Jasha Droppo

Amazon (United States)

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Alex Acero

Apple (United States)

Andreas Stolcke

Amazon (United States)

Dong Yu

Bellevue Hospital Center

D.L. Strenge

Environmental Protection Agency

Xuedong Huang

Microsoft Research (United Kingdom)

Michael L. Seltzer

Meta (United States)

Ariya Rastrow

Amazon (United States)

J.W. Buck

Pacific Northwest National Laboratory

Li Deng

Shanghai First People's Hospital

Geoffrey Zweig

Meta (Israel)

Cooperative Institutions

Amazon (United States)

Microsoft (United States)

Pacific Northwest National Laboratory

Microsoft Research (United Kingdom)

Amazon (Germany)

Battelle

National Technical Information Service

United States Department of Commerce

University of Washington

Office of Scientific and Technical Information

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Single-shell tank constituent rankings for use in preparing waste characterization plans

Jasha Droppo J.W. Buck J.S. Wilbur D.L. Strenge M.D. Freshley

Waste characterization efforts for underground single-shell tanks (SSTs) containing chemical and radioactive mixed wastes at the US Department of Energy's (DOE) Hanford Site are currently under way. As a component of this effort, an analysis was conducted to give a preliminary ranking of possible constituents in the SST waste and provide data for establishing detection limit requirements for the SST characterization effort. These SST constituent rankings were based on a relative comparison using potential human health impacts modeled using a hypothetical exposure scenario. This modeling effort used preliminary estimates of the SST inventories, simplified estimates of constituent release rates and environmental transport, a hypothetical usage location, and a standard Hanford exposure scenario. The results of this evaluation are SST constitutents for each of six groups of SSTs ranked according to their relative potential for impacts. The relative rankings for different recharge rates at the tank farms were nearly equivalent. Sensitivity tests demonstrated that the rankings are influenced by changes in recharge and transport parameters. 45 refs., 18 figs., 2 tabs.

Hanford Site

Characterization

10.2172/5670015

Cite

Citations (4)

Review of calculational models and computer codes for environmental dose assessment of radioactive releases

D.L. Strenge E.C. Watson Jasha Droppo

The development of technological bases for siting nuclear fuel cycle facilities requires calculational models and computer codes for the evaluation of risks and the assessment of environmental impact of radioactive effluents. A literature search and review of available computer programs revealed that no one program was capable of performing all of the great variety of calculations (i.e., external dose, internal dose, population dose, chronic release, accidental release, etc.). Available literature on existing computer programs has been reviewed and a description of each program reviewed is given.

Computer program

10.2172/7346903

Cite

Citations (8)

Recursive noise estimation using iterative stochastic approximation for stereo-based robust speech recognition

Li Deng Jasha Droppo Alex Acero

We present an algorithm for recursive estimation of parameters in a mildly nonlinear model involving incomplete data. In particular, we focus on the time-varying deterministic parameters of additive noise in the nonlinear model. For the nonstationary noise that we encounter in robust speech recognition, different observation data segments correspond to different noise parameter values. Hence, recursive estimation algorithms are more desirable than batch algorithms, since they can be designed to adaptively track the changing noise parameters. One such design based on the iterative stochastic approximation algorithm in the recursive-EM framework is described. This new algorithm jointly adapts time-varying noise parameters and the auxiliary parameters introduced to give a linear approximation of the nonlinear model. We present stereo-based robust speech recognition results for the AURORA task, which demonstrate the effectiveness of the new algorithm compared with a more traditional, MMSE noise estimation technique under otherwise identical experimental conditions.

Stochastic Approximation

10.1109/asru.2001.1034594

Cite

Citations (14)

Assessment of the long-term exposure due to resuspension

T.W. Horst Jasha Droppo C.E. Elderkin C.L. Simpson

Source

Cite

Citations (1)

Ozone field studies adjacent to a hvdc transmission test line

Jasha Droppo O.B. Abbey Debra Glover

Field studies of atmospheric ozone concentrations adjacent to high voltage direct current (hvdc) transmission test lines were conducted at the Bonneville Power Administration's (BPA) hvdc Test Facility at The Dalles, Oregon. The transmission lines were operating at voltages from +- 400 to +- 600 kV during the field studies. The downwind ozone plumes were studied using a roving vertical profiling system. Ambient meteorological and test line parameters were also recorded to allow comparison of predicted and observed ozone plumes. For fair weather conditions no ozone plumes were evident. The absence of identifiable changes in ozone concentrations by the energized lines demonstrates the trivial nature of the ozone concentrations from the energized lines for fair weather conditions. Since corona loss from power lines is largest during precipitation, the ozone production is also largest. At the same time, the natural background atmospheric ozone concentrations are depressed by the scavenging effect of the precipitation. Therefore ozone production rates can best be measured during precipitation periods. With the exception of precipitation cases, the vertical profiles of ozone concentration demonstrated no discernible evidence of ozone plumes from the energized conductors. Ozone plumes, if any, were masked in the natural background ozone variability.

Tropospheric ozone

10.2172/5556286

Cite

Citations (2)

Class-Dependent Time-Frequency Distributions via Operator Theory

International Conference on Acoustics, Speech, and Signal Processing (1997)

John McLaughlin Les Atlas Jasha Droppo

We propose a property for kernel design which results in distributions for each of two classes of signals which maximally separates their energies in the time-frequency plane. Such maximally separated distributions may result in improved classification because the signal representation is optimized to accentuate the differences in signal classes. This is not the case with other time-frequency kernels which are optimized based upon some criteria unrelated to the classification task. Using our operator theory formulation for time-frequency representations, our “maximal separation” criteria takes on a very easily solved form. Analysis of the solution in both the time-frequency and ambiguity planes is given along with an example on discrete signals.

Kernel (algebra)

Instantaneous phase

Operator (biology)

Time–frequency representation

Representation

SIGNAL (programming language)

Source

Cite

Citations (7)

A harmonic-model-based front end for robust speech recognition

Michael L. Seltzer Jasha Droppo Alex Acero

Speech recognition accuracy degrades significantly when the speech has been corrupted by noise, especially when the system has been trained on clean speech. Many compensation algorithms have been developed which require reliable online noise estimates or a priori knowledge of the noise. In situations where such estimates or knowledge is difficult to obtain, these methods fail. We present a new robustness algorithm which avoids these problems by making no assumptions about the corrupting noise. Instead, we exploit properties inherent to the speech signal itself to denoise the recognition features. In this method, speech is decomposed into harmonic and noise-like components, which are then processed independently and recombined. By processing noise-corrupted speech in this manner we achieve significant improvements in recognition accuracy on the Aurora 2 task.

Robustness

10.21437/eurospeech.2003-406

Cite

Citations (17)

Wav2vec-C: A Self-supervised Model for Speech Representation Learning

arXiv (Cornell University) (2021)

Samik Sadhu Di He Che-Wei Huang Sri Harish Mallidi Minhua Wu

Wav2vec-C introduces a novel representation learning technique combining elements from wav2vec 2.0 and VQ-VAE. Our model learns to reproduce quantized representations from partially masked speech encoding using a contrastive loss in a way similar to Wav2vec 2.0. However, the quantization process is regularized by an additional consistency network that learns to reconstruct the input features to the wav2vec 2.0 network from the quantized representations in a way similar to a VQ-VAE model. The proposed self-supervised model is trained on 10k hours of unlabeled data and subsequently used as the speech encoder in a RNN-T ASR model and fine-tuned with 1k hours of labeled data. This work is one of only a few studies of self-supervised learning on speech tasks with a large volume of real far-field labeled data. The Wav2vec-C encoded representations achieves, on average, twice the error reduction over baseline and a higher codebook utilization in comparison to wav2vec 2.0

Representation

Source

Cite

Citations (0)

Guided Contrastive Self-Supervised Pre-Training for Automatic Speech Recognition

2022 IEEE Spoken Language Technology Workshop (SLT) (2023)

Aparna Khare Minhua Wu Saurabhchand Bhati Jasha Droppo Roland Maas

Contrastive Predictive Coding (CPC) is a representation learning method that maximizes the mutual information between intermediate latent representations and the output of a given model. It can be used to effectively initialize the encoder of an Automatic Speech Recognition (ASR) model. We present a novel modification of CPC called Guided Contrastive Predictive Coding (GCPC). Our proposed method maximizes the mutual information between representations from a prior-knowledge model and the output of the model being pre-trained, allowing prior knowledge injection during pre-training. We validate our method on 3 ASR tasks: German, French and English. Our method outperforms CPC pre-training on all three datasets, reducing the Word Error Rate (WER) by 4.44%, 6.55% and 15.43% relative on the German, French and English (Librispeech) tasks respectively, compared to training from scratch, while CPC pre-training only brings 2.96%, 1.01% and 14.39% relative WER reduction respectively.

Word error rate

Predictive coding

Training set

10.1109/slt54892.2023.10022676

Cite

Citations (0)

Parallelizing WFST Speech Decoders (Poster)

Jasha Droppo Charith Mendis Saeed Maleki Madanlal Musuvathi Todd Mytkowicz

Source

Cite

Citations (0)