Phone-level pronunciation scoring and assessment for interactive language learning

This paper investigates a method of automatic pronunciation scoring for use in computer-assisted language learning (CALL) systems. The method utilizes a likelihood-based `Goodness of Pronunciation' (GOP) measure which is extended to include individual thresholds for each phone based on both averaged native confidence scores and on rejection statistics provided by human judges. Further improvements are obtained by incorporating models of the subject's native language and by augmenting the recognition networks to include expected pronunciation errors. The various GOP measures are assessed using a specially recorded database of non-native speakers which has been annotated to mark phone-level pronunciation errors. Since pronunciation assessment is highly subjective, a set of four performance measures has been designed, each of them measuring different aspects of how well computer-derived phone-level scores agree with human scores. These performance measures are used to cross-validate the reference annotations and to assess the basic GOP algorithm and its refinements. The experimental results suggest that a likelihood-based pronunciation scoring metric can achieve usable performance, especially after applying the various enhancements.

Thermal building simulation using the UKCP09 probabilistic climate projections

This article investigates how to use UK probabilistic climate-change projections (UKCP09) in rigorous building energy analysis. Two office buildings (deep plan and shallow plan) are used as case studies to demonstrate the application of UKCP09. Three different methods for reducing the computational demands are explored: statistical reduction (Finkelstein-Schafer [F-S] statistics), simplification using degree-day theory and the use of metamodels. The first method, which is based on an established technique, can be used as reference because it provides the most accurate information. However, it is necessary to automatically choose weather files based on F-S statistic by using computer programming language because thousands of weather files created from UKCP09 weather generator need to be processed. A combination of the second (degree-day theory) and third method (metamodels) requires only a relatively small number of simulation runs, but still provides valuable information to further implement the uncertainty and sensitivity analyses. The article also demonstrates how grid computing can be used to speed up the calculation for many independent EnergyPlus models by harnessing the processing power of idle desktop computers. © 2011 International Building Performance Simulation Association (IBPSA).

Thermal-creep analysis of concrete bridges

The creep effects on sequentially built bridges are analysed by the theory of thermal creep. Two types of analysis are used: time dependent and steady state. The traditional uniform creep analysis is also introduced briefly. Both simplified and parabolic normalising creep-temperature functions are used in the analysis for comparison. Numerical examples are presented, calculated by a computer program based on the theory of thermal creep and using the displacement method. It is concluded that different assumptions within thermal creep can lead to very different results when compared with uniform creep analysis. The steady-state analysis of monolithically built structures can serve as a limit to evaluate total creep effects for both monolithically and sequentially built structures. The importance of the correct selection of the normalising creep-temperature function is demonstrated.

The deposition of small particles from a turbulent air flow in a curved duct

The paper describes an experimental and theoretical study of the deposition of small spherical particles from a turbulent air flow in a curved duct. The objective was to investigate the interaction between the streamline curvature of the primary flow and the turbulent deposition mechanisms of diffusion and turbophoresis. The experiments were conducted with particles of uranine (used as a fluorescent tracer) produced by an aerosol generator. The particles were entrained in an air flow which passed vertically downwards through a long straight channel of rectangular cross-section leading to a 90° bend. The inside surfaces of the channel and bend were covered with tape to collect the deposited particles. Following a test run the tape was removed in sections, the uranine was dissolved in sodium hydroxide solution and the deposition rates established by measuring the uranine concentration with a luminescence spectrometer. The experimental results were compared with calculations of particle deposition in a curved duct using a computer program that solved the ensemble-averaged particle mass and momentum conservation equations. A particle density-weighted averaging procedure was used and the equations were expressed in terms of the particle convective, rather than total, velocity. This approach provided a simpler formulation of the particle turbulence correlations generated by the averaging process. The computer program was used to investigate the distance required to achieve a fully-developed particle flow in the straight entry channel as well as the variation of the deposition rate around the bend. The simulations showed good agreement with the experimental results. © 2012 Elsevier Ltd.

The aeroacoustics of transonic helicopter blades

One of the most important issues facing the helicopter industry today is helicopter noise, in particular transonic rotor noise. It is the main factor limiting cruise speeds, and there is real demand for efficient and reliable prediction methods which can be used in the rotor design process. This paper considers the Ffowcs Williams-Hawkings equation applied to a permeable control surface. The surface is chosen to be as small as possible, while enclosing both the blade and any transonic flow regions. This allows the problematic quadrupole term to always be neglected, and requires only near field CFD input data. It is therefore less computationally intensive than existing prediction methods, and moreover retains the physical interpretation of the sources in terms of thickness, loading and shock-associated noise. A computer program has been developed which implements the permeable surface form of retarded time formulation. The program has been validated and subsequently used to validate an acoustic 2-D CFD code. It is fast and reliable for subsonic motion, but it is demonstrated that it cannot be used at high subsonic or supersonic speeds. A second computer program implementing a more general formulation has also been developed and is presently being validated. This general formulation can be applied at high subsonic and supersonic speeds, except under one specific condition. © 2002 by the author(s). Published by the American Institute of Aeronautics and Astronautics, Inc.

AI in the 21st century - With historical reflections

The discipline of Artificial Intelligence (AI) was born in the summer of 1956 at Dartmouth College in Hanover, New Hampshire. Half of a century has passed, and AI has turned into an important field whose influence on our daily lives can hardly be overestimated. The original view of intelligence as a computer program - a set of algorithms to process symbols - has led to many useful applications now found in internet search engines, voice recognition software, cars, home appliances, and consumer electronics, but it has not yet contributed significantly to our understanding of natural forms of intelligence. Since the 1980s, AI has expanded into a broader study of the interaction between the body, brain, and environment, and how intelligence emerges from such interaction. This advent of embodiment has provided an entirely new way of thinking that goes well beyond artificial intelligence proper, to include the study of intelligent action in agents other than organisms or robots. For example, it supplies powerful metaphors for viewing corporations, groups of agents, and networked embedded devices as intelligent and adaptive systems acting in highly uncertain and unpredictable environments. In addition to giving us a novel outlook on information technology in general, this broader view of AI also offers unexpected perspectives into how to think about ourselves and the world around us. In this chapter, we briefly review the turbulent history of AI research, point to some of its current trends, and to challenges that the AI of the 21st century will have to face. © Springer-Verlag Berlin Heidelberg 2007.

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.

Language model cross adaptation for LVCSR system combination

State-of-the-art large vocabulary continuous speech recognition (LVCSR) systems often combine outputs from multiple sub-systems that may even be developed at different sites. Cross system adaptation, in which model adaptation is performed using the outputs from another sub-system, can be used as an alternative to hypothesis level combination schemes such as ROVER. Normally cross adaptation is only performed on the acoustic models. However, there are many other levels in LVCSR systems' modelling hierarchy where complimentary features may be exploited, for example, the sub-word and the word level, to further improve cross adaptation based system combination. It is thus interesting to also cross adapt language models (LMs) to capture these additional useful features. In this paper cross adaptation is applied to three forms of language models, a multi-level LM that models both syllable and word sequences, a word level neural network LM, and the linear combination of the two. Significant error rate reductions of 4.0-7.1% relative were obtained over ROVER and acoustic model only cross adaptation when combining a range of Chinese LVCSR sub-systems used in the 2010 and 2011 DARPA GALE evaluations. © 2012 Elsevier Ltd. All rights reserved.

Discriminative language model adaptation for Mandarin broadcast speech transcription and translation

This paper investigates unsupervised test-time adaptation of language models (LM) using discriminative methods for a Mandarin broadcast speech transcription and translation task. A standard approach to adapt interpolated language models to is to optimize the component weights by minimizing the perplexity on supervision data. This is a widely made approximation for language modeling in automatic speech recognition (ASR) systems. For speech translation tasks, it is unclear whether a strong correlation still exists between perplexity and various forms of error cost functions in recognition and translation stages. The proposed minimum Bayes risk (MBR) based approach provides a flexible framework for unsupervised LM adaptation. It generalizes to a variety of forms of recognition and translation error metrics. LM adaptation is performed at the audio document level using either the character error rate (CER), or translation edit rate (TER) as the cost function. An efficient parameter estimation scheme using the extended Baum-Welch (EBW) algorithm is proposed. Experimental results on a state-of-the-art speech recognition and translation system are presented. The MBR adapted language models gave the best recognition and translation performance and reduced the TER score by up to 0.54% absolute. © 2007 IEEE.

Language model combination and adaptation using weighted finite state transducers

In speech recognition systems language model (LMs) are often constructed by training and combining multiple n-gram models. They can be either used to represent different genres or tasks found in diverse text sources, or capture stochastic properties of different linguistic symbol sequences, for example, syllables and words. Unsupervised LM adaptation may also be used to further improve robustness to varying styles or tasks. When using these techniques, extensive software changes are often required. In this paper an alternative and more general approach based on weighted finite state transducers (WFSTs) is investigated for LM combination and adaptation. As it is entirely based on well-defined WFST operations, minimum change to decoding tools is needed. A wide range of LM combination configurations can be flexibly supported. An efficient on-the-fly WFST decoding algorithm is also proposed. Significant error rate gains of 7.3% relative were obtained on a state-of-the-art broadcast audio recognition task using a history dependently adapted multi-level LM modelling both syllable and word sequences. ©2010 IEEE.