Page content generation method, page access method, equipment, storage medium and program product

Abstract

The embodiment of the invention provides a page content generation method, a page access method, equipment, a storage medium and a program product. In the embodiment of the invention, the page content is mainly generated by taking the theme content, so that the user is attracted, the online staying time of the user is prolonged, and the shopping guide efficiency and the conversion rate of the application are improved; when the large model is applied to the page content generation process with the theme content as the main part, on one hand, copywriting information of the theme content is generated by means of a copywriting generation model; on the other hand, a background image corresponding to the subject content is generated by means of a text graph model; the page content is generated based on the document information, the background image and the detailed information of the information object, the page corresponding to the subject content is conveniently and rapidly rendered, the mode of producing the page content depending on manpower is converted into model generation, manpower is liberated from page content generation, the cost is lower, the page content generation efficiency is high, and the page content generation efficiency is improved. Therefore, the number of pages taking the theme content as the main part can meet user requirements and market changes.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a page content generation method, a page access method, a device, a storage medium, and a program product. Background Art

[0002] Users' requirements for online shopping experience are increasing day by day. The product display methods of traditional e-commerce applications are gradually unable to meet users' personalized and diversified shopping needs, resulting in users staying online for shorter and shorter time and less ideal purchase conversion rate.

[0003] The theme venue is a new e-commerce scenario that displays product information around a specific theme or event. By centrally displaying highly relevant product information and presenting graphic and text information related to the specific theme or event, it creates an immersive shopping experience for users. It can not only significantly increase the user's online stay time, but also effectively improve shopping guide efficiency and conversion rate.

[0004] At present, the production of theme venue page content mainly relies on manpower. For example, graphic and text information related to specific themes or activities are provided by designers or operators, which results in high costs and low efficiency. As a result, the number of theme venue pages online daily is difficult to meet user needs and respond quickly to market changes. Summary of the invention

[0005] Multiple aspects of the present application provide a page content generation method, a page access method, a device, a storage medium and a program product to reduce the cost of manually building pages, improve the efficiency of page generation, and make the number of pages based on subject content meet user needs and market changes.

[0006] An embodiment of the present application provides a page content generation method, including: determining target subject content adapted to a target application and a target information object adapted to the target subject content; based on the target information object and the target subject content, calling a text generation model to generate text information corresponding to the target subject content; based on the target information object and the target subject content, calling a text generation graph model to generate a background graph corresponding to the target subject content; generating target page content according to the text information, the background graph and detailed information of the target information object, wherein the target page content is used to render a target page corresponding to the target subject content.

[0007] An embodiment of the present application also provides a page access method, including: during the operation of a target application, displaying a specific page, wherein the specific page includes access entry information of the target page; in response to a user's access operation to the access entry information, obtaining target page content, and rendering the target page content to display the target page; wherein the target page content is generated according to the page content generation method.

[0008] An embodiment of the present application also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor is coupled to the memory and is used to execute the computer program to implement the steps in each method provided in the embodiment of the present application.

[0009] An embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor is enabled to implement the steps in the above method.

[0010] An embodiment of the present application also provides a computer program product, which includes a computer program / instructions. When the computer program / instructions are executed by a processor, the processor is enabled to implement the steps in the above method embodiment.

[0011] In the embodiment of the present application, generating page content mainly based on the theme content is conducive to attracting users, increasing the time users stay online, and improving the shopping guide efficiency and conversion rate of the application; wherein, the large model is applied to the process of generating page content mainly based on the theme content, on the one hand, the copy information of the theme content is generated with the help of the copy generation model; on the other hand, the background image corresponding to the theme content is generated with the help of the text-generated graph model; the target page content is generated based on the copy information, the background image and the detailed information of the information object, so as to quickly render the page corresponding to the theme content, convert the method of relying on manpower to produce page content into model generation, liberate manpower from page content production, lower cost and high efficiency of page content generation, so that the number of pages mainly based on theme content can meet user needs and market changes. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0013] Figure 1 A flow chart of a method for generating page content provided by an exemplary embodiment of the present application;

[0014] Figure 2 A schematic diagram of the structure of a page content generation system provided for an exemplary embodiment of the present application;

[0015] Figure 3a and Figure 3b A schematic diagram of a target page generated based on a text generation model and a text image model according to an embodiment of the present application;

[0016] Figure 4a A schematic diagram of a flow chart of a page access method provided in an embodiment of the present application;

[0017] Figure 4b A schematic diagram of the jump process of accessing a target page provided in an embodiment of the present application;

[0018] Figure 5 A schematic structural diagram of an electronic device is provided as yet another exemplary embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0020] It should be noted that, in the case where the embodiments of the present application involve user information, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) are in compliance with relevant laws and standards.

[0021] In addition, it should be noted that, in the case where the embodiments of the present application involve user interaction operations or trigger operations, the user interaction operations or trigger operations involved in the embodiments of the present application include but are not limited to: touch operation, gesture operation, voice operation, head movement operation, eye movement operation and other interactive operations in various ways; among which, touch operation includes but is not limited to: click operation, double-click operation, long press operation, sliding operation, pinch operation or mouse hover operation, etc. Sliding operation includes but is not limited to: straight line sliding, curve sliding, etc.

[0022] Furthermore, it should be noted that, in the case where the embodiments of the present application involve jumping between the first interface and the second interface, the jumping methods involved in the embodiments of the present application include but are not limited to: jumping directly from the first interface to the second interface, jumping from the first interface to the task interface first and jumping to the second interface after completing the corresponding task operation on the task interface; completing the corresponding task operation on the task interface includes but is not limited to: when the task interface is implemented as a game interface, completing the game operation on the game interface; when the task interface is implemented as an identity authentication interface, completing the identity authentication on the identity authentication interface; when the task interface is implemented as a recharge interface, completing the recharge operation on the recharge interface; and so on.

[0023] In view of the technical problems of high cost and low efficiency in manually building venue pages, in an embodiment of the present application, page content is generated mainly based on theme content, which is conducive to attracting users, increasing the length of time users stay online, and improving the shopping guide efficiency and conversion rate of the application; wherein, the large model is applied to the generation process of pages with theme content as the main content, on the one hand, the copy information of the theme content is generated with the help of the copy generation model; on the other hand, the background image corresponding to the theme content is generated with the help of the text-generated graph model; the page content is generated based on the copy information, the background image and the detailed information of the information object, so as to quickly render the page corresponding to the theme content, convert the method of relying on manpower to produce page content into model generation, freeing manpower from page content production, lowering the cost and increasing the efficiency of page content generation, so that the number of pages with theme content as the main content can meet user needs and market changes.

[0024] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0025] Figure 1 The following is a flow chart of a method for generating page content provided by an exemplary embodiment of the present application. Figure 1 As shown, the method includes:

[0026] S101: Determine a target subject content adapted to a target application and a target information object adapted to the target subject content;

[0027] S102: Based on the target information object and the target subject content, calling the copywriting generation model to generate copywriting information corresponding to the target subject content;

[0028] S103: Based on the target information object and the target subject content, the text graph model is called to generate a background graph corresponding to the target subject content;

[0029] S104: Generate target page content according to the copy information, the background image and the detailed information of the target information object, wherein the target page content is used to render a target page corresponding to the target subject content.

[0030] In the embodiment of the present application, the target application may be various types of applications, including but not limited to live broadcast applications, video applications, e-commerce applications, news applications, and social applications. Based on the target application, the target subject content adapted thereto may be determined. The target subject content is the core concept that the target information object needs to present. The target information object may be matched according to the target subject content, which may be one or more. Among them, the target information object may be a collection of information objects expanded around the target subject content, which may be used to present the core concept of the target subject content.

[0031] For example, if the target application is an e-commerce application, the target theme content that matches it can be "Mid-Autumn Festival promotion activities", and the target information object obtained by matching the target theme content can be product information related to the Mid-Autumn Festival, such as a collection of product information related to the Mid-Autumn Festival. For another example, if the target application is a news application, the target theme content that matches it can be various types of news, such as technology news "artificial intelligence", and the target information object obtained by matching the target theme content can be news related to "artificial intelligence".

[0032] In the embodiment of the present application, the content form of the target information object is not limited, and the target information object can be at least one of text content, audio content and video content, and the presentation method of different types of content is not limited. For example, if the target information object is text content, it can be presented directly as text content; or, it can also be presented by converting the text content itself into voice content, and optionally, TTS (Text to Speech) technology can be used to convert the text content.

[0033] In this embodiment, the target information objects are matched to the corresponding target subject contents based on different target subject contents. The target information objects matched to the different target subject contents may be the same or different, and there is no limitation on this. For example, if the type of the target application is an e-commerce application, the target subject content may be related to product promotion, and the target information object matched is product information. For another example, if the type of the target application is a news application, the target subject content is mainly related to various types of news, and the target information objects matched are mainly various types of news. Of course, product information may also be matched, and there is no limitation on this.

[0034] Optionally, in the process of determining the target subject content adapted by the target application, search hotspot information can be mined from the search data of the target application, and the target subject content can be determined based on the mined search hotspot information. Optionally, the mined search hotspot information can be obtained by analyzing the search data obtained from the target application. For example, the target application can provide the user with a search function for information objects. The user can enter a search term to search for information objects that meet their needs based on the search function of the target application, and can perform access operations on the searched information objects. Among them, according to the user's search operation and the user's access operation performed on the searched information objects, the search data corresponding to the user can be obtained. The search data can include but is not limited to the search term, as well as user information, time information of the user initiating the search operation, information of the information object obtained by the search term search, and which information objects the user has performed what type of access operation on (referred to as access operation information), etc.

[0035] Further optionally, in the process of mining search hotspot information based on search data, the search terms in each search data can be segmented, keywords can be extracted from the segmentation results, and the frequency of each keyword in the search data can be counted, so that the search terms with frequencies that meet the frequency conditions (such as the highest frequency) can be determined, and then the search hotspot information can be determined based on the determined search terms. For example, assuming that the search term is "Mid-Autumn Festival", the search term "Mid-Autumn Festival" can be directly used as search hotspot information, or the search term can be described twice, such as "commodity information suitable for Mid-Autumn Festival" as search hotspot information. In another optional implementation, the target application can also obtain which searched information objects the user has performed access operations on, and then obtain an information object set, use the information object set as part of the search data, and mine search hotspot information based on the information object set. The search hotspot information can be information objects that appear frequently in the information object set. For example, taking the e-commerce field as an example, the information object can be various product information. Assuming that "moon cakes" is a relatively popular product information mined, "moon cakes" can be directly used as the search hotspot information, or "Mid-Autumn Festival" derived from "moon cakes" can be used as the search hotspot information.

[0036] Then, the target subject content is determined based on the search hotspot information, and the target information object adapted to the target subject content is determined. On this basis, based on the target information object and the target subject content, the copy generation model is called to generate copy information corresponding to the target subject content. The copy generation model can be a large language model. Optionally, the copy generation model is trained in advance using a variety of context learning (In Context Learning-ICL) methods so that the copy generation model has context learning capabilities, which can be used to perform analogical learning on the context prompt words provided in the input of the model, and infer and generate the expected copy information based on the copy example information in the context prompt words.

[0037] Among them, different types of target subject content have their own context prompt words, and the context prompt words are pre-set according to the types of corresponding target subject content. Inputting the context prompt words corresponding to the type of target subject content into the copy generation model can guide the model to generate copy information corresponding to the target subject content. With the help of the context learning mechanism of the copy generation model, there is no need to retrain the model for different types of target subject content.

[0038] Further, based on the target information object and the target subject content, the text graph model is called to generate a background image corresponding to the target subject content; the text graph model is a machine learning model that takes a natural language description as input and outputs a background image that matches the description; wherein, in this embodiment, the copywriting generation model and the text graph model can be the same or different models. Optionally, the text graph model can also be combined with the copywriting generation model. In this case, the copywriting generation model can be used to convert the input text into a machine description, and the text graph model generates a background image that matches the machine description according to the machine description.

[0039] Further, the target page content can be generated according to the generated copy information, background image and detailed information of the target information object, and the target page content is used to render the target page corresponding to the target theme content. Among them, the detailed information of the target information object includes an explanation of the target information object, covering the properties, functions and other information of the target information object, so as to provide users with more details about the target information object. For example, if the target information object is product information, the detailed information can be detailed information related to the product, including but not limited to the price, specifications, functions and origin of the product, etc.; further, the copy information and the detailed information of the target information object are superimposed on the corresponding position in the background image to obtain the target page content corresponding to the target theme content. It is explained here that the target page content includes but is not limited to: copy information, background image and detailed information of the target information object. Depending on the application scenario, the target page content may also include some other information, such as a logo, an ID of the page content, etc.

[0040] In the embodiment of the present application, page content is generated mainly based on the theme content, which is conducive to attracting users, increasing the time users stay online, and improving the shopping guide efficiency and conversion rate of the application; wherein, the large model is applied to the process of generating page content mainly based on the theme content, on the one hand, the copy information of the theme content is generated with the help of the copy generation model; on the other hand, the background image corresponding to the theme content is generated with the help of the text-generated image model; the target page content is generated based on the copy information, the background image and the detailed information of the information object, so as to render the page corresponding to the theme content, converting the method of relying on manpower to produce page content into model generation, freeing manpower from page content production, lowering the cost and increasing the efficiency of page content generation, and being able to meet user needs and market changes.

[0041] Figure 2 The following is a schematic diagram of a page content generation system provided by an exemplary embodiment of the present application. Figure 2 As shown, the system 200 includes a semantic embedding model 20, a text generation model 21, and a text graph model 24; wherein the semantic embedding model 20 is used to match candidate information objects that are semantically close to the target subject content from candidate information objects as target information objects; the text generation model 21 is used to generate text information corresponding to the target subject content based on the target information object and the target subject content; the text graph model 24 is used to generate a background image corresponding to the target subject content based on the target information object and the target subject content; in this embodiment, the text generation model 21 and the text graph model 24 can be the same or different models, and there is no limitation on this. The same mainly refers to the same structure, and it may not be the same model.

[0042] In this embodiment, the target subject content adapted to the target application can be determined first; wherein, the target subject content can be the search hotspot information mined from the search data of the target application, and the target subject content is determined based on the search hotspot information. Regarding the target application and the target subject content and the method of determining them, reference can be made to the above embodiments and will not be repeated here.

[0043] Further, a target information object adapted to the target subject content is determined. The target information object is obtained by screening candidate information objects, and the candidate information object may be one or more. In this embodiment, there is no limitation on the method of obtaining the candidate information object. For example, an information object library may be maintained for the information object, and the library includes multiple candidate information objects. The candidate information object may be manually uploaded; or it may be obtained from the Internet and then imported into the library after structured processing.

[0044] like Figure 2As shown, the target information object can be determined from the candidate information objects using the semantic embedding model 20. In the determination process, the candidate information object and the target subject content are input into the semantic embedding model 20 for semantic feature extraction, so that the candidate information object and the target subject content are mapped to the vector space to obtain the embedding vector of the candidate information object and the embedding vector of the target subject content. Among them, the embedding vector is used to represent semantic information. In this embodiment, by comparing the similarity between the embedding vector of the candidate information object and the embedding vector of the target subject content, the degree of proximity between the semantic information of the candidate information object and the semantic information of the target subject content can be measured; wherein, the higher the similarity, the closer the semantic information of the candidate information object and the semantic information of the target subject content are; then, the candidate information object whose similarity meets the preset condition is selected from the candidate information object as the target information object; in this embodiment, the preset condition is not limited, for example, it can be selected from the candidate information object with a similarity within a certain range; or, it can also be selected from the candidate information object. The candidate information object with the highest similarity; or, it can also be selected from the candidate information object. A specified number of candidate information objects with similarity within a certain range, and the specified number can be one or more.

[0045] Further, if Figure 2 As shown, the copy generation model 21 is called to generate copy information. In this process, based on the type of target subject content, the copy context prompt words and the copy requirement description information are determined; wherein the copy context prompt words include the input and output of the copy generation model 21 when generating historical copy information, for example, the historical input may include historical commodity information and historical subject content, and the historical output corresponding to the historical input may include historical copy information; wherein the historical input and the historical output may be referred to as the interaction history between the user and the copy generation model, and inputting the interaction history into the copy generation model 21 may guide the model to generate copy information corresponding to the target subject content; the copy requirement description information is used to describe the generation requirements of the copy information, for example, it may be but not limited to the number of generated copy information, the word count information of each copy, the position information, and the format information of the copy information; the word count information of the copy information may be the word limit number; the position information refers to the display position of the copy information superimposed on the background image; the format information refers to the style that the copy information needs to meet, for example, it may be casual, natural, and close to colloquial expression; or, it may also be solemn and rigorous in tone, which is not limited to this.

[0046] Among them, in the process of guiding the output of the copy generation model 21 by the copy context prompt words, the copy generation model 21 can understand the contextual semantics of the input through analogical learning of the historical input and output in the copy context prompt words, which includes vocabulary, syntax and semantic relations, so as to accurately infer the meaning of the text and generate the expected copy information according to the copy context prompt words. With the help of the context learning mechanism of the copy generation model 21, there is no need to retrain the model for different types of target subject content.

[0047] Secondly, the copy context prompt words can also help the copy generation model 21 perceive the context of the input text, which includes the contextual perception of conversation history, article structure, domain knowledge, etc. This context perception capability enables the model to understand the user's intentions and needs based on previous interaction history in multiple rounds of conversations, so that the model can adjust the content and method of the output copy information according to the differences in different copy context prompt words, so as to output copy information that is more in line with the current context.

[0048] In addition, copy context prompts can also improve the language flexibility of the copy generation model, enabling it to adapt to different language expressions and context changes.

[0049] Furthermore, the target information object, target subject content, copy context prompt words and copy requirement description information are input into the copy generation model 21. The copy generation model 21 generates copy information that is adapted to the target information object and meets the copy requirement description information based on the semantic understanding and context perception of the copy context prompt words.

[0050] In this embodiment, the method of generating the copywriting information adapted to the target information object and satisfying the copywriting requirement description information based on the semantic understanding and context perception of the copywriting context prompt words is not limited. An optional implementation is introduced below, but it does not limit this embodiment.

[0051] In this embodiment, the copy generation model 21 first decomposes, vectorizes and positionally encodes the copy context prompt words, target information objects and copy requirement description information; wherein, the input of the copy generation model 21 is first decomposed into a series of tokens by a tokenizer, each token can be a word, a subword or a character; further, each token is vectorized to obtain a corresponding token vector, thereby converting each token into a high-dimensional vector representation, which captures the semantic information of the token; further, each token vector is positionally encoded to obtain multiple token vectors with position information; wherein, the token vector with position information enables the copy generation model to understand the structure and order of the sentence.

[0052] Secondly, when the tag vector with position information is obtained, further, semantic features and context features are extracted from multiple tag vectors with position information based on the causal attention mechanism to obtain the semantic features and context features corresponding to each position information. Optionally, a causal decoder is used to extract semantic features and context features; wherein the causal decoder includes a causal attention layer, a feature extraction layer, a residual connection and a normalization layer; wherein a mask (Masked Self-Attention) is used in the causal attention layer, so that when generating the current tag information, only the tag vectors at the current moment and before are accessed, and the tag vectors in the future are not seen, so that the autoregressive nature of the generation process can be maintained; on the other hand, the causal attention layer also uses the encoder-decoder attention mechanism (Encoder-Decoder Attention) to refer to the context information provided by the encoder during the generation process; after being processed by the causal self-attention layer, a context-enhanced tag vector is obtained; further, the feedforward neural network transforms and extracts features from the context-enhanced tag vector to obtain the semantic features and semantic features of each position information.

[0053] Furthermore, the copy generation model 21 generates copy information based on the semantic features and context features corresponding to each position information, wherein the copy information is obtained by gradually generating tag information in the order of the positions, and when generating the tag information at the next position, all the previously generated tag information will be used as input, and the tag information at the next position will be predicted based on these tag information.

[0054] Optionally, the copy information may include different levels of copy information, and the different levels of copy information may be superimposed on the display position corresponding to the background. For example, the copy information may include two levels: title and subtitle. The title may be superimposed on a prominent position in the background image, such as the center or the top, to attract the user's attention and convey the core concept of the target subject content; the subtitle may be used as an extension of the title to help users quickly understand the details of the target subject content. The subtitle may be superimposed on a relatively minor position in the background image, such as near the title. This may form a hierarchical expression of the copy information, which may help users quickly understand the hierarchy and structure of the target page content. Figure 3a The examples of “banana greening prohibited”, “greenery preferred, not banana greening”, “indoor greening”, “elegant vase”, “banning green ornaments”, “rejecting banana greening in summer”, “flower growing expert”, and “plants bring good mood” shown in the text generation model are examples of text information generated by the text generation model. Figure 3bThe "indoor greenery", "elegant home vase", "forbidden green ornaments", "reject banana green in summer", "elegant home vase" with green background, "high-end flower arrangement", "retro flower arrangement", "vase to decorate the home" and so on are another example of the copywriting information generated by the copywriting generation model.

[0055] Further, if Figure 2 As shown, based on the target information object and the target subject content, the text graph model 24 is called to generate a background graph corresponding to the target subject content. In an optional embodiment, according to the target subject content, the visual attribute prediction model 22 is called to predict the target visual element information adapted to the target subject content based on the visual element information in the historical background graph corresponding to the historical subject content; the visual element information is the visual features constituting the background graph, which may include but is not limited to the atmosphere, color, and style of the background graph, etc., and is not limited to this.

[0056] Among them, the historical theme content is the theme content input in the past, and the historical background picture is the historical background picture corresponding to the historical theme content generated by calling the text graph model in the past; similar to the copywriting generation model 21, the visual attribute prediction model 22 can also be pre-trained using a variety of context learning methods to enable the model to have context learning capabilities, so as to perform analogical learning on the visual element information in the historical background picture corresponding to the historical theme content provided in the input of the model, and predict and generate the visual element information corresponding to the target theme content based on the visual element information.

[0057] The visual attribute prediction model 22 may be a large language model, and may be the same or different model from the text generation model 21 and the text image model 24, without limitation. The same model mainly refers to the same model structure, and may not refer to the same model, and these models perform their respective tasks. Regarding the working principle of the visual attribute model 22, reference may be made to the implementation of the above-mentioned text generation model, which will not be repeated here.

[0058] Further, based on the target visual element information and the target information object, the scene description generation model 23 is called to generate scene description information that is adapted to the target subject content; the scene description information is detailed information describing the background image; it may include scene keywords and scene detail information; wherein, scene keywords refer to the core visual features of the background image, such as core objects, colors, etc., such as the object may be a sofa, the color may be green, etc.; scene detail information refers to scene description information other than scene keywords, such as style, atmosphere, light and shadow, etc.; wherein the style may be realistic, modern or simple; the atmosphere may be warm, comfortable and bright, etc.; then, based on the scene description information, the text map model 24 is called to generate a background image corresponding to the target subject content; the scene description information can help the text map model accurately understand the core visual features, atmosphere and style, etc. in the background image, thereby generating a background image that meets expectations.

[0059] In an optional implementation, based on the scene description information, the text graph model 24 is called to generate a background graph corresponding to the target subject content, which can be implemented based on the following steps:

[0060] Input the scene description information into the text encoder in the text graph model 24 for vector embedding to obtain a semantic vector corresponding to the scene description information; wherein the text encoder in the text graph model 24 has been pre-trained and has the ability to map the scene description information in text form into the semantic space to obtain the corresponding semantic vector;

[0061] Further, the semantic vector is input as a generation condition into the diffusion generator in the Wensheng graph model 24, and the semantic vector is denoised based on the cross attention mechanism and the jump connection mechanism to obtain the feature vector corresponding to the scene description information; wherein, the diffusion generator in the Wensheng graph model 24 has the following capabilities after the pre-model training process: according to the guidance of the input generation condition, a random noise vector is continuously denoised, and finally a feature vector that meets the generation condition requirements is obtained. Based on this model capability, the diffusion generator can be used to more accurately determine the feature vectors corresponding to multiple scene description information. In this embodiment, the semantic vectors corresponding to multiple scene description information are input into the diffusion generator as the generation condition of the diffusion generator, so that the diffusion generator can continuously denoise the random noise vector with the semantic vector corresponding to each scene description information as the guidance, and finally generate the feature vectors corresponding to multiple scene description information more accurately. In this embodiment, the number of times the random noise vector is denoised is not limited, for example, it can be 30-50 times, and the semantic vector corresponding to the scene description information is continuously injected into the noise vector during each denoising process, so as to obtain a feature vector with rich semantic information, which can also be called a latent space vector.

[0062] Then, the feature vector is input into the image decoder of the Wensheng graph model 24 for decoding to obtain the background image corresponding to the target subject content. The image decoder of the Wensheng graph model 24 has been trained in advance and has learned the mapping relationship between the feature vector and the corresponding background image. Based on this, the image decoder of the Wensheng graph model 24 can decode the feature vectors corresponding to the multiple scene description information based on the mapping relationship to obtain the background image corresponding to the target information object. Figure 3a The background image including green plants and sofas shown in is an example of a background image generated by the text image model. The background image includes the copy information listed above as well as various product images, price information, and discount information. Figure 3b The background image shown, including the flower arrangement and the vase, is another example of a background image generated by the text image model. The background image includes the text information listed above as well as various product images, price information, and discount information.

[0063] It is stated here that Figure 2 In order to facilitate the illustration, the branch of selecting the target information object from the candidate information objects using the semantic embedding model 20 is repeatedly displayed. In practical applications, it only needs to be executed once. In addition, when the background image and text information are obtained, the target page content can be generated according to the text information, the background image and the detailed information of the target information object. This part of the content is in Figure 2 Further optionally, generating the target page content according to the text information, the background image and the detailed information of the target information object includes: adding the text information to the background image according to the display position of the text information in the page to obtain the basic page content; adding the detailed information of the target information object to the basic page content according to the layout and style information of the target information object to obtain the target page content.

[0064] Further optionally, based on the target information object and the target subject content, the Wensheng video model is called to generate dynamic picture content corresponding to the target subject content; wherein the dynamic picture content can be an animated picture and / or a video; then the target page content is generated according to the copy information, background image and detailed information of the target information object, including: generating the target page content according to the copy information, background image, dynamic picture content and detailed information of the target information object.

[0065] Further optionally, the target page content is generated according to the text information, background image, dynamic picture content and detailed information of the target information object, including: adding the text information and dynamic picture content to the background image according to their display positions in the page to obtain basic page content; adding the detailed information of the target information object to the basic page content according to the layout and style information of the target information object to obtain the target page content.

[0066] In an embodiment of the present application, after generating the target page content mainly based on the target theme content, the target page content can be published to the target application, so that the user can access the target page corresponding to the target page content through the target application, thereby attracting users through the target page, increasing the user's stay time in the target application, and thus improving the shopping guide efficiency and conversion rate of the target application. There is no limitation on the way to publish the target page content to the target application. For example, the target application can be updated, or the target page content can be published to the target application by patching or plug-in. Further, the embodiment of the present application does not limit the implementation form of the target page corresponding to the target page content in the target application. For example, the access entry information of the target page can be embedded in the existing page of the target application, so that the user can access the target page through the access entry information, wherein the existing page can be a specific page in the target application. Taking e-commerce applications as an example, the specific page can be but not limited to: home page, shopping cart page, product details page, etc.; or, a new tag page can be added in the target application, and the target page can be implemented as the tag page.

[0067] In an optional embodiment, the target page content can be stored on the server side. Based on this, the method also includes: receiving an access request sent by a terminal device running a target application, the access request being used to request access to the target page; sending the target page content to the terminal device so that the terminal device can display the target page by rendering the target page content during the running of the target application.

[0068] In the embodiment of the present application, a page access method is also provided, such as Figure 4a As shown, the method includes:

[0069] 401. During the operation of the target application, a specific page is displayed, and the specific page includes access entry information of the target page;

[0070] 402. In response to the user's access operation to the access portal information, the target page content is obtained, and the target page content is rendered to display the target page, wherein the target page includes a background image, text information carried on the background image, and detailed information of at least one target information object.

[0071] In this embodiment, the target page content is generated according to the page content generation method in the above embodiment. The generation process and beneficial effects of the target page content have been described in detail in the above embodiment, and will not be elaborated here.

[0072] In an optional embodiment, the target application is implemented as an e-commerce application, and the specific page can be the home page of the e-commerce application, or other pages, such as a shopping cart page, or an existing tag page. The specific page includes access entry information of the target page, and the access entry information can be various forms of interface elements that can jump to the target page, such as floating icons, text, links, etc. Figure 4b In the example of the home page of an e-commerce application, a floating icon with the words "Mid-Autumn Festival Venue" is displayed on the home page. The floating icon points to the target page corresponding to the "Mid-Autumn Festival Venue". When a user clicks the floating icon, he or she can jump to the target page corresponding to the "Mid-Autumn Festival Venue". Figure 4b In the figure, the Mid-Autumn Festival venue page is taken as an example. Figure 4b The target page corresponding to the "Mid-Autumn Festival Venue" shown is obtained by rendering the target page content. Figure 4b In the Mid-Autumn Festival venue page, a background picture related to "Mid-Autumn Festival" is displayed, and the background picture displays text information related to "Mid-Autumn Festival", as well as detailed information of various products related to "Mid-Autumn Festival", including product pictures, prices, discount information, etc.

[0073] Figure 4b The Mid-Autumn Festival venue page shown belongs to a new e-commerce scenario that displays product information around a specific theme, namely "Mid-Autumn Festival". It creates an immersive shopping experience for users by centrally displaying highly relevant product information and presenting graphic and text information related to specific themes or activities. It can not only significantly increase the user's online stay time, but also effectively improve shopping guide efficiency and conversion rate.

[0074] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 101 to 103 can be device A; for another example, the execution subject of steps 101 and 102 can be device A, and the execution subject of step 103 can be device B; and so on.

[0075] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel, and the sequence numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0076] Figure 5 FIG. 1 is a schematic diagram of a structure of an electronic device provided by another exemplary embodiment of the present application. Figure 5 As shown, the electronic device includes: a memory 54 and a processor 55.

[0077] The memory 54 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, contact data, phone book data, messages, pictures, videos, etc.

[0078] The memory 54 can be implemented by any type of volatile or nonvolatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0079] The processor 55 is coupled to the memory 54 and is used to execute the computer program in the memory 54, so as to: determine the target subject content adapted to the target application and the target information object adapted to the target subject content; based on the target information object and the target subject content, call the copy generation model to generate the copy information corresponding to the target subject content; based on the target information object and the target subject content, call the text generation model to generate the background image corresponding to the target subject content; generate the target page content according to the copy information, the background image and the detailed information of the target information object, and the target page content is used to render the target page corresponding to the target subject content.

[0080] In an optional embodiment, when determining the target subject content adapted to the target application, the processor 55 is specifically configured to: mine search hotspot information from the search data of the target application, and determine the target subject content based on the search hotspot information.

[0081] In an optional embodiment, when determining the target information object that is compatible with the target subject content, the processor 55 is specifically used to: input the candidate information object and the target subject content into a semantic embedding model for semantic feature extraction to obtain an embedding vector of the candidate information object and an embedding vector of the target subject content; calculate the similarity between the embedding vector of the candidate information object and the embedding vector of the target subject content, and select the target information object from the candidate information objects based on the similarity.

[0082] In an optional embodiment, when the processor 55 calls the copy generation model based on the target information object and the target subject content to generate copy information corresponding to the target subject content, it is specifically used to: determine the copy context prompt words and copy requirement description information based on the type of the target subject content; input the target information object, the target subject content, the copy context prompt words and the copy requirement description information into the copy generation model, and based on the semantic understanding and context perception of the copy context prompt words, generate copy information that is adapted to the target information object and satisfies the copy requirement description information.

[0083] In an optional embodiment, when the processor 55 generates the copy information that is adapted to the target information object and satisfies the copy requirement description information based on the semantic understanding and contextual perception of the copy context prompt words, it is specifically used to: decompose, vectorize and position-encode the copy context prompt words, the target information object and the copy requirement description information to obtain multiple tag vectors with position information; extract semantic features and context features from the multiple tag vectors with position information based on the causal attention mechanism to obtain the semantic features and context features corresponding to each position information; generate the copy information based on the semantic features and context features corresponding to each position information.

[0084] In an optional embodiment, when the processor 55 calls the text graph model based on the target information object and the target subject content to generate a background image corresponding to the target subject content, it is specifically used to: call the visual attribute prediction model according to the target subject content, and predict the target visual element information adapted to the target subject content based on the visual element information in the historical background image corresponding to the historical subject content; call the scene description generation model according to the target visual element information and the target information object to generate scene description information adapted to the target subject content; call the text graph model according to the scene description information to generate a background image corresponding to the target subject content.

[0085] In an optional embodiment, when the processor 55 calls the text graph model according to the scene description information to generate a background image corresponding to the target subject content, it is specifically used to: input the scene description information into the text encoder in the text graph model for vector embedding to obtain a semantic vector corresponding to the scene description information; input the semantic vector as a generation condition into the diffusion generator in the text graph model, and denoise the semantic vector based on the cross attention mechanism and the jump connection mechanism to obtain a feature vector corresponding to the scene description information; input the feature vector into the image decoder of the text graph model for decoding to obtain the background image corresponding to the target subject content.

[0086] In an optional embodiment, the processor 55 is also used to: based on the target information object and the target subject content, call the Wensheng video model to generate dynamic picture content corresponding to the target subject content; generate target page content according to the copy information, the background image and the detailed information of the target information object, including: generating the target page content according to the copy information, the background image, the dynamic picture content and the detailed information of the target information object.

[0087] In an optional embodiment, when the processor 55 generates the target page content based on the text information, the background image and the detailed information of the target information object, it is specifically used to: add the text information to the background image according to the display position of the text information in the page to obtain the basic page content; add the detailed information of the target information object to the basic page content according to the layout and style information of the target information object to obtain the target page content.

[0088] In an optional embodiment, the processor 55 is also used to: receive an access request sent by a terminal device running the target application, the access request being used to request access to a target page; and send target page content corresponding to the target page to the terminal device, so that the terminal device can render the target page content to display the target page during the process of running the target application.

[0089] Further, if Figure 5 As shown, the electronic device also includes: a communication component 56, a display 57, a power component 58, an audio component 59 and other components. Figure 5 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 5 In addition, Figure 5The components in the dashed box are optional components, not mandatory components, and the specific components depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, or an IOT device, or a server device such as a conventional server, a cloud server, or a server array. If the electronic device of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, etc., it can include Figure 5 If the electronic device of this embodiment is implemented as a server device such as a conventional server, a cloud server or a server array, it may not include Figure 5 Components within the dashed box.

[0090] The embodiment of the present application also provides an electronic device, the implementation structure of the electronic device is Figure 5 The implementation structure of the electronic device shown is the same or similar, and can refer to Figure 5 The electronic device provided in this embodiment is implemented as follows: Figure 5 The difference between the electronic devices in the illustrated embodiments mainly lies in that the functions implemented by the processor in the electronic device executing the computer program stored in the memory are different. For the electronic device provided in this embodiment, its processor executes the computer program stored in the memory, which can be used to: display a specific page during the operation of the target application, the specific page includes the access entry information of the target page; in response to the user's access operation to the access entry information, obtain the target page content, render the target page content, and display the target page; wherein the target page is generated according to the page content generation method.

[0091] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor is enabled to implement the steps in the above method.

[0092] An embodiment of the present application also provides a computer program product, which includes a computer program / instructions. When the computer program / instructions are executed by a processor, the processor is enabled to implement the steps in the above method embodiment.

[0093] The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0094] The above-mentioned communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.

[0095] The above-mentioned display includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0096] The power supply assembly provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply assembly is located.

[0097] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (Microphone, MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.

[0098] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-readable storage media (including but not limited to disk storage, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical storage, etc.) containing computer-usable program code.

[0099] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0102] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input / output interface, network interface and memory.

[0103] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0104] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0105] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0106] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for generating page content, characterized in that: include: Determine a target subject content adapted to a target application and a target information object adapted to the target subject content; Based on the target information object and the target subject content, calling a copywriting generation model to generate copywriting information corresponding to the target subject content; Based on the target information object and the target subject content, calling the text graph model to generate a background graph corresponding to the target subject content; The target page content is generated according to the text information, the background image and the detailed information of the target information object, and the target page content is used to render the target page corresponding to the target subject content.

2. The method according to claim 1, characterized in that Determine the target subject content that is suitable for the target application, including: Search hotspot information is mined from the search data of the target application, and the target subject content is determined based on the search hotspot information.

3. The method according to claim 1, characterized in that Determining a target information object that is compatible with the target subject content includes: Inputting the candidate information object and the target subject content into a semantic embedding model to extract semantic features, so as to obtain an embedding vector of the candidate information object and an embedding vector of the target subject content; The similarity between the embedding vector of the candidate information object and the embedding vector of the target subject content is calculated, and the target information object is selected from the candidate information objects according to the similarity.

4. The method according to claim 1, characterized in that: Based on the target information object and the target subject content, calling a copywriting generation model to generate copywriting information corresponding to the target subject content includes: Determine the copy context prompt words and copy requirement description information based on the type of the target subject content; The target information object, the target subject content, the copy context prompt words and the copy requirement description information are input into the copy generation model, and based on the semantic understanding and context perception of the copy context prompt words, copy information that is adapted to the target information object and meets the copy requirement description information is generated.

5. The method according to claim 4, characterized in that Based on the semantic understanding and context perception of the copy context prompt words, copy information that is adapted to the target information object and meets the copy requirement description information is generated, including: Decomposing, vectorizing and position-encoding the copy context prompt words, the target information object and the copy requirement description information to obtain a plurality of tag vectors with position information; Extracting semantic features and contextual features from the multiple tag vectors with position information based on a causal attention mechanism to obtain semantic features and contextual features corresponding to each position information; The copywriting information is generated based on the semantic features and context features corresponding to each position information.

6. The method according to claim 1, characterized in that Based on the target information object and the target subject content, calling the text graph model to generate a background graph corresponding to the target subject content includes: According to the target subject content, calling a visual attribute prediction model, based on the visual element information in the historical background image corresponding to the historical subject content, predicting target visual element information adapted to the target subject content; According to the target visual element information and the target information object, a scene description generation model is called to generate scene description information adapted to the target subject content; The text graph model is called according to the scene description information to generate a background graph corresponding to the target subject content.

7. The method according to claim 6, characterized in that According to the scene description information, calling the text graph model to generate a background graph corresponding to the target subject content includes: Inputting the scene description information into a text encoder in the text graph model for vector embedding to obtain a semantic vector corresponding to the scene description information; Inputting the semantic vector as a generation condition into the diffusion generator in the text graph model, and performing denoising on the semantic vector based on a cross attention mechanism and a skip connection mechanism to obtain a feature vector corresponding to the scene description information; The feature vector is input into the image decoder of the text graph model for decoding processing to obtain a background image corresponding to the target theme content.

8. The method according to claim 1, characterized in that Also includes: Based on the target information object and the target subject content, calling the Wensheng video model to generate dynamic picture content corresponding to the target subject content; Rendering the text information, the background image, and the detailed information of the target information object to generate a target page corresponding to the target subject content includes: The text information, the background image, the dynamic picture content and the detailed information of the target information object are rendered to generate the target page.

9. The method according to any one of claims 1 to 8, characterized in that: Generating target page content according to the copy information, the background image and the detailed information of the target information object includes: According to the display position of the text information in the page, the text information is added to the background image to obtain basic page content; According to the layout and style information of the target information object, the detail information of the target information object is added to the basic page content to obtain the target page content.

10. The method according to any one of claims 1 to 8, characterized in that: Also includes: Receiving an access request sent by a terminal device running the target application, wherein the access request is used to request access to the target page content; The target page content is sent to the terminal device, so that the terminal device renders the target page based on the target page content during the running of the target application.

11. A page access method, characterized in that: include: During the running of the target application, a specific page is displayed, wherein the specific page includes access entry information of the target page; In response to a user's access operation to the access entry information, obtaining target page content, and rendering the target page content to display the target page; Wherein, the target page content is generated according to the method described in any one of claims 1-10.

12. An electronic device, characterized in that: include: Memory and processor; The memory is used to store a computer program, and the processor is coupled to the memory and is used to execute the computer program to implement the steps in the method according to any one of claims 1 to 10 and claim 11.

13. A computer storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is enabled to implement the steps of the method according to any one of claims 1 to 10 and claim 11.

14. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, enables the processor to implement the steps of the method according to any one of claims 1 to 10 and claim 11.

Images