Abstract: Methods, systems, and computer storage media for providing supplementary item information on communications-based compatibility features for a primary item in an item listing platform. In particular, communications-based compatibility features are extracted from user communications about the primary item to identify additional features of the item. In operation, a supplementation engine of an item listing platform, accesses user communications for the primary item. Supplementary item information is generated for the item, where the supplementary item information includes a communications-based compatibility feature associated with the item. The communications-based compatibility feature indicates that the primary item is compatible with the secondary item, a non-standard feature (e.g., a feature not provided in a feature set of the primary item), or a demographic feature (e.g., age).

Abstract: Systems and methods for are provided for predicting impending failure of a database and preemptively initiating mitigating failover actions, for example by shedding connections or redirecting connection requests to an alternate database that can fulfill resources being requested. In an example embodiment, to detect a slow or unstable database, connection wait times are monitored over a rolling window of time intervals, a quantity of intervals in which at least one excessive wait time event occurred are counted during the time window, and if the quantity exceeds a threshold, the database is deemed unavailable, thereby triggering connection adjustments.

Abstract: A system comprising a computer-readable storage medium storing at least one program and a computer-implemented method for simulating the fit of a particular garment on a user using a fit simulation garment. The fit simulation garment is wearable by a user and comprises a plurality of actuators configured to simulate tightness of regions of the particular garment. The method includes receiving a garment fit model defining the fit of the particular garment on the user. The method further includes parsing the garment fit model to generate control data for controlling the fit simulation garment, and transmitting the control data to the fit simulation garment to cause the plurality of actuators to collectively simulate the fit of the particular garment on the user.

Abstract: Search techniques are described that support locating and displaying prospective objects in digital images based on a removed object. A digital image, for instance, is received by a computing device as an input depicting a physical environment with various objects and displayed in a user interface. An object depicted by the digital image is removed. An aspect of the removed object is identified. A search system leverages the aspect to locate a prospective object. The prospective object is configured for display within the digital image. As a result, the digital image is displayed in the user interface as having the configured prospective object depicted in the physical environment.

Abstract: Dynamic link preview generation techniques are described that overcome the challenges of conventional techniques by supporting link preview generation by a content provider system that proves the digital content via a respective network address. In one example, a content provider system, based on a request received from a service provider system, identifies a communication platform of the service provider system that is to be used to communicate the shared link. Upon identifying the communication platform, the content provider system locates customization data that describes how the link preview is to be generated for the communication platform. In response, the content provider system renders digital content available via the network address to generate the link preview.

Abstract: Technologies are shown for network attribution tracking for a multi-legged transaction. In accordance with some aspects, a request is received at a second service from a client device via a first redirect including a first token associated with a first service. The second service sends, to a token service, a token request including the first token, wherein the token request causes the token service to associate a second token with the first token. The second service receives, from the token service, a token response including the second token. The second service provides, to the client device, a second redirect to a third service, the second redirect including the second token. In some aspects, a transaction is attributed to at least the first service and the second service based on the association of the second token with the first token.

Abstract: Techniques are described, as implemented by computing devices, to control access to transactions through use of cosigning based on tokenized reputation scores. This is performed by leveraging a blockchain such that tokenized reputation scores are generated based on amounts of cryptographic reputation tokens associated with blockchain account addresses associated with applicant and co-signer service provider accounts. Transactional functionality is made available to an applicant service provider account having an insufficient tokenized reputation score by using a co-signer service provider account having a sufficient tokenized reputation score to at least partially back an obligation of the applicant service provider for a transaction.

Abstract: An inventory prediction system is described that outputs a predicted inventory item not included in a user's known inventory using a cross-category directional graph that represents item categories as nodes. The inventory prediction system implements a prediction model trained using machine learning to output the predicted inventory item using the graph and at least one item from the user's known inventory. The inventory prediction system is further configured to generate a listing recommendation for the predicted inventory item. To do so, the inventory prediction system implements a logistic regression model trained using machine learning to calculate a probability that the listing recommendation should be generated using attributes of the predicted inventory item and attributes of currently trending items.

Abstract: Technologies are shown for shipping route selection involving receiving sender and recipient shipping information for an item to be shipped and obtaining shipping route options for the item based on the sender and recipient shipping information. The route options are provided for display and selection of an option. A routing data block is created for the item at an address on a blockchain that stores shipping information for each stage of the selected route. A shipping tag is encoded with the blockchain address and attached to the item. The tag can be scanned to obtain the blockchain address and request information for a next stage of the shipping route from the block. The next shipping information from the block is received and utilized to ship the item to a next geolocation. The next shipping information can be determined based on current conditions, such as weather, pricing and availability.

Abstract: A search query and a concept comprising a feature are received from a user. Positive and negative item sets are identified, where items of the positive item set comprise the feature and items of the negative item set do not comprise the feature. A hyperplane located between the positive and negative item sets is generated using a machine learning model. A concept activation vector (CAV) orthogonal to the hyperplane is determined. The CAV is used to produce a modified search query vector. Based on comparing the modified search query vector with item vectors, a second set of search results is provided.

Abstract: Techniques are described, as implemented by computing devices, to control access to transactions through use of tokenized reputation scores. This is performed by leveraging a blockchain such that a tokenized reputation score is generated or calculated based on an amount of reputation tokens associated with a blockchain account address associated with a service provider account, and by making transactional functionality available to the service provider account based on a comparison of a tokenized reputation score affiliated with the service provider account with a threshold score associated with a transaction.

Abstract: Technologies are shown for storing data from a data object in a distributed application architecture and reassembling the data object from the stored data. A first set of data from the data object is stored on a distributed file system, and a second set of data from the data object is stored on a blockchain. In response to a request for the data object, a script from the blockchain is executed to generate a reassembled data object by: obtaining, from the blockchain, metadata to reassemble the data object, obtaining the second set of data from the blockchain and the first set of data from the distributed file system, and generating the reassembled data object using the first set of data and the second set of data based on the metadata.

Abstract: Systems and methods provide increased authentication security by verifying that a device is associated with a human user (as opposed to, for instance, malicious code). In some configurations, a request for authentication is received from a device that includes an identifier of the device (e.g., an IP address of the device). A determination is made that the identifier of the device was not previously associated with a human user. In response to determining that the identifier of the device was not previously associated with a human user, a verification process is initiated to determine that the device is associated with a human user.

Abstract: Technologies are shown for secure management of evaluation data. An evaluation value is received from a source entity. The evaluation value relates to an evaluation entity. Trusted source data secured in one or more data blocks on a first blockchain is searched for an entry corresponding to the source entity. Based on finding the entry corresponding to the source entity in the trusted source data, an evaluation score for the evaluation entity is obtained from the first blockchain or a second blockchain. The evaluation score was previously calculated based on one or more previous evaluation values for the evaluation entity and stored on the first blockchain or the second blockchain. A new evaluation score for the evaluation entity is calculated using the evaluation score and the evaluation value. The new evaluation score for the evaluation entity is securely committed to a new data block on the first blockchain or the second blockchain.

Abstract: Systems and methods for centralized client application management are provided. In an example embodiment, a client state is received from a mobile device at a management device. The client state comprises information about a particular application. A first instruction for the mobile device is generated by one or more hardware processors. The first instruction comprises an instruction to the mobile device to prevent user interaction with a user interface of a particular application. The first instruction is communicated to the mobile device, and the mobile device performs the first instruction to prevent user interaction with the user interface of the particular application.

Abstract: A database node selects one of a number of database mirror hosts to perform a database read query. In theory, mirror hosts store redundant database entries. In practice, some mirror hosts fail to receive or retain some entries, e.g. if a mirror host was down for maintenance, a mirror host was overburdened, etc. The health of a mirror host, quantified as a health score, represents a likelihood that an insert statement was received and retained. Health scores are generated based on health metrics—data periodically retrieved from mirror hosts used to infer how well the database was operating, e.g. a count of entries that were inserted, a count of errors encountered while the entries were inserted, etc. If the database read query specifies entries inserted during a specific time period, the mirror host may be selected based on the health scores representing that time period.

Abstract: Technologies are disclosed herein for enhancing machine learning (“ML”)-based anomaly detection systems using knowledge graphs. The disclosed technologies generate a connected graph that defines a topology of infrastructure components along with associated alarms generated by a ML component. The ML component generates the alarms by applying ML techniques to real-time data metrics generated by the infrastructure components. Scores are computed for the infrastructure components based upon the connected graph. A root cause of an anomaly affecting infrastructure components can then be identified based upon the scores, and remedial action can be taken to address the root cause of the anomaly. A user interface is also provided for visualizing aspects of the connected graph.

Abstract: A machine may be configured to perform image evaluation of images depicting items for sale and to provide recommendations for improving the images depicting the items to increase the sales of the items depicted in the images. For example, the machine accesses a result of a user behavior analysis. The machine receives an image of an item from a user device. The machine performs an image evaluation of the received image based on an analysis of the received image and the result of the user behavior analysis. The performing of the image evaluation may include determining a likelihood of a user engaging in a desired user behavior in relation to the received image. Then, the machine generates, based on the evaluation of the received image, an output that references the received image and indicates the likelihood of a user engaging in the desired behavior.

Abstract: A method and a system process a stream of data in parallel across a plurality of nodes. The log processing system has a log module, a query language module, and a query processing module. The log module receives and organizes the stream of data into a sequential and nested data structure. The query language operator module defines operators that operate on the sequential and nested data structure. The query processing module processes in parallel across a plurality of nodes a query based on an operator on the stream of data.

Abstract: A multimodal embedding modifier generates a modified seed search selection embedding for providing a set of search results. The multimodal embedding modifier enhances the ability and accuracy of identifying a user's true intent when searching the online marketplace. For example, embodiments disclosed herein can allow a user to navigate multiple modalities for an item. In some embodiments, a user may select a search result corresponding to an initial search query, and further modify the selected search result by inputting a modifier (e.g., a textual modifier). The multimodal embedding modifier can be trained using a training dataset including a text embedding, an image embedding, another type of embedding, or a combination thereof.