CN117909564A

CN117909564A - Content speed and super personalization using generated AI

Info

Publication number: CN117909564A
Application number: CN202310963347.5A
Authority: CN
Inventors: O·布拉迪克兹卡; K·坎塔瓦拉; I·罗斯卡; A·达拉比; A·V·科斯汀; A·奇库丽塔
Original assignee: Adobe Systems Inc
Current assignee: Adobe Inc
Priority date: 2022-10-17
Filing date: 2023-08-02
Publication date: 2024-04-19

Abstract

Embodiments of the present disclosure relate to content speed and super-personalization using a generative AI. A method includes receiving a description of content to be generated using a generative model. The received description of the content is associated with a user profile. The method also includes determining semantic terms based on the description of the content. The method also includes generating a user-specific template that includes semantic terms and user preferences associated with the user profile. The method also includes generating content using the generative model based on the user-specific template. The method also includes outputting the content for display on the target user device.

Description

Content speed and super personalization using generated AI

Cross Reference to Related Applications

The present application claims the benefit of U.S. provisional application No.63/416,874 filed on 10 months 17 of 2022 and U.S. non-provisional application No.18/306,408 filed on 24 months 4 of 2023, the disclosures of which are incorporated herein by reference in their entireties.

Background

Content generation speed prevents the expansion of personalized content. Creating personalized high quality content for the target audience requires manual, repetitive, and labor intensive processes. Target audience members include, but are not limited to, age groups, geographic groups, and individual users.

Disclosure of Invention

Described herein are techniques/processes for generating personalized content for a target audience. The personalized content recommendation system utilizes natural language processing, personalization techniques, and generation AI to generate content on a large scale. Generating such large-scale content enables designers to produce super-personalized content for different viewers and purposes.

More specifically, in one or more embodiments, the personalized content recommendation system decomposes the intent/title (or other received description of the content to be generated) into personalized variables. The meta-templates are employed by the personalized content recommendation system to recommend content on a large scale, wherein the meta-templates include personalized variables, control loops, and sub-hints that drive one or more generated AI algorithms. As described herein, the sub-hints may include a free-text description (e.g., a preamble) and one or more personalized variables derived from the input. Using the meta-templates, a user (such as a designer or other content producer) navigates through the space of the multi-level semantic representation to refine the combination of images generated from the sub-hints. Specifically, the user may refine (refine) the personalized variables extracted from the decomposed intent/title (e.g., the colors to be used in the content, the overall emotion of the content, etc.), and provide such refinement as variables into the sub-cues.

Using the meta-template, a user (such as a designer or other content producer) may personalize content for a consumer (user consuming the generated content) by utilizing the consumer profile. In this way, the content is created by a personalized content recommendation system that is consistent with the intent of the user (such as the designer), but dynamically provides personalized content to the consumer. In some embodiments, a layer within a meta-template (or sub-hint) may be used as a search term for a repository of consumers or marketplace content.

In some embodiments, the meta-templates are personalized with respect to a given user (such as a designer) and/or user/consumer group (e.g., user/consumer group of the same age, user/consumer group in the same/similar geographic location, user/consumer group providing content (or consuming content) in the same/similar industry/domain). For example, the personalized content recommendation system may utilize collaborative filtering or other personalization techniques to determine sub-cues for a meta-template for a particular user (or group of users).

Additional features and advantages of exemplary embodiments of the disclosure will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of such exemplary embodiments.

Drawings

The specific embodiments are described with reference to the accompanying drawings, in which:

FIG. 1 shows an illustration of a process for generating personalized content for a target audience in accordance with one or more embodiments;

FIG. 2 illustrates a meta template in accordance with one or more embodiments;

FIG. 3 illustrates an image generated using a personalized content recommendation system in accordance with one or more embodiments;

FIG. 4 illustrates another meta-template in accordance with one or more embodiments;

FIG. 5 illustrates a batch of images generated using a personalized content recommendation system in accordance with one or more embodiments;

FIG. 6 illustrates an example of content generated in accordance with one or more embodiments;

FIG. 7 illustrates an example implementation of a diffusion model in accordance with one or more embodiments;

FIG. 8 illustrates a diffusion process for training a diffusion model in accordance with one or more embodiments;

FIG. 9 illustrates a schematic diagram of a personalized content recommendation system in accordance with one or more embodiments;

FIG. 10 illustrates a flow diagram of a series of acts in a method of generating personalized content for a target audience in accordance with one or more embodiments; and

FIG. 11 illustrates a block diagram of an exemplary computing device in accordance with one or more embodiments.

Detailed Description

One or more embodiments of the present disclosure include a personalized content recommendation system that generates content that is personalized (or super-personalized) for different viewers and purposes. One conventional approach involves manually creating each individual piece of content and adapting the content for purposes in the audience or activity. Such a process limits the content generation speed. Other conventional methods automatically replicate certain content. However, these methods still limit content generation speed when one or more users adjust the copied content by replacing different images, fonts and/or colors, reorganizing the content, etc. Still other conventional methods automatically generate images from descriptions. However, these approaches limit the control and customization of the large-scale content generation process. For example, these conventional methods may produce content of different quality, but there is little or no control by the user to edit the large-scale generation of such content.

To address these and other deficiencies in conventional systems, the personalized content recommendation system of the present disclosure combines natural language processing, generation AI, and personalization techniques to produce personalized content to different viewers, where a viewer may be a group of consumers in a certain age group, a group of consumers in a certain geographic location, a group of consumers with similar favorites, a single user, a group of consumers of the same gender, and so forth. At any time, the user (such as a designer) may edit the personalized variable to generate a new personalized content recommendation.

Allowing users to mass-generate content reduces the computational resources (e.g., power, memory, bandwidth, etc.) expended to adjust or otherwise adjust visual elements of the content to a target audience. In addition, enabling users to edit the generated content reduces computing resources associated with the multiple execution system. For example, the user may refine any one or more of the sub-cues determined using the personalized variable extracted from the user input before the user executes the personalized content recommendation system. Refinement limits the amount of wasted computing resources by allowing a user to make such refinement before the personalized content recommendation system is executed, as opposed to executing the recommendation system and re-executing the recommendation system after refining one or more parameters.

FIG. 1 illustrates a diagram of a process for generating personalized content for a target audience in accordance with one or more embodiments. As shown in fig. 1, an embodiment includes a content recommendation system 100. The content recommendation system 100 includes a text parser 102, a meta-template manager 104, a generative AI module 106, a PCF manager 126, and a meta-template storage 114. The content recommendation system 100 is responsible for generating content on a large scale.

At reference numeral 1, the content recommendation system 100 receives an input 120. The input 120 is a title of the content, a purpose of the content, a user intent, or some description of the content (e.g., consumable content) to be generated. The input 120 may be received in the form of text input, selections from a drop down menu, audio-to-text transcription, and the like. As described herein, the input 120 may be attached with user identification information such that the meta-template manager 104 (described herein) may map the user identification information to a user profile.

At reference numeral 2, the text parser 102 parses or otherwise decomposes the input 120 into one or more personalized variables. Such personalized variables may be semantically related blocks determined from the input 120. A semantic block is a semantically related phrase, term, etc., derived from the input 120 that may be stored as a personalized variable for future use.

In some embodiments, the text parser 102 uses natural language processing techniques to expand the semantic blocks derived from the input 120. For example, the text parser 102 parses the input 120 using any suitable parsing technique (such as a sentence segmenter). The text parser 102 then performs any suitable semantic search/semantic relevance technique to derive personalized variables from the parsed input 120. For example, geographic location, landmarks, style, emotion, persona, color, etc. may be semantic personalization variables derived from input 120 using any suitable semantic search/semantic relevance technique.

In a particular example, the text parser 102 extracts location personalization variables from the input 120. For example, the text parser 102 may utilize any one or more portions of the phonetic marking and entity extraction algorithms. In a different example, the text parser 102 uses a list of predetermined locations and any one or more previously extracted personalization variables to extract landmark personalization variables. For example, the text parser 102 may utilize a language converter model (such as GPT-3) to generate/isolate the most suitable landmarks for locations listed in a predetermined location list.

In some embodiments, the text parser 102 further augments the set of landmarks using vectors of the large word2vec model. Specifically, the word2vec vector includes vector representations of landmarks and corresponding geographic locations. Such a relationship characterizing landmarks and geographic locations results in identifying semantically related categories in the input. For example, people, geographic locations, landmarks, etc. are semantically related personalized variables extracted by the text parser 102. A person refers to a person from a certain location. For example, "german" from germany, "paris" from paris, and the like.

In yet a different example, the text parser 102 uses a zero-sample (zero-shot) classification transformer model (such as BART) that operates on a limited emotion dictionary to extract one or more emotion or theme personalization variables. In another example, the text parser 102 extracts colors as personalization variables using any suitable mechanism. For example, in one implementation, one or more word2vec vectors may use vector similarity running on all nouns in the input 120 to provide an initial list of colors related to the input 120 (e.g., extracted by a voice Part (POS) tag, POS tag NN, NNP, etc.). In another implementation, the colors may be extracted by the text parser 102 as personalized variables using the generated AI. For example, the generation formula AI is used to generate a small image from the input 120. The text parser 102 then extracts dominant colors from the image.

In some embodiments, the personalized variable extracted from the input depends on the meta-template storage 114. At reference numeral 3, the meta-template storage 114 prompts the text parser 102 to extract certain personalized variables from the input. As described herein, the meta-template storage 114 stores template information, such as personalized variables, refinements 112, profile information 116, sub-hints, and the like. The meta-template storage 114 may be a repository (e.g., server, database, etc.) hosted by the personalized content recommendation system 100, hosted by one or more external systems, or some combination (e.g., cloud storage system). The meta-template storage 114 may direct the text parser 102 to extract personalized variables (e.g., profile information 116) corresponding to a particular user and/or company by storing information corresponding to the particular user and/or company. As described herein, profile information 116 includes both end-user (e.g., designer, company, etc.) and consumer user (e.g., target consumer) profiles. Depending on the profile information 116, different personalization variables are extracted from the input. For example, the meta-template storage 114 may use profile information 116 (such as a user identifier, a company identifier, etc.) to associate a user and/or company with a particular meta-template. In some embodiments, profile information 116 is attached to input 120. As described herein, a meta-template is a user-specific template that is populated with personalized variables and corresponding sub-hints extracted from the input 120.

At reference numeral 4, the personalized variables to be extracted from the input are fed to the meta-template manager 104. The meta-template manager 104 organizes templates to be displayed to end users (designers or other content producers). Templates are assemblies of sub-hints that may be determined manually and/or automatically. As described herein, the sub-hints may include free text descriptions provided by the user and/or one or more personalized variables derived from the input 120. Additionally or alternatively, the sub-hint may be part of the input 120. In some implementations, the operation of meta-template manager 104 is the same or similar to the operation of PCF manager 126 described herein.

In some embodiments, the meta-template manager 104 queries the meta-template store 114 at reference numeral 3 for meta-templates associated with the end-user profile (as opposed to the target consumer user profile). Using the user profile information 116, the meta-template manager 104 populates the meta-template with user preferences. The meta-templates associated with different users (such as designers), companies, entities, etc. are stored in the meta-template storage 114. As described herein, the meta-template manager 104 determines a meta-template associated with a user profile in response to an identifier associated with the user. An identifier (such as a user number, user name, email address, phone number, etc.) may be attached to the input 120, associated with the input 120 metadata, etc. The meta-templates may be varied, for example, by sub-hints associated with the user profile identified via profile information 116. In addition, the meta-templates may vary depending on one or more control loops and/or constraints. Such control loops/constraints may refine the use of personalized variables in the meta-template. For example, the constraints may specify a particular subset of personalized variables for use in a particular meta-template. For example, user preferences (obtained via profile information 116) may exclude < emotion > personalization variables from the meta-template. Thus, the meta-template manager 104 does not prompt the text parser for < emotion > personalized variables extracted from the input 120.

At reference numeral 4A, the meta-template manager 104 displays the meta-templates to the user and receives one or more user refinements 112 determined by the end user (e.g., content designer). User refinements 112 are modifications to the semantic blocks determined by text parser 102, addition of one or more sub-hints in the meta-template, deletion of one or more sub-hints in the meta-template, modification of sub-hints in the meta-template, modified intent (e.g., new input 120), modified control loops and/or constraints, addition of control loops and/or constraints, removal of control loops and/or constraints, and the like. For example, the user may navigate the space of personalized variables in the displayed meta-templates as a multi-level semantic representation. The user is provided access to the personalized variables to refine the composition of the image generated from the sub-cues. For example, the user may choose between colors or emotions that have been extracted for insertion into the sub-cues. Additionally or alternatively, the user may also define semantics and/or design rules in a prompt field specific language to configure contrasting colors, make decisions regarding specific conditions in personalized variables, and the like. The user refines 112 the personalized variable and original hint operation fed to the generated AI module 106.

At reference numeral 4B, the meta-template manager 104 displays the meta-templates to the user and receives profile information 116. As described herein, profile information 116 is any information specific to a consumer user (e.g., a target user) and/or an end user (e.g., a designer). For example, profile information 116 associated with the consumer user may include geographic location, name, favorite color, occupation, type of computing device, music preferences, and the like. The meta-template manager 104 incorporates the profile information 116 into the sub-hints. The more information is provided to the generated AI module 106 (in the form of hints based on meta-template sub-hints), the more personalized and specific the generated content produced by the generated AI module will be. In this way, personalized content corresponding to the end user's (designer's) intent with respect to a particular consumer user is generated based on the profile information. For example, an end user (e.g., a content producer) determines a preferred style of content to display to a client, where the content itself is based on the client.

In some embodiments, an end user (e.g., designer) enters profile information 116 (e.g., consumer information) to super-personalize content determined by the personalized content recommendation system 100. In other embodiments, the meta-template manager 104 queries one or more servers, applications, databases, etc. for profile information 116. In these embodiments, the meta-template manager 104 populates the sub-hints with the received profile information. In still other embodiments, the user inputs profile information as part of the input 120. In these embodiments, text parser 102 extracts the profile information such that the extracted profile information may be populated into the sub-hints.

At reference numeral 5, user-specific templates, filled with personalized variables and corresponding sub-cues (e.g., meta-templates), are fed into one or more generated AI modules 106. The generated AI module 106 generates content (e.g., images) for consumption by a consumer (or target user). The generated AI module 106 may be any generated AI module configured to generate content using hints. In some embodiments, the generated content includes text. As described herein, a hint is a meta-template that includes a plurality of sub-hints and corresponding personalized variables. Personalized variables are used in hints for the generative AI module 106, where the sub-layers of the larger synthetic document are configured using a Domain Specific Language (DSL) that includes both variable substitution and basic control loops (e.g., if/else statements, foreach statements, etc.).

The generation-type AI module 106 uses a seed as part of the content generation process performed by the generation-type AI module 106. In some embodiments, the generative AI module 106 creates a seed. The seed is an initialization state for a deterministic process, such as generating an image by the generative AI module 106. Specifically, the seed includes any configuration settings of one or more machines (or one or more virtual machines) executing the generational AI module 106. For example, the configuration settings may include the time at which the generational AI module 106 is being executed to generate content. Additionally or alternatively, the seed may include initialization noise that is noise reduced to generate content according to the template.

The generated AI module 106 may be any artificial intelligence including one or more neural networks. The neural network may include a machine learning model that may be adjusted (e.g., trained) to approximate the unknown function based on training inputs. In particular, the neural network may include a model of interconnected digital neurons that communicate and learn to approximate complex functions and generate outputs based on a plurality of inputs provided to the model. For example, the neural network includes one or more machine learning algorithms. In other words, neural networks are an algorithm that implements deep learning techniques, i.e., machine learning that utilizes a set of algorithms to attempt to model high-level abstractions in data.

In some embodiments, one or more generation layers of the generated AI module 106 can be used as search terms for a repository of user or marketplace content. For example, once a particular image for a beach in a geographic location is generated, the generated image may be used to search one or more databases for an actual beach in the geographic image having the same or similar composition, palette, negative space, etc. I.e., the content determined by the generative AI module, can be used to query semantically related images. As a result, the content output from the personalized content recommendation system 100 can be obtained using the received semantic related images.

At reference numeral 6, content is output from the personalized content recommendation system 100 as output 122. Such content may be transferred to one or more computing devices (e.g., devices of the target user), downstream processors (for subsequent processing), and so forth.

At reference numeral 7, a personalized creative file format (PCF) manager 126 aggregates one or more of the following to create a PCF document (also referred to herein as a PCF file): input 120, personalized variables extracted (e.g., from text parser 102), refined personalized variables (e.g., received from refinements 112), sub-hints generated in meta-templates, refined sub-hints in meta-templates (e.g., received from refinements 112), control loops, constraints, rules, and output 122.PCF documents may be stored in a data store (e.g., database, server, cloud hosted by one or more servers, etc.). As described herein, PCF manager 126 persists (or otherwise stores) information of the personalized content recommendation system 100 pipeline in a PCF document. The personalized content recommendation system 100 pipeline refers to the modules of the personalized content recommendation system 100 (e.g., text parser 102, meta-template storage 114, meta-template manager 104, and generated AI module 106) and the corresponding data associated with each module, resulting in a particular output 122.

In some implementations, the information resident in the PCF document includes the production AI module 106, the seed used in the production AI module 106, one or more algorithms/techniques to accommodate any signal (e.g., text, language, product insertion, or artistic changes, etc.). By storing such data, the PCF document retains the original design or business intent. Specifically, by storing the seed of the generated AI modules, the randomness of each generated AI module becomes predictable or capable of being replicated. In some embodiments, noise used in the diffusion process performed by the generated AI module 106 is also stored in the PCF document.

In some embodiments, other information (e.g., profile information) associated with a particular consumer consuming content is aggregated and stored in the PCF document. By including the profile information in the PCF document, the personalized content recommendation system 100 is able to generate super-personalized content that matches the intent of the user (such as the content producer) and dynamically provide personalized content to a particular consumer (e.g., the target user).

In some embodiments, PCF manager 126 personalizes PCF documents for a particular user (designer or content creator). For example, PCF manager 126 may personalize the display of the PCF document (e.g., actual navigation within the PCF document) based on a state representation previously chosen by the user from the multi-level semantic representation (for the previous sub-hint). For example, PCF manager 126 may employ collaborative filtering in which one or more algorithms are utilized to filter data from users to make personalized recommendations (e.g., inter-user personalization) for users with similar preferences. For example, PCF manager 126 may create a meta-template with populated preferences based on a group of users similar to the target user and their corresponding preferences. Similar users may be users in the same geographic location, users employed by the same company/entity, users of the same gender, users of the same age, etc.

Other personalization techniques (e.g., identifying "recent" user preferences) may be performed by PCF manager 126 to personalize the PCF document for a particular user (e.g., in-user personalization). For example, PCF manager 126 may store the user's editing history (e.g., refinements to sub-hints, controls, etc.) as user-specific preferences. Additionally or alternatively, PCF manager 126 stores specific user preferences associated with the sub-hints.

To achieve user personalization, PCF manager 126 may save a history of personalization variables in the recorded PCF to enable restoration of previous states or navigation between states.

In a first non-limiting example, during a first period of time, the user indicates a preference for a "cheerful" design with a red/orange hue. During a second period of time (e.g., when personalized content recommendation system 100 is being executed to create consumable content), PCF manager 126 presets (or otherwise populates) such user preferences in the meta-template. For example, user preferences may be default (or preferred) when the meta-template is displayed to the user. In this manner, PCF manager 126 persists learned user preferences, profile information, templates, semantically related terms extracted from the input, generated AI module 106 parameters, and the like.

In a second non-limiting example, a user may work in a particular area (e.g., travel). In this way, PCF manager 126 populates the template with sub-hints associated with the travel. For example, the sub-hints may be for geographic locations, landmarks, and the like.

If the user makes any refinements to any information stored in the file (e.g., terms semantically related to descriptions of content, user preferences, etc.), PCF manager 126 may update the file to include such refinements. In some embodiments, PCF manager 126 overwrites any information in the file that has been refined. In other embodiments, PCF manager 126 creates a new file that includes such refinements. In still other embodiments, PCF manager 126 stores such refinements as part of the user edit history in the file.

By persisting the personalized content recommendation system 100 pipeline, the PCF manager tracks the content output at output 122. In some embodiments, PCF manager 126 compares the current pipeline or information obtained from the personalized content recommendation system 100 pipeline (e.g., input 120, generated AI module 106, seed of generated AI module 106, templates populated by meta-template manager 104, etc.) to one or more stored PCF documents. For example, PCF manager 126 may compare the current pipeline to a plurality of recently stored PCF files, a plurality of stored PCF files associated with a user, and so forth. In particular, PCF manager 126 may compare the seed of the current pipelined generated AI module 106, the templates populated by meta-template manager 104, etc. to the pipeline of stored PCF files.

PCF manager 126 determines whether the current pipeline is similar or dissimilar by comparing the information of the current pipeline to one or more PCF files. For example, if the information of the current pipeline differs from the most recent PCF file by a threshold amount (e.g., a threshold number of information fields, where the information fields correspond to the information of the PCF file), PCF manager 126 determines that the current pipeline differs from one or more stored PCF files. By determining that the current pipeline is different, PCF manager 126 ensures that each output 122 is different from the contents of the previous output.

Alternatively, if the information of the current pipeline is similar to the most recent PCF file by a threshold amount, PCF manager 126 determines that the current pipeline is the same as or similar to one or more stored PCF files. By determining that the current pipeline is similar, PCF manager 126 ensures that output 122 is consistent with the content of the previous output.

In some embodiments, PCF manager 126 ensures that the current pipeline is similar or identical to the stored PCF file. For example, PCF manager 126 populates templates of the current pipeline with information obtained during a second time period after the first time period during which profile information is obtained, the user refines, semantically related personalized variables are extracted from input 120, and so forth. In a first non-limiting example, a user may refine a previously stored personalized content recommendation system 100 pipeline by populating the current pipeline with information of PCF files. For example, the user may continue to create personalized content after rest. In a second non-limiting example, by populating the current timeline with information of the PCF file, the user may provide less refinement to the personalized content recommendation system 100 because previously generated and user-approved output would be similarly generated by the personalized content recommendation system 100. For example, the user may have previously adjusted the meta-template and thus need not readjust the surviving meta-template.

FIG. 2 illustrates a meta template in accordance with one or more embodiments. As shown, a user may enter an input 202. As shown, the input may be received by the personalized content recommendation system 100 as free text in a text box. In response to the input 202, the personalized content recommendation system 100 generates and displays a meta-template that is specific to the end user entering the input 202. As described herein, the meta-templates may be pre-populated with sub-hints (e.g., sub-hints 204) based on one or more user preferences, corporate preferences (e.g., a company associated with the user, such as an employer), control loops, constraints, and the like. As shown, the sub-hint 204 describes sunset rich in artistic breath in "< semantic qualifier > < location > and" person from < semantic qualifier > < location > ". As described herein, the information extracted from the input 202 becomes the personalized variable for completing the sub-prompt. Specifically, the semantically related personalized variables extracted from the input 202 are the semantic blocks < emotion > and < location > indicated at 206. If the personalization variable is not extracted from the input, in some embodiments, one or more default user preferences are used to populate the meta-template. As described herein, personalized variables are semantically related phrases, terms, etc. (otherwise referred to herein as semantic blocks).

FIG. 3 illustrates an image generated using the personalized content recommendation system 100 in accordance with one or more embodiments. As shown, the meta-template 302 is populated with personalized variables extracted from the input 202. In the example of fig. 2, the tags < semantic qualifier > and < location > refer to one or more editable lists of personalized variables extracted from the input 202. Such personalized variables have been determined, as shown by meta-template 302. Specifically, sunset is on the beach and the person is a surfer. The user's intent to "visit beautiful California" (depicted in input 202 in FIG. 2) is visualized in image 304 using semantic related terms associated with a person from California (e.g., a surfer) and semantic related terms associated with the location of sunset in California (e.g., a beach).

FIG. 4 illustrates another meta-template in accordance with one or more embodiments. As shown, a user may enter an input 402 in a free text box. In response to the input 402, the personalized content recommendation system 100 generates and displays a meta template. As described herein, the meta-templates may be pre-populated with sub-hints (e.g., sub-hints 404) based on profile information (such as user preferences, company preferences, control loops, constraints, etc.). For example, the user may have modified the template to include the user preferences indicated at 404-1. In particular, the user preference is the creation of a tree associated with the geographic location. As described herein, the information extracted from the input 402 by the text parser 102 becomes a personalized variable for completing one or more sub-hints. As shown, the meta-template includes another user preference 404-2 independent of the input 402 or any semantically derived personalized variable.

FIG. 5 illustrates a collection of images generated using the personalized content recommendation system 100 in accordance with one or more embodiments. As shown, the meta-template 502 is populated with personalized variables extracted from the input 402. Specifically, text parser 102 extracts states as < location > personalization variables based on the input "california" received in input 402. As shown, only a portion of the states (e.g., alabama, alaska, arkinson, california, corrado, connecticut, tara, florida, georgia, hawaii, idaho) are selected from all states. This means that constraints (not shown) are used by the meta-template manager 104 to constrain all states to a set of identified states. In this way, content is generated based only on the set of states identified in the constraint.

As shown, from a single input (e.g., input 402), a large-scale batch of personalized content 504 is created. The generated content 504 is personalized according to each sub-hint. For example, as shown, each generated content describes a unique geographic location and faithfully depicts a description of each sub-hint in the self-hints from the meta-template. For example, user preferences (e.g., indicated at sub-hint 404-1 in FIG. 4) to create a tree corresponding to a particular geographic location are shown in the generated batch of content 504. As shown, pine fills in the "Alaska clear" activity 504-1, cactus fills in the "Aristolochia clear" activity 504-2, and palm fills in the "California clear" activity 504-3. In addition, user preferences to show orange sunset (e.g., indicated by user preferences 404-2 in fig. 4) are displayed in all generated content 504.

FIG. 6 illustrates an example of content generated in accordance with one or more embodiments. Specifically, fig. 6 shows content generated in an email to be distributed in an email application. Although an email is shown and an email application is described, the personalized recommendation system 100 may interface with and provide content to other application(s).

The generated content may originate from PCF documents that are adapted to a particular user profile. As shown, at 602, profile information (e.g., consumer information) is extracted from a PCF file in the form of a particular user email address. Using such profile information, the personalized content recommendation system generates content that may be of interest to the user (e.g., consumer/target user). For example, the target user may live in san francisco or a fan of san francisco, as evidenced by email address "SanFranciscoFan". As a result, at 604, personalized content recommendation system 100 generates an email using the generated san francisco neighbors.

Specifically, the content at 604 describes "neighborhood party (neighbor party)". Such content may be based on one or more personalized variables. In the specific example of fig. 6, the personalized content recommendation system may generate a "neighbor party" due to the < seasonal > personalized variable. For example, a neighbor party may be semantically associated with a particular season of san francisco.

As indicated by 606, the text content may be generated by a personalized content recommendation system. Such content may be editable by an end user (e.g., content producer). In some embodiments, some of the content generated by the personalized content recommendation system 100 is interactive, as shown at 606-1. That is, the personalized content recommendation system 100 may link content to one or more applications, URLs, etc. The user may reconfigure the linked content, interactive buttons corresponding to the linked content, etc.

FIG. 7 illustrates an example implementation of a diffusion model in accordance with one or more embodiments. As described herein, generating the AI may be performed using any suitable mechanism. In some embodiments, such a generated AI is performed using a diffusion model.

The diffusion model is one example architecture for performing the generated AI. Generating the formula AI involves predicting features for a given tag. For example, given a tag (or natural hint describing a "cat"), the generated AI module determines the most likely feature associated with the "cat". Features associated with the tag are determined during training using a back-diffusion process in which noisy images are iteratively denoised to obtain images. In operation, a function is determined to predict noise of potential spatial features associated with a tag.

During training, images (e.g., images of cats) and corresponding tags (e.g., "cats") are used to teach diffusion model features of cues (e.g., tags "cats"). As shown in fig. 7, the input image 702 and text input 712 are transformed into a potential space 720 using image encoder 704 and text encoder 714, respectively. As a result, potential image features 706 and text features 708 are determined from image input 702 and text input 712, respectively. The potential space 720 is the following space: wherein unobserved features are determined such that relationships and other dependencies of such features can be learned. In some embodiments, the image encoder 704 and/or the text encoder 714 are pre-trained. In other embodiments, the image encoder 704 and/or the text encoder are co-trained.

Once the image feature 706 has been determined by the image encoder 704, a forward diffusion process 716 is performed according to a fixed markov chain to inject gaussian noise into the image feature 706. The forward diffusion process 716 is described in more detail with reference to fig. 8. As a result of the forward diffusion process 716, a set of noise image features 710 is obtained.

The text feature 708 and the noise image feature 710 are algorithmically combined in one or more steps (e.g., iterations) of the back-diffusion process 726. The back diffusion process 726 is described in more detail with reference to fig. 8. As a result of performing the back diffusion, image features 718 are determined, wherein such image features 718 should be similar to image features 706. The image features 718 are decoded using an image decoder 722 to predict an image output 724. The similarity between image features 706 and 718 may be determined in any manner. In some embodiments, the similarity between the image input 702 and the predicted image output 724 is determined in any manner. The similarity between the image features 706 and 718 and/or the images 702 and 724 is used to adjust one or more parameters of the back diffusion process 726.

FIG. 8 illustrates a diffusion process for training a diffusion model in accordance with one or more embodiments. The diffusion model may be implemented using any artificial intelligence/machine learning architecture where the input and output dimensions are the same. For example, the diffusion model may be implemented according to a u-net neural network architecture.

As described herein, the forward diffusion process adds noise over a series of steps (iterations t) according to a fixed markov diffusion chain. Subsequently, the back-diffusion process removes noise to learn the back-diffusion process to construct a desired image (based on text input) from the noise. During deployment of the diffusion model, a back-diffusion process is used in the generated AI module to generate images from the input text. In some embodiments, the input image is not provided to the diffusion model.

The forward diffusion process 716 begins with an input (e.g., feature X ₀ indicated by 802). For each time step T (or iteration) up to T iterations, noise is added to feature x such that feature x _T indicated by 810 is determined. As described herein, the features of the injected noise are potential spatial features. The denoising performed during the back diffusion process 726 may be accurate if the noise injected at each step size is small. The noise added to feature X may be described as a markov chain, where the noise profile injected for each time step depends on the previous time step. That is, the forward diffusion process 716 may be expressed mathematically as

The back diffusion process 726 begins with a noise input (e.g., noise signature X _T indicated by 810). At each time step t, noise is removed from the feature. The noise removed from the features may be described as a markov chain, where the noise removed at each time step is the product of the noise removed between the features of the two iterations and a normal gaussian noise distribution. That is, the back-diffusion process 726 may be mathematically represented as a joint probability of a sequence of samples in a Markov chain, where the marginal probability is multiplied by the product of the conditional probabilities of noise added at each iteration in the Markov chain. In other words, the back diffusion process 726 is Wherein p (x _t)＝N(x_t; 0, 1).

FIG. 9 illustrates a schematic diagram of a personalized content recommendation system (e.g., the "personalized content recommendation system" described above) in accordance with one or more embodiments. As shown, personalized content recommendation system 900 may include, but is not limited to, a user interface manager 914, a text parser 902, a meta template manager 904, a generative AI module 906, a PCF manager 908, a neural network manager 912, and a training manager 916. Storage manager 910 includes meta-template storage 918, training data 920, PCF document 922, and user preferences 924.

The personalized content recommendation system 900 includes a user interface manager 914. The user interface manager 914 allows a user to provide input (e.g., title of content, purpose of content, user intent, or some description of the content to be generated, otherwise referred to herein as consumable content). The input also includes refinement of the meta-template. For example, the user may refine the modification of the semantic block determined by the text parser 902, the addition of one or more sub-hints in the meta-template, the deletion of one or more sub-hints in the meta-template, the modification of sub-hints in the meta-template, the modified intent (e.g., new input), the modified control loops and/or constraints, the addition of control loops and/or constraints, the removal of control loops and/or constraints, and the like. Similarly, the input may include profile information, where the profile information may be any information specific to the client or target user. For example, the profile information may include geographic location, name, favorite color, occupation, type of computing device, music preferences, and the like.

As shown in fig. 9, the personalized content recommendation system 900 includes a text parser 902. Text parser 902 parses or otherwise decomposes the user input into one or more personalized variables. The personalized variable is a semantically related variable determined from user input. For example, first text parser 902 parses the input using any suitable parsing technique or techniques. Text parser 902 then performs any suitable semantic search/semantic relevance technique to derive personalized variables from the parsed input. For example, geographic location, landmarks, style, emotion, persona, color, etc. may be semantic personalization variables derived from input using any suitable semantic search/semantic relevance technique.

Personalized variables derived from the input using text parser 902 are fed to meta-template manager 904. The meta-template manager 904 organizes templates to be displayed to a user (designer or other content producer). Templates are assemblies of sub-hints that may be determined manually and/or automatically. As described herein, the sub-hints may include a free-text description and one or more personalized variables derived from the input.

The meta-template manager 904 may query the meta-template storage 918 of the storage manager 910 for meta-templates associated with a user, such as a designer. The meta-templates may vary, for example, by sub-hints associated with a particular user, company, etc. In addition, the meta-templates may vary depending on one or more control loops and/or constraints. Such control loops/constraints may refine the use of personalized variables in the meta-template. For example, the constraints may specify a particular subset of personalized variables for use in a particular meta-template.

The meta-template manager 904 may receive one or more user refinements (e.g., user inputs received by the user interface manager 914). As described herein, user refinement is a modification to a semantic block determined by a text parser, an addition of one or more sub-hints in a meta-template, a deletion of one or more sub-hints in a meta-template, a modification of a sub-hint in a meta-template, a modified intent (e.g., new input), a modified control loop and/or constraint, an addition of a control loop and/or constraint, a removal of a control loop and/or constraint, and the like. Providing the user with access to the personalized variables may refine the composition of the image generated from the sub-cues. For example, the user may choose between colors or emotions that have been extracted for insertion into the sub-cues. Additionally or alternatively, the user may also define semantics and/or design rules in a prompt field specific language to configure contrasting colors, make decisions regarding specific conditions in personalized variables, and the like.

The meta-template manager 904 may also receive profile information (e.g., user input received by the user interface manager 914). As described herein, profile information is any information specific to a consumer or target user. For example, the profile information may include geographic location, name, favorite color, occupation, type of computing device, music preferences, and the like. Given the profile information, the meta-template manager 904 incorporates consumer information into the sub-hints. The more information is provided to the generated AI module 906 (in the form of hints based on meta-template sub-hints), the more personalized and specific the generated content produced by the generated AI module will be. In this way, personalized content corresponding to the user's (designer) intent with respect to a particular consumer (based on profile information) is generated.

As shown in FIG. 9, the personalized content recommendation system 900 also includes a generated AI module 906. It should be appreciated that the plurality of generated AI modules 906 may be executed by the personalized content recommendation system 900. The generated AI module 906 generates content (e.g., an image) to be consumed by a consumer (or target user). The generated AI module 906 may be any generated AI module configured to generate content using hints. As described herein, a hint is a meta-template that includes a plurality of sub-hints and corresponding personalized variables.

As shown in fig. 9, the personalized recommendation system 900 also includes a PCF manager 908.PCF manager 908 groups the extracted one or more of: inputs, personalized variables extracted (e.g., from the template parser 902), refined personalized variables (e.g., received from refinements), sub-hints generated in meta-templates, refined sub-hints in meta-templates (e.g., received from refinements), control loops, rules, and outputs of the generation-type AI module 906. Such information is persisted in the PCF document. In some embodiments, additional information such as the generated AI module 906, one or more algorithms/techniques to accommodate any signal (e.g., text, language, product insertion, or artistic changes), etc., is persisted in the PCF. Additionally or alternatively, information, such as profile information associated with a particular consumer consuming content and/or user preferences, such as content creator preferences, is aggregated by PCF manager 908 and persisted in the PCF document. By storing such data, the PCF retains the original design or business intent.

As shown in FIG. 9, the personalized content recommendation system 900 further comprises a neural network manager 912. The neural network manager 912 can host a plurality of neural networks or other machine learning models, such as the generated AI module 906. The neural network manager 912 may include an execution environment, libraries, and/or any other data needed to execute the machine learning model. In some embodiments, the neural network manager 912 may be associated with dedicated software and/or hardware resources to execute a machine learning model. As discussed, the generated AI module 906 can be implemented as any type of generated AI. In various embodiments, each neural network hosted by the neural network manager 912 may be the same type of neural network or may be a different type of neural network, depending on the implementation. Although depicted in fig. 9 as being hosted by a single neural network manager 912, in various embodiments, the neural network may be hosted in multiple neural network managers and/or as part of different components. For example, the generated AI modules 906 may be hosted by their own neural network manager or other host environment in which the respective neural network executes, or the generated AI modules 906 may be distributed across multiple neural network managers depending on, for example, resource requirements of the generated AI modules 906, and the like.

As shown in FIG. 9, personalized content recommendation system 900 also includes training manager 916. The training manager 916 may teach, instruct, adjust, and/or train one or more neural networks. In particular, training manager 916 may train the neural network based on a plurality of training data. For example, the generative AI module 906 may be trained to perform a back diffusion process. More specifically, training manager 916 may access, identify, generate, create, and/or determine training inputs and utilize the training inputs to train and fine tune the neural network.

As shown in FIG. 9, the personalized content recommendation system 900 further comprises a storage manager 910. The storage manager 910 maintains data for the personalized content recommendation system 900. The storage manager 910 may maintain any type, size, or kind of data as needed to perform the functions of the personalized content recommendation system 900. As shown in fig. 9, the storage manager 910 includes a meta-template 918. The meta template 918 is a compiled template that includes sub-hints, which can be manually determined and/or automatically determined. As described herein, the sub-hints may include a free-text description and one or more personalized variables derived from the input. Additionally or alternatively, the sub-hint may be part of the input. The meta-templates associated with a user (such as a designer) may include user preferences, sub-hint preferences, corporate preferences (e.g., employer company that employed the user designer), control loops, constraints, and so forth. The control loops/constraints may refine the use of personalized variables in the meta-templates. The storage manager 910 also stores a record of historical inputs (such as user preferences 924). Training data 920 is also stored in storage manager 910. Training data 920 includes manually labeled data for supervised learning. Training using supervised learning is part of training performed during semi-supervised learning. Finally, storage manager 910 stores PCF document 922 generated using PCF manager 908. As described herein, PCF documents include one or more of the following: inputs, personalized variables extracted (e.g., from the template parser 902), refined personalized variables (e.g., received from refinements), sub-hints generated in meta-templates, refined sub-hints in meta-templates (e.g., received from refinements), control loops, rules, and content determined by the generative AI module 906.

Each of the components 902-916 of the personalized content recommendation system 900 and their corresponding elements (as shown in fig. 9) may communicate with each other using any suitable communication technology. It will be appreciated that although components 902 through 916 and their corresponding elements are shown as separate in fig. 9, any of components 902 through 916 and their corresponding elements may be combined into fewer components, such as into a single facility or module, separated into more components, or configured to serve different components of a particular embodiment.

The components 902 through 916 and their corresponding elements may include software, hardware, or both. For example, components 902 through 916 and their corresponding elements may include one or more instructions stored on a computer-readable storage medium and executable by a processor of one or more computing devices. The computer-executable instructions of the personalized content recommendation system 900, when executed by one or more processors, may cause a client device and/or a server device to perform the methods described herein. Alternatively, the components 902-916 and their corresponding elements may comprise hardware, such as a dedicated processing device for performing a certain function or group of functions. Additionally, components 902 through 916 and their corresponding elements may comprise a combination of computer-executable instructions and hardware.

Further, the components 902-916 of the personalized content recommendation system 900 may be implemented, for example, as one or more stand-alone applications, one or more modules of an application, one or more plug-ins, one or more library functions, or functions that may be invoked by other applications, and/or a cloud computing model. Accordingly, the components 902 through 916 of the personalized content recommendation system 900 may be implemented as stand-alone applications, such as desktop or mobile applications. Further, the components 902 through 916 of the personalized content recommendation system 900 may be implemented as one or more web-based applications hosted on a remote server. Alternatively or additionally, the components 902-916 of the personalized content recommendation system 900 may be implemented in a suite of mobile device applications (applications) or "apps".

As shown, personalized content recommendation system 900 may be implemented as a single system. In other embodiments, the personalized content recommendation system 900 may be implemented across multiple systems. For example, one or more functions of the personalized content recommendation system 900 may be performed by one or more servers, and one or more functions of the personalized content recommendation system 900 may be performed by one or more client devices.

For example, in one or more embodiments, after a client device accesses a web page or other web application hosted at one or more servers, the one or more servers may provide access to a user interface displayed at the client device, prompting a user for a description of consumable content. The client device may provide a description of the consumable content to one or more servers. Upon receiving the description, one or more servers may automatically perform the above-described methods and processes to identify semantically related terms based on the description of the consumable content and populate the templates with user preferences, consumer information, semantically related terms, controls, and the like. One or more servers can use the populated templates to provide access to a user interface displayed at the client device. The client device may be used to receive user input modifying one or more entries of the sub-hints of the template, thereby refining the template. Upon receiving the modification (or upon receiving a request to perform content generation), one or more servers automatically perform the above-described methods and processes to generate personalized content for the target audience based on the description of the consumable content.

Figures 1 through 9, corresponding text, and examples provide a variety of different systems and devices that allow a user to generate personalized content for a target audience in accordance with one or more embodiments. In addition to the foregoing, embodiments may be described in terms of flow charts including acts and steps in methods for achieving a particular result. For example, FIG. 10 illustrates a flow diagram of an exemplary method in accordance with one or more embodiments. The method described with respect to fig. 10 may be performed with fewer or more steps/acts, or the steps/acts may be performed in a different order. Additionally, the steps/acts described herein may be repeated or performed in parallel with each other or with different instances of the same or similar steps/acts.

FIG. 10 illustrates a flow diagram 1000 of a series of acts in a method of generating personalized content for a target audience in accordance with one or more embodiments. In one or more embodiments, the method 1000 is performed in a digital media environment that includes the personalized content recommendation system 900. Method 1000 is intended to illustrate one or more methods in accordance with the present disclosure and is not intended to limit potential embodiments. Alternative embodiments may include additional, fewer, or different steps than those illustrated in fig. 10.

As shown in FIG. 10, method 1000 includes an act 1002 of receiving a description of content to be generated using a generative model, wherein the received description of content is associated with a user profile. As described herein, the description of the content may include a title of the content, a purpose of the content, a user intent, or some other description of the content to be generated. As described herein, the description of the content may be attached with user identification information such that the user identification information associated with the description of the content may be mapped to a user profile. In some embodiments, different personalization variables are extracted from the description of the content depending on the determined user profile.

Method 1000 includes an act 1004 of determining semantic terms based on the description of the content. As described herein, the description of the consumable content is decomposed into personalized variables using any suitable semantic search/semantic relevance algorithm, where the personalized variables are semantically related blocks determined from the description of the consumable content. Semantic blocks are semantically related terms, etc.

Method 1000 includes an act 1006 of generating a user-specific template that includes semantic terms and user preferences associated with a user profile. As described herein, a template is a compilation of sub-hints that may include a free-text description and one or more personalized variables derived from the description of the consumable content. One or more personalization variables may be automatically populated into the template. As described herein, a meta-template associated with a user profile may be determined in response to an identifier associated with a user. An identifier (such as a user number, user name, email address, phone number, etc.) may be attached to the description of the content, associated with the description of the content in metadata, etc. The meta-templates may be changed, for example, by sub-hints associated with the user profile identified via the user profile. In addition, the meta-templates may vary depending on one or more control loops and/or constraints. Such control loops/constraints may refine the use of personalized variables in the meta-template.

Method 1000 includes an act 1008 of generating content using a generative model based on the user-specific template. As described herein, user-specific templates populated with personalized variables and corresponding sub-cues (e.g., meta-templates) are fed into one or more generated AI modules. The generation AI module generates content (e.g., an image) to be consumed by a consumer (or target user). The generated AI module may be any generated AI module configured to generate content using hints.

Method 1000 includes an act 1010 of outputting the content for display on the target user device. As described herein, personalized content may be transmitted to one or more computing devices.

Embodiments of the present disclosure may include or utilize a special purpose or general-purpose computer, including computer hardware, such as, for example, one or more processors and system memory, as discussed in more detail below. Embodiments within the scope of the present disclosure also include physical and other computer-readable media for carrying or storing computer-executable instructions and/or data structures. In particular, one or more processes described herein may be at least partially implemented as instructions embodied in a non-transitory computer-readable medium and executable by one or more computing devices (e.g., any of the media content access devices described herein). In general, a processor (e.g., a microprocessor) receives instructions from a non-transitory computer-readable medium (e.g., memory, etc.) and executes those instructions, thereby performing one or more processes, including one or more of the processes described herein.

Computer readable media can be any available media that can be accessed by a general purpose or special purpose computer system. The computer-readable medium storing computer-executable instructions is a non-transitory computer-readable storage medium (device). The computer-readable medium carrying computer-executable instructions is a transmission medium. Thus, by way of example, and not limitation, embodiments of the present disclosure may include at least two distinct types of computer-readable media: a non-transitory computer readable storage medium (device) and a transmission medium.

Non-transitory computer readable storage media (devices) include RAM, ROM, EEPROM, CD-ROM, solid state drive ("SSD") (e.g., based on RAM), flash memory, phase change memory ("PCM"), other types of memory, other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired program code means in the form of computer-executable instructions or data structures and that can be accessed by a general purpose or special purpose computer.

A "network" is defined as one or more data links capable of transferring electronic data between computer systems and/or modules and/or other electronic devices. When information is transferred or provided over a network or another communications connection (either hardwired, wireless, or a combination of hardwired or wireless) to a computer, the computer properly views the connection as a transmission medium. The transmission media can include networks and/or data links, which can be used to carry desired program code means in the form of computer-executable instructions or data structures, and which can be accessed by a general purpose or special purpose computer. Combinations of the above should also be included within the scope of computer-readable media.

Furthermore, upon reaching various computer system components, program code means in the form of computer-executable instructions or data structures can be transferred automatically from transmission media to non-transitory computer-readable storage media (devices) (or vice versa). For example, computer-executable instructions or data structures received over a network or data link may be buffered in RAM within a network interface module (e.g., a "NIC") and then ultimately transferred to computer system RAM and/or less volatile computer storage media (devices) at the computer system. Thus, it should be understood that a non-transitory computer readable storage medium (device) can be included in a computer system component that also (or even primarily) utilizes transmission media.

Computer-executable instructions comprise, for example, instructions and data which, when executed at a processor, cause a general purpose computer, special purpose computer, or special purpose processing device to perform a certain function or group of functions. In some embodiments, computer-executable instructions are executed on a general-purpose computer to transform the general-purpose computer into a special-purpose computer that implements the elements of the present disclosure. The computer-executable instructions may be, for example, binary, intermediate format instructions (such as assembly language), or even source code. Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the described features or acts described above. Rather, the described features and acts are disclosed as example forms of implementing the claims.

Those skilled in the art will appreciate that the disclosure may be practiced in network computing environments with many types of computer system configurations, including personal computers, desktop computers, laptop computers, message processors, hand-held devices, multiprocessor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, mobile telephones, PDAs, tablet computers, pagers, routers, switches, and the like. The disclosure may also be practiced in distributed system environments where local and remote computer systems, which are linked (either by hardwired data links, wireless data links, or by a combination of hardwired and wireless data links) through a network, both perform tasks. In a distributed system environment, program modules may be located in both local and remote memory storage devices.

Embodiments of the present disclosure may also be implemented in a cloud computing environment. In this description, "cloud computing" is defined as a model for enabling on-demand network access to a shared pool of configurable computing resources. For example, cloud computing may be employed in the marketplace to provide universal and convenient on-demand access to a shared pool of configurable computing resources. The shared pool of configurable computing resources may be provisioned quickly via virtualization and released with lower management workload or service provider interactions and then scaled accordingly.

Cloud computing models may be composed of various features such as, for example, on-demand self-service, wide network access, resource pools, rapid elasticity, measurement services, and so forth. The cloud computing model may also expose various service models, such as, for example, software as a service ("SaaS"), platform as a service ("PaaS"), and infrastructure as a service ("IaaS"). The cloud computing model may also be deployed using different deployment models, such as private cloud, community cloud, public cloud, hybrid cloud, and so forth. In this specification and in the claims, a "cloud computing environment" is an environment in which cloud computing is employed.

Fig. 11 illustrates, in block diagram form, an exemplary computing device 1100 that may be configured to perform one or more of the processes described above. It will be appreciated that one or more computing devices, such as computing device 1100, may implement a personalized content recommendation system. As shown in fig. 11, the computing device may include a processor 1102, a memory 1104, one or more communication interfaces 1106, a storage device 1108, and one or more I/O devices/interfaces 1110. In some embodiments, computing device 1100 may include fewer or more components than those shown in FIG. 11. The components of computing device 1100 shown in fig. 11 will now be described in more detail.

In a particular embodiment, the processor(s) 1102 include hardware for executing instructions (such as those making up a computer program). As an example and not by way of limitation, to execute instructions, processor(s) 1102 may retrieve (or instruction) instructions from internal registers, internal caches, memory 1104, or storage 1108 and decode and execute them. In various embodiments, the processor(s) 1102 may include one or more Central Processing Units (CPUs), graphics Processing Units (GPUs), field Programmable Gate Arrays (FPGAs), system-on-a-chip (socs), or other processor(s) or combination of processors.

Computing device 1100 includes memory 1104 coupled to processor(s) 1102. Memory 1104 may be used to store data, metadata, and programs for execution by the processor(s). Memory 1104 may include one or more of volatile and nonvolatile memory, such as random access memory ("RAM"), read only memory ("ROM"), solid state disk ("SSD"), flash memory, phase change memory ("PCM"), or other types of data storage. The memory 1104 may be internal memory or distributed memory.

Computing device 1100 may also include one or more communication interfaces 1106. Communication interface 1106 may include hardware, software, or both. The communication interface 1106 may provide one or more interfaces for communication (such as, for example, packet-based communication) between the computing device and one or more other computing devices 1100 or one or more networks. By way of example, and not by way of limitation, communication interface 1106 may include a Network Interface Controller (NIC) or network adapter for communicating with an ethernet or other wire-based network, or a Wireless NIC (WNIC) or wireless adapter for communicating with a wireless network, such as WI-FI. Computing device 1100 may also include a bus 1112. Bus 1112 may include hardware, software, or both that couple components of computing device 1100 to one another.

Computing device 1100 includes a storage device 1108, storage device 1108 including memory for storing data or instructions. By way of example, and not by way of limitation, storage device 1108 may include the non-transitory storage media described above. Storage 1108 may include a Hard Disk Drive (HDD), flash memory, a Universal Serial Bus (USB) drive, or a combination of these or other storage devices. Computing device 1100 also includes one or more input or output ("I/O") devices/interfaces 1110 that are provided to allow a user to provide input to computing device 1100, such as user strokes, to receive output from computing device 1100, and to otherwise communicate data to and from computing device 1100. These I/O devices/interfaces 1110 may include a mouse, a keypad or keyboard, a touch screen, a camera, an optical scanner, a network interface, a modem, other known I/O devices, or a combination of such I/O devices/interfaces 1110. The touch screen may be activated with a stylus or finger.

The I/O devices/interfaces 1110 may include one or more devices for presenting output to a user, including, but not limited to, a graphics engine, a display (e.g., a display screen), one or more output drivers (e.g., a display driver), one or more audio speakers, and one or more audio drivers. In some embodiments, the I/O devices/interfaces 1110 are configured to provide graphical data to a display for presentation to a user. The graphical data may represent one or more graphical user interfaces and/or as any other graphical content that may serve a particular implementation.

In the foregoing specification, embodiments have been described with reference to specific exemplary embodiments thereof. Various embodiments are described with reference to details discussed herein and the accompanying figures illustrate the various embodiments. The above description and drawings are illustrative of one or more embodiments and are not to be construed as limiting. Numerous specific details are described to provide a thorough understanding of the various embodiments.

Embodiments may include other specific forms without departing from the spirit or essential characteristics thereof. The described embodiments are to be considered in all respects only as illustrative and not restrictive. For example, the methods described herein may be performed with fewer or more steps/acts, or the steps/acts may be performed in a different order. Additionally, the steps/acts described herein may be repeated or performed in parallel with each other or in parallel with different instances of the same or similar steps/acts. The scope of the invention is, therefore, indicated by the appended claims rather than by the foregoing description. All changes which come within the meaning and range of equivalency of the claims are to be embraced within their scope.

In the various embodiments described above, unless specifically indicated otherwise, disjunctive language (such as the phrase "at least one of A, B or C") is intended to be understood to mean A, B or C or any combination thereof (e.g., A, B and/or C). Thus, the disjunctive language is not intended nor should it be construed to imply that at least one of the requirements a, at least one of the requirements B, or at least one of the requirements C, respectively, are each present in a given embodiment.

Claims

1. A method, comprising:

receiving a description of content to be generated using a generative model, wherein the received description of content is associated with a user profile;

determining semantic terms based on the description of content;

generating a user-specific template comprising the semantic terms and user preferences associated with the user profile;

Generating the content using the generative model based on the user-specific template; and

The content is output for display on the target user device.

2. The method of claim 1, further comprising:

Receiving a user refinement of the semantic terms that modifies the user-specific template, wherein the user refinement is at least one of: modification of the semantic terms, addition of one or more semantic terms, deletion of one or more semantic terms, modification of the description of content, modification of a control loop, or modification of a constraint; and

The content is generated using the generative model based on a modified user-specific template that includes the user refinement.

3. The method of claim 1, further comprising:

Obtaining target profile information, wherein the target profile information is information specific to a target user; and

Generating the content using the generative model is also based on the target profile information.

4. A method according to claim 3, wherein the target profile information is obtained by querying a database or user input.

5. The method of claim 1, wherein generating the content using the generative model is further based on profile information of a user similar to a target user.

6. The method of claim 1, further comprising:

a semantically related image is obtained from the generated content.

7. The method of claim 6, further comprising:

outputting the semantically related image for display on the target user device.

8. A system, comprising:

A memory component; and

A processing device coupled to the memory component, the processing device to perform operations comprising:

determining semantic terms based on the description of content;

generating content using a generative model based on the user-specific template; and

The content is output for display on the target user device.

9. The system of claim 8, wherein the processing device performs further operations comprising:

Receiving a user refinement of the semantic terms that modifies the user-specific template, wherein the user refinement is at least one of: modifying semantic terms, adding one or more semantic terms, deleting one or more semantic terms, modifying the description of content, modifying a control loop, or modifying a constraint; and

The content is generated using the generative model based on modified user-specific templates including the user refinements.

10. The system of claim 8, wherein the processing device performs further operations comprising:

11. The system of claim 10, wherein the target profile information is obtained by querying a database or user input.

12. The system of claim 8, wherein generating the content using the generative model is further based on profile information of a user similar to a target user.

13. The system of claim 8, wherein the processing device performs further operations comprising:

a semantically related image is obtained from the generated content.

14. The system of claim 13, wherein the processing device performs further operations comprising:

15. A non-transitory computer-readable medium storing executable instructions that, when executed by a processing device, cause the processing device to perform operations comprising:

During a first period of time, creating a file comprising templates, the templates being filled with semantic terms determined based on the description of the content;

Receiving the file during a second time period;

Generating content using a generative model based on the template of the file; and

Outputting the content.

16. The non-transitory computer-readable medium of claim 15, wherein the template of the file further comprises sub-hints associated with user preferences.

17. The non-transitory computer-readable medium of claim 15, wherein the template of the file further comprises sub-hints associated with preferences of a group of users similar to a user.

18. The non-transitory computer-readable medium of claim 15, wherein the template of the file further comprises information specific to a target user.

19. The non-transitory computer-readable medium of claim 15, storing instructions that further cause the processing device to perform operations comprising:

Receiving user refinements associated with the semantic terms during the second time period; and

Content is generated using the generative model based on the templates in the file and also based on the user refinement.

20. The non-transitory computer-readable medium of claim 19, storing instructions that further cause the processing device to perform operations comprising:

Updating the template to include the received user refinement during the second time period.