Abstract: A method is provided, including: storing comments generated in response to a content item served over a network; analyzing the comments to determine features associated with each of the comments; using a scoring model to score each comment based on the comment's corresponding features; receiving a request to serve a subset of the comments; responsive to the request, selecting a ranking of the comments that is one permutation from possible rankings of the comments, wherein selecting the ranking is in accordance with a probability distribution of the possible rankings that is based on the scores of the comments; serving comments identified by the selected ranking over the network to a client device; determining a dwell time on the served comments; applying the dwell time to update the scoring model.

Abstract: Technologies for selectively streaming video based on mapping buffer data to download rates. The technologies can include sampling download rates of a video player prior to or during downloading of a first video segment to the video player. The technologies can include determining a range of download rates of the video player based at least on the sampled download rates. The technologies can include determining an amount of accumulated video data in a buffer of the video player immediately prior to or during the downloading of the first video segment. Also, the technologies can include mapping the determined amount of accumulated video data to a download rate within the determined range of download rates using a mapping function, and determining a second video segment quality for a second video segment based at least on the mapped download rate.

Abstract: One or more computing devices, systems, and/or methods for searching across multiple user interfaces are provided. For example, a search query may be received from a user of a computing environment. A set of user interfaces, hosted by the computing environment, may be identified (e.g., mobile apps of a mobile operating system, browser tabs of a web browser, etc.). Content within the set of user interfaces may be queried using the search query to identify search results of the search query (e.g., webpages, social network posts, emails, or other content matching the search query). In an example, cached content of the user interfaces (e.g., recently accessed emails cached within a local content buffer cache by the web browser) may be searched using the search query for efficient querying of relevant content, such as through local or offline querying. The search results may be provided to the user.

Abstract: Disclosed are systems and methods for improving interactions with and between computers in content generating, searching, hosting and/or providing systems supported by or configured with personal computing devices, servers and/or platforms. The systems interact to identify and retrieve data within or across platforms, which can be used to improve the quality of data used in processing interactions between or among processors in such systems. The disclosed systems and methods provide systems and methods for automatically allowing web browsing in a user interface of an electronic messaging system. The disclosed systems and methods automatically display electronic messages containing item information displayed in an electronic message in place of any web page links associated with the item enabling browsing of the item information from within the electronic messaging system's user interface.

Abstract: Disclosed is a system and method for email management that leverages information derived from automatically generated messages in order to identify types of messages and message content. The disclosed systems and methods apply the information learned from decoding previously received messages to other messages in a user's inbox to fully, or at least partially decode the information included within such messages. The disclosed systems and methods analyze messages received in a user's inbox to detect message specific information corresponding to types of content in the message and the location of such content in the messages. The message specific information is then applied to other newly received or identified messages to learn message specific information about those messages. Based on such learning, information can be extracted from such messages in order to increase a user's experience and increase monetization.

Abstract: Methods and apparatus for using features of images representing content items to improve the presentation of the content items are disclosed. In one embodiment, a plurality of digital images are obtained, where each of the images represents a corresponding one of a plurality of content items Image features of each of the digital images are determined. Additional features including at least one of user features pertaining to a user of a client device or contextual features pertaining to the client device are ascertained. At least a portion of the content items are provided via a network to the client device using features that include or are derived from both the image features of each of the plurality of digital images and the additional features.

Abstract: Disclosed are systems and methods for improving interactions with and between computers in content searching, hosting and/or providing systems supported by or configured with devices, servers and/or platforms. The disclosed systems and methods provide a novel framework for providing users with electronic retrieval capabilities that are activated upon the users' determined locations respective to real-world locations associated with a message providing entity. The disclosed technology combines the previously separate systems of mail extraction, geo-fencing and content delivery (e.g., notification) into a single system that efficiently manages a user's inbox in order to provide the user with content the user has expressly indicated is of interest to that user.

Abstract: At least one label prediction model is trained, or learned, using training data that may comprise training instances that may be missing one or more labels. The at least one label prediction model may be used in identifying a content item's ground-truth label set comprising an indicator for each label in the label set indicating whether or not the label is applicable to the content item.

Abstract: Disclosed are systems and methods utilizing neural contextual bandit for improving interactions with and between computers in content generating, searching, hosting and/or providing systems supported by or configured with personal computing devices, servers and/or platforms. The systems interact to make item recommendations using latent relations and latent representations, which can improve the quality of data used in processing interactions between or among processors in such systems. The disclosed systems and methods use neural network modeling in automatic selection of a number of items for recommendation to a user and using feedback in connection with the recommendation for further training of the model(s).

Abstract: One or more computing devices, systems, and/or methods for hyper parameter optimization for machine learning ensemble generation are provided. For example, one or more base models are trained using diverse sets of hyper parameters, wherein different sets of hyper parameters (e.g., hyper parameters with different values) are used to train different base models. A matrix, populated with predictions from the set of base models, is generated. A machine learning ensemble is generated by processing the matrix utilizing a meta learner.

Abstract: The present teaching relates to providing a query suggestion. In one example, a request is received for query suggestions with respect to a query prefix input by a user. A plurality of query suggestions is determined based on the query prefix and a preceding query input by the user. A degree of popularity of the preceding query is determined. One or more query suggestions are selected from the plurality of query suggestions based on the degree of popularity of the preceding query. The one or more query suggestions are provided as a response to the request.

Abstract: Software for online active learning receives content posted to an online stream at a website. The software converts the content into an elemental representation and inputs the elemental representation into a probit model to obtain a predictive probability that the content is abusive. The software also calculates an importance weight based on the elemental representation. And the software updates the probit model using the content, the importance weight, and an acquired label if a condition is met. The condition depends on an instrumental distribution. The software removes the content from the online stream if a condition is met. The condition depends on the predictive probability, if an acquired label is unavailable.

Abstract: In one embodiment, a user of a client device may selectively share personal data with web companies. The personal data may be shared directly via transmission by the client device to the web companies. Alternatively, the personal data may be shared with the web companies via a personal data sharing system that collects and maintains personal data associated with users. The personal data sharing system may reward the user for sharing their personal data with the web companies.

Abstract: The present teaching relates to a method, system, and programming for providing content. A plurality of content items and publication information related thereto are obtained. For each of the plurality of content items, one or more topics are determined in accordance with a model. The related publication information associated with each content item is analyzed to identify at least one source of a plurality of sources that published the content item. A distribution is generated of each of the plurality of content items with respect to the plurality of sources and the one or more topics of the content item, and a bias of a source with respect to publishing content is identified based on the distributions of the plurality of content items.

Abstract: One or more computing devices, systems, and/or methods for selecting messages for transmission to client devices are provided. A first message associated with a first entity may be received from a first client device. The first message may comprise information associated with the first entity. A user account database, comprising a plurality of user profiles, may be analyzed based upon the first message to select a set of user accounts, from a plurality of user accounts, for transmission of a set of messages associated with the first message. The plurality of user accounts may be subscribed to an information service. Each user profile of the plurality of user profiles may comprise information corresponding to a user account of the plurality of user account. The set of messages may be transmitted to the set of user accounts.

Abstract: One or more computing devices, systems, and/or methods are provided. Event information associated with a plurality of events may be identified. The plurality of events may be associated with first entities corresponding to a first entity type and second entities associated with a second entity type. A first network profile associated with the first entities and the second entities may be generated based upon the event information. An iterative process may be performed to identify a coalition network associated with fraudulent activity. The iterative process may include analyzing the first network profile to identify a first set of entities, of the first entities, that are related to an entity of the second entities, and/or analyzing the first network profile to identify a second set of entities, of the second entities, that are related to the first set of entities. Multiple iterations may be performed to identify the coalition network.

Abstract: One or more computing devices, systems, and/or methods for cross-domain action prediction are provided. Action sequence embeddings are generated based upon a textual embedding and a graph embedding utilizing past user action sequences corresponding to sequences of past actions performed by users across a plurality of domains. An autoencoder is trained to utilize the action sequence embeddings to project the action sequence embeddings to obtain intent space vectors. A service switch classifier is trained using the intent space vectors. In response to the service switch classifier predicting that a current user will switch from a current domain to a next domain, the current user is provided with a recommendation of an action corresponding to the next domain.

Abstract: Disclosed are embodiments for providing batch performance using a stream processor. In one embodiment, a method is disclosed comprising receiving, at a stream processor, an event, the stream processor including a plurality of processing stages; generating, by the stream processor, an augmented event based on the event, the augmented event including at least one additional field not appearing in the event, the additional field generated by an operation selected from the group consisting of a join or dimensional annotation operation; and emitting, by the stream processor, the augmented event to downstream consumer.

Abstract: One or more computing devices, systems, and/or methods for content recommendation based upon continuity and grouping information of attributes are provided herein. User interaction data specifying whether users interacted with content items, user attributes of the users, and content attributes of the content items is obtained. A data structure is populated with the user interaction data. The data structure is modified by inserting a set of sub-fields into the data structure for a user attribute. A sub-field is populated with a value representing an option of the user attribute. The set of sub-fields are an encoding of continuity information and grouping information representing options for the user attribute. The data structure is processed using machine learning functionality to generate a model. The model is utilized to generate a prediction as to whether a user will interact with a content item.

Abstract: The present teaching relates to dynamically generating a card. In one example, a request is received for generating a card to be provided to a user. Dynamic information related to the request is obtained. One or more modules are selected to be put into the card based on the dynamic information. The card is generated based on the selected one or more modules.