Abstract: Various techniques are provided for enabling an iterator (called persistent key-value (KV) iterator) that uses a special key structure embedded with secondary index information for efficient search (or query) of a key-value database (KVDB). In some embodiments, the key structure comprises a plurality of fields categorized into two groups, infrastructure hierarchy fields and object characteristic fields. Keys for the KVDB may be partitioned into one or more levels of hierarchy to generate many smaller key lists for easy search/query of the KVDB. These smaller key lists can be used for concurrent searches/queries.

Abstract: Techniques are disclosed for improving automatic speech recognition (ASR) of code-switched speech, such as for Hindi-English, using two distinct approaches. The first approach combines limited, in-domain, code-switched data with monolingual data and implements on-the-fly code switch simulation during training. The second approach leverages language models (LMs) to generate domain-specific, code-switched utterances through few-shot prompting, followed by text-to-speech (TTS) synthesis to create training data. The first approach, using language-specific prompts and on-the-fly code mixing, achieves significant improvements with minimal in-domain data. The second approach, utilizing synthetic data generated through language model prompting and TTS, achieves a relative improvement over the pretrained model without using real, in-domain data. Both approaches are adaptable to other speech recognition systems and language pairs.

Abstract: The present disclosure relates to techniques for a dynamic and unified performance evaluation of a machine-learning (ML) pipeline, incorporating a configuration of ML models, based on a unified performance metric. The techniques involve accessing data comprising a set of data samples, where each data sample corresponds to one or more data types, including images, text, audio, or structured data elements. The data may be input into the ML pipeline, including ML models combined in the specific configuration (e.g., series, parallel), with model types corresponding to classification or regression. The decision boundary of each ML model is dynamically adjusted by evaluating relevant evaluation metrics. The ML pipeline may process data to generate a result based on the adjusted decision boundaries. The unified performance metric may be generated dynamically based on the model types, configuration, and their respective contributions.

Abstract: A system replicates source data from a source dataset stored in a source data repository to a destination dataset stored in a destination data repository. The system executes a sampling process that includes obtaining, from the source dataset, a source data sample corresponding to source data previously transmitted to a destination entity for replication of the source dataset. The system generates and transmits a control message to the destination entity that includes the source data sample. The destination entity receives the control message and executes a verification process that includes comparing the source data sample to the destination dataset to verify consistency between the source data sample and the destination dataset. The source dataset and the destination dataset are classified as consistent based at least in part on verifying consistency between the source data sample and the destination dataset.

Abstract: Techniques are disclosed for improving automatic speech recognition (ASR) of code-switched speech, such as for Hindi-English, using two distinct approaches. The first approach combines limited, in-domain, code-switched data with monolingual data and implements on-the-fly code switch simulation during training. The second approach leverages language models to generate domain-specific, code-switched utterances through few-shot prompting, followed by text-to-speech (TTS) synthesis to create training data. The first approach, using language-specific prompts and on-the-fly code mixing, achieves significant improvements with minimal in-domain data. The second approach, utilizing synthetic data generated through language model prompting and TTS, achieves a relative improvement over the pretrained model without using real, in-domain data. Both approaches are adaptable to other speech recognition systems and language pairs.

Abstract: Techniques are disclosed for a scalable message broker service in a publisher-subscriber (Pub-Sub) messaging system. A message broker service executing in a computing environment of a distributed computing system can obtain quality of service metrics of the Pub-Sub messaging system. The message broker service can include a plurality of message broker instances. The message broker service can determine, based at least in part on the quality of service metrics, a number of message broker instances to send messages to subscribing clients. The message broker service can then determine whether the number of message broker instances exceeds the plurality of message broker instances and execute the number of message broker instances.

Abstract: Disclosed is an approach to use an intermediate OS (operating system) user to prevent lateral movement of cloud service operators among databases that share a database binary. Some approaches use an on-demand secure communications channel to access a cloud-related resource that is associated with a customer. This creates on a temporary basis the infrastructure (including the intermediate OS user) that is needed to allow the operational access to the customer system, which can then be destroyed once it is no longer needed.

Abstract: A method of providing access to a hardware security module (HSM) partition may include receiving request for access to the HSM partition from a client device. The request may include a leaf certificate signed with a public key associated with a user and a secret key associated with the client device. The method may include verifying the request using the leaf certificate and a trust anchor certificate signed with a public key associated with the client device. The method may include a first connection between the HSM partition and the client device. The method may include verifying the request using the leaf certificate and an authentication certificate stored on the HSM partition. The method may include establishing a second connection between the client device and the HSM partition such that the computing system is isolated from the second connection.

Abstract: A computer program product, system, and computer implemented method for a framework for hierarchically allocating resources and monitoring resource usage of cloud database runtime entities. The approach leverages control groups mechanisms provided by underlying operating systems for managing resource allocations to processes and collecting metrics using control groups at a granularity that is conducive to efficient operation and management of modern database systems such as consolidated databases (CDBs). In some embodiments, the present approach provides the ability to manage the isolation of resource between groups by enforcing separation between resource such as cores, memory regions, and the like. Another way in which the present approach can improve management of CDB instances is by allowing for the database to manage creation of additional groups (e.g.

Abstract: Some embodiments relate to techniques including: accessing a prompt that is associated with a logical reasoning problem; generating one or more logical steps of a logical response based on the prompt by using a generative artificial intelligence (GenAI) model; for each of the one or more logical steps of the logical response: querying one or more theorem solvers using the logical response and the logical reasoning problem; receiving, in response to the query, one or more feedback from the one or more theorem solvers corresponding to the logical response of the logical reasoning problem; and updating the logical response based on the one or more feedback; determining that the logical response is complete and verified from the one or more theorem solvers; and outputting, in response to determining that the logical response is complete and verified, the logical response on a user interface.

Abstract: Techniques are described herein for merging multiple smart card application files into a single, consolidated file that may be used by a smart card runtime environment to execute multiple applications. The techniques may reduce the load size of installed application code on a smart card by bundling applications and libraries together into an optimized file. As a result, smart card platforms may have more space available to execute the applications at runtime and/or to install additional applications. Embodiments herein may further provide flexibility on defining access controls over resources for which the code is not known. When application files are merged, packages and libraries that were previously public may be made private within the merged application file to restrict external access to unknown code in the bundle.

Abstract: Techniques for multilingual batch alignment training and/or fine-tuning for language models are included herein. The system generates and/or accesses data points from batches of data having semantically equivalent content across languages and/or semantically equivalent prompts across languages. The prompts include queries related to a topic or a predictive language task. The data points include one or more sets of resources that are semantic equivalents in different languages. Data points for which a resource enables a correct response are used as positive training data points for reinforcement learning. Data points for which a resource does not enable a correct response are considered as negative training data points for reinforcement learning. The data points are used for alignment training and/or other reinforcement learning for generative language models.

Abstract: Disclosed are methods, systems, and articles of manufacture for implementing a standby database with a secure subsetting technique. One or more sensitivity criteria for a change may be determined in a primary database. A tagged redo record may be determined for the change, wherein the tagged redo record pertains to sensitive information in the change and the one or more sensitivity criteria. The change may be constructed on a standby database with data subsetting based at least in part upon the tagged redo record and a replication mode of the standby database.

Abstract: A method of generating natural language outputs may include accessing a model of a system, where the system may be represented by a hierarchy of nodes in a data structure, and nodes in the hierarchy of nodes may include time series of data. The method may also include identifying a time series represented by a node in the data structure that will generate a future anomaly; accessing a template corresponding to a type of the time series; populating semantic tags in the template using data from the time series; sending a phrase from the template to a natural language model; receiving a plurality of similar phrases from the natural language model; selecting one of the plurality of similar phrases and replacing the phrase in the template; and causing language from the template to be displayed on a display device.

Abstract: Techniques for natural language processing include accessing an input string comprising a natural language utterance and a database schema representation for a database; providing the natural language utterance to a first encoder to generate one or more embeddings of the natural language utterance; providing the database schema representation to the first encoder to generate one or more embeddings of the database schema representation; encoding, by a second encoder, relations between elements in the database schema representation and words in the natural language utterance based on the one or more embeddings of the natural language utterance and the one or more embeddings of the database schema representation; and generating a logical form for the natural language utterance based on the encoded relations, the one or more embeddings of the natural language utterance, and the one or more embeddings of the database schema representation.

Abstract: In accordance with an embodiment, described herein is a system and method for providing multi-hub and/or multi-table datasets with a computing environment such as, for example, a business intelligence environment, database, data warehouse, or other type of environment that supports data analytics. An analysis can be used to query data to provide information in the form of tables, graphs, pivot tables, or other data views. A hub table operates as fact table and carries the data metrics for analysis, enabling a user to join two tables, the data from which can be further transformed or enriched to prepare it for analysis.

Abstract: Techniques for monitoring the health of services of a system are disclosed. A system determines that a detected alarm is associated with a service feature, and the service feature is associated with a service of a cloud environment. The system computes a health metric for the service based at least on the detected alarm that is associated with the service feature. Additionally, the system generates a visual representation that includes the health metric for display on a service health interface.

Abstract: Systems, methods, and machine-readable media may place workloads of source systems in a migration of data and applications from the source systems to target systems. Data relating to a first number of source nodes and a second number of target nodes may be received and analyzed. A migration plan that specifies placement of workloads from source systems into target systems may be created. The placement may include packing a pluggable environment or a clustered environment. The packing the pluggable environment or the clustered environment may include: when the second number of target nodes is greater than the first number of source nodes, placing workload from the source nodes to the target nodes; and, when the second number of target nodes is less than the first number of source nodes, the workload from the source nodes is not placed to the target nodes.

Abstract: Described are improved systems, computer program products, and methods for efficiently accessing an ordered sequence in a clustered database environment. The approach includes receiving multiple requests for ordered sequence numbers from a plurality of entities. These multiple entries respectively pertaining to the multiple requests may be stored in a data structure. An entry may be selected from the multiple entries in the data structure; and one or more ordered sequence numbers may be provided to a requesting entity for the entry with a distributed lock mechanism based at least in part upon sequence number reservation or a local entity priority.

Abstract: The present disclosure relates to an automatic assignment of new entities in an existing hierarchical structure by leveraging retrieval-augmented generation (RAG) that improves the performance of a generative artificial intelligence (AI) model by generating a suggested hierarchical structure. For an automatic assignment, a user input may be received that may include a hierarchical structure and a query for adding one or more new entities to the existing hierarchical structure. For each new embedding vector associated with the query for adding a new entity, a retrieval-augmented generation may generate a context from the hierarchical structure based on a relevance to the query, and an additional context from a knowledge base. The retrieval-augmented generation may input a human-readable prompt, combining query with the context, additional context and a system prompt, to the generative AI model such as a large language model for generating the suggested hierarchical structure.

Abstract: Techniques are described for mapping a namespace to a compartment. An example method includes receiving, by a manager instance and from a pod, a first request for a token. The manager instance can transmit, to a token issuance service, a second request for the token. The token issuance service can identify a mapping object that maps the namespace to the resource. The token issuance service can transmit, to an identity service, the mapping object and a third request for the token. The identity service can identify a compartment of the customer tenancy based at least in part on the compartment identifier, the compartment managing the resource. The identity service can determine whether the pod has permission to access the resource. The identity service can generate the token based at least in part on the mapping object and the policy. The identity service can transmit the token to the pod.

Abstract: Various embodiments of the present technology generally relate to systems and methods for providing a privacy engine. In an example, the privacy engine converts, on a prover-side, confidential information into a plurality of tuples for a verification process and embeds the plurality of tuples with an embedding element. The prover-side privacy engine then generates a commitment using the tuples and transmits the commitment to a verifier-side privacy engine. Responsive to receiving the commitment, the verifier-side privacy engine verifies that the commitment includes the embedded element and generates a challenge to the commitment based on the commitment containing the embedded element. The verifier-side privacy engine then transmits the challenge to the prover-side privacy engine, which responsively generates a response containing the embedded element based on the challenge.

Abstract: A method may include receiving a primary document including one or more selection boxes, one or more text lines, and one or more annotations. The method may include determining, a class based on the annotations. The method may include identifying the one or more selection boxes and one or more text lines of the primary document. The method may include generating a graph representing the one or more selection boxes and the one or more text lines. The method may include mapping each of the one or more selection boxes to a respective text line of the one or more text lines of the graph based at least in part on one or more characteristics associated with the selection boxes. The method may include generating a key-value pair associated with each of the one or more text lines and generating a document model of the primary document.

Abstract: A reconciliation engine and data reconciliation techniques are disclosed that partially use machine learning and natural language processing techniques to overcome challenges associated with traditional manual reconciliation engines and data reconciliation techniques without compromising the accuracy and handle the complexity of the features. In one aspect, a computer-implemented method is provided that includes: accessing a source and target table, performing intra-reconciliation on the source table, grouping the source and target table based on system configurations, reconciling data is iterated over each group, where the reconciling data includes: finding, using a machine learning model, a closest match for each group, and outputting the closest match for each group.

Abstract: Novel techniques are described for routing of overlay packets within overlay networks in a cloud environment. A network device, located in the data path between a compute instance in an overlay network that is the source of a packet and a compute instance in the overlay network that is the intended destination of the packet, is able to route the packet using only special encoded information included in the packet's header when the packet is received by the network device. The special encoded information is in the form of a special encoded address (e.g., an encoded IP address) that is included in a field of the packet's header. The special encoded address encodes various different pieces of information that are used by the network devices in the data path from the source compute instance to the destination compute instance to route the packet in the overlay network.

Abstract: Techniques for drilling back to an original audio clip in virtual assistant initiated lists and reminders are disclosed. The system may receive audio input comprising a first request. Based on the first request, the system may schedule an action to be performed by the virtual assistant platform. The system stores at least a portion of the audio input and a mapping between the action and at least the portion of the audio input. The system performs the action. Subsequent to performing the action, the system receives a second request for audio playback of the first request corresponding to the action. The system retrieves at least the portion of the audio input based on the mapping between the action and at least the portion of the audio input, and plays at least the portion of the audio input comprising the first request.

Abstract: Techniques for evaluating a language model are disclosed herein. One or more agents and/or one or more language models are used to generate responses to queries in conversations. For example, during chat session conversations, the agent and/or model used to generate responses to queries in the conversation is sometimes changed. Various metrics are collected for the conversations for which a prompt, agent or model change occurs and/or the conversations for which a change to the prompt, agent and/or model does not occur. Metrics such as success rate, escalation rate, query repetition, direct feedback, or other feedback is collected and/or aggregated for groups of users and/or topics of conversations. One or more models used to generate responses in the conversations are evaluated and/or selected based on the metrics.

Abstract: Disclosed is a system for positioning fiber and electronics cables within a server room, which includes a wire-pulley system which includes a wire operably coupled to a first pulley wheel and a second pulley wheel. The wire is looped around the first and second pulley wheel such that a point in the wire is laterally movable between the first and second pulley wheels when the first and second pulley wheels are rotated. The system for positioning cables includes a cable carrier which is removably coupled to the point in the wire. The cable carrier includes a first panel, and a second panel hingedly coupled to the first panel. The first panel and second panel each include a plurality of receiving slots, where the receiving slots are configured to removably receive a distal end of various cables.

Abstract: Techniques discussed herein relate to generating synthetic training data with which a multi-speaker Text-To-Speech model may be trained. Text and an audio sample of a speaker may be provided as input to a voice generation system (VGS). The VGS may generate, based on audio features of the audio sample, machine-generated audio comprising a synthetic voice providing the text as spoken words. A synthetic training example including the machine-generated audio may be generated and combined with synthetic training data examples corresponding to a second speaker (and/or with training data examples comprising recordings of a second speaker). A multi-speaker Text-To-Speech model may be trained with a training set comprising the synthetic training data examples to generate subsequent audio that provides spoken words of input text in a synthetic voice that is generated to replicate features associated with one of a plurality of speakers comprising the first speaker and the second speaker.

Abstract: Systems, methods, and computer-readable media are provided for intelligent code integration, deployment, and analysis. An example intelligent code analysis system is configured to access and maintain an issue knowledge base storing information about code issues and/or issue resolutions associated with one or more software pipelines managed by the system. The system may process a natural language request about one or more code issues, issue resolutions, and/or sets of code. The system may additionally or alternatively analyze code issues to determine sets of code that should be newly associated with the issues, and/or analyze sets of code to determine issues that should be newly associated with the sets of code. The system generates a prompt that includes content from the natural language request, information about code issues, and/or information about sets of code.

Abstract: Techniques are disclosed herein for automatically curating preference alignment training (PAT) data and subsequently using the PAT data to train a machine-learning model used to perform natural language-to-query language tasks. A first PAT dataset is generated using a first trained LLM in an execution-based data generation technique. A second PAT dataset is generated using a second trained LLM in a LLM-based data generation technique. The second PAT dataset is populated with synthetic data points generated by the second trained LLM. Each data point of first PAT dataset and the second PAT dataset comprises a natural language query and an executable and non-executable query language statement. The first PAT dataset and the second PAT dataset can be combined and filtered to generate a PAT dataset that is optimized for training a machine-learning model to overcome one or more weaknesses exhibited when generating query language statements responsive to natural language questions.

Abstract: Techniques for predicting render times are disclosed. A system accesses auxiliary rendered outputs that were generated from a coarse render pass of a 3D scene description using a set of hardware resources. The system accesses system performance metrics of the set of hardware resources associated with the coarse render pass. The system encodes visual feature representations from the auxiliary rendered outputs using a convolutional neural network. The system also encodes the system performance metrics and a set of rendering configuration parameters associated with a target rendering operation. The system generates, using a multi-modal fusion model, a fused representation of the encoded visual feature representations, the encoded system performance metrics, and the encoded first set of rendering configuration parameters, using an attention-based mechanism that weights contributions of the inputs to the model. The system predicts a render time for the target rendering operation based on the fused representation.

Abstract: A method implements generating points-to relations using language models. The method involves executing a preprocessing session using one or more language models to process source code using one or more preprocessing prompts to generate metadata from the source code. The method further involves executing an analysis session using the one or more language models to process the metadata using an analysis prompt to generate a points-to report for the source code from the metadata. The method further involves executing an evaluation session using the one or more language models to process the points-to report and normalized code from the metadata using an evaluation prompt to generate a revised points-to report. The revised points-to report includes a confidence score for an entry in the points-to report.

Abstract: Systems, methods, and computer-readable media are provided for intelligent code integration, deployment, and analysis. An example intelligent code analysis system is configured to access and maintain an issue knowledge base storing information about code issues and/or issue resolutions associated with one or more software pipelines managed by the system. The system may process a natural language request about one or more code issues, issue resolutions, and/or sets of code. The system may additionally or alternatively analyze code issues to determine sets of code that should be newly associated with the issues, and/or analyze sets of code to determine issues that should be newly associated with the sets of code. The system generates a prompt that includes content from the natural language request, information about code issues, and/or information about sets of code.

Abstract: A system that trains a machine learning model to determine capture levels for target computations is disclosed. The system utilizes training data encompassing attribute sets and information levels for various computation types. For a form field value computation, the system determines associated attributes. The trained model processes these attributes to establish an appropriate information storage level. Based on this level, the system selects a relevant subset of information related to the computation or its result. The system then stores this selected subset in association with the computed value. The system uses feedback to retrain the model to enhance its performance.

Abstract: Techniques for populating the fields of a form are disclosed. A machine learning model may be trained to predict associations between fields. The trained machine learning model may be applied to a plurality of forms to predict an association between a first field in a first form type and a second field in a second form type. Upon receiving a value for a first field in a first form of a first form type and based on the predicted association, the system may populate a second field of a second form of the second form type based on the value.

Abstract: Method includes: accessing text, where spans are identified within the text and include one or more pairs of target spans and one or more mid-context spans; generating embedding representations of tokens associated with each target span, tokens associated with the entity types of each target span, and tokens associated with each mid-context span; generating, for each target span, entity-focused span embedding representation based on embedding representations of tokens associated with each target span and embedding representations of tokens associated with entity type of target span; generating, for each mid-context span, mid-context embedding representation based on the embedding representations of tokens associated with each mid-context span; and generating probability distribution of each relation of set of relations based on entity-focused span embedding representations of subject span and object span that are included in each target pair and mid-context embedding representation for mid-context span appear

Abstract: Techniques for detected reuse of a public key of a public/private cryptographic key pair within a cloud environment are disclosed. A database maintained within a cloud environment is accessed, the database storing a plurality of public keys. Each public key has a corresponding attribute, and the database further stores a plurality of attributes corresponding to the plurality of public keys. Each public key is associated with either (i) a corresponding user account of a corresponding tenancy of the cloud environment or (ii) a compute instance hosted within a corresponding tenancy of the cloud environment. A number of times a first public key or a first attribute corresponding to the first public key occurring within the database is determined. Responsive to the number of times being equal to or greater than a threshold number, the first public key is tagged as being reused for at least the threshold number of times.

Abstract: Techniques are disclosed herein for automatically curating preference alignment training (PAT) data and subsequently using the PAT data to train a machine-learning model used to perform natural language-to-query language tasks. A first PAT dataset is generated using a first trained LLM in an execution-based data generation technique. A second PAT dataset is generated using a second trained LLM in a LLM-based data generation technique. The second PAT dataset is populated with synthetic data points generated by the second trained LLM. Each data point of first PAT dataset and the second PAT dataset comprises a natural language query and an executable and non-executable query language statement. The first PAT dataset and the second PAT dataset can be combined and filtered to generate a PAT dataset that is optimized for training a machine-learning model to overcome one or more weaknesses exhibited when generating query language statements responsive to natural language questions.

Abstract: A change tracking and delta generation (CTDG) system servicing multiple clients is disclosed for enabling the delta generation during replication to be proportional to the number of changes between two snapshots instead of the size of the snapshots. In some embodiments, CTDG can track changes to a file system, where change-tracking can be dynamically switched between two sub-CT modes, file level or directory level. Change-tracking (CT) keys may be generated accordingly and stored in a data structure to be used for delta generation during replications for clients. In some embodiments, change-tracking may be enabled or disabled for different clients. CTDG can coordinate the CT enablement/disablement to ensure replications can use the information in CT keys properly.

Abstract: Techniques for detecting inter-tenancy exfiltration of data in a cloud environment are disclosed. A log that includes information associated with receipt of a service message at a gateway within a cloud environment is accessed. Based on the log, (i) originating information of the service message (such as identification of an originating tenancy of the service message) and (ii) target information of the service message (such as identification of a target tenancy of the service message) are determined. The originating information and the target information are compared. A mismatch between the originating information of the service message and the target information of the service message is detected. In response to the detected mismatch between the originating information of the service message and the target information of the service message, information indicative of the detected mismatch is caused to be presented at a user interface.

Abstract: Techniques are described to determine whether an input utterance is unrelated to a set of skill bots associated with a master bot. In some embodiments, a system described herein includes a training system and a master bot. The training system trains a classifier of the master bot. The training includes accessing training utterances associated with the skill bots and generating training feature vectors from the training utterances. The training further includes generating multiple set representations of the training feature vectors, where each set representation corresponds to a subset of the training feature vectors, and configuring the classifier with the set representations. The master bot accesses an input utterance and generates an input feature vector. The master bot uses the classifier to compare the input feature vector to the multiple set representations so as to determine whether the input feature falls outside and, thus, cannot be handled by the skill bots.

Abstract: Techniques described herein relate to systems and methods of data storage, and more particularly to providing layering of file system functionality on an object interface. In certain embodiments, file system functionality may be layered on cloud object interfaces to provide cloud-based storage while allowing for functionality expected from a legacy applications. For instance, POSIX interfaces and semantics may be layered on cloud-based storage, while providing access to data in a manner consistent with file-based access with data organization in name hierarchies. Various embodiments also may provide for memory mapping of data so that memory map changes are reflected in persistent storage while ensuring consistency between memory map changes and writes. For example, by transforming a ZFS file system disk-based storage into ZFS cloud-based storage, the ZFS file system gains the elastic nature of cloud storage.

Abstract: Techniques for data communication protocols for data transmission and software assurance in a cloud environment are disclosed. Data is received from a first component and at a first adapter using a first communication protocol. The data is transmitted, by the first adapter and to a proxy, using a second communication protocol. The proxy audits the data, and (i) blocks passage of the data, or (ii) forwards the data to a second adapter using the second communication protocol, to cause the second adapter to transmit the data to a second component using the first communication protocol. The proxy is within a tenancy of a cloud environment. The first adapter and the first component are within a first section of the cloud environment. The second adapter and the second component are within a second section of the cloud environment. Neither of the first or second sections is within the tenancy.

Abstract: Various embodiments of the present technology generally relate to systems and methods for validating 5G sessions at a binding support function (BSF) based on subscription-based notifications. In certain embodiments, a BSF system may comprise one or more processors, and a memory having stored thereon instructions that, when executed, cause the one or more processors to create a binding record associated with a 5G communication session, receive a subscription request from a Policy Control Function (PCF) for a session revalidation event associated with the 5G communication session, and initiate a revalidation timer based on the subscription request. The BSF system may send a notification request to the PCF based on the revalidation timer expiring, and delete the binding record based on a response to the notification request.

Abstract: Techniques are described for Transaction Guard to impose at-most-once execution by generating and using the database's native transaction identifier, DB XID. In an implementation, DB XID is unique within a (pluggable) database instance (with local undo) and uniquely identifies a transaction in the database. The Transaction Guard that is extended to use native transaction information determines the commit outcome using the native transaction identifier of the transaction instead of relying on the persistence of the Logical Transaction Identifier (LTXID) in a separate table. Using the native transaction identifier, the Transaction Guard significantly improves performance by eliminating the extra write(s) incurred during commit operations.

Abstract: Techniques are disclosed herein for training and deploying a named entity recognition model. The techniques include implementing a nested labeling scheme for named entities within the training data and then training a machine learning model on the training data The techniques further include extracting an entity hierarchy for a predicted class based on a hierarchical template associated with a composite label, where the predicted class is representative of multiple named entity classes comprising at least a parent class and a child class associated with the composite label. The techniques further include increasing the volume of training data via data mining for sequence tags in a language corpus and then training a machine learning model on the training data.

Abstract: Systems, methods, and other embodiments associated with a data quality framework for a machine learning pipeline are described. In one embodiment, a method includes receiving an input dataset prior to training a machine learning model with the input dataset. The input dataset, having data records in column tabular form, is analyzed to determine a data quality of the input dataset prior to training the machine learning model. A data quality score is generated that represents an overall quality of the input dataset, where the data quality score is generated based on a combined ensemble of at least two factors selected from (i) an abundance factor, (ii) a completeness factor, and (iii) a dimension efficiency factor. Based on the data quality score, the input dataset is either permitted or prohibited from continuing in the machine learning pipeline, and corrective actions may be performed.

Abstract: Techniques are disclosed for generating machine learning models that are insensitive to drift. A system trains a machine learning model using a divergent training dataset including synthesized data points simulating drift. The system can evaluate the machine learning models in terms of accuracy, latency, efficiency, and other metrics. Based on the evaluation, the system can select a machine learning model least susceptible to drift.

Abstract: A novel system is described for performing hierarchical named entity recognition (“HNER”) processing that includes identifying categories at different hierarchical levels for a named entity. The HNER system uses a novel architecture comprising an encoder model and a system of trained machine learning (ML) models to perform the HNER processing, where each trained model in the system of ML models corresponds to a particular hierarchical level, and each model is trained to extract one or more named entities and predict a category for each extracted named entity for the corresponding hierarchical level. Novel techniques are also described for training the various models in HNER system including an encoder model and models in the system of models.