Abstract: An online system presents a sponsored content page to a user in conjunction with a model serving system. The online system accesses a content page for a food item and identifies one or more sponsorship opportunities at the content page. The online system identifies one or more candidate sponsors for each sponsorship opportunity. The online system selects a bidding sponsor for the sponsorship opportunity from the one or more candidate sponsors and a candidate item associated with the bidding sponsor as a sponsored item. The online system provides a content page, a description of the sponsored item, and a request to generate a sponsored content page for the sponsorship opportunity to a model serving system. The online system receives a sponsored content page generated by a machine-learning language model at the model serving system and presents the sponsored content page to a user.

Abstract: An online system obtains a target food from an order for a user and alcohol preferences from an order purchase history. The online system generates a prompt for a machine learning model to request alcohol candidates based on the target food category. The prompt includes the alcohol preferences, and requests for each alcohol candidate, a pairing score indicating how well the target food category pairs with the alcohol candidate and a user preference score indicating how well the alcohol candidate aligns with the alcohol preferences. The online system receives as output the candidate alcohol items. Each alcohol candidate has the pairing score, the user preference score, and a textual reason for scores. The online system matches at least one alcohol item from a catalog with each alcohol candidate. A subset of alcohol items is presented to the user as a carousel.

Abstract: Embodiments relate to utilizing a language model to automatically generate a novel recipe with refined content, which can be offered to a user of an online system. The online system generates a first prompt for input into a large language model (LLM), the first prompt including a plurality of task requests for generating initial content of a recipe. The online system requests the LLM to generate, based on the first prompt input into the LLM, the initial content of the recipe. The online system generates a second prompt for input into the LLM, the second prompt including the initial content of the recipe and contextual information about the recipe. The online system requests the LLM to generate, based on the second prompt input into the LLM, refined content of the recipe. The online system stores the recipe with the refined content in a database of the online system.

Abstract: An online system provides a support application including a chatbot application. One or more tools may each be configured to access external data. The interface system hosts an agent powered by an underlying large language model. The online system receives a user query via the chatbot application. For at least one or more iterations, the online system performs steps to provide a prompt to the LLM that specifies at least the user query, contextual information, a list of available tools, or a request to output an action. The system parses the response from the LLM to extract a selected action and action inputs for the selected action. The system triggers execution of a respective tool that corresponds to the selected action with the action inputs. The system generates a response to the user query and transmits the response to the client device.

Abstract: An online concierge system compensates pickers who fulfill orders including one or more items based in part on weights of the items included in an order. Because the online concierge system does not physically possess the items that are obtained, the online concierge system cannot directly weigh the items and weights specified for items in a catalog from a retailer may be inaccurate. To more accurately determine weights of items, the online concierge system trains a weight prediction model to estimate an item's weight from attributes of the item and uses the output of the weight prediction model to determine compensation to a picker. The weight prediction model may output a predicted weight of an item or a classification of the item as heavy or light. Where discrepancies are found between a predicted weight and the catalog weight of an item, additional information about the item is obtained.

Abstract: An online concierge system trains a fine-tuned generative image model for distinct categories of items based on a generative image model that takes a textual query as input and outputs and an associated image. Training of the fine-tuned generative image model is additionally based on a small set of representative images associated with the various categories, as well as textual tokens associated with the categories. Once trained, the fine-tuned generative image model can be used to generate realistic representative images for items in a database of the online concierge system that are lacking associated images. The fine-tuned model permits the generation of different variants of an item, such as different quantities or amounts, different packaging or packing density, and the like.

Abstract: An online system performs a task in conjunction with the model serving system or the interface system. The system generates a first prompt for input to a machine-learned language model, which specifies contextual information and a first request to generate a theme. The system provides the first prompt to a model serving system for execution by the machine-learned language model, receives a first response, and generates a second prompt. The second prompt specifies the theme and a second request to generate a third prompt for input to an image generation model that includes a third request to generate one or more images of one or more items associated with the theme. The system receives the third prompt by executing the model on the second prompt, provides the third prompt to the image generation model, and receives one or more images for presentation.

Abstract: An online system performs an inference task in conjunction with the model serving system or the interface system to generate customized recipes for users. The online system identifies a plurality of popular recipes based on historical user search data. The online system uses the collection of popular recipes to generate customized recipes for users based on user data and retailer data. The online system presents a customized recipe to the user, which may include items required to fulfill the recipe, a list of retailers at which the items are available for purchase, and instructions to combine the items. The online system collects user ratings and feedback on customized recipes to calculate a quality score. The online system may use the quality score to rank the customized recipes.

Abstract: A computer system uses clustering and a large language model (LLM) to normalize attribute tuples for items stored in a database of an online system. The online system collects attribute tuples, each attribute tuple comprising an attribute type and an attribute value for an item. The online system initially clusters the attribute tuples into a first plurality of clusters. The online system generates prompts for input into the LLM, each prompt including a subset of attribute tuples grouped into a respective cluster of the first plurality. Based on the prompts, the LLM generates a second plurality of clusters, each cluster including one or more attribute tuples that have a common attribute type and a common attribute value. The online system maps each attribute tuple to a respective normalized attribute tuple associated with each cluster. The online system rewrites each attribute tuple in the database to a corresponding normalized attribute tuple.

Abstract: The present disclosure relates to systems and methods for automatically associating additional content with sports games. The systems and methods obtain game schedule data for a plurality of sports games, and for each scheduled game, the systems and methods select articles published within a timeframe of the game date of the scheduled game. The systems and methods identify sports articles associated with the sports game based on entities and event phrases extracted from the sports articles matching the schedule game data. The systems and methods classify the sports articles into a plurality of clusters and use the clusters to associate the sports articles to the game schedule data and store the associated sports articles with the game schedule data.

Abstract: Examples of the present disclosure describe systems and methods that provide a pipeline to generate personalized queries that are associated with and based on a user's interests determined from a user's past searches, on an Internet search engine, and/or the content the user viewed from the past searches. The suggested queries can be shown in a user interface component associated with the user interface of the search engine and before the user enters anything, such as a new Internet search. This pre-population of searches associated with a user's interests gives an opportunity to the user to try these queries without manually entering in a search string.

Abstract: A method of selectively providing different types of search engine results to different searchers is provided. A browsing history for each of a plurality of unique identities is logged. A unique identity is associated with a rich segment experience responsive to the browsing history for the unique identity satisfying correlation criteria of the rich segment experience. The rich segment experience is configured to present curated segment-specific information with other search results on a search result web page. Responsive to receiving a search query from the unique identity previously associated with the rich segment experience, the rich segment experience is presented with other search results on the search result web page. Responsive to receiving the search query from a different unique identity not previously associated with the rich segment experience, other search results are presented without the rich segment experience on the search result web page.

Abstract: Aspects of the present disclosure are directed to providing a rich content experience based on information received from unstructured content. A plurality of information items may be obtained from a plurality of data source, where each information item includes unstructured content. The plurality of information items may be provided to a trained machine learning model, where the model is trained with training data that includes information items and corresponding labeled entities for a plurality of historical events. In examples, a formatted request may be received, where the formatted request is associated with one or more labeled entities associated with the trained machine learning model. The trained machine learning model may identify multiple entities from the unstructured content based on the formatted request associated with the one or more labeled entities. In examples, each identified entity of the multiple identified entities is stored as structured content responsive to the formatted request.

Abstract: The present concepts relate to an improved personalized search engine that can generate personalized rankings of search results in view of individual user's personal preferences and interests. Information about a segment of online content is collected. Certain activities by a user are tracked, including search queries submitted by the user, search results clicked on by the user, and/or web pages browsed by the user. From these activities, the user's preferences relating the segment are inferred using the collected segment information. When the user conducts a search directed to the segment, certain search results that the user is more likely to be interested in, based on the user's preferences, are ranked higher to generate the personalized rankings.

Abstract: The present disclosure relates to systems and methods for automatically associating additional content with sports games. The systems and methods obtain game schedule data for a plurality of sports games, and for each scheduled game, the systems and methods select articles published within a timeframe of the game date of the scheduled game. The systems and methods identify sports articles associated with the sports game based on entities and event phrases extracted from the sports articles matching the schedule game data. The systems and methods classify the sports articles into a plurality of clusters and use the clusters to associate the sports articles to the game schedule data and store the associated sports articles with the game schedule data.

Abstract: The present disclosure relates to systems, devices, and methods for identifying structured data for any webpage when a user requests the webpage, or the webpage loads in a browser. The systems, devices, and methods extract a portion of the webpage content and use the webpage content to determine the domain of the webpage, extract entities from the webpage content, query one or more datastores with content for structured data based on the domain and the extracted entities and present the structured data with the webpage.

Abstract: The present disclosure relates to systems, devices, and methods for identifying additional content for an article. The systems, devices, and methods may identify a domain for the articles and content and may use machine learning models to classify the articles and the content into categories using smart tags for the domain. The systems, devices, and methods may convert the articles and the content into document vectors using a pre-trained domain specific language model and generate a relevance score for the articles and the content using the document vectors. The systems, devices, and methods may generate a list of predicted matches that includes content that is similar to the article based on the relevance score. The systems, devices, and methods may filter the list of predicted matches based on a temporal proximity to generate a list of additional content for the article.

Abstract: A method of selectively providing different types of search engine results to different searchers is provided. A browsing history for each of a plurality of unique identities is logged. A unique identity is associated with a rich segment experience responsive to the browsing history for the unique identity satisfying correlation criteria of the rich segment experience. The rich segment experience is configured to present curated segment-specific information with other search results on a search result web page. Responsive to receiving a search query from the unique identity previously associated with the rich segment experience, the rich segment experience is presented with other search results on the search result web page. Responsive to receiving the search query from a different unique identity not previously associated with the rich segment experience, other search results are presented without the rich segment experience on the search result web page.

Abstract: A method of selectively providing different types of search engine results to different searchers is provided. A browsing history for each of a plurality of unique identities is logged. A unique identity is associated with a rich segment experience responsive to the browsing history for the unique identity satisfying correlation criteria of the rich segment experience. The rich segment experience is configured to present curated segment-specific information with other search results on a search result web page. Responsive to receiving a search query from the unique identity previously associated with the rich segment experience, the rich segment experience is presented with other search results on the search result web page. Responsive to receiving the search query from a different unique identity not previously associated with the rich segment experience, other search results are presented without the rich segment experience on the search result web page.

Abstract: Aspects of the present disclosure are directed to providing a rich content experience based on information received from unstructured content. A plurality of information items may be obtained from a plurality of data source, where each information item includes unstructured content. The plurality of information items may be provided to a trained machine learning model, where the model is trained with training data that includes information items and corresponding labeled entities for a plurality of historical events. In examples, a formatted request may be received, where the formatted request is associated with one or more labeled entities associated with the trained machine learning model. The trained machine learning model may identify multiple entities from the unstructured content based on the formatted request associated with the one or more labeled entities. In examples, each identified entity of the multiple identified entities is stored as structured content responsive to the formatted request.