Combining information sources for confidence estimation with CRF models

Obtaining accurate confidence measures for automatic speech recognition (ASR) transcriptions is an important task which stands to benefit from the use of multiple information sources. This paper investigates the application of conditional random field (CRF) models as a principled technique for combining multiple features from such sources. A novel method for combining suitably defined features is presented, allowing for confidence annotation using lattice-based features of hypotheses other than the lattice 1-best. The resulting framework is applied to different stages of a state-of-the-art large vocabulary speech recognition pipeline, and consistent improvements are shown over a sophisticated baseline system. Copyright © 2011 ISCA.

Phase perfection in zinc Blende and Wurtzite III-V nanowires using basic growth parameters.

Controlling the crystallographic phase purity of III-V nanowires is notoriously difficult, yet this is essential for future nanowire devices. Reported methods for controlling nanowire phase require dopant addition, or a restricted choice of nanowire diameter, and only rarely yield a pure phase. Here we demonstrate that phase-perfect nanowires, of arbitrary diameter, can be achieved simply by tailoring basic growth parameters: temperature and V/III ratio. Phase purity is achieved without sacrificing important specifications of diameter and dopant levels. Pure zinc blende nanowires, free of twin defects, were achieved using a low growth temperature coupled with a high V/III ratio. Conversely, a high growth temperature coupled with a low V/III ratio produced pure wurtzite nanowires free of stacking faults. We present a comprehensive nucleation model to explain the formation of these markedly different crystal phases under these growth conditions. Critical to achieving phase purity are changes in surface energy of the nanowire side facets, which in turn are controlled by the basic growth parameters of temperature and V/III ratio. This ability to tune crystal structure between twin-free zinc blende and stacking-fault-free wurtzite not only will enhance the performance of nanowire devices but also opens new possibilities for engineering nanowire devices, without restrictions on nanowire diameters or doping.

The rocking response of large flexible structures to earthquakes

The rocking response of structures subjected to strong ground motions is a problem of 'several scales'. While small structures are sensitive to acceleration pulses acting successively, large structures are more significantly affected by coherent low frequency components of ground motion. As a result, the rocking response of large structures is more stable and orderly, allowing effective isolation from the ground without imminent danger of overturning. This paper aims to characterize and predict the maximum rocking response of large and flexible structures to earthquakes using an idealized structural model. To achieve this, the maximum rocking demand caused by different earthquake records was evaluated using several ground motion intensity measures. Pulse-type records which typically have high peak ground velocity and lower frequency content caused large rocking amplitudes, whereas non-pulse type records caused random rocking motion confined to small rocking amplitudes. Coherent velocity pulses were therefore identified as the primary cause of significant rocking motion. Using a suite of pulse-type ground motions, it was observed that idealized wavelets fitted to velocity pulses can adequately describe the rocking response of large structures. Further, a parametric analysis demonstrates that pulse shape parameters affect the maximum rocking response significantly. Based on these two findings, a probabilistic analysis method is proposed for estimating the maximum rocking demand to pulse-type earthquakes. The dimensionless demand maps, produced using these methods, have predictive power in the near-field provided that pulse period and amplitude can be estimated a priori. Use of this method within a probabilistic seismic demand analysis framework is briefly discussed. © 2013 Springer Science+Business Media Dordrecht.