Abstract: In an example embodiment, a machine learning training pipeline is introduced that continuously monitors and processes training data having multiple transaction types using a sliding window, adding labels as they are available for the various different types of transactions in the training data. The processed training data, with the appropriate labels added, can then be utilized by any machine learning model that is being onboarded using the pipeline, without any specialized setup being necessary. Further, even if new data is added to the pipeline to aid in the training of a new model (such as data regarding a new payment type), this new data can be processed quickly and added to the existing data without requiring specialized processes by the entity requesting the new machine learning model. This allows the actual training of the new machine learning model to be accomplished very quickly, and deployment to be accomplished even faster.

Abstract: Disclosed are various embodiments for detecting security breaches using watchdog transaction accounts. A security agent can initiate a purchase with a first electronic commerce system and provide a watchdog transaction account as payment for the purchase. The security agent can then store a record of the purchase which includes a merchant identifier for the merchant and the watchdog transaction account. Subsequently, a transaction authorization system can determine that authorization for a transaction with second electronic commerce system failed. If the transaction authorization system determines that the account used in the transaction with the second electronic commerce system, then the transaction authorization system can determine that the first electronic commerce system has suffered a security breach.

Abstract: Disclosed are various embodiments for detecting security breaches using watchdog transaction accounts. A security agent can initiate a purchase with a first electronic commerce system and provide a watchdog transaction account as payment for the purchase. The security agent can then store a record of the purchase which includes a merchant identifier for the merchant and the watchdog transaction account. Subsequently, a transaction authorization system can determine that authorization for a transaction with second electronic commerce system failed. If the transaction authorization system determines that the account used in the transaction with the second electronic commerce system, then the transaction authorization system can determine that the first electronic commerce system has suffered a security breach.

Abstract: Systems and methods for fraud management using a distributed database are disclosed. The system may receive a payment request and generate a payment request hash by cryptographically processing the payment request using a hashing algorithm. The system may invoke a fraud reporting smart contract by passing the payment request hash and a public blockchain address to the fraud reporting smart contract. The system may query a local blockchain database to locate a fraud report matching the payment request hash to determine whether the payment request has been previously reported as fraud. In response to the payment request hash not matching the fraud report, the fraud reporting smart contract is configured to write the payment request hash to the blockchain as a second fraud report.

Abstract: Techniques for enabling an artificial intelligence system to infer grounded intent from user input, and automatically suggest and/or execute actions associated with the predicted intent. In an aspect, core task descriptions are extracted from actionable statements identified as containing grounded intent. A machine classifier receives the core task description, actionable statements, and user input to predict an intent class for the user input. The machine classifier may be trained using unsupervised learning techniques based on weakly labeled clusters of the core task description extracted over a training corpus. The core task description may include verb-object pairs.

Abstract: Automatic extraction and application of conditional tasks from content is provided. A conditional task system includes a classifier that is trained and used to identify conditional tasks and to learn appropriate times and methods to engage a user for reminding the user about conditional tasks. The conditional task system includes components for enabling an automated detection of a conditional task, extracting of attributes that characterize a condition associated with the task, using information about the condition to determine how to monitor for satisfaction of the condition, determining when and how to engage the user about the task, and notifying the user at an appropriate time and using an appropriate method when the condition is satisfied.

Abstract: Examples of the present disclosure describe systems and methods for improving the recommendations provided to a user by a recommendation system using viewed content as implicit feedback. In some aspects, attention models are created/updated to infer the user attention of a user that has viewed or is viewing content on a computing device. The attention model may be used to convert inferences of user attention into inferences of user satisfaction with the viewed content. The inferences of user satisfaction may be used to generate inferences of fatigue with the viewed content. The inferences of user satisfaction and inferences of user fatigue may then be used as implicit feedback to improve the content selection, content triggering and/or content presentation by the recommendation system. Other examples are also described.

Abstract: One or more techniques and/or systems are provided for ranking recommendations within a set of recommendations. For example, a set of locational relevance boundaries may be generated and/or configured for ranking the set of recommendation. For example, a locational relevance boundary may adjust a rank of a recommendation using a rank influence (e.g., a linear function, a step function, a numerical value, and/or any other function used to increase, decrease, or assign a value to the rank based upon a current location of the user). The locational relevance boundary may be applied based upon the current location of the user corresponding to one or more threshold distances from a target recommendation location. For example, a logarithmic function may be applied to a rank of a theater recommendation when the user is less than 1.2 miles from the theater. Ranked recommendations may be provided to the user.

Abstract: A user profile is constructed from implicit user signals, such as prior searches, as well as from explicit user signals. The user profile informs which types of information is proactively obtained for the user. The proactively obtained information is ranked so that only the highest ranked information is proactively presented to the user. The proactive presentation of information to the user can take the form of tiles, or other like graphical elements, that can comprise images with text indicative of information that is being proactively presented to the user. User action directed to a tile can result in the presentation of further information.

Abstract: Systems, methods, and computer-readable storage media for facilitating discoverability of user interface elements associated with search engine homepages are provided. When users navigate to a search engine homepage having a background image associated therewith and perform a specified action with respect thereto (for instance, moving a mouse pointer over the image, clicking on the image, or tapping on the image), a plurality of user interface elements is surfaced. Each user interface element is associated with information with which the user may be presented upon interaction with the element. Upon the user ceasing to perform the specified action (or upon the expiration of a predefined period of time), the user interface elements cease to be presented and the user is able to view the search engine homepage without the embedded information.

Abstract: Systems, methods, and computer-readable storage media are provided for integrating third party content with search engine homepages. Users indicate preferences for content to be presented in association with all of part of a search engine homepage such that, upon requesting presentation of the search engine homepage, the preferred content is presented. Generally, the preferred content is provided by a third-party content provider. User preferences may be indicated via a user interface control that permits selection of preferred content from a plurality of content options. Upon receiving the user preference, the search engine requests presentation of the preferred content from the third-party provider each time the user requests presentation of the search engine homepage.

Abstract: A method for subscribing to a portion of a plurality of software modules allows a subscriber to have access to the portion of the plurality of software modules. The method includes the steps of identifying the subscriber, providing the plurality of software modules to the subscriber and receiving the selections from the subscriber. The method then associates each of the selections with the portion of the plurality of software modules. Each of the portion of plurality of software modules is then tagged to correlate the portion of the plurality of software modules to the subscriber. The method then stores these tags such that the subscriber can access the portion of the plurality of software modules when desired.

Abstract: A method is disclosed for performing an action associated with a security in a market that has a plurality of conditions. The method includes the steps of identifying the security. Once the security is identified, an action is selected which will be taken at a certain time. A portion of the plurality of conditions are chose. Parameters are set for each of the conditions that are to be met. The action is then performed with regard to that security when the parameters for each of the conditions is met. The method provides for the monitoring of several securities and several conditions for each of the securities. The conditions that are being watched for each security may be different and may be edited on the fly independently of the conditions being watched for the other securities, which continue to be monitored even during the editing of the other security.