Abstract: A data object has a lock and a condition indicator associated with it. Based at least partly on detecting a first setting of the condition indicator, a reader stores an indication that the reader has obtained read access to the data object in an element of a readers structure and reads the data object without acquiring the lock. A writer detects the first setting and replaces it with a second setting, indicating that the lock is to be acquired by readers before reading the data object. Prior to performing a write on the data object, the writer verifies that one or more elements of the readers structure have been cleared.

Abstract: Techniques for identity management, and more particularly, to techniques for dynamically assigning membership to users in the system based on dynamic rules. In one aspect a computer-implement method is provided that breaks down the processing from a single large thread or operation into multiple minutest level threads or operations and makes use of event driven architecture used in distributed environments such as a cloud environment, to achieve a scalable model and can work seamlessly for multi-tenant applications. Every sub problem is assigned to a dedicated set of subscribers on a messaging service for processing.

Abstract: A manager file system (MFS), implemented using an OS process, exposes a path in the native file system as the root of its file system (e.g., mount point). The MFS informs the OS kernel that any file system operation whose path contains the mount point as its prefix, should be handed over to the process by invoking an appropriate interface method that the process implements as part of its file system implementation. The MFS manages one or more applications and leverages the authentication mechanism of the OS to provide fined-grained access control of MFS managed applications.

Abstract: Methods, systems, and computer readable media for validating subscriber entities against spoofing attacks in a communications network are disclosed. One method includes receiving, by a proxy element, a subscriber authentication request message that is sent from an access and mobility management function (AMF) and is directed to a unified data management (UDM) function in a home network, creating, by the proxy element, a record entry including a subscriber identifier and a public land mobile network (PLMN) identifier contained in the subscriber authentication request message, wherein the record entry is stored in an authentication registry database prior to forwarding the subscriber authentication request message to the UDM function.

Abstract: In accordance with an embodiment, described herein is system and method for use with software application development environments, for determining attribution associated with licensed software code. The system can receive as input an indication of a source application codebase for a software application, determine dependencies on third-party modules, libraries, or other software code, and output associated license attributions for that codebase. Optionally, the system can combine and/or de-duplicate multiple sets of attributions associated with a source application codebase, or generate a set of differences between two versions of attributions, to illustrate changes between the versions. In accordance with an embodiment, if the system detects potential licensing issues associated with the use of third-party modules, libraries, or other software code, then an appropriate notification can be generated.

Abstract: A shared-nothing database system is provided in which parallelism and workload balancing are increased by assigning the rows of each table to “slices”, and storing multiple copies (“duplicas”) of each slice across the persistent storage of multiple nodes of the shared-nothing database system. When the data for a table is distributed among the nodes of a shared-nothing system in this manner, requests to read data from a particular row of the table may be handled by any node that stores a duplica of the slice to which the row is assigned. For each slice, a single duplica of the slice is designated as the “primary duplica”. All DML operations (e.g. inserts, deletes, updates, etc.) that target a particular row of the table are performed by the node that has the primary duplica of the slice to which the particular row is assigned. The changes made by the DML operations are then propagated from the primary duplica to the other duplicas (“secondary duplicas”) of the same slice.

Abstract: Techniques for using enhanced logit values for classifying utterances and messages input to chatbot systems in natural language processing. A method can include a chatbot system receiving an utterance generated by a user interacting with the chatbot system and inputting the utterance into a machine-learning model including a series of network layers. A final network layer of the series of network layers can include a logit function. The machine-learning model can map a first probability for a resolvable class to a first logit value using the logit function. The machine-learning model can map a second probability for a unresolvable class to an enhanced logit value. The method can also include the chatbot system classifying the utterance as the resolvable class or the unresolvable class based on the first logit value and the enhanced logit value.

Abstract: Techniques discussed herein relate to implementing a distributed computing cluster (the “cluster”) including a plurality of edge devices (e.g., devices individually configured to selectively execute within an isolated computing environment). One edge device may be configured to operate as a head node of the cluster at a given time. A request for virtual resources of the cluster may be received from a user device and directed to the first edge device of the cluster. The first edge device may determine it is not operating as a head node of the cluster. The first edge device may determine that a second edge device of the cluster is operating as the head node. In response, the first edge device may forward the request to the second edge device, wherein forwarding the request to the second edge device causes the second request to be processed by the cluster.

Abstract: Client data corresponding to a single client can be stored. A decision to proceed with a defragmentation of two or more particular core groups can be determined. A destination core group to receive client data can be identified. At least one additional core group can be initiated to manage interim data write requests from the single client. Each of the two or more particular core groups can be closed to data write requests. A defragmentation of the two or more particular core groups can be performed by merging into the destination core group. The destination core group can be availed to data read and delete requests. The two or more fragmented core groups can be reallocated for other uses. The additional core groups continue to receive read and write requests. The destination core groups further can be merged and defragmented further.

Abstract: Techniques for managing lifecycles of sets of foreign resources are disclosed, including: opening, in a runtime environment configured to use a garbage collector to manage memory in a heap, a memory session; allocating a first subset of a set of foreign memory to a memory segment associated with the memory session, the foreign memory including off-heap memory that is not managed by the garbage collector; encountering, in the runtime environment, an instruction to close the memory session; responsive to encountering the instruction to close the memory session, deallocating the subset of the set of foreign memory.

Abstract: Techniques discussed herein relate to managing aspects of provisioning cloud-computing edge devices. In some embodiments, a user request specifying one or more resources to be provisioned at a cloud-computing edge device may be received by a cloud-computing device operated by a cloud-computing provider. An ephemeral credential may be generated for the edge device and used to determine whether access control policies exist that enable the edge device to access the resource(s). If the resource(s) are accessible, a manifest may be generated in accordance with the user request. The manifest may specify a configuration for the cloud-computing edge device and includes information related to the resource. Operations may be subsequently executed to provision the cloud-computing edge device in accordance with the manifest. However, if the resource(s) are not accessible, the user request may be denied.

Abstract: A parametric constant resolves to different values in different contexts, but a single value within a particular context. An anchor constant is a parametric constant that allows for a degree of parametricity for an API point. The context for the anchor constant is provided by a caller to the API point. The anchor constant resolves to an anchor value that records specialization decisions for the API point within the provided context. Specialization decisions may include type restrictions, memory layout, and/or memory size. The anchor value together with an unspecialized type of the API point result in a specialized type of the API point. A class object representing the specialized type is created. The class object may be accessible to the caller, but the full value of the anchor value is not accessible to the caller. The API point is executed based on the specialization decisions embodied in the anchor value.

Abstract: Techniques for modifying queries in a set of nested queries are disclosed. A graphical user interface displays a query detail region alongside a nested query display region. The graphical user interface includes functionality to provide for modification of queries in the nested set of queries. Based on a selection by a user, a query modification tool promotes a query attribute from a child query to one or more parent queries. Based on another selection by a user associated with one query in the set of nested queries, the system deletes an attribute from each query in the set of nested queries. Responsive to a selection to create multiple conditions for a query rule, the system modifies the functionality of the user interface to enable entry of multiple condition characteristics. Based on a further selection, the system creates the multiple conditions for the query rule.

Abstract: A continuous query identified by an archived relation of an application may be received. A physical query plan may be generated for the continuous query and/or a child operator plan may be identified. A buffer operator may be inserted between the child operator and its parent operator and/or the event information from the child operator may be stored in the buffer operator. A query operator may be identified in the physical query plan. It may be determined that the query operator identified in the physical query plan is a same type as a second query operator in a second continuous query concurrently executing. Further, a combined query plan may be generated based at least in part on the query operator identified in the physical query plan being the same type as the second query operator in the second continuous query.

Abstract: Disclosed is a system, method, and computer program product for implementing a log analytics method and system that can configure, collect, and analyze log records in an efficient manner. Machine learning-based classification can be performed to classify logs. This approach is used to group logs automatically using a machine learning infrastructure.

Abstract: Techniques for noise data augmentation for training chatbot systems in natural language processing. In one particular aspect, a method is provided that includes receiving a training set of utterances for training an intent classifier to identify one or more intents for one or more utterances; augmenting the training set of utterances with noise text to generate an augmented training set of utterances; and training the intent classifier using the augmented training set of utterances. The augmenting includes: obtaining the noise text from a list of words, a text corpus, a publication, a dictionary, or any combination thereof irrelevant of original text within the utterances of the training set of utterances, and incorporating the noise text within the utterances relative to the original text in the utterances of the training set of utterances at a predefined augmentation ratio to generate augmented utterances.

Abstract: Techniques are described for a hierarchical caching mechanism enabling efficient cross-region replications. In some embodiments, replication-related information (e.g., key-value pairs) is stored in a particular layout in a binary tree (B-tree) of a file system for replication processing. A hierarchy of caches storing a first type of information (e.g., crypto keys associated with iNodes) may be arranged to match the particular layout in the B-tree to enable efficient parallel processing of a second type of information (e.g., files, file data, or symbolic links), where the replication-related information in the B-tree is partitioned into multiple key ranges for parallel processing. In some embodiments, the caches in different hierarchies may be shared by different parallel-processing key ranges and replication jobs in a file system.

Abstract: Techniques for perspective-preserving seamless application switching are disclosed. A system may display a first interface using a first application. The first interface includes interface elements representing a plurality of objects. The system may detect a zoom-in command, received by the first application, requesting a particular zoom level for a first interface element, corresponding to a first object in the first plurality of objects. The system may determine that the requested zoom level exceeds a threshold. Responsive to determining that the requested zoom level exceeds the threshold, the system may display, using a second application, a second interface corresponding to the first object. The second interface may include one or more of: (a) characteristics associated with the first object that were not displayed by the first application, or (b) user input elements for executing operations associated with the first object that were not displayed by the first application.

Abstract: A computer-implemented method includes: accessing a plurality of datasets, where each dataset of the plurality of datasets includes training examples; selecting datasets that include the training examples in a source language and a target language; and sampling, based on a sampling weight that is determined for each of the selected datasets, the training examples from the selected datasets to generate the training batches; training an ML model for performing at least a first task using the training examples of the training batches, by interleavingly inputting the training batches to the ML model; and outputting the trained ML model configured to perform the at least the first task on input utterances provided in at least one among the source language and the target language. The sampling weight is determined for each of the selected datasets based on one or more attributes common to the training examples of the selected dataset.

Abstract: Approaches of reassigning a home region from a first data center to a second data center as requested by a customer are described herein. The home region is able to implement write operations to a domain, whereas other data centers cannot implement write operations to the domain. The customer can request the home region being reassigned to another data center such that the customer can utilize the other data center to implement write operations to the domain.

Abstract: The present disclosure is related to techniques for converting a natural language utterance to a logical form query and deriving a natural language interpretation of the logical form query. The techniques include accessing a Meaning Resource Language (MRL) query and converting the MRL query into a MRL structure including logical form statements. The converting includes extracting operations and associated attributes from the MRL query and generating the logical form statements from the operations and associated attributes. The techniques further include translating each of the logical form statements into a natural language expression based on a grammar data structure that includes a set of rules for translating logical form statements into corresponding natural language expressions, combining the natural language expressions into a single natural language expression, and providing the single natural language expression as an interpretation of the natural language utterance.

Abstract: Systems and methods for reducing latency of probing operations of remotely located linear hash tables are described herein. In an embodiment, a system receives a request to perform a probing operation on a remotely located linear hash table based on a key value. Prior to performing the probing operation, the system dynamically predicts a number of slots for a single read of the linear hash table to minimize total cost for an average probing operation. The system determines a hash value based on the key value and determines a slot of the linear hash table to which the hash value corresponds. After predicting the number of slots, the system issues an RDMA request to perform a read of the predicted number of slots from the linear hash table starting at the slot to which the hash value corresponds.

Abstract: Systems and methods for implementing extensible attributes in ETL are disclosed. In some examples, attributes configured at a source file may be extracted from the source file. The extracted attributes can be mapped to a target column of a data warehouse table, and then a dynamic ETL script may be generated. The dynamic script may be executed to move data associated with the attributes to an appropriate new column of the data warehouse.

Abstract: In accordance with an embodiment, described herein are systems and methods for use with an analytic applications environment, for determination of recommendations and alerts in such environments. A data pipeline or process can operate in accordance with an analytic applications schema adapted to address particular analytics use cases or best practices, to receive data from a customer's (tenant's) enterprise software application or data environment, for loading into a data warehouse instance. When provided as part of a software-as-a-service (SaaS) or cloud environment, the data sourced from a plurality of organizations can be aggregated, to leverage information gleaned from the collective or shared data. The system can be used to generate semantic alerts, including obtaining permission from; and analyzing the collective data of; the plurality of organizations, to determine operational advantages indicated by the data, and providing alerts associated with those operational advantages.

Abstract: The present invention relates to machine learning (ML) explainability (MLX). Herein are local explanation techniques for black box ML models based on coalitions of features in a dataset. In an embodiment, a computer receives a request to generate a local explanation of which coalitions of features caused an anomaly detector to detect an anomaly. During unsupervised generation of a new coalition, a first feature is randomly selected from features in a dataset. Which additional features in the dataset can join the coalition, because they have mutual information with the first feature that exceeds a threshold, is detected. For each feature that is not in the coalition, values of the feature are permuted in imperfect copies of original tuples in the dataset. An average anomaly score of the imperfect copies is measured. Based on the average anomaly score of the imperfect copies, a local explanation is generated that references (e.g. defines) the coalition.

Abstract: For a communication channel having a first endpoint in a customer on-premise network and a second endpoint on a primary host machine in a cloud service provider infrastructure, the primary host machine determines a change in a state information of the communication channel and identifies a backup host machine for the communication channel. The primary host machine causes the change in the state information to be replicated to the backup host machine, wherein the replicated state information stored by the backup host machine is usable by the backup host machine after a failover causes the backup host machine to become the second endpoint of the communication channel.

Abstract: Systems and methods for using queue pair 1 (QP1) for receiving multicast based announcements in multiple partitions in a high performance computing. In accordance with an embodiment, by extending the scope of QP1 to also include receiving and sending multicast packets in any partition defined for the port, it is possible to implement generic MC based announcement and discovery without requiring the complexity of unique QPs for individual partitions, nor any update of QP configuration as a consequence of change of partition membership.

Abstract: The present embodiments relate to implementing change data on no-master NoSQL data stores. An optimized node can be identified from a plurality of NoSQL data storage nodes and a specialized node can be connected (e.g., collocated) to the optimized node. The specialized node can maintain change data capture (CDC) data provided by client nodes in a hash map that can be used as a point of truth for coordinating CDC data across the plurality of NoSQL data storage nodes. The plurality of NoSQL data storage nodes can identify and coordinate all read/write data obtained from multiple client devices in a geographically separated large-scale (e.g., planet scale) system to identify change data in a distributed data store. The specialized data can provide read data to devices in the large-scale system to reconcile inconsistencies in change data across nodes in the large-scale system.

Abstract: A data catalog system is disclosed that provides capabilities for uniquely identifying and retrieving data entities stored in diverse data sources managed by an organization. The data catalog system includes capabilities for generating a unique external identifier for a data entity (e.g., a data asset or a data object) by identifying a set of immutable configuration parameters associated with the data asset and identifying a set of data object attributes that uniquely identify data objects within the data asset. The generated unique external identifiers are stored as part of the metadata harvested by the data catalog system. The external identifiers are used to enforce a single representation of the data assets and the data objects in the data catalog system. The external object identifiers are used to perform data lookups and reconcile states of data entities during the metadata harvesting process.

Abstract: Techniques are disclosed for applying a trained machine learning model to incoming voice communications to determine whether the voice communications are genuine or not genuine. The trained machine learning model may identify vocal attributes within the target call and use the identified attributes, and the training, determine whether the target call is genuine or not genuine. An applied trained machine learning model may include multiple different types of trained machine learning models, where each of different types of machine learning models is trained and/or configured for a different function within the analysis.

Abstract: A distributed computing system is described that leverages a nearline storage layer to minimize the downtime required for bootstrapping a new computing cluster in the distributed computing system. The system executes a computing cluster comprising a set of computing nodes and determines a set of one or more data segments to be written to a nearline storage system. The system writes the data segments to the nearline storage system. In certain examples, the system receives a request to create a second computing cluster and responsive to the request, bootstraps the second computing cluster using the set of data segments stored on the nearline storage system. The system additionally leverages the nearline storage layer to accelerate query processing by the computing nodes of a computing cluster.

Abstract: A type restriction contextually modifies an existing type descriptor. The type restriction is imposed on a data structure to restrict the values that are assumable by the data structure. The type restriction does not cancel or otherwise override the effect of the existing type descriptor on the data structure. Rather the type restriction may declare that a value of the data structure's type is forbidden for the data structure. Additionally or alternatively, the type restriction may declare that an element count allowable for a data structure's type is forbidden for the data structure. Type restriction allows optionality (where only a singleton value for a data structure is allowed), empty sets (where no value for a data structure is allowed), and multiplicity (where only a limited element count for a data structure) to be injected into a code set independent of data type. Type restriction allows certain optimizations to be performed.

Abstract: In accordance with an embodiment, described herein is a system and method for loading and transforming data to a cloud database, using a hybrid set-based extract, load, transform approach. During the loading and transforming of large amounts of data, from one or more data sources, a data synchronization application sends small batches of data, in parallel streams, to the database, which operates as a transformation engine to load the data. Each stream can be associated with a separate staging table in the database. For each staging table, the subset of data therein is transformed and loaded, before the next subset is processed. A transform and merge process operating at the database can then be used to transform and merge the data, from each of the staging tables, to the target table.

Abstract: Systems, methods, and other embodiments for sorting-based assignment to optimize order fulfillment with short supply. One embodiment includes identifying a set of one or more distributions within a series of days. Sorting all demands of a first priority level occurring in the series of days by ascending order of size of the demand. For each demand of the first priority level in sorted order, (a) selecting an initial distribution during which the demand is scheduled to be fulfilled from the set of distributions, and (b) generating an indication that the demand can cannot be completely fulfilled, based on the size of the demand and the amount of supply available. Automatically arranging a graphical representation of a schedule of supplies and met and unmet demands for the series of days in a graphical user interface based at least on the indications generated for each demand.

Abstract: Techniques for processing top-K queries are provided. In one technique, a database statement is received that requests top-K results related to a database object and that indicates two columns thereof: a first column by which to partition a result set and a second column by which to order the result set. A buffer is generated. For each of multiple rows in the database object: a first key value that associated with a first value in the first column of said each row is identified; a second key value that associated with a second value in the second column of said each entry is identified; a slot in the buffer is identified based on the first key value and the second key value; and the slot in the buffer may be updated based on the second key value. A response to the database statement is generated based on the buffer.

Abstract: Techniques are described for implementing a container environment where each pod within the container environment is provided with a unique IP address and a virtual communication device such as an IPvlan device. Communications from source pods are directly routed to destination pods within the container environment by one or more virtualized network interface cards (VNICs) utilizing the unique IP addresses of the destination pods, without the need for bridging and encapsulation. This reduces a size of data being transmitted and also eliminates a compute cost necessary to perform encapsulation of data during transmission.

Abstract: A computer-implemented method includes obtaining, from text corpus including article-summary pairs in a plurality of languages, a plurality of article-summary pairs in a target language among the plurality of languages, to form an article-summary pairs dataset in which each article corresponds to a summary; inputting articles from the article-summary pairs to a machine learning model; generating, by the machine learning model, embeddings for sentences of the articles; extracting, by the machine learning model, keywords from the articles with a probability that varies based on lengths of the sentences, respectively; outputting, by the machine learning model, the keywords; applying a maximal marginal relevance algorithm to the extracted keywords, to select relevant keywords; and generating a keyword-text pairs dataset that includes the relevant keywords and text from the articles, the text corresponding to the relevant keywords in each of keyword-text pairs of the keyword-text pairs dataset.

Abstract: Techniques for configuring an enriched data metrics pipeline (DMP) include: obtaining node configuration data associated with an unenriched DMP for nodes of a heterogeneous computing platform, including (a) a first kind of node including an application programming interface (API) for obtaining unenriched data associated with the first kind of node and (b) a second kind of node including an API for obtaining unenriched data associated with the second kind of node; the unenriched DMP being configured to provide node-level unenriched data to a service according to a first schedule; the service being configured to generate node-level data metrics based on the unenriched data; based on the node configuration data, configuring an enriched DMP to provide node-level enriched data to the service according to a second schedule that is less frequent than the first schedule; the service being configured to generate fleet-level data metrics based on the node-level enriched data.

Abstract: A method includes preparing a base model using an input model pretrained on at least three languages different from each other and a base vocabulary including words corresponding to two languages among the at least three languages, where the preparing the base model includes constraining the input model to the words included in the base vocabulary; training the base model using a first enhanced training dataset generated from public data, to generate a text summarization model; training the base model using a second enhanced training dataset generated from the first enhanced training dataset, to generate a text generation model; and training the base model using a third enhanced training dataset that is generated using the second enhanced training dataset and the text summarization model, to generate a next sentence generation model.

Abstract: Techniques are described for providing a multi-cloud control plane (MCCP) in a first cloud infrastructure (included in a first cloud environment provided by a first cloud services provider) that enables services and/or resources provided in the first cloud infrastructure to be utilized by users of a second cloud environment, where the second cloud environment is different than the first cloud environment. The multi-cloud infrastructure enables a user associated with an account with a second cloud services provider to use, from the second cloud infrastructure, a first service from the set of one or more cloud services. The multi-cloud infrastructure creates a link between the account with the second cloud service provider and a tenancy created in the first cloud infrastructure for enabling using the first service by the user.

Abstract: Techniques are described herein for analyzing and tuning database workloads to optimize application performance. In some embodiments, a workload analyzer identifies a captured workload that includes a set of database queries executed within a particular timeframe. The workload analyzer compares the workload within one or more other workloads executed within a previous timeframe to determine differences between the different workloads. For example, the workload analyzer may identify changes in the distributions of queries, including how many queries are unchanged, missing, and/or new. The workload analyzer may further detect changes in the performance of individual queries. The workload analyzer may determine the overall performance impact of such changes on the total workload. Based on the analysis, the workload analyzer may generate reports, alerts, tuning advice, and/or recommendations to boost performance.

Abstract: Techniques for managing task assignments to workers in a work environment are disclosed. A system identifies one or more workers with qualifications that match recommended qualifications to perform a task in a work environment. The system applies a trained machine learning model to task performance data associated with the worker, such as a past history of tasks performed and statistics associated with the performance of the task. The machine learning model generates a prediction of downstream effects associated with assigning the task to the user. The downstream effects include delays and performance improvements on subsequent tasks performed by the worker, as well as effects on tasks performed by other workers, at work centers in the work environment.

Abstract: Method includes populating fake value for each of entities, to generate string of fake entity values that correspond to entities; inserting sentinel token between adjacent fake values included in the string to generate first input data; generating, by natural language generation model, natural language sentences based on first input data, natural language sentences including one or more fake values from the string; analyzing natural language sentences to determine whether any fake value from the string is missing; based on the fake value missing, summarizing, using text summarization model, natural language sentences to generate text summary; concatenating the text summary with the fake value, to generate second input data; and generating, by a next sentence generation model, additional natural language sentence using the second input data, the additional natural language sentence including the fake value.

Abstract: Techniques for managing digital messages to and from a proxy message address are disclosed. A system receives a message directed to a particular destination address. The system replaces any source address included in the message with a proxy address. When the system receives a reply to the message, the reply is directed to the proxy address. The system analyzes message data to identify a target address for the reply message. The system identifies contextual data associated with the reply message. The system transmits the reply message, and the contextual data, to the target address.

Abstract: The present disclosure relates to a system and techniques for resolving dangling references resulting from a dependency relationship between computing resource objects uncovered during a harvesting process. In embodiments, a harvester application adds computing resource objects associated with a client to a resource collection as those computing resource objects are identified. Dependencies are identified as each computing resource object is added to the resource collection, which are resolved only if the computing resource objects associated with those dependencies have already been added to the resource collection. If the computing resource objects associated with the dependencies have not already been added to the resource collection, then the dependency is added to an observer pool. Observer modules are configured to check each computing resource object as it is processed during the harvest process in order to match those computing resource objects to unresolved dependencies.

Abstract: Techniques for implementing a qualification-based task management system in a work environment are disclosed. When a user logs in to a work center terminal, a system identifies a set of pending tasks that need to be completed. The system filters the tasks available to the user based on the user's qualifications and the equipment present at the work center. When the system identifies a task for which there is not a set of users with matching qualifications, the system applies a machine learning model to the task parameters to identify candidate users or recommended qualifications for performing the task.

Abstract: The present embodiments relate to data processing model recommendation and review of a portion of data using a recommended model. A model catalog executing on a cloud infrastructure (CI) system can parse data from an obtained dataset identifying aspects of the dataset. The parsed data from the dataset can be compared with a plurality of potential models stored in a domain ontology store of the model catalog to identify one or more recommended models. Review output data can be generated using the dataset and any of the recommended models. The review output data resulting from the recommended model can be provided to the client for the client to either accept or reject the model.

Abstract: An expense report generation system receives transaction authorization data from a card issuer and compares the data with expense report generation criteria to determine whether to generate an expense report, prior to settlement of the transaction, based on the authorization data. The expense report generation system evaluates additional data obtained from other data sources including contextual information of the employee, transaction authorization, location, and other employees to generate the expense report. The expense report generation system subsequently updates the generated expense report based on updated transaction authorization data and/or transaction settlement data. The expense report generation system trains and uses a machine learning model for efficiency and accuracy in generating expense reports from transaction authorization data while reducing an employee's burden in manually inputting expense information and the approval process burden.

Abstract: Techniques are disclosed herein for integrating document question answering in an artificial intelligence-based platform, such as a chatbot system. The techniques include receiving a query from a user, rewriting the query to include one or more specific descriptors, computing an embedding vector for the rewritten query, retrieving one or more textual passages from a document store utilizing the embedding vector for the rewritten query, determining one or more answers to the rewritten query within the one or more textual passages, and returning the one or more answers.

Abstract: In one or more embodiments, a software service allows software providers to implement machine learning (ML) features into products offered by the software providers. Each ML feature may be referred to as an encapsulated ML application, which may be defined and maintained in a central repository, while also being provisioned for each user of the software provider on an as-needed basis. Advantageously, embodiments allow for a central definition for an ML application that encapsulates data science and processing capabilities and routines of the software provider. This central ML application delivers a ML deployment pipeline template that may be replicated multiple times as separate, tailored runtime pipeline instances on a per-user basis. Each runtime pipeline instance accounts for differences in the specific data of each user, resulting in user-specific ML models and predictions based on the same central ML application.