Abstract: Systems and methods for generating embeddings for nodes of a corpus graph are presented. More particularly, operations for generation of an aggregated embedding vector for a target node is efficiently divided among operations on a central processing unit and operations on a graphic processing unit. With regard to a target node within a corpus graph, processing by one or more central processing units (CPUs) is conducted to identify the target node's relevant neighborhood (of nodes) within the corpus graph. This information is prepared and passed to one or more graphic processing units (GPUs) that determines the aggregated embedding vector for the target node according to data of the relevant neighborhood of the target node.

Abstract: Described are systems and methods to determine body types of individuals represented in content items. The determined body types may be associated with the content items to facilitate indexing, filtering, diversifying, etc. of the content items based on the determined body types. In exemplary implementations, a corpus of content items may be associated with an embedding vector that includes a representation of the content item. The embedding vectors associated with each content item can be provided as inputs to a trained machine learning model, which can process the embedding vectors to determine one or more body types of individuals represented in each content item while eliminating the need for performing image pre-processing prior to determination of the body type(s) presented in the content item.

Abstract: System and methods are presented for associating a user-posted content item with an interest node of an interest taxonomy. A corpus of content items and an interest taxonomy are maintained. The interest taxonomy comprises interest nodes organized in a hierarchical organization, each node having a text label descriptive of the interest node. Additionally, the content items of the corpus are associated with one or more interest nodes of the interest taxonomy. Upon receiving a user-posted content item, feature sets of the received content item are generated, these feature sets based on features and/or aspects of the received content item. After generating at least one feature set, the at least one feature set is provided to an interest prediction model that generates candidate interest nodes for the user-posted content item. At least some of the candidate interest nodes are associated with the user-posted content item in the corpus.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for analyzing content. One of the methods includes obtaining a collection of content items, each content item including an image and associated metadata; generating a collection of attributes for each content item by analyzing the image and the associated metadata; assigning one or more attributes to each content item; generating a plurality of queries from the collection of attributes; and populating a plurality of collections, each collection having a topic corresponding to a query, wherein populating the collections comprises evaluating a relevance of each query-content item pair based on an overlap between attributes associated with the query and attributes assigned to the content item.

Abstract: Described are systems and methods for providing a content generation service that may be configured to generate content, such as digital images and the like, based on both a textual input and supplemental information. The content generation service may include one or more trained machine learning models that may be configured to generate an image based on a textual input and one or more supplemental information input(s). The supplemental information may include, for example, user information, content information, etc. and can provide context, preferences, or any other additional information beyond the textual input that may be used to generate the content. Further, the content generation service and/or the user may be able to assign weights to each of the textual input and the supplemental information input in connection with the generation of the content.

Abstract: According to aspects of the disclosed subject matter, a taste graph comprising likely content collection nodes with corresponding likely digital content items is generated through one or more analyses of a corpus of content collections that is maintained by the online content service. As should be understood, this corpus of content collections is comprised of a plurality of curated content collections, with each content collection comprising a plurality of digital content items. With this taste graph available, as a user generates (or in response to a user generating) a content collection of digital content items, reference can be made to the taste graph to identify one or more digital content items that may be added to the content collection, where the one or more digital content items have a probabilistic likelihood of being complimentary and/or compatible with the other digital content items of the content collection.

Abstract: Described are systems and methods for determining and/or generating a queue of content items to improve the playback experience of the content items for a user. The content items may be obtained by a client device from an online service in response to a query, a request for content items, etc. Relevance rankings and/or scores associated with the content items may be replaced and/or augmented with a playability score, which can represent a quality of the playback experience associated with the content item. The playability score may be aggregated with the relevance and/or user engagement score to determine an overall playback score for each content item. The content items may be ranked, ordered, arranged, and/or presented in accordance with the overall playback scores associated with the content items to facilitate an improved playback experience for a user associated with the client device.

Abstract: A computer system extracts product data from a website and correlates product records from multiple sources to one another as corresponding to the same product. A website is crawled efficiently by rendering webpages using a virtual browser that ignores blacklisted elements, extracts data from objects without rendering, and suppressing retrieval of remote resources. Data is extracted according to engine control statements including a selector and extractor. A website may be crawled repeatedly and changes in extracted data may be detected and flagged. Engine control statements may be automatically changed in response to detecting a change in the configuration of the website. Images of product records may be correlated with one another by first comparing text of the product records and selecting images for comparison based on composition. Images are compared using a machine learning model. Images determined to be similar may be presented to a human for a correlation decision.

Abstract: Described is a system and method for enabling dynamic selection of a search input. For example, rather than having a static search input box, the search input may be dynamically positioned such that it encompasses a portion of displayed information. An image segment that includes a representation of the encompassed portion of the displayed information is generated and processed to determine an object represented in the portion of the displayed information. Additional images with visually similar representations of objects are then determined and presented to the user.

Abstract: Disclosed are systems and methods to enhance user engagement and personalization on digital platforms by employing Generative Artificial Intelligence (GenAI) techniques to suggest new nodes for a taxonomy. The process involves a GenAI technique to extract hypernyms from search query terms and a Retrieval-Augmented Generation (RAG) technique to expand a seed concept (i.e., node in a taxonomy), retrieve hyponyms (i.e., sub-categories) and come up with new node suggestions. This structured approach enables the automated extension of a taxonomy that enhances access to items maintained by an online service. The taxonomy nodes can be mapped to items, improving search accuracy and user interaction.

Abstract: Described are systems and methods for providing a multi-tasked trained machine learning model that may be configured to generate product embeddings from multiple types of product information. The exemplary product embeddings may be generated for a corpus of products (e.g., products included in a product catalog, etc.) based on both image information and text information associated with each respective product. Accordingly, the generated product embeddings may be compatible with learned representations of the different types of product information (e.g., image information, text information, etc.) and may be used to create a product index, which can be used to determine and serve product recommendations in connection with multiple different recommendation services that may be configured to receive different types of inputs (e.g., a single image, multiple images, text-based information, etc.).

Abstract: Disclosed are systems and methods for unified content item ranking across multiple output channels that provide personalized content item recommendations while adapting to different output channel contexts. A machine learning model receives four sets of data for each candidate content item: user engagement history data, content item popularity data, content item information data, and contextual relevance data. The contextual relevance data includes both a query context associated with an output channel and a content relevance based on existing content. The model generates engagement scores for multiple types of engagement actions, which are then weighted according to output channel-specific parameters to produce final ranking scores.

Abstract: Disclosed are systems and methods that connect an input image, such as an individual product image, to lifestyle images that include multiple products, and then to one or more complementary product images. For example, when a user provides or selects a product image, the disclosed implementations determine one or more lifestyle images of multiple products that include the product or include another, visually similar, product. Still further, the disclosed implementations may also determine, for each of the multiple products in the lifestyle image, a set of complementary product images that may work well together and with the product in the input image.

Abstract: System and methods are presented for associating a user-posted content item with an interest node of an interest taxonomy. A corpus of content items and an interest taxonomy are maintained. The interest taxonomy comprises interest nodes organized in a hierarchical organization, each node having a text label descriptive of the interest node. Additionally, the content items of the corpus are associated with one or more interest nodes of the interest taxonomy. Upon receiving a user-posted content item, feature sets of the received content item are generated, these feature sets based on features and/or aspects of the received content item. After generating at least one feature set, the at least one feature set is provided to an interest prediction model that generates candidate interest nodes for the user-posted content item. At least some of the candidate interest nodes are associated with the user-posted content item in the corpus.

Abstract: Described are systems and methods for determining and/or generating a queue of content items to improve the playback experience of the content items for a user. The content items may be obtained by a client device from an online service in response to a query, a request for content items, etc. Relevance rankings and/or scores associated with the content items may be replaced and/or augmented with a playability score, which can represent a quality of the playback experience associated with the content item. The playability score may be aggregated with the relevance and/or user engagement score to determine an overall playback score for each content item. The content items may be ranked, ordered, arranged, and/or presented in accordance with the overall playback scores associated with the content items to facilitate an improved playback experience for a user associated with the client device.

Abstract: Systems and methods for determining one or more topics that may be associated and/or grounded with a node of a taxonomy to facilitate the creation and/or modification of the taxonomy. The one or more topics can be determined by generating two layers of associations and/or groundings. In a first layer, tokens can be associated and/or grounded in a corpus of queries, and in a second layer, topics can be associated and/or grounded in the tokens. The topics can then be associated with and/or grounded in nodes of a taxonomy which can facilitate access to content items stored and maintained by an online service. Further, in exemplary implementations where the content items include associations and/or mappings to the corpus of queries, the nodes of the taxonomy (which are associated with one or more topics) can be transitively mapped to the content items.

Abstract: This disclosure describes systems and methods for matching user provided images that include representations of items with sellers of those items. A management service, as described herein, may provide a web site where users can post images, view images, share images, correspond with other users, etc. The management service may identify items represented in the images and determine one or more sellers that offer those items for sale. When another user requests to view the user provided image, the image, seller information identifying the seller determined to sell the item represented in the image, and/or a purchase control that may be selected by a user to initiate a purchase with the seller is presented.

Abstract: This disclosure describes, in part, systems and methods that enable users to manage, search for, share and discover objects based on a context of the object from the user's perspective. The same object may have vastly different meanings (context) to different individuals based on how they experience the object. Rather than managing objects solely based on information about the object, the implementations described allow users to specify a context for the object and manage objects based on that context. In addition, external sources may provide supplemental information about objects and/or representations of objects.