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.

An Empirical Study on Robustness to Spurious Correlations using Pre-trained Language Models

Abstract: Recent work has shown that pre-trained language models such as BERT improve robustness to spurious correlations in the dataset. Intrigued by these results, we find that the key to their success is generalization from a small amount of counterexamples where the spurious correlations do not hold. When such minority examples are scarce, pre-trained models perform as poorly as models trained from scratch. In the case of extreme minority, we propose to use multi-task learning (MTL) to improve generalization. Our experiments on natural language inference and paraphrase identification show that MTL with the right auxiliary tasks significantly improves performance on challenging examples without hurting the in-distribution performance. Further, we show that the gain from MTL mainly comes from improved generalization from the minority examples. Our results highlight the importance of data diversity for overcoming spurious correlations.

COVID-19 Risk Assessment Tool: Dual application of risk communication and risk governance

Abstract: Risk awareness is the best way to prevent and slow-down the transmission of the COVID-19 pandemic. Risk awareness is achieved through communication of risk assessment. Effective risk communication is an important measure to control the infodemic. Most risk assessment tools focus on either tracking the affected patients or diagnosing a probable health condition through symptoms. RIKA India introduces an innovative Risk Assessment Tool which goes beyond the symptom detection and patient tracking. It includes four factors in assessment of risk: Health, Behaviour, Exposure and Social Policy. Each of these four factors have sub-factors which help to assess the overall risk in a more comprehensive way and also present it to the user in a simplified way. The paper discusses the importance of the Risk Assessment Tool for awareness generation and decision making. Further, the datasets generated through the tool have been analysed to understand the key intervention areas for COVID-19 response and management.

Gesture detection systems

Abstract: The amount of power and processing needed to enable gesture input for a computing device can be reduced by utilizing one or more gesture sensors. A gesture sensor can have a lower resolution but larger pixel pitch than conventional cameras. The lower resolution can be achieved in part through skipping or binning pixels in some embodiments. The low resolution enables a global shutter to be used with the gesture sensor. The gesture sensor can be connected to an illumination controller for synchronizing illumination from a device emitter with the global shutter. In some devices, the gesture sensor can be used as a motion detector, enabling the gesture sensor to run in a low power state unless there is likely gesture input to process. At least some processing and circuitry is included with the gesture sensor such that functionality can be performed without accessing a central processor or system bus.

An Evaluation of Empirical Bayes’s Estimation of Value-Added Teacher Performance Measures:

Abstract: Empirical Bayes’s (EB) estimation has become a popular procedure used to calculate teacher value added, often as a way to make imprecise estimates more reliable. In this article, we review the theo...