Abstract: Techniques for dynamically developing a contextual set of prompts based on relevant aspects extracted from s set of training data. One technique includes obtaining training data comprising text examples and associated labels, extracting aspects from the training data, generating prompting templates based on the training data and the extracted aspects, concatenating each of the text examples with the respective generated prompting template to create prompting functions, training a machine learning language model on the prompting functions to predict a solution for a task, where the training is formulated as a masked language modeling problem with blanks of the prompting templates being set as text labels and expected output for the task being set as specified solution labels, and the training learns or updates model parameters of the machine learning language model for performing the task. The machine learning language model is provided with the learned or updated model parameters.

Abstract: Various techniques can include accessing a master tree that was generated using a plurality of protocol definitions. The plurality of protocol definitions can identifies an ordered set of actions and specifies, for each sequential pair of actions in the ordered set of actions, an action-advancement condition that identifies a criterion for advancing across the sequential pair of actions in the ordered set of actions so as to trigger a later of the sequential pair of actions. A master tree includes a set of dynamic nodes and a set of static nodes. The technique can include accessing a partial protocol definition that includes at least one action. The technique can include generating an auto-completion of the partial protocol definition using the master tree, at least some of the dynamic-node weights, and at least some of the static-node weights. The technique can output a representation of an auto-completed protocol definition.

Abstract: Methods and systems are disclosed for automated semantic tagging that include detecting a particular thread executed by a processor and identifying a root process of the particular thread. An object-process link may be by linking an object that executed code that called the particular thread to the root process. A thread list of thread definitions of the object may be identified. A particular thread definition that corresponds to the particular thread can be mapped. Resource types to be consumed upon executing an instance of the thread instantiated from the particular thread definition can be identified and the corresponding values of the resource types can be determined. A process specification can be generated that encapsulates the thread definition, resource types and values so as to reproduce a state of the root process at a point in which the particular thread executed.

Abstract: Techniques are disclosed for eliding load and store barriers while maintaining garbage collection invariants. Embodiments described herein include techniques for identifying an instruction, such as a safepoint poll, that checks whether to pause a thread between execution of a dominant and dominated access to the same data field. If a poll instruction is identified between the two data accesses, then a pointer for the data field may be recorded in an entry associated with the poll instruction. When the thread is paused to execute a garbage collection operation, the recorded information may be used to update values associated with the data field in memory such that the dominated access may be executed without any load or store barriers.

Abstract: Techniques for modifying hierarchal-structured data of one hierarchal data structure based on a modification to another hierarchal data structure are disclosed. A system determines that a modification has been made, or is requested to be made, to a particular hierarchal data structure. The system analyzes a set of rules to determine whether the modification of the hierarchal data structure triggers another modification to an additional hierarchal data structure. The additional hierarchal data structure includes different nodes, or nodes arranged in a different hierarchal structure, than the particular hierarchal data structure. The system modifies the additional hierarchal data structure based on the rule. The modification of the additional hierarchal data structure is different than the modification to the particular hierarchal data structure.

Abstract: Techniques for runtime customization and management of user interfaces for enterprise applications are disclosed. In some embodiments, the techniques include tools, systems, and processes for building frontend, technology-independent user interfaces. A system may consume user input files that are language dependent and generate a metadata file that is language independent. The metadata file may be updated to customize or otherwise modify a user interface of an application at runtime. An event signal may be raised to notify the application when a new user interface is available. In response to detecting the signal, the application may refresh the user interface, including the application page being currently viewed, to modify the set of user interface components that are rendered and displayed to the end user. The system may store metadata for multiple versions of the user interface to allow users to quickly roll back to prior versions of a user interface.

Abstract: Techniques are disclosed for dynamic time-based custom model generation as part of infrastructure-as-a-service (IaaS) environment. A custom model generation service may receive a set of training data and a time-based constraints for training a machine learning model. The custom model generation service may subsample the training data and generate a set of optimized tuned hyperparameters for a machine learning model to be trained using the subsampled training data. An experimental interval time of training is determined and the machine learning model is trained on the subsampled training data according to the optimized tuned hyperparameters over a set of training intervals similar to the experimental time interval. A customized machine learning model trained in the time-based constraint is output. The hyperparameter tuning may be performed using a modified mutating genetic algorithm for a set of hyperparameters to determine the optimized tuned hyperparameters prior to the training.

Abstract: The present disclosure relates to systems and methods for an intelligent assistant (e.g., a chatbot) that can be used to enable a user to generate a machine learning system. Techniques can be used to automatically generate a machine learning system to assist a user. In some cases, the user may not be a software developer and may have little or no experience in either machine learning techniques or software programming. In some embodiments, a user can interact with an intelligent assistant. The interaction can be aural, textual, or through a graphical user interface. The chatbot can translate natural language inputs into a structural representation of a machine learning solution using an ontology. In this way, a user can work with artificial intelligence without being a data scientist to develop, train, refine, and compile machine learning models as stand-alone executable code.

Abstract: A computing device may access a target code for implementing an application. The device may identify addresses for one or more functions or one or more variables associated with the target code. The device may generate an interval tree comprising a root node and one or more function nodes. The device may in response to the target code invoking a function or variable: generate an intercept function configured to intercept communication between the target code and a call address for the at least one of the one or more functions or the one or more variables invoked by the target code. The device may intercept data communicated between the target code and the call address. The device may store the intercepted data as a function node in the interval tree. The device may transmit the interval tree to a user device.

Abstract: Techniques are described herein for probabilistic monitoring of high-frequency, low-latency database queries. In some embodiments, a probabilistic query monitoring system periodically samples active database sessions. For example, the system may generate sample data every one second or at some other sampling rate for each database session that is currently active. The sample data may include a mapping between query identifiers to sample counter values that are extracted at different sample intervals. The system may then estimate performance metrics for the set of active database based on the counter values sampled across consecutive sample intervals.

Abstract: Systems, devices, and methods discussed herein are directed to intelligently adjusting the set of worker nodes within a computing cluster. By way of example, a computing device (or service) may monitor performance metrics of a set of worker nodes of a computing cluster. When a performance metric is detected that is below a performance threshold, the computing device may perform a first adjustment (e.g., an increase or decrease) to the number of nodes in the cluster. Training data may be obtained based at least in part on the first adjustment and utilized with supervised learning techniques to train a machine-learning model to predict future performance changes in the cluster. Subsequent performance metrics and/or cluster metadata may be provided to the machine-learning model to obtain output indicating a predicted performance change. An additional adjustment to the number of worker nodes may be performed based at least in part on the output.

Abstract: Disclosed is an improved approach to implement the XA architecture, which permits a TM to call back to a RM via an AP. This approach avoids the need for the TM to maintain authentication/authorization credentials for each RM that will be contacted by the TM. This approach therefore also eliminates the need to establish a direct connection between the TM and the RM. An approach is also provided to integrate a traditional XA application with an XA transaction.

Abstract: One example embodiment is adapted for use in facilitating collection and transport of memory devices, e.g., drives, motherboards, etc., e.g., for the purposes of subsequent destruction. The example embodiment has an outer bin, also called a garage or outer enclosure, within which is placed an inner bin with wheels. Once closed and locked, the outer bin has a door that can be opened so drives or other memory devices can be inserted. The drives then fall into the inner bin and, once sufficiently filled, the inner bin be easily wheeled out from the outer bin. The inner bin has its own locking lid that is secured when transporting the inner bin with the drives, e.g., to a destruction facility or area. Other features are provided as described below. Enhanced synergistic security features include use of multi-point locking latches to secure doors and lids, metal construction of the bins, and so on. The bins are sized to facilitate efficient collection and transport of potentially sensitive data center media.

Abstract: Discussed herein are techniques that utilize hierarchical locality information of host machines included in a cluster network for the execution of general workloads. Hierarchical locality information for each host machine of a plurality of host machines is stored. The hierarchical locality information for a host machine identifying, for each locality of a plurality of localities, location information for the locality. Responsive to receiving a request requesting execution of a workload, the hierarchical locality information for the plurality of host machines is obtained and provided (e.g., to a customer) in response to the request.

Abstract: Discussed herein are techniques that utilize locality information of host machines included in a cluster network for the execution of graphical processing unit based workloads. For each host machine of a plurality of host machines, locality information for the host machine is stored therein. The locality information for a host machine identifies a rack comprising the host machine. Responsive to receiving a request requesting execution of a workload, one or more host machines of the plurality of host machines are identified as being available for executing the workload. For each of the one or more host machines, the locality information for the host machine is obtained. Further, linkage information of the one or more host machines is identified. The locality information and the linkage information of the one or more host machines is provided in response to the request.

Abstract: Techniques for creating a custom metric type to be added to a set of metrics generated by a data monitoring platform at run-time are disclosed. A system receives values defining properties of a custom metric type based on a custom metric template and a custom schema template. The system generates an instruction set, based on the values associated with the custom metric template, for generating the custom metric type on an executing data monitoring system. The system validates the instruction set and the custom schema to verify that the definitions for the custom metric type and the custom schema may be executed by the data monitoring system. The system adds the custom metric type, at run-time, to a set of metrics generated by the data monitoring system.

Abstract: A framework for managing authorization for performance of actions with a computing system. For example, techniques for performing authorization of users and/or clients for access to an infrastructure service provided by a cloud servicer provider (CSP) and/or for performance of actions with the infrastructure service.

Abstract: Techniques for modifying a query based on a data index of nodes in a data set are disclosed. A system modifies queries based on query terms associated with indexed data. The system modifies queries to include query terms based on indexed data or to obtain values for query terms that are not associated with indexed node properties. The system adds query terms, that reference indexed data, to a query in response to determining that none of a query's terms reference indexed data. The system derives values for query terms that are not associated with indexed node properties using a logical or mathematical formula. The system traverses parent nodes of a child node to identify values for query terms that are not associated with the child node in a data index, but are inherited from a parent node.

Abstract: A framework for managing credentials for access to a secured entity of an infrastructure service. For example, techniques for maintaining credentials for access to the secured entity within a trusted environment while utilizing the credentials for performance of actions within the infrastructure service.

Abstract: Techniques for implementing user interfaces, systems, and processes for multidimensional clustering and analysis are described herein. In one aspect, an application or cloud service receives a request to cluster a set of records where the request identifies a first set of one or more dimensions to use for clustering and a second set of one or more dimensions to analyze for correlation patterns. Responsive to receiving the request to cluster the set of records, the system generates clusters based at least in part on variances in the first set of one or more dimensions, wherein each cluster includes at least one record from the set of records. The system may generate, for each respective cluster, an analytic result that identifies how strongly the second set of one or more dimensions correlate to the respective cluster. The system may present the clusters and analytic results for further processing.

Abstract: In some aspects, a server device may identify one or more services of a cloud infrastructure via a management layer. The server device may determine service information and configuration information for the one or more services. The server device may generate an environment model based at least in part on the service information and the configuration information, the environment model providing information on relationship between one or more components of the cloud infrastructure. The server device may determine one or more threats to the one or more services based at least in part on analyzing the environment model and accessing a threat information database. The server device may generate a threat model that lists the one or more threats to the one or more services. The server device may generate one or more recommendations for the cloud infrastructure based at least on the threat model.

Abstract: Techniques for executing operations associated with sub-sets of queries in a set of nested queries are disclosed. A graphical user interface displays individual sub-queries as separate visual representations. Based on detecting a selection, a system executes a selected query in the set of nested queries, without executing parent queries in the set of nested queries. Based on detecting another selection, the system exports a selected query without exporting parent queries of the selected query. The system copies the metadata describing the selected query and transfers the metadata to another set of nested queries. Based on detecting another selection, the system displays a results count for a selected query. The system displays a number representing the number of results returned by the selected query, without displaying the query results.

Abstract: A spreadsheet user interface shows polymorphic records. The records include global fields, discriminator fields, and context-sensitive fields. A context-sensitive field is available or unavailable dependent on a discriminator field value. Record fields associate with spreadsheet columns, regardless of field availability. User confusion is prevented by marking a record's unavailable context-sensitive fields as unavailable and visually distinguishing them from available fields. A user-requested operation spanning multiple rows and columns is supported by testing whether a source cell and source data is available, a destination cell is available, and skipping the operation for those records whose source and/or destination cell is required but unavailable. An embodiment supports interfacing with a single record in a spreadsheet form by applying global form templates associated with global fields, and context-dependent form templates associated with specific discriminator field values.

Abstract: A computer-implemented method interfaces with a remote attachment from a spreadsheet program on a client device. A first area in the spreadsheet associates with a top-level object in a web server, and a second area associates with a descendant object. On a first user interaction with a cell in the second area, a user interface item is shown, allowing uploading and/or downloading the attachment. When the user interacts to upload or download, the program uploads (or queues for upload) a selected attachment and modifies attachment metadata in the second area, or downloads the attachment based on attachment metadata in the second area. Communication between the spreadsheet program and the web server may be REST compliant.

Abstract: Example embodiments employ a selective memory swapping system for selectively placing non-volatile memory devices of a computer system offline, e.g., for background updating, and online, for use by a computer system, whereby the background updating process includes a mechanism for performing forensics analysis and updating of offline memory devices while an alternate memory device is usable by a user of the first computer system.

Abstract: Systems, devices, and methods discussed herein enable the provisioning of one or more virtual resources within a cloud-computing environment based at least in part on user-defined constraints. By way of example, a request to provision one or more virtual resources within the cloud-computing environment may be received. The request may include a set of user-defined constraints. Placement data indicating a mapping of a particular virtual resource to a particular hardware resource may be obtained based at least in part on the set of user-defined constraints. The virtual resource(s) may be provisioned within the cloud-computing environment in accordance with the set of user-defined constraints based at least in part on utilization of the placement data.

Abstract: The present disclosure relates to chatbot systems, and more particularly, to techniques for obtaining data items for input to a chatbot. In certain embodiments, a chatbot system includes a component that can be invoked by a chatbot in the chatbot system to obtain data items needed by the chatbot. The component can be invoked based on a reference to the component in a dialog flow definition configured for the chatbot. The reference to the component can indicate a composite entity that the component will use to determine how the data items are obtained from a user. The composite entity acts as a container for the data items and may be configured separately from the dialog flow definition of the chatbot. The data items can be obtained based on rules specified in a composite entity definition configured for the composite entity.

Abstract: Using a computer system, an instruction is received to define or modify a permission constraint corresponding to one or more files. A permission-instruction data set representing the permission constraint is stored in a data store. Subsequent to storing the permission-instruction data, a user request to access a particular file is intercepted. The data store is queried to determine whether any pending permission-instruction data set corresponds to the particular file. In response to the query, it is determined that the permission-instruction data set corresponds to the particular file. A permission constraint of the particular file is added or modified based on the permission-instruction data set. Based on the modified or added permission constraint, it is determined whether and/or an extent to which the user request is authorized. A response to the user request based on the determination as to whether and/or an extent to which the user request is authorized.

Abstract: Techniques are provided for improved training of a machine learning model using lexical dropout. A machine learning model and a training data set are accessed. The training data set can include sample utterances and corresponding labels. A dropout parameter is identified. The dropout parameter can indicate a likelihood for dropping out one or more feature vectors for tokens associated with respective entities during training of the machine learning model. The dropout parameter is applied to feature vectors for tokens associated with respective entities. The machine learning model is trained using the training data set and the dropout parameter to generate a trained machine learning model. The use of the trained the machine learning model is facilitated.

Abstract: Techniques are described for using a single application to interact with multiple separate realms simultaneously while maintaining data security boundaries. For example, a web browser may be used to access and interact with the multiple separate secure realms while maintaining data security boundaries between the systems. Multiple concurrent sessions may be established for a user between the web browser and multiple realms. Separate sets of security credentials (e.g., credentials used for authentication and authorization purposes) may be used to establish the sessions and for operations performed in the realms via the sessions. The application can also execute logic (e.g., via machine-executable code or instructions) for automating operations performed in the realms, such as, automating the initiation of a certain operation in one realm based upon a response received from another realm, causing operations to be initiated in two different realms such that the operations overlap in the time; and the like.

Abstract: Techniques disclosed herein relate generally to constructing a customized knowledge graph. In one embodiment, entities and relations among entities are extracted from a user dataset based on certain rules to generate a seed graph. Large-scale knowledge graphs are then traversed using a finite state machine to identify candidate entities and/or relations to add to the seed graph. A priority function is used to select entities and/or relations from the candidate entities and/or relations. The selected entities and/or relations are then added to the seed graph to generate the customized knowledge graph.

Abstract: The present disclosure generally relates to systems and methods for efficiently and dynamically integrating a chatbot into an application service. More specifically, the present disclosure relates to systems and methods that dynamically build custom components that enable the chatbot to communicate and facilitate the use of the application service directly from the chatbot application. Accordingly, the systems and methods described herein provide the benefit of integrating the chatbot application into an application service without having the application developer write code for the integration, thereby enhancing the computational efficiency of the integration and the reliability of the application service when new services or service functionality are added or modified.

Abstract: Techniques are disclosed to provide automated remediation of services within a distributed computing system. A remediation service may receive an indication to evaluate a level of health of a first service in a distributed computing system. In response, the remediation service may determine a dependency of the first service on a second service. The remediation service can use a monitor plugin corresponding to the second service to obtain a health metric of the second service. Based on the health metric, the remediation service can determine whether the second service is unhealthy. If the second service is unhealthy, the remediation service can perform a remediation task for the second service.

Abstract: An instruction is received to define or modify a permission constraint corresponding to one or more file. A period of time is determined, where the period is one during which the defined or modified permission constraint is to apply to the one or more files. It is determined that the one or more files are to be deleted following completion of the period of time. A permission-instruction data set representing the permission constraint and a flag indicating that the one or more files are to be deleted following completion of the period of time are stored in a data store. It is detected that the period of time has ended. The one or more files are deleted in response to the detection.

Abstract: Techniques are provided for named entity recognition using a gazetteer incorporated with a neural network. An utterance is received from a user. The utterance is input into a neural network comprising model parameters learned for named entity recognition. The neural network generates a first representation of one or more named entities based on the utterance. A gazetteer is searched based on the input utterance to generate a second representation of one or more named entities identified in the utterance. The first named entity representation is combined with the second named entity representation to generate a combined named entity representation. The combined named entity representation is output for facilitating a response to the user.

Abstract: Techniques for predicting anomalies in forecasted time-series data are disclosed. A system. A system predicts whether a monitored computing system will experience anomalies by comparing forecasted values associated with components in the monitored computing system to threshold values. The system utilizes time-series machine learning models to forecast workloads of computing resources in the monitored computing system. The system trains and tests multiple different versions of a time-series model and selects the most accurate version to generate forecasts for a particular workload in the computing system. The system compares the forecasts to threshold values to predict anomalies. Based on detecting anomalies, the system generates recommendations for remediating predicted anomalies.

Abstract: A model-based approach to determining an optimal configuration for a data center may use an environmental chamber to characterize the performance of various data center configurations at different combinations of temperature and altitude. Telemetry data may be recorded from different configurations as they execute a stress workload at each temperature/altitude combination, and the telemetry data may be used to train a corresponding library of models. When a new data center is being configured, the temperature/altitude of the new data center may be used to select a pre-trained model from a similar temperature/altitude. Performance of the current configuration can be compared to the performance of the model, and if the model performs better, a new configuration based on the model may be used as an optimal configuration for the data center.

Abstract: Using a computer system, an instruction is received to define or modify a permission constraint corresponding to one or more files. A permission-instruction data set representing the permission constraint is stored in a data store. Subsequent to storing the permission-instruction data, a user request to access a particular file is intercepted. The data store is queried to determine whether any pending permission-instruction data set corresponds to the particular file. In response to the query, it is determined that the permission-instruction data set corresponds to the particular file. A permission constraint of the particular file is added or modified based on the permission-instruction data set. Based on the modified or added permission constraint, it is determined whether and/or an extent to which the user request is authorized. A response to the user request based on the determination as to whether and/or an extent to which the user request is authorized.

Abstract: A bill verification system for verifying bill records associated with an entity account is disclosed. The system receives a request to verify a bill record associated with an entity account. The system identifies a bill level model to be used for verifying the bill record and detects based on the bill level model, bill level anomaly information for the bill record. The system additionally identifies a bill line level model to be used for verifying one or more bill lines in the bill record and detects based on the bill line level model, bill line level anomaly information for the bill record. The system then aggregates the bill level anomaly information and the bill line level anomaly information to generate a bill verification report for the bill record. The system provides the bill verification report as a response to the request received to verify the bill record.

Abstract: Techniques for selecting and training candidate time-series models to forecast computational workloads are disclosed. A system creates a candidate set of time-series models for forecasting computing workloads by filtering the sets of parameter values to a number that meets system performance specifications. The system selects different sets of parameter values for different candidate models based on analyzing correlogram data. The system identifies in the correlogram data a set of one or more correlation values that (a) meet or exceed a threshold value, and (b) meet a distance criteria from the threshold value. The system trains the candidate set of time-series models with a training data set. The system selects the best-performing time-series model to generate forecasts for a particular computing resource in a computing system.

Abstract: A Representational State Transfer (REST)-Simple Message Transfer Protocol (SMTP) protocol gateway (RSPG) is disclosed that includes capabilities for accessing and processing REST Application Programming Interfaces (APIs) using SMTP. The RSPG receives a first email message, extracts information from the received first email message and constructs a REST API call based on the extracted information. The RSPG invokes the REST API call against a REST endpoint and receives a response generated from execution of the REST API call. The RSPG generates a second email message based on the response and causes the second email message to be communicated to an intended recipient of the response of the REST API call.

Abstract: Embodiments implement a secure connector framework at a cloud infrastructure. Embodiments receive one or more notebook profiles from an on-premises system corresponding to a first cloud customer, the on-premises system comprising at least one of one or more datasets, one or more models, or one or more libraries, the notebook profiles comprising permission sets that specify a level of access to the datasets, the models and the libraries, the notebook profiles corresponding to an on-premises machine learning (“ML”) notebook. Embodiments transform the received notebook profiles into a cloud policy set for sharing the datasets, the models and the libraries. Embodiments then transmit and receive corresponding data from the first cloud customer to a second cloud customer, the transmitted and received data based on the cloud policy set.

Abstract: A method may include obtaining, for an application, application dependency specifications, identifying vulnerable components using the application dependency specifications and a list of known vulnerable components, selecting, for a vulnerable component, candidate dependency specifications each specifying a version ID for a component, selecting, for a candidate dependency specification, an upgraded version ID for a component, verifying, using an application dependency graph generated from the application dependency specifications, that upgrading the candidate dependency specification to the upgraded version ID removes a dependency on the vulnerable component, and recommending, for the application, an upgrade solution including upgrading the candidate dependency specification to the upgraded version ID.

Abstract: A data catalog system is described that includes capabilities for automatically identifying and classifying sensitive information stored in data objects associated with various data sources. The data catalog system identifies a data object associated with a data asset stored in a data catalog metadata repository and computes a sensitivity score for the data object based on a set of one or more sensitive data identification techniques. The system determines a set of enrichment labels for the data object based on the sensitivity score computed for the data object. The enrichment labels are used to further qualify, enrich, or classify the data objects identified as containing sensitive information. For instance, the enrichment labels may identify a set of custom properties to be assigned to a data object, identify glossary terms to be applied to the data object or the enrichment labels may identify tags to be assigned to the data object.

Abstract: Techniques for preprocessing data assets to be used in a natural language to logical form model based on scalable search and content-based schema linking. In one particular aspect, a method includes accessing an utterance, classifying named entities within the utterance into predefined classes, searching value lists within the database schema using tokens from the utterance to identify and output value matches including: (i) any value within the value lists that matches a token from the utterance and (ii) any attribute associated with a matching value, generating a data structure by organizing and storing: (i) each of the named entities and an assigned class for each of the named entities, (ii) each of the value matches and the token matching each of the value matches, and (iii) the utterance, in a predefined format for the data structure, and outputting the data structure.

Abstract: Techniques are disclosed herein for synthesizing synthetic training data to facilitate training a natural language to logical form model. In one aspect, training data can be synthesized from original under a framework based on templates and a synchronous context-free grammar. In one aspect, training data can be synthesized under a framework based on a probabilistic context-free grammar and a translator. In one aspect, training data can be synthesized under a framework based on tree-to-string translation. In one aspect, the synthetic training data can be combined with original training data in order to train a machine learning model to translate an utterance to a logical form.

Abstract: Techniques for configuring one or more applications based on a detected wakefulness state of a user are disclosed. A system trains and applies a machine learning model to wakefulness data to compute a wakefulness state of a user. The system obtains the wakefulness data from wearable devices worn by the user and environmental devices in a user's environment. The system configures applications and/or devices based on the computed wakefulness state of the user. The system configures the ability of devices or applications to generate visual, audible, or tactile notifications in response to determining that a user is awake or asleep.

Abstract: Techniques are disclosed herein for synthesizing synthetic training data to facilitate training a natural language to logical form model. In one aspect, training data can be synthesized from original under a framework based on templates and a synchronous context-free grammar. In one aspect, training data can be synthesized under a framework based on a probabilistic context-free grammar and a translator. In one aspect, training data can be synthesized under a framework based on tree-to-string translation. In one aspect, the synthetic training data can be combined with original training data in order to train a machine learning model to translate an utterance to a logical form.

Abstract: Techniques are disclosed for training a model, using multi-task learning, to transform natural language to a logical form. In one particular aspect, a method includes accessing a first set of utterances that have non-follow-up utterances and a second set of utterances that have initial utterances and associated one or more follow-up utterances and training a model for translating an utterance to a logical form. The training is a joint training process that includes calculating a first loss for a first semantic parsing task based on one or more non-follow-up utterances from the first set of utterances, calculating a second loss for a second semantic parsing task based on one or more initial utterances and associated one or more follow-up utterances from the second set of utterances, combining the first and second losses to obtain a final loss, and updating model parameters of the model based on the final loss.

Abstract: Described herein are dialog systems, and techniques for providing such dialog systems, that are suitable for use on standalone computing devices. In some embodiments, a dialog system includes a dialog manager, which takes as input an input logical form, which may be a representation of user input. The dialog manager may include a dialog state tracker, an execution subsystem, a dialog policy subsystem, and a context stack. The dialog state tracker may generate an intermediate logical form from the input logical form combined with a context from the context stack. The context stack may maintain a history of a current dialog, and thus, the intermediate logical form may include contextual information potentially missing from the input logical form. The execution subsystem may execute the intermediate logical form to produce an execution result, and the dialog policy subsystem may generate an output logical form based on the execution result.