application), news application, food delivery application, etc.

[0064] Server 20 can be used to provide background services for clients of target applications (such as social applications) in terminal device 10. Server 20 can be a single server, a server cluster consisting of multiple servers, or a cloud computing service center. In some embodiments, server 20 includes a background server 21 for the client of the aforementioned target application (such as a social application) and an inference server 22. The background server 21 receives an object recommendation request sent by terminal device 10 and, in response to the request, retrieves recommendation information from a recommendation information database (such as an advertising database). Then, the inference server 22 infers the emojis to be added and their placement based on the recommendation information retrieved by the background server 21. Finally, the background server 21 adds an emoji (a first emoji) to the recommendation information based on the inference result of the inference server 22, obtaining simulated comment information and returning it to terminal device 10. Terminal device 10 can then display native recommendation information with added emojis in the comment interface.

[0065] Terminal device 10 and server 20 can communicate via a network, such as a wireless or wired network.

[0066] Furthermore, in this embodiment, the implementation form of the target application is not limited. For example, it can be an application that needs to be downloaded and installed, a mini-program that does not require installation, a web application, etc.

[0067] Please refer to Figure 2 The diagram illustrates a flowchart of an information display method according to an embodiment of this application. The execution entity for each step of the method is a computer device, such as terminal device 10 (e.g., a client of a target application in terminal device 10). The method includes at least one of the following steps 210 to 230.

[0068] Step 210: In response to the operation of viewing the comment information of the first displayed content, obtain at least one comment information of the first displayed content, and simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes the recommendation information of the recommended object and a first emoticon added to the recommendation information, the first emoticon being determined based on the recommendation information.

[0069] The first display content is any display content (in the target application) that supports (user) comments. For example, the first display content includes video content (such as short videos), information content (such as news), social content posted by user accounts, product content, event content, etc. This application does not limit the specific form of the first display content.

[0070] In some embodiments, the operation of viewing comment information of the first displayed content is an operation on a comment control, which is used to trigger the operation of viewing comment information of the first displayed content. It should be noted that, in the embodiments of this application, the operation on the displayed content (such as controls, simulated comment information, etc.) mentioned can be any operation on the displayed content, such as clicking, long pressing (an operation where the pressing time exceeds a set threshold, the set threshold is set by the technician as needed, such as 0.3s, 0.5s, 1s, etc.), swiping, etc., which will not be described in detail in the following embodiments.

[0071] For example, the first displayed content includes video content, and the method further includes: displaying a video playback interface, in which the video content and the aforementioned comment control are displayed. Step 210 includes: in response to an operation on the comment control, obtaining at least one comment from the first displayed content, and simulated comment information displayed synchronously with the comment information.

[0072] It should be noted that the operation of viewing the comment information of the first displayed content is not limited to being triggered by the aforementioned comment control. For example, if the first displayed content is video content, the operation of viewing the comment information of the first displayed content could also be clicking on the video content to pause the video content. Or, for example, if the first displayed content is video content and the comment information is bullet comments displayed on the video content, then the operation of viewing the comment information of the first displayed content could be playing the video content. This application does not limit the specific form of the operation of viewing the comment information of the first displayed content.

[0073] Comment information for the first displayed content is used to comment on the first displayed content, and it can include text, emojis, images, etc.

[0074] In some embodiments, comments on the first displayed content are posted by a user account (of the target application).

[0075] The recommendation information of the recommended object is used to recommend the recommended object. The recommended object is any object that users can learn about, obtain or use. For example, the recommended object includes products, merchants, accounts, activities, applications, enterprises, etc.

[0076] In some embodiments, the recommendation information includes text recommendation information, which is used to recommend the recommended objects in natural language text, and the first emoji is determined based on the text recommendation information.

[0077] For example, the copywriting recommendation message might look like this: "XX product, direct from the manufacturer! No middlemen, cheap and affordable." As you can see, the copywriting recommendation message includes not only the text itself but also punctuation marks to indicate pauses and tone, thus providing fluent and readable natural language text.

[0078] In the above embodiment, the first emoji is determined based on the text recommendation information. Therefore, the first emoji can match the semantics and emotions conveyed by the text recommendation information, thus integrating into the recommendation information more harmoniously.

[0079] In some embodiments, the first emoji is added to the header of the text recommendation. For example, please refer to... Figure 3 In scenario 1, the first emoji 49 is added to the header of the copy recommendation information 48 (XX product, factory direct! No middlemen, cheap and affordable).

[0080] In some embodiments, the first emoji is added to the end of the text recommendation. For example, please refer to... Figure 3 In scenario 2, the first emoji 49 is added to the end of the copy recommendation information 48.

[0081] In some embodiments, the first emoji is added between the characters included in the text recommendation information. For example, please refer to... Figure 3 In scenario 3, the first emoji 49 is added between the characters included in the copy recommendation information 48.

[0082] In the above embodiment, the first emoji is added to the text recommendation information (at the beginning, end, or middle of the text recommendation information). This addition method does not change the original text of the text recommendation information, which not only ensures the integrity of the text recommendation information, but also enhances the emotional expression of the text recommendation information through the first emoji, making the text recommendation information closer to the comment text posted by the user, and ultimately improving the acceptance and click-through rate of the imitation comment information.

[0083] In some embodiments, the position of the first emoji within the text recommendation information is determined based on the text recommendation information itself. Therefore, the first emoji can be added to a suitable position within the text recommendation information, allowing it to enhance the emotional expression of the text recommendation information and promote its native feel, while not affecting the original semantic expression of the text recommendation information. This avoids adding the first emoji affecting the original punctuation of the text recommendation information, thereby improving the user's reading experience.

[0084] Step 220: Display the comment interface of the first displayed content, showing at least one comment and a simulated comment.

[0085] In some embodiments, the display style (or style) of the pseudo-comment information is consistent with the display style (or style) of the comment information, and is therefore referred to as pseudo-comment information.

[0086] In some embodiments, the display style (or style) of the pseudo-comment information is consistent with that of the comment information, which is reflected in the fact that the text recommendation information contained in the pseudo-comment information is consistent with the size, color and font of the characters in the comment information.

[0087] In some embodiments, the display style (or style) of the pseudo-comment information is consistent with that of the comment information, which is reflected in the different items displayed sequentially in the comment interface by the pseudo-comment information and the comment information.

[0088] In some embodiments, please refer to Figure 4 Step 220 includes step 222: displaying a comment interface for the first displayed content, in which at least two items are displayed in sequence, and each item displays a comment or a pseudo-comment.

[0089] An entry refers to a sub-item listed according to its content, and each entry occupies an independent area in the comment interface.

[0090] Displaying at least two items sequentially in the comment interface can mean displaying at least two items from top to bottom or from left to right; this application does not limit this.

[0091] In some embodiments, each entry displays a comment and the identifier of the user account that posted the comment, or it displays the identifier of the provider of the fake comment and recommendation information. That is to say, both the comment and the fake comment have corresponding identifiers (user account identifier and provider identifier), which also reflects the consistency of their display styles.

[0092] User account identification information is used to indicate the user account, including the user account name, avatar, number, etc. Provider identification information is used to indicate the provider, including the provider's name, trademark, avatar, etc.

[0093] For example, please refer to Figure 5 The first displayed content is video content 41. In the comment interface 42 of video content 41, at least two entries 43 are displayed from top to bottom. Each entry 43 displays a comment message 44 and the identification information 45 of the user account that posted the comment message 44. Alternatively, it displays pseudo-comment information 46 and the identification information 47 of the provider of the recommendation information. The pseudo-comment information 46 includes text recommendation information 48 and a first emoticon 49 added to the text recommendation information 48.

[0094] In the above embodiment, the pseudo-comment information and the comment information are distributed among the items displayed sequentially in the comment interface. Therefore, the pseudo-comment information can be displayed to the user in the form of comment information, reducing the awkwardness of directly displaying recommendation information and improving the nativeness of the pseudo-comment information.

[0095] In some embodiments, when the number of comment messages is greater than 1, the pseudo-comment message is displayed between any two comment messages, and the display style of the pseudo-comment message is consistent with the display style of the comment message.

[0096] In some embodiments, the entry containing the pseudo-comment information is located between any two entries containing comment information, and the display style of the pseudo-comment information is consistent with the display style of the comment information.

[0097] For example, please refer to Figure 5When the number of comment information 44 is greater than 1, imitation comment information 46 is displayed between the two comment information 44, and the display style of imitation comment information 46 is consistent with that of comment information 44. For example, the font, color, and size of the copywriting recommendation information 48 contained in imitation comment information 46 are consistent with the font, color, and size of the characters in comment information 44. Imitation comment information 46 and comment information 44 are located in different entries 43 displayed in sequence in the comment interface 42. In the entry 43 where comment information 44 is located, the identification information 45 of the user account that published comment information 44 is displayed, thereby indicating the publisher of comment information 44. Similarly, in the entry 43 where imitation comment information 46 is located, the identification information 47 of the provider of recommendation information is displayed, thereby indicating the publisher of imitation comment information 46.

[0098] In the above embodiment, when the number of comment messages is greater than one, the pseudo-comment message is displayed between any two comment messages, and the display style of the pseudo-comment message is consistent with the display style of the comment messages, that is, the two have the same style. Therefore, the pseudo-comment message can be naturally integrated into the comment messages in the comment interface, reducing interference with the user's attention and improving the native effect of the recommendation information.

[0099] In some embodiments, the copywriting recommendations in the simulated comment information support interactive features.

[0100] In some embodiments, in response to an operation on the copywriting recommendation information, the updated copywriting recommendation information and a second emoji added to the updated copywriting recommendation information are displayed in the comment interface, the second emoji being determined based on the updated copywriting recommendation information.

[0101] In some embodiments, the updated copywriting recommendation information is still used to recommend the aforementioned recommended objects.

[0102] For example, please refer to Figure 5 In response to an action (such as clicking) on ​​the copy recommendation information 48, the updated copy recommendation information and the second emoji added to the updated copy recommendation information are displayed in the entry 43 where copy recommendation information 48 is located. Figure 5 (Not shown in the image).

[0103] In the above embodiment, users can change the text recommending the above-mentioned recommended objects by operating on the text recommendation information, thereby improving the interactivity of the text recommendation information.

[0104] In some embodiments, please refer to Figure 6 In item 43, where comment information 44 is located, there is also reply information 51 for comment information 44. Similarly, copywriting recommendation information 48 can also be designed to support user account replies.

[0105] In some embodiments, in response to the operation of replying to copywriting recommendation information, reply information for copywriting recommendation information is displayed in the comment interface.

[0106] In some embodiments, in response to an operation on the copywriting recommendation information, a reply bar and a reply control are displayed. The reply bar is used to input reply information on the copywriting recommendation information, and the reply control is used to confirm the reply to the copywriting recommendation information. If reply information on the copywriting recommendation information is entered in the reply bar, in response to an operation on the reply control, the reply information on the copywriting recommendation information is displayed in the comment interface.

[0107] Step 230: In response to the operation on the simulated comment information, display the associated interface of the recommended object.

[0108] The associated interfaces of a recommended item are any interfaces related to that item. For example, if the recommended item is a product, the associated interfaces include the product purchase interface and the product details interface. Similarly, if the recommended item is an activity, the associated interfaces include the activity participation interface. And if the recommended item is an application, the associated interfaces include the application download interface.

[0109] In some embodiments, the recommendation information further includes redirection recommendation information, which is used to redirect to the associated interface of the recommended object. The redirection recommendation information may be an image, link, video, etc., and this application does not limit this. Step 230 includes: in response to an operation on the redirection recommendation information, displaying the associated interface of the recommended object.

[0110] For example, please refer to Figure 5 In addition to the text recommendation information 48, the recommendation information in the imitation comment information 46 also includes the jump recommendation information 50. In response to the click operation on the jump recommendation information 50, the associated interface of the recommended object is displayed.

[0111] In the above embodiments, the recommendation information also provides separate jump recommendation information for jumping to the associated interface of the recommended object, which can reduce the probability of users accidentally entering the associated interface and avoid the influence of fake comment information on users' reading and interaction with real comment information.

[0112] The technical solution provided in this application, upon receiving an operation to view comment information of the first displayed content, acquires both the comment information and simulated comment information of the first displayed content, and displays both synchronously in the comment interface of the first displayed content. The simulated comment information includes not only the recommendation information of the recommended object, but also a first emoticon determined based on that recommendation information. Therefore, when object recommendations are performed in the comment interface using the above method, the recommendation information will include a matching emoticon (the first emoticon), resulting in richer displayed content.

[0113] In addition, the above method targets the comment interface, using emojis, a commonly used element, to increase the sense of authenticity. Without changing the original recommendation information (text recommendation information), adding emojis can create pseudo-comment information, thereby improving the integration of recommendation information and comment information. This maintains the integrity of the recommendation information while allowing it to be well integrated into users' daily communication, achieving higher acceptance and click-through rates.

[0114] The above embodiments described the process of displaying simulated comment information. The following embodiments will provide a more detailed explanation of the process of providing the simulated comment information. The following embodiments are interchangeable with the above embodiments; for any details not described in detail in one embodiment, please refer to the other embodiment.

[0115] Please refer to Figure 7 The diagram illustrates a flowchart of an information provision method according to an embodiment of this application. The execution entity for each step of the method is a computer device, such as server 20. The method includes at least one of the following steps 610 to 640.

[0116] Step 610: After receiving the object recommendation request sent by the client, obtain the recommendation information of the recommended object. The object recommendation request is a request sent by the client after receiving the operation to view the comment information of the first displayed content.

[0117] An object recommendation request is used to request the execution of object recommendations. In some embodiments, the object recommendation request is the same as a comment retrieval request sent by the client to retrieve comment information. In some embodiments, the object recommendation request and the comment retrieval request are different requests.

[0118] In some embodiments, the server includes a backend server for the client of the target application described above. Upon receiving an object recommendation request from the client, the backend server obtains recommendation information for the recommended objects.

[0119] In some embodiments, please refer to Figure 8After receiving an object recommendation request from the client, the backend server 21 retrieves at least two recommendation information entries from the recommendation information database 23 (such as an advertising database), and then sorts and filters these entries to obtain the adopted recommendation information. This application does not limit the specific method of sorting and filtering the at least two recommendation information entries. For example, the at least two recommendation information entries may have pre-configured priorities, and the recommendation information with the highest priority may be selected for adoption. Alternatively, they may be sorted by click-through rate, and the recommendation information with the highest click-through rate may be selected for adoption. The recommendation information database 23 may be deployed on the backend server 21 or on another device; this application does not limit its deployment in this regard.

[0120] Step 620: Determine the first emoji based on the recommendation information.

[0121] In some embodiments, the recommendation information includes text recommendation information, which is used to recommend the aforementioned recommended objects in natural language text. Please refer to... Figure 9 Step 620 includes at least one of the following sub-steps 622 to 624.

[0122] Sub-step 622: Input the copy recommendation information into the language model, and output the first code from the language model. The first code is used to indicate the first emoji from at least two emojis.

[0123] In some embodiments, please refer to Figure 8 The server also includes an inference server 22, which is different from the aforementioned backend server 21. The inference server 22 is equipped with the language model described above. The backend server 21 extracts copywriting recommendation information from the recommendation information and sends the copywriting recommendation information to the inference server 22. The inference server 22 inputs the copywriting recommendation information into the language model, and the language model outputs a first code. The inference server 22 then sends the first code to the backend server 21.

[0124] At least two emojis are supported for use in this application. Because there are a large number of emojis (such as emoticons), but users only frequently use a limited number, emojis can be sorted according to their usage frequency (e.g., their frequency of appearance in short video comment sections), and the emojis with the highest usage frequency can be identified as those supported for use in generating simulated comment information. For example, the top 20 most frequently used emojis can be identified as at least two emojis.

[0125] A language model refers to an artificial intelligence model with natural language processing capabilities. To facilitate the language model's inference about emojis, at least two supported emojis can be pre-encoded. For example, the aforementioned 20 emojis can be encoded as 1-20, with a one-to-one correspondence between emojis and numbers. For an example, please refer to... Figure 10 The emoticon [smile] corresponds to the number 1, the emoticon [pout] corresponds to the number 2, the emoticon [curious] corresponds to the number 3, the emoticon [dazed] corresponds to the number 4, ..., and the emoticon [okay] corresponds to the number 20. In this example, if the first code output by the language model is 1, then the first emoticon [smile] can be indicated from the above 20 emoticons.

[0126] In some embodiments, please refer to Figure 11 Sub-step 622 includes the following sub-step 623.

[0127] Sub-step 623: Input the copy recommendation information into the language model, and the language model outputs a first code and a second code. The second code is used to indicate the position in the copy recommendation information where the first emoji can be added.

[0128] In some embodiments, the inference server inputs the text recommendation information into the language model, and the language model outputs a first code and a second code. The inference server then sends the first code and the second code to the backend server.

[0129] In some embodiments, text recommendation information is input into a language model, which outputs a no-emoji code. This no-emoji code indicates that no emojis should be added to the text recommendation information. For example, please refer to... Figure 10 If the language model outputs (two) codes 0, since code 0 corresponds to no emoji, it indicates that no emoji should be added to the copy recommendation information.

[0130] In some embodiments, copy recommendation information and contextual information are input into a language model, and the language model outputs a first encoding and a second encoding.

[0131] The context information includes at least one of the following: type information of the first displayed content, account information of the provider of the first displayed content, and an emoji library, which is used to record at least two emojis and their respective association information.

[0132] The type information of the first displayed content is used to indicate the type of the first displayed content. For example, if the first displayed content is video content, the type information of the first displayed content can indicate the type of the video content, such as indicating that it is a comedy, documentary or animation.

[0133] The account information of the provider of the first displayed content is used to indicate the account of the provider of the first displayed content. For example, if the first displayed content is video content, the account information of the provider of the first displayed content may include the identifier (ID), avatar, name, etc. of the video publishing account that published the video content.

[0134] In some embodiments, the emoji library records the codes corresponding to at least two emojis.

[0135] The associated information of the above emojis is used to indicate at least one of the following: the type of content to which the emojis are applicable, the account that publishes the content to which the emojis are applicable, the sentiment of the text recommendation information to which the emojis are applicable, and the theme of the text recommendation information to which the emojis are applicable.

[0136] For example, when the displayed content is a video, the associated information for the emoji [laughing] indicates that the applicable video type is comedy video, the applicable account for the displayed content is the account of a comedy video publisher, the applicable copywriting recommendation message has a lighthearted and humorous tone, and the applicable copywriting recommendation message theme is recommending entertainment products.

[0137] The association information of the aforementioned emojis can be obtained through machine learning by using artificial intelligence models to summarize the comment information posted by user accounts, or it can be obtained by manual data analysis. This application does not limit the method of obtaining the aforementioned association information.

[0138] In the above embodiment, the language model also refers to contextual information when reasoning about the first emoji and its position. The contextual information includes the type information of the first displayed content, the account information of the provider of the first displayed content, the emoji library, etc. In this way, the contextual learning ability of the language model can be used to improve the accuracy of its reasoning, so that the emotion expressed by the first emoji is consistent with the first displayed content and the copywriting recommendation information. This improves the rationality and naturalness of the copywriting recommendation information with the first emoji added in the comment interface of the first displayed content.

[0139] Sub-step 624: Obtain the first emoji based on the first code.

[0140] In some embodiments, the backend server retrieves a first emoji from at least two emojis based on a first encoding.

[0141] In the above embodiment, the language model determines the first emoji based on the copy recommendation information. Therefore, the first emoji can match the semantics and emotions conveyed by the copy recommendation information, thus integrating into the recommendation information more harmoniously.

[0142] Step 630: Add the first emoji to the recommendation information to obtain pseudo-comment information, wherein the pseudo-comment information is used to be displayed synchronously with at least one comment of the first displayed content.

[0143] In some embodiments, the backend server adds the first emoji to the recommendation information to obtain pseudo-comment information.

[0144] In some embodiments, please refer to Figure 11 Step 630 includes the following sub-step 632.

[0145] Sub-step 632: Add the first emoji to the position indicated by the second code in the copy recommendation information to obtain the pseudo-comment information.

[0146] In some embodiments, for a text recommendation message containing N characters, if the second code is i (N is an integer greater than 1, and i is a positive integer less than or equal to N), it indicates that the position for adding the first emoji in the text recommendation message is before the i-th character (between the (i-1)-th character and the i-th character). If the second code is N+1, it indicates that the position for adding the first emoji in the text recommendation message is after the N-th character.

[0147] In other words, in some embodiments, for text recommendation information including N characters, if the second code is i, then the first emoji is added before the i-th character (between the (i-1)-th character and the i-th character), and if the second code is N+1, then the first emoji is added after the N-th character.

[0148] For example, consider the following copywriting recommendation message: "XX product, direct from the manufacturer! No middlemen, cheap and affordable," which has 19 characters. If the second code is 1, then the copywriting recommendation message with the first emoji would look like this: Figure 3 As shown in case 1. If the second code is 11, then add the following text recommendation information with the first emoji: Figure 3 As shown in case 3. If the second code is 20, then add the following text recommendation information with the first emoji: Figure 3 As shown in case 2.

[0149] In the above embodiment, the position of the first emoji in the text recommendation information is determined by the language model based on the text recommendation information. Therefore, the first emoji can be added to a suitable position in the text recommendation information, so that it can enhance the emotional expression of the text recommendation information and promote the nativeization of the text recommendation information, while not affecting the expression of the original semantics of the text recommendation information. This avoids the addition of the first emoji affecting the original punctuation of the text recommendation information, thereby improving the user's reading experience.

[0150] Step 640: Send simulated comment information to the client.

[0151] In some embodiments, please refer to Figure 8 The backend server 21 sends simulated comment information to the client.

[0152] In some embodiments, a pseudo-comment message and at least one comment message are sent to the client.

[0153] In some embodiments, after receiving an information update request sent by the client (the information update request is a request sent by the client after receiving an operation on the copywriting recommendation information), the updated copywriting recommendation information is obtained, and a second emoticon is determined based on the updated copywriting recommendation information. The second emoticon is added to the updated copywriting recommendation information, and the updated copywriting recommendation information with the added second emoticon is sent to the client. For the method of determining the second emoticon based on the updated copywriting recommendation information, please refer to the method of determining the first emoticon based on the copywriting recommendation information in the above embodiments, which will not be repeated here.

[0154] The technical solution provided in this application, upon receiving an object recommendation request sent by the client (triggered by the operation of viewing comment information of the first displayed content), acquires the recommendation information of the recommended object and adds a first emoticon determined based on the recommendation information to the recommendation information to obtain pseudo-comment information. This pseudo-comment information is then displayed synchronously with the comment information of the first displayed content. Therefore, when object recommendation is performed in the comment interface (i.e., the interface used to present comment information) using the above method, the recommendation information will include a matching emoticon (the first emoticon), resulting in richer displayed content.

[0155] The training process of the above language model will be described in the following embodiments.

[0156] Please refer to Figure 12 The language model is trained through the following steps 1110-1130.

[0157] Step 1110: Obtain the original language model.

[0158] In some embodiments, the original language model is the large language model described above.

[0159] Step 1120: Adjust the output layer structure of the original language model so that the adjusted language model can output two-dimensional vectors.

[0160] The output layer is the layer in an artificial intelligence model that maps received features into output information.

[0161] For example, please refer to Figure 13The original language model is a large language model, which includes a tokenized embedding layer 121, a positional embedding layer 122, a dropout layer 123, M Transformer layers 124 (M is a positive integer), a final layer normalization layer 125, and a linear output layer 126. The structure of the linear output layer 126 is adjusted so that the dimension of its output vector is changed from 50257 to 2, thereby enabling the adjusted language model to output two-dimensional vectors and avoid outputting redundant data.

[0162] It should be noted that, Figure 13 The language model shown is merely an example; the actual language model construction can be changed arbitrarily according to requirements, and this application does not limit it.

[0163] Step 1130: Train the adjusted language model using the training data to obtain the language model. The training data includes the comment text posted by the user account and the tag information of the emojis added by the user account in the comment text. The tag information of the emojis is used to indicate the emojis and their positions in the comment text.

[0164] The text of a user's comment is the text displayed in the user's comment section.

[0165] In some embodiments, the emoji's label information is a two-dimensional label vector, in which one element indicates the emoji from at least two emojis, and the other element indicates the emoji's position in the comment text.

[0166] In some embodiments, comment information on content displayed in the target application is collected, and based on the collected comment information, the comment text and tag information of the emojis added in the comment text are extracted as training data. This process can be called sample engineering.

[0167] For example, please refer to Figure 14 The original comment was “[Staring blankly] This scenery is so beautiful, it's so pretty.” After sample engineering, training data can be obtained. This training data includes the comment text “This scenery is so beautiful, it's so pretty” and label information [1, 4]. In the label information, 1 indicates that the emoticon in the comment is located before the first character of the comment text. (Please refer to the example below.) Figure 10 The 4 in the tag information indicates that the emoticon in the comment information is [staring blankly].

[0168] It should be noted that, in this application, the technology for obtaining comment information from user accounts, when applied to specific products or technologies in the embodiments of this application, should comply with the requirements of national laws and regulations in the process of data collection, use and processing. Before collecting users' historical data, the information processing rules should be communicated and the individual consent of the target object should be obtained (or there should be a legal basis). Users' historical data should be processed in strict accordance with the requirements of laws and regulations and personal information processing rules, and technical measures should be taken to ensure the security of related data.

[0169] In some embodiments, step 1130 includes at least one of the following sub-steps 1132 to 1136 ( Figure 12 (Not shown in the image).

[0170] Sub-step 1132: Freeze all parameters in the adjusted language model except for the parameters of at least two layers to be adjusted, where the at least two layers to be adjusted include the output layer and are connected in series.

[0171] Once frozen, the other parameters mentioned above will not be changed during the language model training process (parameter tuning process).

[0172] For example, please refer to Figure 13 Freeze the parameters of all layers in the adjusted language model except for the parameters of the Mth Transformer layer 124, the final normalization layer 125, and the linear output layer 126.

[0173] Sub-step 1134: Input the comment text into the adjusted language model to obtain the prediction vector. The prediction vector is used to indicate the predicted emoji and the position in the comment text predicted by the adjusted language model for adding the predicted emoji. The predicted emoji is the emoji predicted by the adjusted language model to be added to the comment text.

[0174] In some embodiments, one element of the prediction vector is used to indicate the predicted emoji from at least two emojis, and another element is used to indicate the position in the comment text predicted by the adjusted language model for adding the predicted emoji.

[0175] For example, the comment text "This scenery is so beautiful, so gorgeous" is input into the adjusted language model, and the adjusted language model outputs a prediction vector [1, 4], where 1 in the prediction vector indicates that the adjusted language model predicts that the position in the comment text where the predicted emoji can be added is before the first character of the comment text (see example...). Figure 10 The 4 in the prediction vector indicates that the adjusted language model predicts the emoji to be added to the comment text is [staring blankly].

[0176] In some embodiments, sample context information and comment text are input into the adjusted language model to obtain prediction vectors.

[0177] The sample context information includes at least one of the following: the type of content being commented on by the comment text, the account information of the provider of the content being commented on by the comment text, and the emoji library.

[0178] Sub-step 1136: Based on the predicted vectors and emoji label information, adjust the parameters of at least two layers to be adjusted to obtain the language model.

[0179] In some embodiments, a first loss is calculated based on the predicted vector and the label information of the emoji. The first loss is used to reflect the difference between the predicted vector and the label information of the emoji. With the goal of minimizing the first loss, the parameters of at least two layers to be adjusted are adjusted to obtain a language model.

[0180] In some embodiments, the vector distance between the predicted vector and the label information of the emoji is calculated to obtain the first loss. The vector distance can be Euclidean distance, Manhattan distance, cosine distance, etc., and this application does not limit it.

[0181] In the above embodiment, during the training of the language model, some parameters are frozen, and only the parameters of the last few layers of the language model, including the output layer, are adjusted. This allows the language model to learn the ability to predict emojis and their locations while controlling the amount of parameters adjusted, thereby controlling the overall training time.

[0182] It should be noted that in the above embodiments, the training process of the language model is described using one training dataset. Those skilled in the art should understand that more training data can be used to train the language model in the same way to achieve better performance. This application does not limit the amount of training data.

[0183] In addition, when the original language model is a large language model, since the large language model itself has natural language processing capabilities, the process of training the language model using the above training data can also be called a fine-tuning process.

[0184] The technical solution provided in this application is that the language model is trained based on the comment text published by the user account and the tag information of the emojis added by the user account in the comment text. Therefore, the language model can learn the way and habit of real users to add emojis to the comment text, thereby improving the naturalness and originality of the language model in reasoning in the recommendation information, and thus ensuring the natural integration of the style of the simulated comment information and the comment information.

[0185] As user comments and emoji usage habits may change over time, in some implementations, it is necessary to periodically update the training data and retrain the language model to improve its performance in performing emoji prediction tasks.

[0186] In some embodiments, the above information providing method further includes: obtaining user feedback information, which reflects the effect achieved by recommending the recommended objects through simulated review information.

[0187] User feedback information may include the click-through rate of simulated comment information, the user conversion rate obtained through simulated comment information, etc. This application does not limit the specific content of user feedback information.

[0188] The user feedback information is also used to determine the weights of the training data corresponding to at least two emojis in the dataset used to train the language model. The training data corresponding to the emojis includes the comment text posted by the user account, as well as the tag information of the emojis added by the user account in the comment text.

[0189] For example, the weights of the training data corresponding to at least two emojis each have initial values ​​(e.g., they are all equal). If the effect of recommending the above-mentioned recommended objects by imitating comment information is not ideal (e.g., the user feedback information is less than the first threshold set by the technician), the weight of the training data corresponding to the first emoji in the dataset is reduced. If the effect of recommending the above-mentioned recommended objects by imitating comment information is good (e.g., the user feedback information is greater than the first threshold set by the technician), the weight of the training data corresponding to the first emoji in the dataset is increased.

[0190] For example, user feedback information is used to determine the optimal training strategy for the language model (the selection strategy for the weights of the training data corresponding to at least two emojis in the aforementioned dataset). For instance, the language model is trained on a first-class dataset, and a control language model trained on a second-class dataset is configured in the A / B test. The weights of the training data corresponding to at least two emojis in the first-class dataset are different from those in the second-class dataset. Based on the statistical user feedback information, technicians can analyze which weight is better, the weights of the training data corresponding to the first-class dataset and the weights of the training data corresponding to the second-class dataset, and thus use the better weight selection strategy in the dataset used to train the language model. For example, if the overall effect of the user feedback information obtained by performing the inference task using the aforementioned language model is better than the overall effect of the user feedback information obtained by performing the inference task using the aforementioned control language model, then the weights of the training data corresponding to at least two emojis in the first-class dataset will continue to be used to construct the dataset used for training the language model next.

[0191] In the above embodiments, the user feedback information obtained is used to determine the weights of the training data corresponding to at least two emojis in the dataset used to train the language model. This allows for continuous optimization of the language model's training strategy based on user feedback, enabling the language model to select and infer emojis in accordance with user preferences, ultimately improving the effectiveness of object recommendation through simulated comment information.

[0192] Please refer to Figure 15 The diagram illustrates an information provision scheme provided in one embodiment of this application.

[0193] The solution includes the following steps:

[0194] 1. Data Collection and Analysis: Collect comment information on content displayed in the target application and perform data analysis on this comment information. Summarize the distribution of emojis in different types of displayed content and comments under different accounts.

[0195] 2. Build an emoji library: Record at least two emojis that can be used and the types of content, emotions, and themes they are applicable to in the emoji library.

[0196] 3. Sample Engineering: Select the comment information to be used to construct the samples, and construct training data based on the encodings of at least two emojis recorded in the emoji library.

[0197] 4. Model fine-tuning: Train the language model using the constructed training data and deploy the language model online.

[0198] 5. Recommendation information processing: Based on the encoding inferred from the language model, emojis are added to the recommendation information to obtain pseudo-comment information, which is then sent to the client.

[0199] 6. Performance monitoring and optimization: Continuously collect comment information from user accounts and regularly fine-tune the language model to maintain the native-likeness of the simulated comment information when user preferences or business logic change.

[0200] Through the aforementioned technical means, this application achieves the addition of emojis to recommended information based on the analysis of user comment information, which has a high degree of user engagement and a natural display effect, and can significantly improve the user experience and user conversion rate brought by comment section advertising.

[0201] The technical solutions provided in this application can be applied to various scenarios involving comment interfaces. In the following embodiments, the application of this solution in short video scenarios will be exemplarily described.

[0202] Please refer to Figure 16 The diagram illustrates a flowchart of an information display method for short video scenarios according to an embodiment of this application. The method includes at least one of the following steps: 1510 to 1530.

[0203] Step 1510: In response to the operation of viewing the comment information of the short video, obtain at least one comment information of the short video, and simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes the recommendation information of the recommended object and a first emoji added to the recommendation information, the first emoji being determined based on the recommendation information.

[0204] Short videos refer to frequently pushed video content that is played on new media platforms, suitable for viewing in mobile and short leisure time. New media platforms are dissemination platforms that utilize digital technology to provide information and services to users through channels such as computer networks, wireless communication networks, and satellites, as well as terminals such as computers, mobile phones, and digital televisions.

[0205] Step 1520: Display the comment section of the short video, showing at least one comment and a simulated comment.

[0206] Step 1530: In response to the operation on the simulated comment information, display the associated interface of the recommended object.

[0207] In some embodiments, displaying at least one comment and a simulated comment in the comment section includes:

[0208] The comment section displays at least two entries in sequence, with each entry containing one comment or a pseudo-comment.

[0209] In some embodiments, when the number of comment messages is greater than 1, the pseudo-comment message is displayed between any two comment messages, and the display style of the pseudo-comment message is consistent with the display style of the comment message.

[0210] In some embodiments, the recommendation information includes text recommendation information, which is used to recommend recommended objects in natural language text, and the first emoji is determined based on the text recommendation information.

[0211] In some embodiments, the first emoji is added to the beginning of the copy recommendation information; or, the first emoji is added to the end of the copy recommendation information; or, the first emoji is added between the characters included in the copy recommendation information.

[0212] In some embodiments, the position of the first emoji in the text recommendation information is determined based on the text recommendation information.

[0213] In some implementations, the recommendation information also includes redirection recommendation information, which is used to redirect to the associated interface of the recommended object; in response to an operation on the pseudo-comment information, the associated interface of the recommended object is displayed, including: in response to an operation on the redirection recommendation information, the associated interface of the recommended object is displayed.

[0214] In some embodiments, in response to an operation on the copywriting recommendation information, an updated copywriting recommendation information and a second emoji added to the updated copywriting recommendation information are displayed in the comment section, the second emoji being determined based on the updated copywriting recommendation information.

[0215] Please refer to Figure 17 The diagram illustrates a flowchart of an information provision method for short video scenarios according to an embodiment of this application. The method includes at least one of the following steps: 1610 to 1640.

[0216] Step 1610: After receiving the object recommendation request sent by the client, obtain the recommendation information of the recommended object. The object recommendation request is a request sent by the client after receiving the operation of viewing the comment information of the short video.

[0217] Step 1620: Determine the first emoji based on the recommendation information.

[0218] Step 1630: Add the first emoji to the recommendation information to obtain pseudo-comment information, wherein the pseudo-comment information is used to be displayed synchronously with at least one comment information of the short video.

[0219] Step 1640: Send simulated comment information to the client.

[0220] In some embodiments, the recommendation information includes text recommendation information, which is used to recommend recommended objects in natural language text.

[0221] Determining the first emoji based on recommendation information includes: inputting text recommendation information into a language model, outputting a first code from the language model, the first code being used to indicate the first emoji from at least two emojis; and obtaining the first emoji based on the first code.

[0222] In some embodiments, inputting copywriting recommendation information into a language model and outputting a first code from the language model includes: inputting copywriting recommendation information into a language model and outputting a first code and a second code from the language model, wherein the second code is used to indicate the position in the copywriting recommendation information for adding a first emoticon; adding the first emoticon to the recommendation information to obtain pseudo-comment information includes: adding the first emoticon to the position indicated by the second code in the copywriting recommendation information to obtain pseudo-comment information.

[0223] In some embodiments, inputting text recommendation information into a language model and having the language model output a first code and a second code includes: inputting text recommendation information and context information into a language model and having the language model output a first code and a second code; the context information includes at least one of the following: short video type information, short video provider account information, and an emoji library, wherein the emoji library is used to record at least two emojis and their respective association information; wherein the association information of the emojis is used to indicate at least one of the following: the type of short video to which the emoji is applicable, the short video publishing account to which the emoji is applicable, the sentiment of the text recommendation information to which the emoji is applicable, and the theme of the text recommendation information to which the emoji is applicable.

[0224] In some embodiments, the language model is trained by the following steps: obtaining the original language model; adjusting the output layer structure of the original language model so that the adjusted language model is used to output two-dimensional vectors; and training the adjusted language model using training data to obtain the language model, wherein the training data includes the comment text posted by the user account, and the tag information of the emojis added by the user account in the comment text, the tag information of the emojis being used to indicate the emojis and their positions in the comment text.

[0225] In some embodiments, training an adjusted language model using training data to obtain a language model includes: freezing all parameters in the adjusted language model except for the parameters of at least two layers to be adjusted, wherein the at least two layers to be adjusted include an output layer and are sequentially connected in series; inputting comment text into the adjusted language model to obtain a prediction vector, the prediction vector being used to indicate the predicted emoji, and the position in the comment text predicted by the adjusted language model for adding the predicted emoji, wherein the predicted emoji is the emoji predicted by the adjusted language model to be added to the comment text; and adjusting the parameters of at least two layers to be adjusted based on the prediction vector and the label information of the emoji to obtain the language model.

[0226] In some embodiments, user feedback information is obtained, which is used to reflect the effect of recommending objects by mimicking comment information;

[0227] The user feedback information is also used to determine the weights of the training data corresponding to at least two emojis in the dataset used to train the language model. The training data corresponding to the emojis includes the comment text posted by the user account, as well as the label information of the emojis added by the user account in the comment text. The label information of the emojis is used to indicate the emojis and their positions in the comment text.

[0228] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0229] Please refer to Figure 18 This diagram illustrates a block diagram of an information display device according to an embodiment of this application. The device has the function of implementing the above-described information display method; this function can be implemented in hardware or by hardware executing corresponding software. The device can be a computer device or can be installed within a computer device. The device 1700 may include: an acquisition module 1710, a comment display module 1720, and an association display module 1730.

[0230] The acquisition module 1710 is configured to, in response to an operation of viewing comment information of the first displayed content, acquire at least one comment of the first displayed content, and simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes recommendation information of the recommended object and a first emoticon added to the recommendation information, the first emoticon being determined based on the recommendation information.

[0231] The comment display module 1720 is used to display the comment interface of the first displayed content, and to display the at least one comment and the simulated comment in the comment interface.

[0232] The associated display module 1730 is used to display the associated interface of the recommended object in response to an operation on the simulated comment information.

[0233] In some embodiments, the comment display module 1720 is configured to sequentially display at least two entries in the comment interface, each entry displaying one comment message or one pseudo-comment message.

[0234] In some embodiments, when the number of comment information is greater than 1, the pseudo-comment information is displayed between any two of the comment information, and the display style of the pseudo-comment information is consistent with the display style of the comment information.

[0235] In some embodiments, the recommendation information includes text recommendation information, which is used to recommend the recommended object in natural language text, and the first emoji is determined based on the text recommendation information.

[0236] In some embodiments, the first emoji is added to the beginning of the text recommendation information; or, the first emoji is added to the end of the text recommendation information; or, the first emoji is added between the characters included in the text recommendation information.

[0237] In some embodiments, the position of the first emoji in the text recommendation information is determined based on the text recommendation information.

[0238] In some embodiments, the recommendation information further includes jump recommendation information, which is used to jump to the associated interface of the recommended object.

[0239] The associated display module 1730 is used to display the associated interface of the recommended object in response to the operation of the jump recommendation information.

[0240] In some embodiments, the comment display module 1720 is further configured to, in response to an operation on the copywriting recommendation information, display updated copywriting recommendation information and a second emoticon added to the updated copywriting recommendation information in the comment interface, the second emoticon being determined based on the updated copywriting recommendation information.

[0241] Please refer to Figure 19 This diagram illustrates a block diagram of an information providing apparatus according to an embodiment of this application. The apparatus has the function of implementing the aforementioned information providing method; this function can be implemented in hardware or by hardware executing corresponding software. The apparatus can be a computer device or can be installed within a computer device. The apparatus 1800 may include: an acquisition module 1810, a determination module 1820, an addition module 1830, and a sending module 1840.

[0242] The acquisition module 1810 is used to acquire recommendation information of the recommended object after receiving an object recommendation request sent by the client. The object recommendation request is a request sent by the client after receiving the operation of viewing the comment information of the first displayed content.

[0243] The determination module 1820 is used to determine the first emoji based on the recommendation information.

[0244] The module 1830 is used to add the first emoji to the recommendation information to obtain pseudo-comment information, wherein the pseudo-comment information is used to be displayed synchronously with at least one comment on the first displayed content.

[0245] The sending module 1840 is used to send the simulated comment information to the client.

[0246] In some embodiments, the recommendation information includes text recommendation information, which is used to recommend the recommended object in natural language text.

[0247] The determining module 1820 is used to input the copy recommendation information into a language model, and the language model outputs a first code, which is used to indicate the first emoji from at least two emojis; and to obtain the first emoji based on the first code.

[0248] In some embodiments, the determining module 1820 is used to input the copywriting recommendation information into the language model, and the language model outputs the first code and the second code, wherein the second code is used to indicate the position in the copywriting recommendation information for adding the first emoji; the adding module 1830 is used to add the first emoji to the position indicated by the second code in the copywriting recommendation information to obtain the simulated comment information.

[0249] In some embodiments, the determining module 1820 is used to input the copywriting recommendation information and context information into the language model, and the language model outputs the first encoding and the second encoding; the context information includes at least one of the following: type information of the first displayed content, account information of the provider of the first displayed content, and an emoji library, wherein the emoji library is used to record the at least two emojis and their respective association information; wherein the association information of the emojis is used to indicate at least one of the following: the type of displayed content to which the emojis are applicable, the account to which the displayed content to which the emojis are applicable, the sentiment of the copywriting recommendation information to which the emojis are applicable, and the theme of the copywriting recommendation information to which the emojis are applicable.

[0250] In some embodiments, the language model is trained by the following steps: obtaining an original language model; adjusting the output layer structure of the original language model so that the adjusted language model is used to output two-dimensional vectors; training the adjusted language model using training data to obtain the language model, wherein the training data includes comment text posted by a user account, and tag information of emojis added by the user account in the comment text, the tag information of the emojis being used to indicate the emojis and their positions in the comment text.

[0251] In some embodiments, training the adjusted language model using training data to obtain the language model includes: freezing all parameters in the adjusted language model except for the parameters of at least two layers to be adjusted, the at least two layers to be adjusted including the output layer, and the at least two layers to be adjusted being sequentially connected in series; inputting the comment text into the adjusted language model to obtain a prediction vector, the prediction vector being used to indicate a predicted emoji, and the position in the comment text predicted by the adjusted language model for adding the predicted emoji, the predicted emoji being an emoji predicted by the adjusted language model to be added to the comment text; and adjusting the parameters of the at least two layers to be adjusted based on the prediction vector and the label information of the emoji to obtain the language model.

[0252] In some embodiments, the acquisition module 1810 is further configured to acquire user feedback information, which reflects the effect achieved by recommending the recommended object through the simulated comment information; wherein, the user feedback information is further configured to determine the weight of the training data corresponding to the at least two emojis in the dataset used to train the language model, wherein the training data corresponding to the emojis includes the comment text posted by the user account, and the tag information of the emojis added by the user account in the comment text, wherein the tag information of the emojis is used to indicate the emojis and the positions of the emojis in the comment text.

[0253] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0254] Please refer to Figure 20 The diagram illustrates a structural block diagram of a computer device provided in one embodiment of this application.

[0255] Typically, computer device 1900 includes a processor 1901 and a memory 1902.

[0256] Processor 1901 may include one or at least two processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 1901 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), and PLA (Programmable Logic Array). Processor 1901 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1901 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the screen. In some embodiments, processor 1901 may also include an AI processor, which is used to handle computational operations related to machine learning.

[0257] The memory 1902 may include one or at least two computer-readable storage media, which may be tangible and non-transitory. The memory 1902 may also include high-speed random access memory and non-volatile memory, such as one or at least two disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1902 stores a computer program that is loaded and executed by the processor 1901 to implement the above-described information display method or the above-described information provision method.

[0258] Those skilled in the art will understand that Figure 20 The structure shown does not constitute a limitation on the computer device 1900, and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0259] In some embodiments, a computer-readable storage medium is also provided, wherein a computer program is stored in the storage medium, the computer program being loaded and executed by a processor to implement the above-described information display method or the above-described information provision method.

[0260] Optionally, the computer-readable storage medium may include: ROM (Read-Only Memory), RAM (Random-Access Memory), SSD (Solid State Drives), or optical disc, etc. The random access memory may include ReRAM (Resistance Random Access Memory) and DRAM (Dynamic Random Access Memory).

[0261] In some embodiments, a computer program product is also provided, the computer program product including a computer program stored in a computer-readable storage medium, the processor reading from the computer-readable storage medium and executing the computer program to implement the above-described information display method or the above-described information provision method.

[0262] It should be noted that this application may display prompts, pop-ups, or output voice prompts before and during the collection of user data. These prompts are used to inform the user that their data is being collected. This ensures that the application only begins the data collection process after receiving confirmation from the user regarding the prompt or pop-up; otherwise (i.e., without confirmation from the user), the data collection process ends, and no user data is collected. In other words, all user data collected by this application (including comments posted by user accounts) is processed strictly in accordance with the requirements of relevant national laws and regulations, obtaining informed consent or separate consent from the data subject, and subsequent data use and processing are conducted within the scope of laws, regulations, and the data subject's authorization.

[0263] It should be understood that "at least two" as mentioned herein refers to two or more. Furthermore, the step numbers described herein are merely illustrative of one possible order of execution. In some other embodiments, the steps may not be executed in the order shown in the figures, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the figures. This application does not limit this practice.

[0264] The above are merely exemplary embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application shall be included within the protection scope of this application.

Claims

1. An information providing method characterized by comprising: The method includes: After receiving an object recommendation request from the client, the recommendation information of the recommended object is obtained. The object recommendation request is a request sent by the client after receiving the operation of viewing the comment information of the first displayed content. The first emoji is determined based on the recommended information; The first emoji is added to the recommendation information to obtain pseudo-comment information, wherein the pseudo-comment information is used to be displayed synchronously with at least one comment on the first displayed content; The simulated comment information is sent to the client.

2. The method of claim 1, wherein, The recommendation information includes text recommendation information, which is used to recommend the recommended object in natural language text; The step of determining the first emoji based on the recommendation information includes: The text recommendation information is input into a language model, and the language model outputs a first code, which is used to indicate the first emoji from at least two emojis. The first emoji is obtained based on the first encoding.

3. The method of claim 2, wherein, The step of inputting the text recommendation information into a language model and having the language model output a first code includes: The text recommendation information is input into the language model, and the language model outputs the first code and the second code, wherein the second code is used to indicate the position in the text recommendation information for adding the first emoji; Adding the first emoji to the recommendation information to obtain pseudo-comment information includes: The first emoji is added to the position indicated by the second encoding in the text recommendation information to obtain the simulated comment information.

4. The method of claim 3, wherein, The step of inputting the text recommendation information into a language model and having the language model output the first code and the second code includes: The text recommendation information and context information are input into the language model, and the language model outputs the first code and the second code. The context information includes at least one of the following: The first displayed content includes type information, account information of the provider of the first displayed content, and an emoji library, wherein the emoji library is used to record the at least two emojis and their respective association information; The associated information of the emoticon is used to indicate at least one of the following: the type of content to which the emoticon is applicable, the account to which the content to which the emoticon is applicable is published, the emotion of the text recommendation information to which the emoticon is applicable, and the theme of the text recommendation information to which the emoticon is applicable.

5. The method according to claim 3 or 4, characterized in that, The language model is trained through the following steps: Obtain the original language model; The output layer structure of the original language model is adjusted so that the adjusted language model can output two-dimensional vectors; The adjusted language model is trained using training data to obtain the language model, wherein the training data includes comment text posted by a user account, and tag information of emojis added by the user account in the comment text, the tag information of the emojis being used to indicate the emojis and their positions in the comment text.

6. The method of claim 5, wherein, The step of training the adjusted language model using training data to obtain the language model includes: Freeze all parameters in the adjusted language model except for the parameters of at least two layers to be adjusted, the at least two layers to be adjusted including the output layer, and the at least two layers to be adjusted are connected in series. The comment text is input into the adjusted language model to obtain a prediction vector. The prediction vector is used to indicate the predicted emoji and the position in the comment text predicted by the adjusted language model for adding the predicted emoji. The predicted emoji is the emoji predicted by the adjusted language model to be added to the comment text. Based on the predicted vector and the emoji label information, the parameters of the at least two layers to be adjusted are adjusted to obtain the language model.

7. The method according to any one of claims 3 to 6, characterized in that, The method further includes: Obtain user feedback information, which reflects the effect of recommending the recommended object through the simulated comment information; The user feedback information is further used to determine the weights of the training data corresponding to the at least two emojis in the dataset used to train the language model. The training data corresponding to the emojis includes the comment text posted by the user account and the tag information of the emojis added by the user account in the comment text. The tag information of the emojis is used to indicate the emojis and their positions in the comment text.

8. An information display method characterized by comprising: The method includes: In response to an operation to view comment information of the first displayed content, at least one comment of the first displayed content is obtained, as well as simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes recommendation information of the recommended object and a first emoticon added to the recommendation information, the first emoticon being determined based on the recommendation information; A comment interface is displayed that shows the first displayed content, and the comment interface displays the at least one comment and the simulated comment. In response to the operation on the simulated comment information, the associated interface of the recommended object is displayed.

9. The method of claim 8, wherein, The display of the at least one comment and the simulated comment information in the comment interface includes: The comment interface displays at least two entries in sequence, and each entry displays one comment or one pseudo-comment.

10. The method of claim 9, wherein, When the number of comments is greater than 1, the pseudo-comment information is displayed between any two comments, and the display style of the pseudo-comment information is consistent with the display style of the comments.

11. The method according to any one of claims 8 to 10, characterized in that, The recommendation information includes text recommendation information, which is used to recommend the recommended object in natural language text, and the first emoji is determined based on the text recommendation information.

12. The method according to claim 11, characterized in that, The first emoji is added to the header of the text recommendation information; or, The first emoji is added to the end of the text recommendation information; or, The first emoji is added between the characters included in the text recommendation information.

13. The method according to claim 11 or 12, characterized in that, The position of the first emoji in the text recommendation information is determined based on the text recommendation information.

14. The method according to any one of claims 11 to 13, characterized in that, The recommendation information also includes jump recommendation information, which is used to jump to the associated interface of the recommended object; The step of displaying the associated interface of the recommended object in response to the operation on the simulated comment information includes: In response to the operation on the jump recommendation information, the associated interface of the recommended object is displayed.

15. The method according to any one of claims 11 to 14, characterized in that, The method further includes: In response to an operation on the copywriting recommendation information, the updated copywriting recommendation information and a second emoticon added to the updated copywriting recommendation information are displayed in the comment interface. The second emoticon is determined based on the updated copywriting recommendation information.

16. An information providing apparatus characterized by comprising: The device includes: The acquisition module is used to acquire the recommendation information of the recommended object after receiving the object recommendation request sent by the client. The object recommendation request is a request sent by the client after receiving the operation of viewing the comment information of the first displayed content. The determining module is used to determine the first emoji based on the recommendation information; An add module is used to add the first emoji to the recommendation information to obtain simulated comment information, wherein the simulated comment information is used to be displayed synchronously with at least one comment on the first displayed content; The sending module is used to send the simulated comment information to the client.

17. An information display device, characterized by comprising: The device includes: The acquisition module is used to acquire at least one comment of the first displayed content in response to the operation of viewing the comment information of the first displayed content, as well as simulated comment information displayed synchronously with the comment information, wherein the simulated comment information includes the recommendation information of the recommended object and a first emoticon added to the recommendation information, the first emoticon being determined based on the recommendation information; A comment display module is used to display a comment interface for the first displayed content, and to display the at least one comment and the simulated comment information in the comment interface; The associated display module is used to display the associated interface of the recommended object in response to the operation on the simulated comment information.

18. A computer device, comprising: The computer device includes a processor and a memory, the memory storing a computer program that is loaded and executed by the processor to implement the method as claimed in any one of claims 1 to 7, or to implement the method as claimed in any one of claims 8 to 15.

19. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is loaded and executed by a processor to implement the method as claimed in any one of claims 1 to 7, or to implement the method as claimed in any one of claims 8 to 15.

20. A computer program product, characterised in that, The computer program product includes a computer program stored in a computer-readable storage medium, which a processor reads from and executes to implement the method as claimed in any one of claims 1 to 7, or to implement the method as claimed in any one of claims 8 to 15.

Images