A cluster-voting approach for speaker diarization and linking of Australian broadcast news recordings

We present a clustering-only approach to the problem of speaker diarization to eliminate the need for the commonly employed and computationally expensive Viterbi segmentation and realignment stage. We use multiple linear segmentations of a recording and carry out complete-linkage clustering within each segmentation scenario to obtain a set of clustering decisions for each case. We then collect all clustering decisions, across all cases, to compute a pairwise vote between the segments and conduct complete-linkage clustering to cluster them at a resolution equal to the minimum segment length used in the linear segmentations. We use our proposed cluster-voting approach to carry out speaker diarization and linking across the SAIVT-BNEWS corpus of Australian broadcast news data. We compare our technique to an equivalent baseline system with Viterbi realignment and show that our approach can outperform the baseline technique with respect to the diarization error rate (DER) and attribution error rate (AER).

Engaging a multiple-track approach to building capacity for 21st Century engineering, opportunities and challenges for rapid curriculum renewal

‘Complexity’ is a term that is increasingly prevalent in conversations about building capacity for 21st Century professional engineers. Society is grappling with the urgent and challenging reality of accommodating seven billion people, meeting needs and innovating lifestyle improvements in ways that do not destroy atmospheric, biological and oceanic systems critical to life. Over the last two decades in particular, engineering educators have been active in attempting to build capacity amongst professionals to deliver ‘sustainable development’ in this rapidly changing global context. However curriculum literature clearly points to a lack of significant progress, with efforts best described as ad hoc and highly varied. Given the limited timeframes for action to curb environmental degradation proposed by scientists and intergovernmental agencies, the authors of this paper propose it is imperative that curriculum renewal towards education for sustainable development proceeds rapidly, systemically, and in a transformational manner. Within this context, the paper discusses the need to consider a multiple track approach to building capacity for 21st Century engineering, including priorities and timeframes for undergraduate and postgraduate curriculum renewal. The paper begins with a contextual discussion of the term complexity and how it relates to life in the 21st Century. The authors then present a whole of system approach for planning and implementing rapid curriculum renewal that addresses the critical roles of several generations of engineering professionals over the next three decades. The paper concludes with observations regarding engaging with this approach in the context of emerging accreditation requirements and existing curriculum renewal frameworks.

Automated 3D mapping & shape analysis of the lateral ventricles via fluid registration of multiple surface-based atlases

We developed and validated a new method to create automated 3D parametric surface models of the lateral ventricles, designed for monitoring degenerative disease effects in clinical neuroscience studies and drug trials. First we used a set of parameterized surfaces to represent the ventricles in a manually labeled set of 9 subjects' MRIs (atlases). We fluidly registered each of these atlases and mesh models to a set of MRIs from 12 Alzheimer's disease (AD) patients and 14 matched healthy elderly subjects, and we averaged the resulting meshes for each of these images. Validation experiments on expert segmentations showed that (1) the Hausdorff labeling error rapidly decreased, and (2) the power to detect disease-related alterations monotonically improved as the number of atlases, N, was increased from 1 to 9. We then combined the segmentations with a radial mapping approach to localize ventricular shape differences in patients. In surface-based statistical maps, we detected more widespread and intense anatomical deficits as we increased the number of atlases, and we formulated a statistical stopping criterion to determine the optimal value of N. Anterior horn anomalies in Alzheimer's patients were only detected with the multi-atlas segmentation, which clearly outperformed the standard single-atlas approach.

fMRI evidence of word frequency and strength effects in recognition memory

We used event-related fMRI to investigate the neural correlates of encoding strength and word frequency effects in recognition memory. At test, participants made Old/New decisions to intermixed low (LF) and high frequency (HF) words that had been presented once or twice at study and to new, unstudied words. The Old/New effect for all hits vs. correctly rejected unstudied words was associated with differential activity in multiple cortical regions, including the anterior medial temporal lobe (MTL), hippocampus, left lateral parietal cortex and anterior left inferior prefrontal cortex (LIPC). Items repeated at study had superior hit rates (HR) compared to items presented once and were associated with reduced activity in the right anterior MTL. By contrast, other regions that had shown conventional Old/New effects did not demonstrate modulation according to memory strength. A mirror effect for word frequency was demonstrated, with the LF word HR advantage associated with increased activity in the left lateral temporal cortex. However, none of the regions that had demonstrated Old/New item retrieval effects showed modulation according to word frequency. These findings are interpreted as supporting single-process memory models proposing a unitary strength-like memory signal and models attributing the LF word HR advantage to the greater lexico-semantic context-noise associated with HF words due to their being experienced in many pre-experimental contexts.

Labeling white matter tracts in hardi by fusing multiple tract atlases with applications to genetics

Accurate identification of white matter structures and segmentation of fibers into tracts is important in neuroimaging and has many potential applications. Even so, it is not trivial because whole brain tractography generates hundreds of thousands of streamlines that include many false positive fibers. We developed and tested an automatic tract labeling algorithm to segment anatomically meaningful tracts from diffusion weighted images. Our multi-atlas method incorporates information from multiple hand-labeled fiber tract atlases. In validations, we showed that the method outperformed the standard ROI-based labeling using a deformable, parcellated atlas. Finally, we show a high-throughput application of the method to genetic population studies. We use the sub-voxel diffusion information from fibers in the clustered tracts based on 105-gradient HARDI scans of 86 young normal twins. The whole workflow shows promise for larger population studies in the future.

Automatic population HARDI white matter tract clustering by label fusion of multiple tract atlases

Automatic labeling of white matter fibres in diffusion-weighted brain MRI is vital for comparing brain integrity and connectivity across populations, but is challenging. Whole brain tractography generates a vast set of fibres throughout the brain, but it is hard to cluster them into anatomically meaningful tracts, due to wide individual variations in the trajectory and shape of white matter pathways. We propose a novel automatic tract labeling algorithm that fuses information from tractography and multiple hand-labeled fibre tract atlases. As streamline tractography can generate a large number of false positive fibres, we developed a top-down approach to extract tracts consistent with known anatomy, based on a distance metric to multiple hand-labeled atlases. Clustering results from different atlases were fused, using a multi-stage fusion scheme. Our "label fusion" method reliably extracted the major tracts from 105-gradient HARDI scans of 100 young normal adults. © 2012 Springer-Verlag.

Predicting white matter integrity from multiple common genetic variants

Several common genetic variants have recently been discovered that appear to influence white matter microstructure, as measured by diffusion tensor imaging (DTI). Each genetic variant explains only a small proportion of the variance in brain microstructure, so we set out to explore their combined effect on the white matter integrity of the corpus callosum. We measured six common candidate single-nucleotide polymorphisms (SNPs) in the COMT, NTRK1, BDNF, ErbB4, CLU, and HFE genes, and investigated their individual and aggregate effects on white matter structure in 395 healthy adult twins and siblings (age: 20-30 years). All subjects were scanned with 4-tesla 94-direction high angular resolution diffusion imaging. When combined using mixed-effects linear regression, a joint model based on five of the candidate SNPs (COMT, NTRK1, ErbB4, CLU, and HFE) explained ∼ 6% of the variance in the average fractional anisotropy (FA) of the corpus callosum. This predictive model had detectable effects on FA at 82% of the corpus callosum voxels, including the genu, body, and splenium. Predicting the brain's fiber microstructure from genotypes may ultimately help in early risk assessment, and eventually, in personalized treatment for neuropsychiatric disorders in which brain integrity and connectivity are affected.

Multiple stages classification of Alzheimer’s disease based on structural brain networks using generalized low rank approximations (GLRAM)

To classify each stage for a progressing disease such as Alzheimer’s disease is a key issue for the disease prevention and treatment. In this study, we derived structural brain networks from diffusion-weighted MRI using whole-brain tractography since there is growing interest in relating connectivity measures to clinical, cognitive, and genetic data. Relatively little work has usedmachine learning to make inferences about variations in brain networks in the progression of the Alzheimer’s disease. Here we developed a framework to utilize generalized low rank approximations of matrices (GLRAM) and modified linear discrimination analysis for unsupervised feature learning and classification of connectivity matrices. We apply the methods to brain networks derived from DWI scans of 41 people with Alzheimer’s disease, 73 people with EMCI, 38 people with LMCI, 47 elderly healthy controls and 221 young healthy controls. Our results show that this new framework can significantly improve classification accuracy when combining multiple datasets; this suggests the value of using data beyond the classification task at hand to model variations in brain connectivity.

Estimating the under-reporting of road crash injuries to police using multiple linked data collections

The reliance on police data for the counting of road crash injuries can be problematic, as it is well known that not all road crash injuries are reported to police which under-estimates the overall burden of road crash injuries. The aim of this study was to use multiple linked data sources to estimate the extent of under-reporting of road crash injuries to police in the Australian state of Queensland. Data from the Queensland Road Crash Database (QRCD), the Queensland Hospital Admitted Patients Data Collection (QHAPDC), Emergency Department Information System (EDIS), and the Queensland Injury Surveillance Unit (QISU) for the year 2009 were linked. The completeness of road crash cases reported to police was examined via discordance rates between the police data (QRCD) and the hospital data collections. In addition, the potential bias of this discordance (under-reporting) was assessed based on gender, age, road user group, and regional location. Results showed that the level of under-reporting varied depending on the data set with which the police data was compared. When all hospital data collections are examined together the estimated population of road crash injuries was approximately 28,000, with around two-thirds not linking to any record in the police data. The results also showed that the under-reporting was more likely for motorcyclists, cyclists, males, young people, and injuries occurring in Remote and Inner Regional areas. These results have important implications for road safety research and policy in terms of: prioritising funding and resources; targeting road safety interventions into areas of higher risk; and estimating the burden of road crash injuries.

Grassland productivity limited by multiple nutrients

Terrestrial ecosystem productivity is widely accepted to be nutrient limited1. Although nitrogen (N) is deemed a key determinant of aboveground net primary production (ANPP)2,3, the prevalence of co-limitation by N and phosphorus (P) is increasingly recognized4,​5,​6,​7,​8. However, the extent to which terrestrial productivity is co-limited by nutrients other than N and P has remained unclear. Here, we report results from a standardized factorial nutrient addition experiment, in which we added N, P and potassium (K) combined with a selection of micronutrients (K+μ), alone or in concert, to 42 grassland sites spanning five continents, and monitored ANPP. Nutrient availability limited productivity at 31 of the 42 grassland sites. And pairwise combinations of N, P, and K+μ co-limited ANPP at 29 of the sites. Nitrogen limitation peaked in cool, high latitude sites. Our findings highlight the importance of less studied nutrients, such as K and micronutrients, for grassland productivity, and point to significant variations in the type and degree of nutrient limitation. We suggest that multiple-nutrient constraints must be considered when assessing the ecosystem-scale consequences of nutrient enrichment.

Non-parametric consistency test for multiple-sensing-modality data fusion

Fusing data from multiple sensing modalities, e.g. laser and radar, is a promising approach to achieve resilient perception in challenging environmental conditions. However, this may lead to \emph{catastrophic fusion} in the presence of inconsistent data, i.e. when the sensors do not detect the same target due to distinct attenuation properties. It is often difficult to discriminate consistent from inconsistent data across sensing modalities using local spatial information alone. In this paper we present a novel consistency test based on the log marginal likelihood of a Gaussian process model that evaluates data from range sensors in a relative manner. A new data point is deemed to be consistent if the model statistically improves as a result of its fusion. This approach avoids the need for absolute spatial distance threshold parameters as required by previous work. We report results from object reconstruction with both synthetic and experimental data that demonstrate an improvement in reconstruction quality, particularly in cases where data points are inconsistent yet spatially proximal.

Service-learning through multiple client-sponsored projects in an MBA marketing course

This paper outlines an approach for teaching Marketing Principles in an MBA course through service-learning to enable adult learners to connect the lectures’ marketing content to a real-world marketing project. During the course, 40 students in groups of four to five individuals were involved in eight different client-sponsored marketing projects executed simultaneously. The rationale, planning and management of this approach utilised current research on service-learning, living cases and client-sponsored projects in marketing education. The experimental curriculum design is presented in a timeline that mirrors the preparation and management of the group projects and the considerations to be taken into account when initiating and facilitating the projects. Reflections from this iteration of the service-learning design suggest the importance of: detailed project planning, the involvement of students in choosing the projects, the introduction of forms and feedback loops, the role of the instructor in facilitating the students and managing expectations, and the role of the company representative in supporting the groups.

Multiple sulfur isotope analyses support a magmatic model for the volcanogenic massive sulfide deposits of the Teutonic Bore volcanic complex, Yilgarn Craton, Western Australia

We report sensitive high mass resolution ion microprobe, stable isotopes (SHRIMP SI) multiple sulfur isotope analyses (32S, 33S, 34S) to constrain the sources of sulfur in three Archean VMS deposits—Teutonic Bore, Bentley, and Jaguar—from the Teutonic Bore volcanic complex of the Yilgarn Craton, Western Australia, together with sedimentary pyrites from associated black shales and interpillow pyrites. The pyrites from VMS mineralization are dominated by mantle sulfur but include a small amount of slightly negative mass-independent fractionation (MIF) anomalies, whereas sulfur from the pyrites in the sedimentary rocks has pronounced positive MIF, with ∆33S values that lie between 0.19 and 6.20‰ (with one outlier at −1.62‰). The wall rocks to the mineralization include sedimentary rocks that have contributed no detectable positive MIF sulfur to the VMS deposits, which is difficult to reconcile with the leaching model for the formation of these deposits. The sulfur isotope data are best explained by mixing between sulfur derived from a magmatic-hydrothermal fluid and seawater sulfur as represented by the interpillow pyrites. The massive sulfide lens pyrites have a weighted mean ∆33S value of −0.27 ± 0.05‰ (MSWD = 1.6) nearly identical with −0.31 ± 0.08‰ (MSWD = 2.4) for pyrites from the stringer zone, which requires mixing to have occurred below the sea floor. We employed a two-component mixing model to estimate the contribution of seawater sulfur to the total sulfur budget of the two Teutonic Bore volcanic complex VMS deposits. The results are 15 to 18% for both Teutonic Bore and Bentley, much higher than the 3% obtained by Jamieson et al. (2013) for the giant Kidd Creek deposit. Similar calculations, carried out for other Neoarchean VMS deposits give value between 2% and 30%, which are similar to modern hydrothermal VMS deposits. We suggest that multiple sulfur isotope analyses may be used to predict the size of Archean VMS deposits and to provide a vector to ore deposit but further studies are needed to test these suggestions.

Aspect-based opinion mining from product reviews using conditional random fields

Product reviews are the foremost source of information for customers and manufacturers to help them make appropriate purchasing and production decisions. Natural language data is typically very sparse; the most common words are those that do not carry a lot of semantic content, and occurrences of any particular content-bearing word are rare, while co-occurrences of these words are rarer. Mining product aspects, along with corresponding opinions, is essential for Aspect-Based Opinion Mining (ABOM) as a result of the e-commerce revolution. Therefore, the need for automatic mining of reviews has reached a peak. In this work, we deal with ABOM as sequence labelling problem and propose a supervised extraction method to identify product aspects and corresponding opinions. We use Conditional Random Fields (CRFs) to solve the extraction problem and propose a feature function to enhance accuracy. The proposed method is evaluated using two different datasets. We also evaluate the effectiveness of feature function and the optimisation through multiple experiments.

Decisions and actions of distracted drivers at the onset of yellow light

Driving on an approach to a signalized intersection while distracted is relatively risky, as potential vehicular conflicts and resulting angle collisions tend to be relatively more severe compared to other locations. Given the prevalence and importance of this particular scenario, the objective of this study was to examine the decisions and actions of distracted drivers during the onset of yellow lights. Driving simulator data were obtained from a sample of 69 drivers under baseline and handheld cell phone conditions at the University of Iowa – National Advanced Driving Simulator. Explanatory variables included age, gender, cell phone use, distance to stop-line, and speed. Although there is extensive research on drivers’ responses to yellow traffic signals, the examinations have been conducted from a traditional regression-based approach, which do not necessary provide the underlying relations and patterns among the sampled data. In this paper, we exploit the benefits of both classical statistical inference and data mining techniques to identify the a priori relationships among main effects, non-linearities, and interaction effects. Results suggest that the probability of yellow light running increases with the increase in driving speed at the onset of yellow. Both young (18–25 years) and middle-aged (30–45 years) drivers reveal reduced propensity for yellow light running whilst distracted across the entire speed range, exhibiting possible risk compensation during this critical driving situation. The propensity for yellow light running for both distracted male and female older (50–60 years) drivers is significantly higher. Driver experience captured by age interacts with distraction, resulting in their combined effect having slower physiological response and being distracted particularly risky.