Patents.

Training models for IOT devices

A model training service of a provider network receives data from edge devices of a remote network. The model training service analyzes the received data. The model training service may also analyze global data from other edge devices of other remote networks.

Techniques to determine document recognition errors

Error recovery for conversational systems

Discovery, extraction, and recommendation of talent-screening questions

Methods, systems, and computer programs are presented for automatically generating phrase-based talent-screening questions. One method includes analyzing job descriptions to generate ngrams. Each ngram comprises one or more words.

Minimizing service restart by optimally resizing service pools

A method, computer program product, and system for optimizing service pools supporting resource sharing and enforcing SLAs, to minimize service restart. A computer processor determines a first resource to be idle, wherein a service instance continues to occupy the first resource that is idle. The processor adds the first resource to a resource pool, wherein the service instance continues to occupy the first resource as a global standby service instance on the first resource. The processor receives a request for a resource, wherein the request for the resource includes a global name associated with a service that corresponds to the global standby service instance, and the processor allocates, from the resource pool, the first resource having the global standby service instance, based on the request for the resource that includes the global name associated with the service corresponding to the global standby service instance.

Risk mitigation for a cryptoasset custodial system using a hardware security key

An approval request is transmitted for a cryptoasset transaction in accordance with a policy stored in a hardware security module ("HSM").

Cryptoasset custodial system with different rules governing access to logically separated cryptoassets and proof-of-stake blockchain support

Methods, and systems for secure storage and retrieval of information, such as private keys, useable to control access to a blockchain, include: receiving a request to take an action with respect to a vault of multiple different vaults in a cryptoasset custodial system, and each of the multiple...

Agent persona grounded chit-chat generation framework

Recommendation of recipes to a user of an online concierge system based on items included in an order by the user

Cryptoasset custodial system with different cryptographic keys controlling access to separate groups of private keys

Methods, systems, and apparatus, including medium-encoded computer program products, for secure storage and retrieval of information, such as private keys, useable to control access to a blockchain, include, in at least one aspect, a method including: identifying for an action an associated...

Quasi-recurrent Neural Network Based Encoder-decoder Model

The technology disclosed provides a quasi-recurrent neural network (QRNN) encoder-decoder model that alternates convolutional layers, which apply in parallel across timesteps, and minimalist recurrent pooling layers that apply in parallel across feature dimensions.

Vehicle sensor system and method of use

A system and method for collecting and processing sensor data for facilitating and/or enabling autonomous, semi-autonomous, and remote operation of a vehicle, including: collecting surroundings at one or more sensors, and determining properties of the vehicle surroundings from the collected sensor data.

Determining whether to take an action by applying a metric calculated using natural language processing tokens

Embodiments described herein are directed to computer-implemented methods, systems, and computer program products for calculating a metric using natural language processing tokens. A non-limiting example of the computer-implemented method includes parsing, by a processing device, user content using a natural language processing technique to extract tokens. The method further includes filtering, by the processing device, the tokens relating to a natural language processing criterion. The method further includes calculating, by the processing device, a metric based at least in part on the filtered tokens. The method further includes determining, by the processing device, whether to take an action by applying the metric to a set of rules. The method further includes taking the action responsive to determining to take the action.

Secure distribution of device key sets over a network

A system is provided for distribution of device key sets over a network in a protected software environment (PSE). In the system, a client device includes a connection interface for receiving a crypto hardware (CH) token.

Performing customized data compaction for efficient parallel data processing amongst a set of computing resources

Described is a system and method for compacting data into customized (e.g. optimal) file sizes for processing by computing resources. The mechanism may leverage various computing resources such as a cluster computing frameworks combined with a stream processing platform to efficiently process the activity data. For example, activity data of an organization may be processed by a set of jobs (or sub-jobs) as part of a data stream by a set of distributed computing resources. In order to efficiently process such data, the mechanism may compact the data into customized (e.g. optimal) file sizes. For example, the customized file sizes may provide an optimal (or near optimal) amount of data to be processed by each job, for example, to improve performance.

Construction of global internet message threads

Methods, systems, and devices supporting global message threads are described. A thread server may receive a communication message associated with a first set of user identifiers (e.g., the author and recipients of the message) and may extract a thread identifier from the message. The thread identifier may associate the message with one or more additional messages (e.g., from the same thread of messages), where the one or more additional messages may be associated with different user identifiers. The thread server may upsert the received communication message to a stored set of messages associated with a root indicating the thread identifier. If the thread server receives a query indicating the thread identifier from a user, the thread server may retrieve all of the messages associated with the thread identifier from persistent memory based on the query, including messages that are not associated with the user identifier for the querying user.

Cache and I/O management for analytics over disaggregated stores

Methods, systems, apparatuses, and computer program products are provided for prefetching data. A workload analyzer may identify job characteristics for a plurality of previously executed jobs in a workload executing on a cluster of one or more compute resources. For each job, identified job characteristics may include identification of an input dataset and an input bandwidth characteristic for the input dataset. A future workload predictor may identify future jobs expected to execute on the cluster based at least on the identified job characteristics. A cache assignment determiner may determine a cache assignment that identifies a prefetch dataset for at least one of the future jobs. A network bandwidth allocator may determine a network bandwidth assignment for the prefetch dataset. A plan instructor may instruct a compute resource of the cluster to load data to a cache local to the cluster according to the cache assignment and the network bandwidth assignment.

Action response selection based on communication message analysis

Methods, systems, and devices for analyzing communication messages (e.g., emails) and selecting corresponding actions are described.

Resource lifecycle optimization in disaggregated data centers

Embodiments for component lifecycle optimization in a disaggregated computing environment. A monitoring and machine learning process is performed to learn a respective lifecycle of different resource types as the different resource types are assigned to respective workloads. The monitoring and machine learning process is used to develop a set of learned failure patterns for determining a mitigation action to perform as new faults are encountered within each of the different resource types while executing the respective workloads. The mitigation action is performed to optimize a remaining lifecycle of respective ones of the different resource types according to the set of learned failure patterns.

Method, medium, and system for automated hardware device replacement

A method and system for replacing a hardware device is provided. The method includes detecting current utilization attributes of an Internet of things (IOT) device connected to a server device through a network. Recommended utilization attributes of the IOT device are retrieved and analyzed with ...

Automated visual inspection of syringes

A processor receives an image of a syringe. After identifying a background and foreground of the image, where the foreground indicates pixels that may be associated with a defect, the processor subtracts the background to generate an updated image with an accentuated foreground. The processor applies a bounding box to a group of pixels in the foreground and inputs the bounding box into a classifier. The classifier outputs a label indicating whether the syringe is defective.

Automated language assessment for web applications using natural language processing

Determining whether to take an action by applying a metric calculated using natural language processing tokens

Embodiments described herein are directed to computer-implemented methods, systems, and computer program products for calculating a metric using natural language processing tokens. A non-limiting example of the computer-implemented method includes parsing, by a processing device, user content using a natural language processing technique to extract tokens. The method further includes filtering, by the processing device, the tokens relating to a natural language processing criterion. The method further includes calculating, by the processing device, a metric based at least in part on the filtered tokens. The method further includes determining, by the processing device, whether to take an action by applying the metric to a set of rules. The method further includes taking the action responsive to determining to take the action.

Interactive Gaming System

Distributed notebook kernels in a containerized computing environment

Distributing kernels in a containerized computing environment includes executing, using computer hardware, a notebook server in a first container, wherein the notebook server is configured to communicate with a gateway in a second container, in response to a request for a kernel from the notebook server, the gateway requesting, using the computer hardware, a new container including the kernel from a container manager, instantiating, using the computer hardware, the new container including the kernel within a selected computing node of a plurality of computing nodes, publishing, using the computer hardware, communication port information for the new container to the gateway, and exchanging electronic messages, using the computer hardware, between the notebook server and the kernel through the gateway using the communication port information for the new container.

Document parsing using multistage machine learning

A system for document parsing includes an interface and a processor. The interface is configured to receive an image. The processor is configured to determine text boxes. Determining text boxes uses multiple zoom levels of the image. The processor is further configured to determine labels from the strings and provide the labels.

A machine learning framework and method for using the same

A machine learning framework and method for using the same are described. In one embodiment, the method for processing data with a machine learning framework comprises creating a plurality of features as independent features, each feature being based on one or more events.

Systems and methods for hierarchical multi-label contrastive learning

Embodiments described herein provide a hierarchical multi-label framework to learn an embedding function that may capture the hierarchical relationship between classes at different levels in the hierarchy. Specifically, supervised contrastive learning framework may be extended to the hierarchical multi-label setting. Each data point has multiple dependent labels, and the relationship between labels is represented as a hierarchy of labels. The relationship between the different levels of labels may then be learnt by a contrastive learning framework.

Health provider matching service

Data model dualization

A data model can be generated by dualizing another data model. In other words, a first data model can be transformed into a second data model, wherein the second data model is a mathematical dual of the first data model. For example, a non-relational data model can be generated by dualizing a relational data model.

Automated transparent distribution of updates to server computer systems in a fleet

A deployment subsystem provides updates to an application and/or software executed by server computer systems. The update is included in an update script. Execution of the update script by a server computer system causes the server computer system to execute a set of checks.

Monitoring dynamic quality of service based on changing user context

One embodiment provides a method for monitoring context-dependent quality of service in a shared computing environment that includes detecting, by a processor, a change in context. Context includes user context and external context, and user context comprises physical activity, mood, engagement levels and type of events. Prior assigned service classes are updated to updated service classes based on the change in context. Service level agreement (SLA) statistics for each assigned service class are aggregated and collected. Each assigned service class includes at least one SLA based on aggregate services received by individual users in that assigned service class, and aggregating SLA statistics is based on a statistical function.

Contact information extraction and identification

A database server may identify mentioned names in a body of a message and extract the names using name identification heuristics and algorithms, using a distributed connection graph to identify contacts.

Recognition of biases in data and models

A method of training a predictive model to predict a likely field value for one or more user selected fields within an application. The method comprises providing a user interface for user selection of the one or more user selected fields within the application; analyzing a pre-existing, user provided data set of objects; training, based on the analysis, the predictive model; and determining, for each of the user selected fields, whether the data or model exhibits bias.

Disaster recovery and replication in disaggregated datacenters

Embodiments for disaster recovery in a disaggregated computing system. Memory resources are allocated at a secondary, disaster recovery site for data received from a primary site. The data from the primary site is continuously replicated to the allocated memory resources at the disaster recovery site without requiring any compute resources to be attached to the allocated memory resources. Responsive to determining a disaster recovery failover is in progress, the compute resources are assigned to the allocated memory resources for performing a failover workload, and the failover workload is executed at the disaster recovery site.

Systems and methods for robust speech recognition using generative adversarial networks

Described herein are systems and methods for a general, scalable, end-to-end framework that uses a generative adversarial network (GAN) objective to enable robust speech recognition. Encoders trained with the proposed approach enjoy improved invariance by learning to map noisy audio to the same embedding space as that of clean audio. Embodiments of a Wasserstein GAN framework increase the robustness of seq-to-seq models in a scalable, end-to-end fashion. In one or more embodiments, an encoder component is treated as the generator of GAN and is trained to produce indistinguishable embeddings between labeled and unlabeled audio samples. This new robust training approach can learn to induce robustness without alignment or complicated inference pipeline and even where augmentation of audio data is not possible.

Using unstructured input to update heterogeneous data stores

Approaches to using unstructured input to update heterogeneous data stores include receiving unstructured text input, receiving a template for interpreting the unstructured text input, identifying, using an entity classifier, entities in the unstructured text input, identifying one or more potential parent entities from the identified entities based on the template, receiving a selection of a parent entity from the one or more potential parent entities, identifying one or more potential child entities from the identified entities based on the template and the selected parent entity, receiving a selection of a child entity from the one or more potential child entities, identifying an action item in the unstructured text input based on the identified entities and the template, determining, using an intent classifier, an intent of the action item, and updating a data store based on the dete

Systems and Methods for Identifying Sentiment in Text Strings

Identifying relevant messages in a conversation graph

A method, system, and non-transitory computer readable medium for identifying a relevant conversation. The method can include: identifying a conversation graph including identifiers of messages posted by a set of authoring accounts; identifying a relevance score for each authoring account.

Wearable word counter

Dense video captioning

Searching archived data

Reactive coincidence

An outer event stream can be modeled to represent duration for an event with an inner event stream. The inner event stream can be generated to represent duration of at least one event from the outer event stream. By modeling the outer event stream to include duration, coincidence between two or more events can be determined. More specifically, the modeling of the outer event stream enables use of operators to identify coincidence between events from event streams.

Autonomous vehicle interface system

In embodiments of an autonomous vehicle interface system, system nodes are each implemented as a distributed node for independent data processing of low-level sensor data and/or high-level system data. The high-level system data is abstracted from the low-level sensor data, providing invariance to system configuration in higher-level processing algorithms. The autonomous vehicle interface system includes at least one real-time bus for data communications of the low-level sensor data and the high-level system data between the system nodes. The system also includes an application programming interface (API) configured for access by the system nodes to the low-level sensor data and the high-level system data that is published on the real-time bus and accessible via the API.

Techniques and architectures for managing disparate heterogeneous cloud-based resources

Techniques and architectures for data modeling and management. Data modeling services are provided to agents within multiple different operating environments of a computing environment having at least one database stored on one or more physical memory devices.

User-interface for developing applications that apply machine learning

Methods, systems and devices used to visualize data: a computing system receives a request from a user to visualize stored data and generates interactive visualizations supporting machine-learning application development.

System and method for securing applications through an application-aware runtime agent

A system and method for securing an application through an application-aware runtime agent can include: acquiring a code profile, instrumenting the application with a runtime agent according to the code profile, enforcing the runtime agent on the execution of the application, and responding to the runtime agent. Enforcing the runtime agent on the execution of the application can include monitoring the execution flow, which comprises of monitoring the utilization of the controls through the execution of the application; detecting a threat, which comprises identifying a section of the execution flow as a potential security threat; and regulating the execution flow to prevent or ameliorate the security threat. Responding to the runtime agent can include responding to the security threat and providing a user interface that may output runtime agent diagnostics and trigger alerts.

Speech detection and speech recognition

A system configured to recognize text represented by speech may determine that a first portion of audio data corresponds to speech from a first speaker and that a second portion of audio data corresponds to speech from the first speaker and a second speaker. Features of the first portion are compared to features of the second portion to determine a similarity therebetween. Based on this similarity, speech from the first speaker is distinguished from speech from the second speaker and text corresponding to speech from the first speaker is determined.

Systems and methods for interest-driven business intelligence systems including geo-spatial data

Table data persistence

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for a distributed parallel processing database system that persists table data in memory to a distributed file system. A distributed parallel processing database system persists table data in memory to a distributed file system. A parameter of creating a database table specifies that data records in the database table and history of changes to the data records can be stored in memory as well as in the distributed file system. When the database table is populated or otherwise modified, data records and the history in memory are evicted to the distributed file system as log files and removed from memory. The log files can be designated as write-only, where the data records, once written, cannot be read by structured query language (SQL) queries, or as read-write, where the data records, once written, can be read by SQL queries.