Abstract: Techniques are disclosed herein for adaptive training data augmentation to facilitate training named entity recognition (NER) models. Adaptive augmentation techniques are disclosed herein that take into consideration the distribution of different entity types within training data. The adaptive augmentation techniques generate adaptive numbers of augmented examples (e.g., utterances) based on the distribution of entities to make sure enough numbers of examples for minority class entities are generated during augmentation of the training data.

Abstract: Systems and methods identify whether an input utterance is suitable for providing to a machine learning model configured to generate a query for a database. Techniques include generating an input string by concatenating a natural language utterance with a database schema representation for a database; providing the input string to a first machine learning model; based on the input string, generating, by the first machine learning model, a score indicating whether the natural language utterance is translatable to a database query for the database and should be routed to a second machine learning model, the second machine learning model configured to generate a query for the database based on the natural language utterance; comparing the score to a threshold value; and responsive to determining that the score exceeds the threshold value, providing the natural language utterance or the input string to the second machine learning model.

Abstract: Techniques are disclosed herein for converting a natural language utterance to an intermediate database query representation. An input string is generated by concatenating a natural language utterance with a database schema representation for a database. Based on the input string, a first encoder generates one or more embeddings of the natural language utterance and the database schema representation. A second encoder encodes relations between elements in the database schema representation and words in the natural language utterance based on the one or more embeddings. A grammar-based decoder generates an intermediate database query representation based on the encoded relations and the one or more embeddings. Based on the intermediate database query representation and an interface specification, a database query is generated in a database query language.

Abstract: A system for analyzing security threat changes of proposed changes to an infrastructure environment. For example, system and approaches for determining actions to be performed based on security threat changes corresponding to proposed changes to the infrastructure environment is disclosed.

Abstract: Techniques for generating formatting tags for textual content obtained from a source electronic document are disclosed. A system parses a digital file to obtain information about characters in an electronic document. The system applies tags to text generated based on the textual content of the electronic document by creating segments of textually-consecutive characters and applying corresponding text formatting style tags to the segments. The system further identifies segments of text overlapping bounding boxes in the electronic document. The system generates textual content including a segment of text and a corresponding hyperlink associated with the segment of text. The system further generates textual content by selectively applying line breaks from the source electronic document in the textual content.

Abstract: Techniques are disclosed herein for calibrating confidence scores of a machine learning model trained to translate natural language to a meaning representation language. The techniques include obtaining one or more raw beam scores generated from one or more beam levels of a decoder of a machine learning model trained to translate natural language to a logical form, where each of the one or more raw beam scores is a conditional probability of a sub-tree determined by a heuristic search algorithm of the decoder at one of the one or more beam levels, classifying, by a calibration model, a logical form output by the machine learning model as correct or incorrect based on the one or more raw beam scores, and providing the logical form with a confidence score that is determined based on the classifying of the logical form.

Abstract: Techniques for automatically preventing execution plan regressions are provided. In one technique, in a first user database session, in response to receiving a first database statement, a first execution plan is generated and, while executing the first execution plan, first performance data that indicates one or more first performance metrics of executing the first execution plan is recorded. In response to receiving a second database statement, where the first execution plan may be used to generate a result for the second database statement, a second execution plan is generated and second performance data that indicates one or more second performance metrics of executing the second execution plan is recorded. A comparison between the first performance data and the second performance data is performed. Based on the comparison, it is determined whether the second execution plan will be stored for future use to process a database statement.

Abstract: Upon receiving a query at a local database server from a client directed to a LOB stored at a remote database server, a remote LOB locator is created comprising a local LOB locator indicating the LOB's location and a database link associated with the remote database server. The remote LOB locator is communicated to the client, which submits a request to the local database server to perform an operation directed to the LOB. The request is forwarded to the remote database server using the remote LOB locator, where the operation is performed. This mechanism is extended to work with sharded databases. Character set conversions are minimized when returning data to the client from the remote database server. Remote LOB locators may serve as bind variables for DML statements directed to local or remote tables. Local LOB locators may serve as bind variables for DML statements directed to remote tables.

Abstract: Systems, methods, and other embodiments associated with performing maintenance of a hardware node and relocating (live migrating) virtual machine (VM) instances in a multi-tenant, hyper scale, public Cloud service are described. In one embodiment, the method includes finding a suitable hardware node for hosting one or more VM instances from a hardware node selected for maintenance, preparing it for hosting VMs, and moving (live migrating) all virtual machines from the hardware node to the hosting hardware node.

Abstract: Example embodiments facilitate creation of efficient tree structures (and strategic modification of and manipulation of existing tree structures), such that the resulting tree structures enable rapid computing operations, including, but not limited to, rapid relational set operations (e.g., joining trees, comparing tree structures, determining structural similarities of trees, adding tree nodes, removing or replacing tree nodes, accessing tree data, updating tree data, and so on). An example method relates to a method and/or system for creating and manipulating tree data structures (also simply called trees herein) as, for example, in relational databases. One embodiment uses compact bit-wise path encoding that stores structural data related to tree branches to which a current node belongs.

Abstract: Disclosed herein are techniques for using a generative adversarial network (GAN) to train a semantic parser of a dialog system. A method described herein involves accessing seed data that includes seed tuples. Each seed tuple includes a respective seed utterance and a respective seed logical form corresponding to the respective seed utterance. The method further includes training a semantic parser and a discriminator in a GAN. The semantic parser learns to map utterances to logical forms based on output from the discriminator, and the discriminator learns to recognize authentic logical forms based on output from the semantic parser. The semantic parser may then be integrated into a dialog system.

Abstract: A method for providing updated network slice information to a network slice selection function (NSSF includes registering, by a network slice management function (NSMF) with a network function (NF) repository function (NRF), an NF profile corresponding to the NSMF and subscribing, by the NSMF with the NRF, for status updates corresponding to NF instances belonging to network slice instances created by the NSMF. The method further includes receiving, by the NSMF from the NRF, a notification message including one or more network traffic load level updates related to at least one of the NF instances, processing the one or more network traffic load level updates to generate network slice instance configuration information for at least one of the network slice instances, and providing, by the NSMF, the network slice instance configuration information to a NSSF managing the at least one of the network slice instances.

Abstract: A feedback button is configured to receive a feedback expression characterizing a user experience and submit information in a user interface with a single action. When the user hovers an input cursor near the feedback button, icons/text may be displayed that indicate different types of feedback expressions that may be submitted. As the user moves the cursor back and forth over the button, the icons for the different feedback expressions may be emphasized indicating locations where the button may be clicked to submit specific feedback. When the user is satisfied with the displayed feedback expression, the user may click the feedback button at the current location to submit the feedback expression and perform the final submit action for other data in the user interface.

Abstract: In an embodiment, multiple computers cooperate to retrieve content from tables in a relational database. Each table contains respective rows. Each row contains a vertex of a graph. Many high-degree vertices are identified. Each high-degree vertex is connected to respective edges in the graph. A count of the edges of each high-degree vertex exceeds a degree threshold. A central computer detects that all vertices in a high-degree subset of tables are high-degree vertices. Based on detecting the high-degree subset of tables, multiple vertices of the graph that are not in the high-degree subset of tables are replicated. Within local storage capacity limits of the computers, this degree-based replication may be supplemented with other vertex replication strategies that are schema based, content based, or workload based. This intelligent selective replication maximizes system throughput by minimizing graph data access latency based on data locality.

Abstract: A computer measures for each column in many rows, a respective frequency of statements that filter the column in a workload of database statements, a respective count of distinct values used for filtration on the column in each statement individually, a respective frequency of each of the counts of distinct values used for filtration on the column across all of the database statements, and a respective value range of the column for each of many storage zones. A respective efficiency is measured for each of many distinct interleaved sorts. Each interleaved sort uses a respective distinct subset of the columns. Each interleaved sort is based on portions of each of the values for each row in a sampled subset of rows in each column of the subset of the columns of the interleaved sort. Efficiency measurement is based on frequencies of statements, value ranges of columns for each storage zone, and frequencies of counts of distinct values.

Abstract: Techniques are described for communications in an L2 virtual network. In an example, the L2 virtual network includes a plurality of L2 compute instances hosted on a set of host machines and a plurality of L2 virtual network interfaces and L2 virtual switches hosted on a set of network virtualization devices. An L2 virtual network interface emulates an L2 port of the L2 virtual network. Access control list (ACL) information applicable to the L2 port is sent to a network virtualization device that hosts the L2 virtual network interface.

Abstract: A system can host APIs for a plurality of different tenants and receive requests from many different client devices. As requests are received, an associated tenant can be identified, and a router can determine if a container instance is available to service the request. A container instance may be an empty container instance including an internal endpoint, a Web server, and a runtime environment. An empty container instance can be unassociated with a particular tenant. To associate a container instance with a tenant, a data store, such as a key-value data store can retrieve configuration files that turn the agnostic container instance into a container instance that is associated with particular tenant and includes configuration code to perform the requisite API functions. The pool of empty and populated containers can be managed efficiently.

Abstract: A manager file system (MFS) runs as a user space file system. The MFS, implemented using an OS process, exposes a mount point as a communication endpoint to the single process. Mounting, unmounting, and changing configuration of individual database file systems (DBFSs) are done by overloading extended attributes on the mount point. The MFS services all DBFSs mounted at different mount points registered to the single process of the MFS and ensures optimal resource utilization among the DBFSs in the single process while guaranteeing resource isolation.

Abstract: Techniques are described for executing machine learning models trained for specific operators with feature values that are based on the actual execution of a workload set. The machine learning models generate an estimate of benefit gain/cost for executing operations on data portions in the alternative encoding format. Such data potions may be sorted based on the estimated benefit, in an embodiment. Using cost estimation machine learning models for memory space, the data portions with the most benefits that comply with the existing memory space constraints are recommended and/or are automatically encoded into the alternative encoding format.

Abstract: Event processing techniques for updating a database in real time based on events in a continuous event stream are disclosed. The techniques can update the database to incorporate information from thousands of received events per second. The events can include metrics measuring milestones for an organizational process defined by a user. Moreover, multiple streams can include metrics from many tenants concurrently. The techniques include receiving a plurality of events corresponding to a time interval and identifying that two events share a value of a key attribute. The techniques then include causing to be generated an in-memory combined database record comprising a combination of first and second attribute information from the two events and the value of the key attribute. In response to a trigger signal, the techniques include transmitting a single message, based at least in part on the combined database record, to update the database.