Word boundary modelling and full covariance gaussians for Arabic Speech-to-Text systems

This paper describes recent improvements to the Cambridge Arabic Large Vocabulary Continuous Speech Recognition (LVCSR) Speech-to-Text (STT) system. It is shown that wordboundary context markers provide a powerful method to enhance graphemic systems by implicit phonetic information, improving the modelling capability of graphemic systems. In addition, a robust technique for full covariance Gaussian modelling in the Minimum Phone Error (MPE) training framework is introduced. This reduces the full covariance training to a diagonal covariance training problem, thereby solving related robustness problems. The full system results show that the combined use of these and other techniques within a multi-branch combination framework reduces the Word Error Rate (WER) of the complete system by up to 5.9% relative. Copyright © 2011 ISCA.

Combustion rumble prediction with integrated computational-fluid-dynamics/ low-order-model methods

The pressure oscillation within combustion chambers of aeroengines and industrial gas turbines is a major technical challenge to the development of high-performance and low-emission propulsion systems. In this paper, an approach integrating computational fluid dynamics and one-dimensional linear stability analysis is developed to predict the modes of oscillation in a combustor and their frequencies and growth rates. Linear acoustic theory was used to describe the acoustic waves propagating upstream and downstream of the combustion zone, which enables the computational fluid dynamics calculation to be efficiently concentrated on the combustion zone. A combustion oscillation was found to occur with its predicted frequency in agreement with experimental measurements. Furthermore, results from the computational fluid dynamics calculation provide the flame transfer function to describe unsteady heat release rate. Departures from ideal one-dimensional flows are described by shape factors. Combined with this information, low-order models can work out the possible oscillation modes and their initial growth rates. The approach developed here can be used in more general situations for the analysis of combustion oscillations. Copyright © 2012 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

Use of contexts in language model interpolation and adaptation

Language models (LMs) are often constructed by building multiple individual component models that are combined using context independent interpolation weights. By tuning these weights, using either perplexity or discriminative approaches, it is possible to adapt LMs to a particular task. This paper investigates the use of context dependent weighting in both interpolation and test-time adaptation of language models. Depending on the previous word contexts, a discrete history weighting function is used to adjust the contribution from each component model. As this dramatically increases the number of parameters to estimate, robust weight estimation schemes are required. Several approaches are described in this paper. The first approach is based on MAP estimation where interpolation weights of lower order contexts are used as smoothing priors. The second approach uses training data to ensure robust estimation of LM interpolation weights. This can also serve as a smoothing prior for MAP adaptation. A normalized perplexity metric is proposed to handle the bias of the standard perplexity criterion to corpus size. A range of schemes to combine weight information obtained from training data and test data hypotheses are also proposed to improve robustness during context dependent LM adaptation. In addition, a minimum Bayes' risk (MBR) based discriminative training scheme is also proposed. An efficient weighted finite state transducer (WFST) decoding algorithm for context dependent interpolation is also presented. The proposed technique was evaluated using a state-of-the-art Mandarin Chinese broadcast speech transcription task. Character error rate (CER) reductions up to 7.3 relative were obtained as well as consistent perplexity improvements. © 2012 Elsevier Ltd. All rights reserved.

IPLR: An online resource for Greek word-level and sublexical information

We present a new online psycholinguistic resource for Greek based on analyses of written corpora combined with text processing technologies developed at the Institute for Language & Speech Processing (ILSP), Greece. The "ILSP PsychoLinguistic Resource" (IPLR) is a freely accessible service via a dedicated web page, at http://speech.ilsp.gr/iplr. IPLR provides analyses of user-submitted letter strings (words and nonwords) as well as frequency tables for important units and conditions such as syllables, bigrams, and neighbors, calculated over two word lists based on printed text corpora and their phonetic transcription. Online tools allow retrieval of words matching user-specified orthographic or phonetic patterns. All results and processing code (in the Python programming language) are freely available for noncommercial educational or research use. © 2010 Springer Science+Business Media B.V.

Semi-implicit second order finite volume jet stream computations

Semi-implicit, second order temporal and spatial finite volume computations of the flow in a differentially heated rotating annulus are presented. For the regime considered, three cyclones and anticyclones separated by a relatively fast moving jet of fluid or "jet stream" are predicted. Two second order methods are compared with, first order spatial predictions, and experimental measurements. Velocity vector plots are used to illustrate the predicted flow structure. Computations made using second order central differences are shown to agree best with experimental measurements, and to be stable for integrations over long time periods (> 1000s). No periodic smoothing is required to prevent divergence.

Discriminative spoken language understanding using word confusion networks

Current commercial dialogue systems typically use hand-crafted grammars for Spoken Language Understanding (SLU) operating on the top one or two hypotheses output by the speech recogniser. These systems are expensive to develop and they suffer from significant degradation in performance when faced with recognition errors. This paper presents a robust method for SLU based on features extracted from the full posterior distribution of recognition hypotheses encoded in the form of word confusion networks. Following [1], the system uses SVM classifiers operating on n-gram features, trained on unaligned input/output pairs. Performance is evaluated on both an off-line corpus and on-line in a live user trial. It is shown that a statistical discriminative approach to SLU operating on the full posterior ASR output distribution can substantially improve performance both in terms of accuracy and overall dialogue reward. Furthermore, additional gains can be obtained by incorporating features from the previous system output. © 2012 IEEE.

High-order DES and zonal LES of a labyrinth seal

Hybrid numerical large eddy simulation (NLES) and detached eddy simulation (DES) methods are assessed on a labyrinth seal geometry. A high sixth order discretization scheme is used and is validated using a test case of a two dimensional vortex. The hybrid approach adopts a new blending function and along with DES is initially validated using a simple cavity flow. The NLES method is also validated outside of RANS zones. It is found that there is very little resolved turbulence in the cavity for the DES simulation. For the labyrinth seal calculations the DES approach is problematic giving virtually no resolved turbulence content. It is seen that over the tooth tips the extent of the LES region is small and is likely to be a strong contributor to excessive flow damping in these regions. On the other hand the zonal Hamilton-Jacobi approach did not suffer from this trait. In both cases the meshes used are considered to be hybrid RANS-LES adequate. Fortunately (or perhaps unfortunately) the DES profiles are in agreement with the time mean experimental measurements. It is concluded that for an inexperienced CFD practitioner this could have wider implications particularly if transient results such as unsteady loading are desired. Copyright © 2012 by the American Institute of Aeronautics and Astronautics, Inc.

Temporal difference models describe higher-order learning in humans.

The ability to use environmental stimuli to predict impending harm is critical for survival. Such predictions should be available as early as they are reliable. In pavlovian conditioning, chains of successively earlier predictors are studied in terms of higher-order relationships, and have inspired computational theories such as temporal difference learning. However, there is at present no adequate neurobiological account of how this learning occurs. Here, in a functional magnetic resonance imaging (fMRI) study of higher-order aversive conditioning, we describe a key computational strategy that humans use to learn predictions about pain. We show that neural activity in the ventral striatum and the anterior insula displays a marked correspondence to the signals for sequential learning predicted by temporal difference models. This result reveals a flexible aversive learning process ideally suited to the changing and uncertain nature of real-world environments. Taken with existing data on reward learning, our results suggest a critical role for the ventral striatum in integrating complex appetitive and aversive predictions to coordinate behaviour.

Shape signifiers for control of a low-order compressor model

Rotating stall and surge, two instability mechanisms limiting the performance of aeroengines compressors, are studied on the third-order Moore-Greitzer model. The skewness of the compressor characteristic, a single parameter shape signifier, is shown to determine the key qualitative properties of feedback control.

Using sub-word-level information for confidence estimation with conditional random field models

The task of word-level confidence estimation (CE) for automatic speech recognition (ASR) systems stands to benefit from the combination of suitably defined input features from multiple information sources. However, the information sources of interest may not necessarily operate at the same level of granularity as the underlying ASR system. The research described here builds on previous work on confidence estimation for ASR systems using features extracted from word-level recognition lattices, by incorporating information at the sub-word level. Furthermore, the use of Conditional Random Fields (CRFs) with hidden states is investigated as a technique to combine information for word-level CE. Performance improvements are shown using the sub-word-level information in linear-chain CRFs with appropriately engineered feature functions, as well as when applying the hidden-state CRF model at the word level.