Danny Wyatt, Tanzeem Choudhury, Jeff Bilmes
In this paper, we present a novel probabilistic framework for recovering global, latent social network structure from local, noisy observations. We extend curved exponential random graph models to...
Towards the Automated Social Analysis of Situated Speech Data (2009)
We present an automated approach for studying fine-grained details of social interaction and relationships. Specifically, we analyze the conversational characteristics of a group of 24 individuals...
Structure Learning on Large Scale Common Sense Statistical Models of Human State (2009)
William Pentney, Matthai Philipose, Jeff Bilmes
Research has shown promise in the design of large scale common sense probabilistic models to infer human state from environmental sensor data. These models have made use of mined and preexisting...
Average-Case Active Learning with Costs (2009)
Guillory, Andrew, Bilmes, Jeff
We analyze the expected cost of a greedy active learning algorithm. Our analysis extends previous work to a more general setting in which different queries have different costs. Moreover, queries may...
A Factored Language Model of Quantized Pitch and Duration (2008)
This paper investigates a novel statistical approach to music classification that utilizes recent technology developed in the domain of natural language processing. Specifically, we investigate the...
1 Introduction Creating Social Network Models from Sensor Data (2008)
Danny Wyatt, Tanzeem Choudhury, Jeff Bilmes
Complex macro-social phenomena can arise from simple micro-level behavior without any global coordination
Necessary Intransitive Likelihood-Ratio Classifiers (2008)
In pattern classification tasks, errors are introduced because of differences between the true generative model and the one obtained via model estimation. Using likelihood-ratio based classification,...
Generalized Graphical Abstractions for Statistical Machine Translation (2008)
We introduce a novel framework for the expression, rapid-prototyping, and evaluation of statistical machine-translation (MT) systems using graphical models. The framework extends dynamic Bayesian...
Disambiguating Speech Commands using Physical Context (2008)
Katherine M. Everitt, Susumu Harada, Jeff Bilmes, James A. L
Speech has great potential as an input mechanism for ubiquitous computing. However, the current requirements necessary for accurate speech recognition, such as a quiet environment and a...
Jeff Bilmes, Gang Ji, Marina Meilă
In this work, we introduce an information-theoretic based correction term to the likelihood ratio classification method for multiple classes. Under certain conditions, the term is sufficient for...
Mukund Narasimhan, Nebojsa Jojic, Jeff Bilmes
We show that Queyranne’s algorithm for minimizing symmetric submodular functions can be used for clustering with a variety of different objective functions. Two specific criteria that we consider...
Generalized Graphical Abstractions for Statistical Machine Translation (2008)
We introduce a novel framework for the expression, rapid-prototyping, and evaluation of statistical machine-translation (MT) systems using graphical models. The framework extends dynamic Bayesian...
Speech, Signal and Language Interpretation Lab, University of Washington, (2008)
Amarnag Subramanya, Jeff Bilmes
We propose a new set of features based on the temporal statistics of the spectral entropy of speech. We show why these features make good inputs for a speech detector. Moreover, we propose a back-end...
UNCERTAINTY IN TRAINING LARGE VOCABULARY SPEECH RECOGNIZERS (2008)
Amarnag Subramanya, Chris Bartels, Jeff Bilmes, Patrick Nguyen
We propose a technique for annotating data used to train a speech recognizer. The proposed scheme is based on labeling only a single frame for every word in the training set. We make use of the...
Jeff Bilmes, Gang Ji, Marina Meilă
In this work, we introduce an information-theoretic based correction term to the likelihood ratio classification method for multiple classes. Under certain conditions, the term is sufficient for...
Pomona College, Swarthmore College, (2008)
Katrin Kirchhoff, Jeff Bilmes, Sourin Das, Nicolae Duta, Melissa Egan, Gang Ji, ...
of Defense, � SRI Although Arabic is currently one of the most widely spoken languages in the world, there has been relatively little speech recognition research on Arabic compared to other...
Mukund Narasimhan, Nebojsa Jojic, Jeff Bilmes
We show that Queyranne’s algorithm for minimizing symmetric submodular functions can be used for clustering with a variety of different objective functions. Two specific criteria that we consider...
Necessary Intransitive Likelihood-Ratio Classifiers (2008)
In pattern classification tasks, errors are introduced because of differences between the true model and the one obtained via model estimation. Using likelihood-ratio based classification, it is...
OOV DETECTION BY JOINT WORD/PHONE LATTICE ALIGNMENT (2008)
Hui Lin, Jeff Bilmes, Dimitra Vergyri, Katrin Kirchhoff
We propose a new method for detecting out-of-vocabulary (OOV) words for large vocabulary continuous speech recognition (LVCSR) systems. Our method is based on performing a joint alignment between...
The PHiPAC v1.0 Matrix-Multiply Distribution. (2007)
Jeff Bilmes Krste, Jeff Bilmes, Krste Asanovi C, Chee-whye Chin, Jim Demmel, Krste Asanovic
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. We have developed a methodology whereby near-peak...
Heinz W. Schmidt, Jeff Bilmes, The Sather, Language Com
6 ACKNOWLEDGEMENTS 15 classes and a symbolic debugger. We are working on higher-level user interface abstractions and extended language tools such as an interpreter. Another direction of research is...
To Appear NIPS 2001 Intransitive Likelihood-Ratio Classifiers (2007)
Jeff Bilmes, Gang Ji, Marina Meil A
We introduce an information-theoretic based correction term to the likelihood ratio classification method for multiple classes. This term makes the class comparisons intransitive and we use several...
A Privacy-Sensitive Approach to Modeling Multi-Person Conversations (2007)
In this paper we introduce a new dynamic Bayesian network that separates the speakers and their speaking turns in a multi-person conversation. We protect the speakers' privacy by using only...
Consensus ranking under the exponential model (2007)
Marina Meilă, Kapil Phadnis, Arthur Patterson, Jeff Bilmes
We analyze a popular exponential model over rankings called the generalized Mallows model. Estimating the central ranking (or consensus ranking) of this model from data is NP-hard. We obtain the...
Conversation detection and speaker segmentation in privacy-sensitive situated speech data (2007)
Danny Wyatt, Tanzeem Choudhury, Jeff Bilmes
We present privacy-sensitive methods for (1) automatically finding multi-person conversations in spontaneous, situated speech data and (2) segmenting those conversations into speaker turns. The...
Learning large scale common sense models of everyday life (2007)
Recent work has shown promise in using large, publicly available, hand-contributed commonsense databases as joint models that can be used to infer human state from day-to-day sensor data. The...
Learning large scale common sense models of everyday life (2007)
Recent work has shown promise in using large, publicly available, hand-contributed commonsense databases as joint models that can be used to infer human state from day-to-day sensor data. The...
A privacy-sensitive approach to modeling multi-person conversations (2007)
In this paper we introduce a new dynamic Bayesian network that separates the speakers and their speaking turns in a multi-person conversation. We protect the speakers ’ privacy by using only...
Factored Language Models Tutorial (2007)
Katrin Kirchhoff, Jeff Bilmes, Kevin Duh, Katrin Kirchhoff, Jeff Bilmes, Kevin Duh
The Factored Language Model (FLM) is a flexible framework for incorporating various information sources, such as morphology and part-of-speech, into language modeling. FLMs have so far been...
Alvin Raj, Amarnag Subramanya, Dieter Fox, Jeff Bilmes
Recent advances in wearable sensing and computing devices and in fast probabilistic inference techniques make possible the fine-grained estimation of a person’s activities over extended periods of...
Graphical model representations of word lattices (2006)
Gang Ji, Jeff Bilmes, Jeff Michels, Katrin Kirchhoff, Chris Manning
We introduce a method for expressing word lattices within a dynamic graphical model. We describe a variety of choices for doing this, including a technique to relax the time information associated...
Graphical model representations of word lattices (2006)
Gang Ji, Jeff Bilmes, Jeff Michels, Katrin Kirchhoff, Chris Manning
We introduce a method for expressing word lattices within a dynamic graphical model. We describe a variety of choices for doing this, including a technique to relax the time information associated...
An online adaptive filtering algorithm for the Vocal Joystick (2006)
Xiao Li, Jonathan Malkin, Susumu Harada, Jeff Bilmes, Richard Wright, James L
This paper introduces a novel adaptive direction filtering algorithm in the Vocal Joystick (VJ) setting that utilizes context information and applies real-time inference in a continuous space. The VJ...
An online adaptive filtering algorithm for the Vocal Joystick (2006)
Xiao Li, Jonathan Malkin, Susumu Harada, Jeff Bilmes, Richard Wright, James L
This paper introduces a novel adaptive direction filtering algorithm in the Vocal Joystick (VJ) setting that utilizes context information and applies real-time inference in a continuous space. The VJ...
Most automatic speech recognition (ASR) systems express probability densities over sequences of acoustic feature vectors using Gaussian or Gaussian-mixture hidden Markov models. In this chapter, we...
Alvin Raj, Amarnag Subramanya, Dieter Fox, Jeff Bilmes
Recent advances in wearable sensing and computing devices and in fast probabilistic inference techniques make possible the fine-grained estimation of a person’s activities over extended periods of...
Mukund Narasimhan, Nebojsa Jojic, Jeff Bilmes
We show that Queyranne's algorithm for minimizing symmetric submodular functions can be used for clustering with a variety of different objective functions. Two specific criteria that we...
DBN multistream models for Mandarin toneme recognition (2005)
Xin Lei, Gang Ji, Tim Ng, Jeff Bilmes, Mari Ostendorf
A toneme in Mandarin Chinese is a tonal phone which consists of a base phone (main vowel) and a tone. To capture both, most recognition systems use two feature streams: the standard MFCCs for the...
Optimal sub-graphical models (2005)
Mukund Narasimhan, Jeff Bilmes
We investigate the problem of reducing the complexity of a graphical model (G, PG) by finding a subgraph H of G, chosen from a class of subgraphs H, such that H is optimal with respect to...
In this paper, we compare both discriminative and generative parameter learning on both discriminatively and generatively structured Bayesian network classifiers. We use either maximum likelihood...
Genetic triangulation of graphical models for speech and language processing (2005)
Chris Bartels, Kevin Duh, Jeff Bilmes, Katrin Kirchhoff
Graphical models are an increasingly popular approach for speech and language processing. As researchers design ever more complex models it becomes crucial to find triangulations that make inference...
In this paper, we compare both discriminative and generative parameter learning on both discriminatively and generatively structured Bayesian network classifiers. We use either maximum likelihood...
Sitting at the intersection between statistics and machine learning, Dynamic Bayesian Networks have been applied with much success in many domains, such as speech recognition, vision, and...
Speech feature smoothing for robust ASR (2005)
In this paper, we evaluate smoothing within the context of the MVA (mean subtraction, variance normalization, and ARMA filtering) post-processing scheme for noise-robust automatic speech recognition....
A Comparison of Floating Point and Logarithmic Number Systems for FPGAs (2005)
Jeff Bilmes, Michael Haselman, Michael Haselman, Michael Haselman
and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the final examining committee have been made. Committee Members:
Sitting at the intersection between statistics and machine learning, Dynamic Bayesian Networks have been applied with much success in many domains, such as speech recognition, vision, and...
Backoff Model Training using Partially Observed Data: Application to Dialog Act Tagging (2005)
Dialog act (DA) tags are useful for many applications in natural language processing and automatic speech recognition. In this work, we introduce hidden backoff models (HBMs) where a large...
Graphical model approach to pitch tracking (2004)
Xiao Li, Jonathan Malkin, Jeff Bilmes
Many pitch trackers based on dynamic programming require meticulous design of local cost and transition cost functions. The forms of these functions are often empirically determined and their...
Bilmes. Statistical models for empirical search-based performance tuning (2004)
Richard Vuduc, James W. Demmel, Jeff Bilmes
Achieving peak performance from the computational kernels that dominate application performance often requires extensive machine-dependent tuning by hand. Automatic tuning systems have emerged in...
Bilmes. Statistical models for empirical search-based performance tuning (2004)
Richard Vuduc, James W. Demmel, Jeff Bilmes
Achieving peak performance from the computational kernels that dominate application performance often requires extensive machine-dependent tuning by hand. Automatic tuning systems have emerged in...
PAC-learning bounded tree-width graphical models (2004)
Mukund Narasimhan, Jeff Bilmes
We show that the class of strongly connected graphical models with treewidth at most k can be properly efficiently PAC-learnt with respect to the Kullback-Leibler Divergence. Previous approaches to...
DBN based multi-stream models for speech (2003)
Yimin Zhang, Qian Diao, Shan Huang, Wei Hu, Chris Bartels, Jeff Bilmes
We propose dynamic Bayesian network (DBN) based synchronous and asynchronous multi-stream models for noise-robust automatic speech recognition. In these models, multiple noise-robust features are...
Necessary Intransitive Likelihood-Ratio Classifiers (2003)
In pattern classification tasks, errors are introduced because of differences between the true model and the one obtained via model estimation.
Necessary Intransitive Likelihood-Ratio Classifiers (2003)
In pattern classification tasks, errors are introduced because of differences between the true model and the one obtained via model estimation.
Necessary Intransitive Likelihood-Ratio Classifiers (2002)
Gang Ji, Jeff Bilmes, Gang Ji, Jeff Bilmes
In any pattern classification task, errors are introduced because of the difference between the true generative model and the one obtained via model estimation. One approach to solve this problem...
Since their inception over thirty years ago, hidden Markov models (HMMs) have have become the predominant methodology for automatic speech recognition (ASR) systems — today, most state-of-the-art...
This paper describes the Graphical Models Toolkit (GMTK), an open source, publically available toolkit for developing graphical-model based speech recognition and general time series systems....
Robust Splicing Costs And Efficient Search With (2002)
Bmm Models For, Ivan Bulyko, Mari Ostendorf, Jeff Bilmes
With the growing popularity of corpus-based methods for concatenative speech synthesis, a large amount of interest has been placed on borrowing techniques from the ASR community. This paper explores...
Low-Resource Noise-Robust Feature Post-Processing On Aurora 2.0 (2002)
We present a highly effective and extremely simple noiserobust front end based on novel post-processing of standard MFCC features. It performs remarkably well on the Aurora 2.0 noisydigits database...
This paper describes the Graphical Models Toolkit (GMTK), an open source, publically available toolkit for developing graphical-model based speech recognition and general time series systems....
GMTK: The Graphical Models Toolkit (2002)
This document describes the use of the graphical models toolkit GMTK, and its supporting programs, which are a software package for graphical-model based speech recognition written by Jeff Bilmes and...
The 2001 GMTK-Based Spine ASR System (2002)
Özgür Cetin, Harriet J. Nock, Katrin Kirchhoff, Jeff Bilmes, Mari Ostendorf
This paper provides a detailed description of the University of Washington automatic speech recognition (ASR) system for the 2001 DARPA SPeech In Noisy Environments (SPINE) task. Our system makes...
The 2001 GMTK-based SPINE ASR system (2002)
Özgür Çetin, Harriet J. Nock, Katrin Kirchhoff, Jeff Bilmes, Mari Ostendorf
This paper provides a detailed description of the University of Washington automatic speech recognition (ASR) system for the 2001 DARPA SPeech In Noisy Environments (SPINE) task. Our system makes...
Intransitive likelihood-ratio classifiers (2001)
Jeff Bilmes, Gang Ji, Marina Meil A
In this work, we introduce an information-theoretic based correction term to the likelihood ratio classification method for multiple classes. Under certain conditions, the term is sufficient for...
Statistical Models for Automatic Performance Tuning (2001)
Richard Vuduc, James W. Demmel, Jeff Bilmes
Achieving peak performance from library subroutines usually requires extensive, machine-dependent tuning by hand. Automatic tuning systems have emerged in response, and they typically operate by (1)...
Hidden-articulator Markov models: performance improvements and robustness to noise (2000)
Matt Richardson, Jeff Bilmes, Chris Diorio
A Hidden-Articulator Markov Model (HAMM) is a Hidden Markov Model (HMM) in which each state represents an articulatory configuration. Articulatory knowledge, known to be useful for speech recognition...
Mixed-Memory Markov Models For Automatic Language Identification (2000)
Katrin Kirchhoff, Sonia Parandekar, Jeff Bilmes
Automatic language identification (LID) continues to play an integral part in many multilingual speech applications. The most widespread approach to LID is the phonotactic approach, which performs...
Buried Markov Models for Speech Recognition (1999)
Matt Richardson, Jeff Bilmes, Chris Diorio
In traditional speech recognition using Hidden Markov Models (HMMs), each state represents an acoustic portion of a phoneme. We explore the concept of an articulator based HMM, where each state...
The PHiPAC v1.0 matrix-multiply distribution (1998)
Jeff Bilmes, Krsteasanović Y, Jim Demmel X
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. We have developed a methodology whereby near-peak...
The PHiPAC v1.0 Matrix-Multiply Distribution. (1998)
Jeff Bilmes, Krste Asanovic, Chee-Whye Chin, Jim Demmel
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. We have developed a methodology whereby near-peak...
The PHiPAC v1.0 matrix-multiply distribution (1998)
Krste Asanović, Jim Demmel, Jeff Bilmes, Jeff Bilmes, Krsteasanović K, Chee-whye Chin, ...
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. We have developed a methodology whereby near-peak...
Jeff Bilmes, Krste Asanovićy, Chee-whye Chinz, Jim Demmelx
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. We have developed a methodology whereby near-peak...
Using Phipac To Speed Error Back-Propagation Learning (1997)
Jeff Bilmes, Krste Asanovic, Chee-whye Chin, Jim Demmel
We introduce PHiPAC, a coding methodology for developing portable high-performance numerical libraries in ANSI C. Using this methodology, we have developed code for optimized matrix multiply...
Jeff Bilmes, Krste Asanovic, Chee-Whye Chin, Jim Demmel
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. Wehave developed a methodology whereby near-peak...
Empirical Observations of Probabilistic Heuristics for the Clustering Problem (1997)
Jeff Bilmes, Amin Vahdat, Windsor Hsu, Eun-Jin Im
We empirically investigate a number of strategies for solving the clustering problem under the minimum variance error criterion. First, we compare the behavior of four algorithms, 1) randomized...
Jeff Bilmes, Krste Asanovic, Jim Demmel, Dominic Lam, Chee-Whye Chin
BLAS3 operations have great potential for aggressive optimization. Unfortunately, they usually need to be hand-coded for a speci#c machine and compiler to achieve near-peak performance. Wehave...
Stochastic Perceptual Speech Models with Durational Dependence (1996)
Jeff Bilmes, Nelson Morgan, Su-lin Wu, Hervé Bourlard
In [6], we develop statistical model of speech recognition where emphasis is placed on the perceptually-relevant and information-rich portion of the speech signal. In that model, speech is viewed as...
Jeff Bilmes, Krste Asanovic, Chee-Whye Chin, Jim Demmel
Modern microprocessors can achieve high performance on linear algebra kernels but this currently requires extensive machine-specific hand tuning. We have developed a methodology whereby near-peak...
A Parallel Object-Oriented System for Realizing Reusable and Efficient Data Abstractions (1993)
Chu-Cheow Lim, Franco Mazzanti, Stephan Murer, Steve Omohundro, Thomas Rauber, ...
ions and Applications 182 4.1 Workbag : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 182 4.1.1 Overview : : : : : : : : : : : : : : : : : : : : : : : : : : : :...
Neurocomputing on the RAP (1992)
Nelson Morgan, Nelson Morgan, James Beck, James Beck, Phil Kohn, Phil Kohn, ...
In 1989 we designed and implemented a Ring Array Processor (RAP) for fast execution of our continuous speech recognition training algorithms, which have been dominated by connectionist calculations....
Ring Array Processor (RAP): Software Architecture (1991)
The design and implementation of software for the Ring Array Processor (RAP), a high performance parallel computer, involved development for three hardware platforms: Sun SPARC workstations, Heurikon...