Abstract: Examples relate to a system including a processor that can perform certain operations. The operations can include obtaining input text. The operations also can include generating a set of vectors from an ensemble of machine-learning models based on the input text. The ensemble of machine-learning models can include a pre-trained language model configured to determine capitalization for mixed cases and acronyms, a pre-trained named entity recognition (NER) model configured to determine capitalization for general proper nouns, and a question-answer NER (QA-NER) model configured to determine capitalization for brand names. The QA-NER model can include a transformer language model and a linear layer. The operations additionally can include generating corrected sentence-case text by modifying capitalization of the input text based on the set of vectors and outputting the corrected sentence-case text on a draft advertisement user interface. Other embodiments are described.

Abstract: Systems and methods for retrieving and providing images are disclosed. An example system receives, from a user device, a request for an image. The system determines, using a machine-learning model, search embeddings based on the request; filters image data based on the request to identify a filtered set of the image data; and obtains a subset of the image embeddings corresponding to the filtered set of the image data. The system further determines based on a comparison of the search embeddings and the subset of the image embeddings, recommended image data, and causes presentation of the recommended image data at the user device. In response to selection of the recommended image data, the system provides the recommended image data to the user device.

Abstract: Systems and methods of item-specific keyword recommendation are disclosed. An item data structure including an item title is received and at least one item embedding is generated by applying a first trained semantic mapping model to the item title. The first trained semantic mapping model includes a first semantic mapping framework. The at least one item embedding is compared to a set of keyword embeddings representative of a set of platform-relevant keywords and a set of item-specific recommended keywords is selected from the set of platform-relevant keywords based on a similarity between the at least one item embedding and each embedding in the set of keyword embeddings. The item title is modified to include at least one of the set of item-specific recommended keywords and an interface including the modified item title is generated.

Abstract: A method including obtaining one or more predicted shelves corresponding to the keyword query. The method additionally can include generating linked categorical facets corresponding to the one or more predicted shelves based on shelf-categorical facet linkages. The method further can include generating, using fuzzy matching, candidate shelf-specific facets based on shelf-specific facet representation mappings and the linked categorical facets. The method additionally can include determining one or more shelf-specific facets from the candidate shelf-specific facets based on facet information in the candidate shelf-specific facets. The one or more shelf-specific facets can correspond to one or more shelves of the one or more predicted shelves. The method further can include outputting the one or more shelves and one or more respective shelf-specific facets of the one or more shelf-specific facets that correspond to each of the one or more shelves. Other embodiments are described.

Abstract: Systems and methods for generating a machine learning model to support multiple tasks for semantic retrieval, embedding and classification are disclosed. In some embodiments, a disclosed method includes: obtaining a training data set generated based on search related data and item related data associated with a website; training, based on the training data set, a machine learning model using a two-tower structure to generate an optimized set of model parameters, wherein the optimized set of model parameters minimizes a total loss function computed based on at least one of: a query classification loss, an item classification loss, and a similarity loss; and transmitting the trained machine learning model to be utilized for at least one of the following tasks: semantic item retrieval, query classification, item classification, query clustering, and item clustering.

Abstract: A method can be implemented via execution of computing instructions configured to run at a processor. The method can include: receiving, from a user, an image of a product; receiving, from the user, user-submitted logo alt text describing a brand of the product in the image; receiving, from the user, user-submitted image alt text describing the image; extracting brand information from the user-submitted logo alt text; extracting product information from the user-submitted image alt text; generating an instruction prompt that includes the brand information, as extracted, and the product information, as extracted; and generating a recommended image alt text describing the image and including the brand information, as extracted, and the product information, as extracted, by querying a multimodal generative artificial intelligence (multimodal GenAI) model with the instruction prompt. Other embodiments are disclosed.

Abstract: This disclosure relates to computer vision and generative artificial intelligence (AI) techniques for generating a target image set based on content included in a source image. The target image set comprises a plurality of target images, each of which is compliant with or more target display specifications for an electronic platform. The target images can be generated by an image adaptation network that comprises various AI models, including one or more saliency model, one or more generative models, one or more scene detection models, and/or one or more segmentation models.

Abstract: Systems and methods of generating keyword-specific content are disclosed. A request for including a keyword is received and the keyword is classified as one of catalog related or unrelated. When the keyword is catalog related, the keyword is categorized in a category associated with the catalog and at least one term in the keyword is categorized in a facet category associated with the catalog. A content template is obtained. The content template is a category specific template when the keyword is catalog related and a generic template when the keyword is catalog unrelated. The category specific template is populated with the at least one term at a position associated with the one of the plurality of facet categories. Responsive content including the category specific template populated with the at least one term when the keyword is catalog related and the generic template when the keyword is catalog unrelated, is transmitted.

Abstract: Systems and methods of item-specific keyword recommendation are disclosed. An item data structure including an item title is received and at least one item embedding is generated by applying a first trained semantic mapping model to the item title. The first trained semantic mapping model includes a first semantic mapping framework. The at least one item embedding is compared to a set of keyword embeddings representative of a set of platform-relevant keywords and a set of item-specific recommended keywords is selected from the set of platform-relevant keywords based on a similarity between the at least one item embedding and each embedding in the set of keyword embeddings. The item title is modified to include at least one of the set of item-specific recommended keywords and an interface including the modified item title is generated.

Abstract: Systems and methods for generating a machine learning model to support multiple tasks for semantic retrieval, embedding and classification are disclosed. In some embodiments, a disclosed method includes: obtaining a training data set generated based on search related data and item related data associated with a website; training, based on the training data set, a machine learning model using a two-tower structure to generate an optimized set of model parameters, wherein the optimized set of model parameters minimizes a total loss function computed based on at least one of: a query classification loss, an item classification loss, and a similarity loss; and transmitting the trained machine learning model to be utilized for at least one of the following tasks: semantic item retrieval, query classification, item classification, query clustering, and item clustering.

Abstract: Systems and methods of generating keyword-specific content are disclosed. A request for including a keyword is received and the keyword is classified as one of catalog related or unrelated. When the keyword is catalog related, the keyword is categorized in a category associated with the catalog and at least one term in the keyword is categorized in a facet category associated with the catalog. A content template is obtained. The content template is a category specific template when the keyword is catalog related and a generic template when the keyword is catalog unrelated. The category specific template is populated with the at least one term at a position associated with the one of the plurality of facet categories. Responsive content including the category specific template populated with the at least one term when the keyword is catalog related and the generic template when the keyword is catalog unrelated, is transmitted.

Abstract: Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform determining a respective interaction metric between at least two respective keywords of a plurality of keywords; constructing, using each respective interaction metric, as determined, a graph comprising a plurality of nodes connected by at least one edge; identifying one or more clusters of nodes using the graph; optimizing each cluster of nodes of the one or more clusters of nodes; and facilitating altering a graphical user interface (GUI) of an electronic device based upon a cluster of nodes of the one or more clusters of nodes, as identified and optimized. Other embodiments are disclosed herein.

Abstract: A method including generating, using a semantic embedding generation machine learning model, one or more respective shelf embedding vector representations for each of one or more browse shelves based on a respective shelf name for the each of the one or more browse shelves. The method also can include obtaining a keyword. The method additionally can include generating, using the semantic embedding generation machine learning model, a keyword embedding vector representation based on the keyword. The method further can include determining a respective similarity score between the keyword embedding vector representation and each of the one or more respective shelf embedding vector representations for each of the one or more browse shelves. The method additionally can include determining whether any of the respective similarity scores for the one or more respective shelf embedding vector representations across the one or more browse shelves exceed a predetermined threshold value.

Abstract: A method including obtaining one or more predicted shelves corresponding to the keyword query. The method additionally can include generating linked categorical facets corresponding to the one or more predicted shelves based on shelf-categorical facet linkages. The method further can include generating, using fuzzy matching, candidate shelf-specific facets based on shelf-specific facet representation mappings and the linked categorical facets. The method additionally can include determining one or more shelf-specific facets from the candidate shelf-specific facets based on facet information in the candidate shelf-specific facets. The one or more shelf-specific facets can correspond to one or more shelves of the one or more predicted shelves. The method further can include outputting the one or more shelves and one or more respective shelf-specific facets of the one or more shelf-specific facets that correspond to each of the one or more shelves. Other embodiments are described.

Abstract: Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform: receiving a set of keywords from a graphical user interface of an electronic device of a user; pre-processing at least one keyword of the set of keywords; receiving a hierarchical categorization; pre-processing at least one category of the hierarchical categorization; determining a respective similarity between each of the at least one keyword of the set of keywords and each of the at least one category of the hierarchical categorization; determining a respective confidence level of a most likely category in the hierarchical categorization for each of the at least one keyword of the set of keywords using the respective similarity between each of the at least one keyword of the set of keywords and each of the at least one category of the hierarchical categorization; ranking each of the at least one keyword of the set of k

Abstract: Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform: receiving a set of keywords from a graphical user interface of an electronic device of a user; pre-processing at least one keyword of the set of keywords; receiving a hierarchical categorization; pre-processing at least one category of the hierarchical categorization; determining a respective similarity between each of the at least one keyword of the set of keywords and each of the at least one category of the hierarchical categorization; determining a respective confidence level of a most likely category in the hierarchical categorization for each of the at least one keyword of the set of keywords using the respective similarity between each of the at least one keyword of the set of keywords and each of the at least one category of the hierarchical categorization; ranking each of the at least one keyword of the set of k

Abstract: Systems and methods including one or more processors and one or more non-transitory storage devices storing computing instructions configured to run on the one or more processors and perform determining a respective interaction metric between at least two respective keywords of a plurality of keywords; constructing, using each respective interaction metric, as determined, a graph comprising a plurality of nodes connected by at least one edge; identifying one or more clusters of nodes using the graph; optimizing each cluster of nodes of the one or more clusters of nodes; and facilitating altering a graphical user interface (GUI) of an electronic device based upon a cluster of nodes of the one or more clusters of nodes, as identified and optimized. Other embodiments are disclosed herein.