Abstract: A server receives a plurality of product data entries from a plurality of retailer computing systems. Each product data entry includes a product identifier uniquely identifying a corresponding physical product and descriptive data of the corresponding physical product. A subset of the plurality of product data entries having a same product identifier is determined. An embedding vector representative of a product data entry in the subset is pairwise compared with each of respective embedding vectors representative of other product data entries in the subset other than the product data entry to compute respective vector similarity metrics. A pooled semantic similarity metric for the product data entry based on the computed respective vector similarity metrics. It is determined whether the product data entry is an outlier in the subset based on the pooled semantic similarity metric. A notification is transmitted to a client device of a user based on the determination.

Abstract: An online system generates a set of genomic representations, each including multiple genes, in which each gene represents users assigned to a control or test group for performing a test. A metric is identified based on a treatment associated with the test group and a score for each representation is computed based on a difference between two values, in which each value is based on the metric associated with users assigned to the test or control group. A propagation process is executed by identifying representations having at least a threshold score, propagating genes included in the representations to an additional set of representations through recombination and/or mutation, and computing the score for each additional representation. The propagation process is repeated for each additional set of representations until stopping criteria are met and a representation is selected based on scores associated with one or more representations.

Abstract: An online concierge system uses a cumulative incrementality score to evaluate the performance of incrementality models used by the online concierge system to identify users for treatment. The online concierge system applies an incrementality model to a set of examples to generate predicted incrementality scores for the examples. The online concierge system ranks the examples based on the predicted incrementality scores for the examples and groups the examples based on their rankings. The online concierge system iteratively computes cumulative incrementality scores for each grouping based on the examples of each grouping, and computes a final cumulative incrementality score for the incrementality model based on each of the cumulative incrementality scores.

Abstract: An online concierge system trains a classification model as a domain adversarial neural network from training data labeled with source classes from a source domain that do not overlap with target classes from a target domain output by the classification model. The online concierge system maps one or more source classes to a target class. The classification model extracts features from an image, classifies whether an image is from the source domain or the target domain, and predicts a target class for an image from the extracted features. The classification model includes a gradient reversal layer between feature extraction layers and the domain classifier that is used during training, so the feature extraction layers extract domain invariant features from an image.

Abstract: An automated checkout system modifies received images of machine-readable labels to improve the performance of a label detection model that the system uses to decode item identifiers encoded in the machine-readable labels. For example, the automated checkout system may transform subregions of an image of a machine-readable label to adjust for distortions in the image's depiction of the machine-readable label. Similarly, the automated checkout system may identify readable regions within received images of machine-readable labels and apply a label detection model to those readable regions. By modifying received images of machine-readable labels, these techniques improve on existing computer-vision technologies by allowing for the effective decoding of machine-readable labels based on real-world images using relatively clean training data.

Abstract: A variation testing system environment for performing variation testing of web pages and applications is disclosed. The variation testing system applies a weighted consistent hash function to user attributes of users to assign the users to a variant of a web page that is undergoing experimentation. The usage of the weighted consistent hash function allows for a stable experimental population.

Abstract: An online system may generate numerous search records in response to searches requested by users. The online system may use a specific way to sample the historical search records to reduce biases in sampling. For example, the online system retrieves historical query records associated with an item query engine. The set of historical query records includes a plurality of search phrases. A historical query record is associated with a search phrase and a list of items returned by the item query engine. The online system determines the search frequencies for the search phrases. The online system stratifies the historical query records into a plurality of bins according to the search frequencies of the search phrases. The online system samples the historical query records from the plurality of bins to collect a representative set of historical query records and outputs the representative set of historical query records for rating.

Abstract: Systems and methods for scoring promotions are provided. A set of training offers are received, which include combinations of variable values. These combinations of variable values are converted into a vector value. The offers are paired and the vectors subtracted from one another, resulting in a pair vector. Metrics for the success of offers is collected, and are subtracted from one another for the paired offers to generate a raw score. This raw score is then normalized using the pair vector. The normalized scores are utilized to generate a model for the impact any variable value has on offer success, which may then be applied, using linear regression, to new offers to generate an expected level of success. The new scored offers are ranked and the top-ranked offers are selected for inclusion in a promotional campaign.

Abstract: An online concierge system generates recipe embeddings for recipes including multiple items and user embeddings for users, with the recipe embeddings and user embeddings in a common latent space. To generate the user embeddings and the recipe embeddings, a model includes separate layers for a user model outputting user embeddings and for a recipe model outputting recipe embeddings. When training the model, a weight matrix generates a predicted dietary preference type for a user embedding and for a recipe embedding and adjusts the user model or the recipe model based on differences between the predicted dietary preference type and a dietary preference type applied to the user embedding and to the recipe embedding. Additionally cross-modal layers generate a predicted user embedding from a recipe embedding and generate a predicted recipe embedding from a user embedding that are used to further refine the user model and the recipe model.

Abstract: Disclosed herein relates to a self-checkout anti-theft vehicle system, comprising: a self-checkout vehicle having a plurality of sensors and components implemented thereon, the self-checkout vehicle being used by shoppers for storing selected merchandises in a retail environment; and a centralized computing device. The centralized computing device is configured to: obtain information related to each merchandise selected and placed into the self-checkout vehicle by a shopper by exchanging data with the plurality of sensors and components via a first communication network, identify each merchandise via a second, different communication network based at least upon the information obtained from the plurality of sensors and components, and process payment information of each merchandise.

Abstract: An online concierge system generates an item graph connecting item nodes with attribute nodes of the items. When the online concierge system receives a search query to identify one or more items from a customer, the online concierge system parses the search query into combinations of terms and identifies item nodes and attribute nodes related to the search query. The online concierge system may determine that no item nodes meet presentation criteria. The online concierge system may determine that a reformulated search query has a higher conversion probability than the search query received from the customer. The online concierge system reformulates the search query. The online concierge system selects item nodes as search results. The online concierge system transmits the search results to the customer.

Abstract: Disclosed are visual recognition and sensor fusion weight detection system and method. An example method includes: tracking, by a sensor system, objects and motions within a selected area of a store; activating, by the sensor system, a first computing device positioned in the selected area in response to detecting a presence of a customer within the selected area; identifying, by the sensor system, the customer and at least one item carried by the customer; transmitting, by the sensor system, identifying information of the customer and the at least one item to a computing server system via a communication network; measuring, by the first computing device, a weight of the at least one item; transmitting, by the first computing device, the weight to the computing server system via the communication network; and generating, by the computing server system, via the communication network, transaction information of the at least one item.

Abstract: An online recommendation system can choose recipes to recommend to a customer based on a set of ingredients the customer is inferred to have on hand (a customer pantry model). For example, the recommendation system can look at recent or historical purchases made by the customer and determine what items the customer still has available based on an assumed shelf life for the purchased items. Using the customer pantry model, the recommendation system selects recipes based on overlapping ingredients between recipe's ingredient lists and ingredients available to the customer (including the customer pantry model and their current shopping cart). In some implementations, the recommendation system first selects a set of candidate recipes based on the overlap, then selects the final set of recipes to recommend based on a score optimization (for example, performed using a machine learning model).

Abstract: A valid combination of products is constructed from a set of products identified by a retailing backend system. The valid combination satisfies a condition of an offer. A product in the combination of product is a physical product located in a physical store. A social sentiment towards the product in social media data of a shopper is evaluated, and a rating of the product is computed using the evaluation. A location of the product in the store is obtained from the retailing backend system. The location of the product is overlaid on a view of the store, where the view is from a perspective of a location of the shopper, the shopper being physically situated in the store.

Abstract: An online concierge identifies orders to shoppers, allowing shoppers to select orders for fulfillment. The online concierge system may generate batches that include multiple orders, allowing a shopper to select a batch to fulfill multiple orders. As orders are continuously being received, delaying identification of orders to shoppers may allow greater batching of orders. To allow greater opportunities for batching, the online concierge system estimates a benefit for delaying identification of an order by different time intervals and predicts an amount of time to fulfill the order. The online concierge system then delays assigning orders for which there is a threshold benefit for delaying and selects a time interval for delaying identification of the order that does not result in greater than a threshold likelihood of a late fulfillment of the order.

Abstract: A method for targeted advertisement includes transmitting a pre-filter to the user device, responsive to contextual information from a user device, to determine, using a processor, one or more inferences based on physical browsing information, collected at the user device, in compliance with one or more privacy policies of the user. The method also includes receiving one or more inferences determined by the pre-filter from the user device and transmitting one or more targeted advertisements to the user device based on one or more inferences.

Abstract: A warehouse from which shoppers fulfill orders for an online concierge system maintains an online concierge system-specific portion for which the online concierge system specifies placement of items in regions. To place items in the online concierge system-specific portion, the online concierge system accounts for co-occurrences of different items in orders and measures of similarity between different items. From the co-occurrences of items, the online concierge system generates an affinity graph. The online concierge system also generates a colocation graph based on distances between different regions in the online concierge system-specific portion. Using an optimization function with the affinity graph and the colocation graph, the online concierge system selects regions within the online concierge system-specific portion for different items to minimize an amount of time for shoppers to obtain items in the online concierge-system specific portion.

Abstract: An online concierge system generates a graph connecting items with attributes of the items and other items. Hence, the graph includes nodes corresponding to attributes and nodes corresponding to items, with an item connected to attributes of the item in the graph. Example attributes include a brand, a category, a department, or any other suitable information about the item. When the online concierge system receives a search query to identify one or more items from a customer, the online concierge system parses the search query into combinations of terms and compares different combinations of terms to the graph to determine connections between different combinations of terms in the graph. Based on measures of connectedness between combinations of terms and connections in the graph, items are identified from one or more combinations of terms. Information about the identified items is presented to the customer.

Abstract: When an online system receives an order from a customer, the online system can fulfill the order (for example, using a picker to acquire and deliver the ordered items) and use audio verification to verify delivery. When using audio verification, a customer or picker's mobile device plays verification audio including a verification code specific to the order. If the mobile devices (and, by proxy, the customer and picker) are nearby (for example, within 10 feet), the other party's mobile device can detect the verification audio through a microphone and decode the verification code from the captured audio. In some implementations, audio verification can also be performed using a smart doorbell or smart home without the presence of the customer.

Abstract: An online concierge system uses a machine learning click through rate model to select promoted items based on user embeddings, item embeddings, and search query embeddings. Embeddings obtained by an embedding model may be used as inputs to the click through rate model. The embedding model may be trained using different actions to score the strength of a customer interaction with an item. For example, a customer purchasing an item may be a stronger signal than a customer placing an item in a shopping cart, which in turn may be a stronger signal than a customer clicking on an item. The online concierge system generates a ranking of candidate promoted items based on the search query and using the click through rate model. Based on the ranking, the online concierge system displays promoted items along with the organic search results to the customer.