Amazon.com

About: Amazon.com is a company organization based out in Seattle, Washington, United States. It is known for research contribution in the topics: Computer science & Service (business). The organization has 13363 authors who have published 17317 publications receiving 266589 citations.

Transfer Learning for Neural Semantic Parsing

Abstract: The goal of semantic parsing is to map natural language to a machine interpretable meaning representation language (MRL). One of the constraints that limits full exploration of deep learning technologies for semantic parsing is the lack of sufficient annotation training data. In this paper, we propose using sequence-to-sequence in a multi-task setup for semantic parsing with focus on transfer learning. We explore three multi-task architectures for sequence-to-sequence model and compare their performance with the independently trained model. Our experiments show that the multi-task setup aids transfer learning from an auxiliary task with large labeled data to the target task with smaller labeled data. We see an absolute accuracy gain ranging from 1.0% to 4.4% in in our in-house data set and we also see good gains ranging from 2.5% to 7.0% on the ATIS semantic parsing tasks with syntactic and semantic auxiliary tasks.

Deferred and off-loaded rendering of selected portions of web pages to incorporate late-arriving service data

Abstract: A web server system that incorporates service data into web pages defers the task of generating a portion of a web page associated with late-arriving service data, and offloads this task to the web browser that requested the page. Upon receiving a page request that involves a service request to a potentially slow service, the web server system generates and serves a “deferred rendering”, version of the web page that lacks the data from this service, but which includes most or all of the other visual elements of the page. Upon receiving the requested service data, the web server system forwards this data to the browser, which executes a page update handler (preferably a JavaScript function) to update the page with the service data. The data may be added near the top of the visual page, within a display object that initially serves as a placeholder for the service data.

Security protocols for low latency execution of program code

Abstract: A system for providing security mechanisms for secure execution of program code is described. The system may be configured to maintain a plurality of virtual machine instances. The system may be further configured to receive a request to execute a program code and allocate computing resources for executing the program code on one of the virtual machine instances. One mechanism involves executing program code according to a user-specified security policy. Another mechanism involves executing program code that may be configured to communicate or interface with an auxiliary service. Another mechanism involves splitting and executing program code in a plurality of portions, where some portions of the program code are executed in association with a first level of trust and some portions of the program code are executed with different levels of trust.

Processing event messages for user requests to execute program code

Abstract: A service manages a plurality of virtual machine instances for low latency execution of user codes. The service can provide the capability to execute user code in response to events triggered on an auxillary service to provide implicit and automatic rate matching and scaling between events being triggered on the auxiliary service and the corresponding execution of user code on various virtual machine instances. An auxiliary service may be configured as an event triggering service to detect events and generate event messages for execution of the user codes. The service can request, receive, or poll for event messages directly from the auxiliary service or via an intermediary message service. Event messages can be rapidly converted to requests to execute user code on the service. The time from processing the event message to initiating a request to begin code execution is less than a predetermined duration, for example, 100 ms.

Key Characteristics of a Container Orchestration Platform to Enable a Modern Application

Abstract: As compute evolves from bare metal to virtualized environments to containers towards serverless, the efficiency gains have enabled a wide variety of use cases. Organizations have used containers to run long running services, batch processing at scale, control planes, Internet of Things, and Artificial Intelligence workloads. Further, methodologies for software as a service, such as twelve-factor app, emphasize a clean contract with the underlying operating system and maximum portability between execution environments.1 In this paper, we address a set of capabilities required of a container orchestration platform to embody the design principles as illustrated by twelve factor app design. This paper also provides a non-exhaustive and prescriptive guide to identifying and implementing key mechanisms required in a container orchestration platform. We will cover capabilities such as cluster state management and scheduling, high availability and fault tolerance, security, networking, service discovery, continuous deployment, monitoring, and governance.