Abstract: System and methods are described for generating training data for a natural language search system, the training data including synthetic user queries and associated structured query language (SQL) statements. A method includes generating an abstract query representation from a probabilistic context-free grammar (PCFG), selecting a sample record from a database using the abstract query representation, generating a tagged user query from the abstract query representation and the sample record, and generating the one or more SQL statements from the tagged user query and one or more SQL templates.

Abstract: A method and system for expressive conversion of more expressive metadata to less expressive metadata while maintaining lossless conversion including receiving a request for a second system metadata, retrieving a first system metadata, generating the second system metadata in response to the request for the second system metadata and the first system metadata wherein the second system metadata includes a first subset of the first system metadata, generating a site metadata mapping in response to the second system metadata and the first system metadata wherein the site metadata mapping includes a second subset of the first system metadata not included in the second subset of the first system metadata, storing the site metadata mapping in a memory, and transmitting the second system metadata to a client application.

Abstract: Embodiments described herein provide natural language processing (NLP) systems and methods that utilize energy-based models (EBMs) to compute an exponentially-weighted energy-like term in the loss function to train an NLP classifier. Specifically, noise contrastive estimation (NCE) procedures are applied together with the EBM-based loss objectives for training the NLPs.

Abstract: System and methods are described for performing sequences of computations in an infrastructure-agnostic manner. In one implementation, a method comprises: receiving a dispatch request for executing a user-defined pipeline; computing a performance metric based on the dispatch request; and determining, based at least partially on the performance metric, whether to execute the user-defined pipeline locally by the pipeline engine or transmit the dispatch request back to the network adapter.

Abstract: Methods, systems, and devices for distributed hyperparameter tuning and load balancing are described. A device (e.g., an application server) may generate a first set of combinations of hyperparameter values associated with training a mathematical model. The mathematical model may include a machine learning model, an optimization model, or any combination. The device may identify a subset of combinations from the first set of combinations that are associated with a computational runtime that exceeds a first threshold and may distribute the subset of combinations across a set of machines. The device may then test each of the first set of combinations in a parallel processing operation to generate a first set of validation error values and may test a second set of combinations of hyperparameter values using an objective function that is based on the first set of validation error values.

Abstract: Techniques for providing a suggested message to a user of a communication platform are described herein. The communication platform can receive a request, from a first user, to suggest a message (e.g., suggested message) to a second user with whom the first user is connected via the communication platform. The request can include contents of the suggested message and a user identifier associated with the second user. In some examples, the request can additionally include a suggested virtual space via which the suggested message is to be published, a suggested time for delivery, and/or the like. The communication platform can cause a presentation of the suggested message to the second user, such as via a user interface. In response to receiving an approval of the suggested message via the user interface, the communication platform can cause a publication thereof in association with a user account of the second user.

Abstract: Techniques and structures to facilitate anomaly detection, including receiving performance metric data at a database system having a plurality of data point values, processing the plurality of data point values via an optimized Extreme Studentized Deviate (ESD) test to detect one or more anomalous data point values within the plurality of data point values, including storing a first set of values representing the largest values of the plurality of data point values and a second set of values representing the smallest values of the plurality of data point values, receiving a first of the plurality of data point values, determining whether the first data point value is to be included in the first set of values or the second set of values and determining that the first data point value is not an anomalous data point value upon a determination that the first data point value is not to be inserted into the first set of values or the second set of values and generating an incident alert upon a detection of one or m

Abstract: A method for sequence-to-sequence prediction using a neural network model includes generating an encoded representation based on an input sequence using an encoder of the neural network model and predicting an output sequence based on the encoded representation using a decoder of the neural network model. The neural network model includes a plurality of model parameters learned according to a machine learning process. At least one of the encoder or the decoder includes a branched attention layer. Each branch of the branched attention layer includes an interdependent scaling node configured to scale an intermediate representation of the branch by a learned scaling parameter. The learned scaling parameter depends on one or more other learned scaling parameters of one or more other interdependent scaling nodes of one or more other branches of the branched attention layer.

Abstract: A user device (e.g., a computing device, a smart device, a mobile device, a laptop, a tablet, a set-top box, a display device, etc.) may generate and/or display a scrollable user interface (e.g., a configuration form/interface, a web form, a web page, an application page, a data form, etc.) that includes data fields. A data reflection element for a data field that is displayed based on proximity of an interactive element to the data field may be generated. Based on an interaction with the data reflection element via the interactive element, data input to the data field may be displayed in a field of a data reflection window of the scrollable user interface. An indication element that indicates correspondence between the data field and the field of the data reflection window may also be displayed.

Abstract: Disclosed herein are system, method, and computer program product embodiments for a query processing system. An embodiment operates by receiving a query associated with a table. An ordered index with entries that include key values that uniquely identifies a corresponding record in the table is determined. A last key value is initialized to store a minimum value for a data type of the key value. The query is executed as a union-all of a plurality of queries. While a count of selected records is less than or equal to a threshold quantity and at least one entry remains in the ordered index, an ordered next entry with a key value subsequent to the last key value in the index is identified, the corresponding record is selected from the table, the last key value is updated, the query is executed, and the count is incremented.

Abstract: Techniques for performing dynamic profile reconciliation in a communication system. A server computing system may identify two or more user profiles corresponding to a same user that is associated with the communication system. The server computing system may integrate historical data associated with the two or more user profiles into a single user profile (e.g., a destination user profile). The historical data may include messages sent and/or received via the two or more user profiles, files associated with the messages, metadata corresponding to each message, and the like. Responsive to combining the two or more user profiles into the single user profile, the server computing system may delete at least one of the two or more user profiles, storing the combined data in association with the single user profile.

Abstract: Embodiments described herein provides a training mechanism that transfers the knowledge from a trained BERT model into a much smaller model to approximate the behavior of BERT. Specifically, the BERT model may be treated as a teacher model, and a much smaller student model may be trained using the same inputs to the teacher model and the output from the teacher model. In this way, the student model can be trained within a much shorter time than the BERT teacher model, but with comparable performance with BERT.

Abstract: Embodiments described herein provide safe policy improvement (SPI) in a batch reinforcement learning framework for a task-oriented dialogue. Specifically, a batch reinforcement learning framework for dialogue policy learning is provided, which improves the performance of the dialogue and learns to shape a reward that reasons the invention behind human response rather than just imitating the human demonstration.

Abstract: Provided is a group-based communication interface configured to allow users of the interface to communicate within group-based communication channels and across group-based communication channels to provide relevant information to other users efficiently and effectively. Group-based messaging communications across the plurality of group-based communication channels can be channeled to specific receiver channels providing relevant information to users of the group-based communication interface thereby increasing the efficiency and effectiveness of group-based messaging communications and the lifetime of the system.

Abstract: A system and apparatus for enhancing the functionality and utility of an authentication process for web applications is disclosed.

Abstract: A request to contact a service provider may be received from a client machine. The request may be associated with an identity claim and including a service identifier. The identity claim may be validated via a distributed identity service that includes a plurality of identity nodes in communication via a network. Validating the identity claim may include determining a designated network identifier associated with a distributed identity account shared among the plurality of identity nodes. A service query that includes the service identifier and the designated network identifier may be sent to a plurality of customer relations management services. A communication session may be established between a service provider remote computing system and the client machine. The service provider may store customer relations management information at a designated one of the plurality of customer relations management services.

Abstract: Multi-workspace shared communication channels described. A communication channel of a group-based communication platform can be associated with multiple group identifiers to associate the communication channel with at least two workspaces, wherein a first group identifier corresponds to a first workspace within which a first group communicates and a second group identifier corresponds to a second workspace within which a second group. The communication channel can be associated with multiple organization identifiers to associate the communication channel with at least two organizations, wherein a first organization identifier corresponds to a first organization and a second organization identifier corresponds to a second organization.

Abstract: Systems, methods, and devices facilitate synchronization of content between instances of data records. Servers may be configured to generate a data record comprising a dynamic synchronization data object, receive an input associated with the data record, and determine a plurality of synchronization parameters including a plurality of expansion parameters and a plurality of permission parameters associated with the dynamic synchronization data object. The servers may be configured to determine a plurality of update parameters based on the synchronization parameters, the update parameters identifying if a modification included in the input should be applied to contents of the dynamic synchronization data object. The servers may also be configured to update the data record and the contents of the dynamic synchronization data object based, at least in part, on the determined update parameters, the updating including pushing the updated contents to at least one of the plurality of target data records.

Abstract: A method for deep reinforcement learning using a neural network model includes receiving a distribution including a plurality of related tasks. Parameters for the reinforcement learning neural network model is trained based on gradient estimation associated with the parameters using samples associated with the plurality of related tasks. Control variates are incorporated into the gradient estimation by automatic differentiation.

Abstract: Database systems and methods are provided for providing record data to an application extension associated with a native application at a client device. One method involves a cross-platform service at the client device receiving a plurality of requests for data for one or more records at the database system from the native application, aggregating a subset of the plurality of requests into a batched request, providing the batched request to the database system over a network, receiving a batched response to the batched request from the database system over the network, and for each request of the subset of the plurality of requests of the batched request, mapping a respective data set of the batched response to a respective request and providing the respective data set to the application extension in response to the respective request via the native application.