Listening like a speech-training app: Expert and non-expert listeners' goodness ratings of children's speech.

Speech training apps are being developed that provide automatic feedback concerning children's production of known target words, as a score on a 1-5 scale. However, this 'goodness' scale is still poorly understood. We investigated listeners' ratings of 'how many stars the app should provide as feedback' on children's utterances, and whether listener agreement is affected by clinical experience and/or access to anchor stimuli. In addition, we explored the association between goodness ratings and clinical measures of speech accuracy; the Percentage of Consonants Correct (PCC) and the Percentage of Phonemes Correct (PPC). Twenty speech-language pathologists and 20 non-expert listeners participated; half of the listeners in each group had access to anchor stimuli. The listeners rated 120 words, collected from children with and without speech sound disorder. Concerning reliability, intra-rater agreement was generally high, whereas inter-rater agreement was moderate. Access to anchor stimuli was associated with higher agreement, but only for non-expert listeners. Concerning the association between goodness ratings and the PCC/PPC, correlations were moderate for both listener groups, under both conditions. The results indicate that the task of rating goodness is difficult, regardless of clinical experience, and that access to anchor stimuli is insufficient for achieving reliable ratings. This raises concerns regarding the 1-5 rating scale as the means of feedback in speech training apps. More specific listener instructions, particularly regarding the intended context for the app, are suggested in collection of human ratings underlying the development of speech training apps. Until then, alternative means of feedback should be preferred.

Lahjoita puhetta: a large-scale corpus of spoken Finnish with some benchmarks.

The Donate Speech campaign has so far succeeded in gathering approximately 3600 h of ordinary, colloquial Finnish speech into the Lahjoita puhetta (Donate Speech) corpus. The corpus includes over twenty thousand speakers from all the regions of Finland and from all age brackets. The primary goals of the collection were to create a representative, large-scale resource to study spontaneous spoken Finnish and to accelerate the development of language technology and speech-based services. In this paper, we present the collection process and the collected corpus, and showcase its versatility through multiple use cases. The evaluated use cases include: automatic speech recognition of spontaneous speech, detection of age, gender, dialect and topic and metadata analysis. We provide benchmarks for the use cases, as well downloadable, trained baseline systems with open-source code for reproducibility. One further use case is to verify the metadata and transcripts given in this corpus itself, and to suggest artificial metadata and transcripts for the part of the corpus where it is missing.

Automatic segmentation of hyperreflective foci in OCT images.

BACKGROUND AND OBJECTIVE: The leading cause of vision loss in the Western World is Age-related Macular Degeneration (AMD), but together with modern medicines, tracking the number of Hyperreflective Foci (HF) on Optical Coherence Tomography (OCT) images should assist the treatment of patients. Here, we developed a framework based on deep learning for the automatic segmentation of HF in OCT images. METHODS: We collected OCT images and annotated them, then these images underwent image preprocessing, and feature extraction steps. Using the prepared data we trained different types of Conventional-, Deep- and Convolutional Neural Networks to perform the task of the automatic segmentation of HF. RESULTS: We evaluated the various Neural Networks, by performing HF segmentation of clinical data belonging to patients, whose data were excluded from the training process. The results suggest that our systems can achieve reasonably high Dice Coefficient values, and they are comparable with (i.e., in most cases above 95%) the similarity between manual annotations performed by different physicians. CONCLUSION: From the results, it can be concluded that neural networks can be used to accurately segment HF in OCT images. The results are sufficiently accurate for us to incorporate them into the next phase of the research, building a decision support system for everyday clinical practice.

Hydrogen bond network topology in liquid water and methanol: a graph theory approach.

Networks are increasingly recognized as important building blocks of various systems in nature and society. Water is known to possess an extended hydrogen bond network, in which the individual bonds are broken in the sub-picosecond range and still the network structure remains intact. We investigated and compared the topological properties of liquid water and methanol at various temperatures using concepts derived within the framework of graph and network theory (neighbour number and cycle size distribution, the distribution of local cyclic and local bonding coefficients, Laplacian spectra of the network, inverse participation ratio distribution of the eigenvalues and average localization distribution of a node) and compared them to small world and Erdos-Rényi random networks. Various characteristic properties (e.g. the local cyclic and bonding coefficients) of the network in liquid water could be reproduced by small world and/or Erdos-Rényi networks, but the ring size distribution of water is unique and none of the studied graph models could describe it. Using the inverse participation ratio of the Laplacian eigenvectors we characterized the network inhomogeneities found in water and showed that similar phenomena can be observed in Erdos-Rényi and small world graphs. We demonstrated that the topological properties of the hydrogen bond network found in liquid water systematically change with the temperature and that increasing temperature leads to a broader ring size distribution. We applied the studied topological indices to the network of water molecules with four hydrogen bonds, and showed that at low temperature (250 K) these molecules form a percolated or nearly-percolated network, while at ambient or high temperatures only small clusters of four-hydrogen bonded water molecules exist.

Solution structure of NaNO3 in water: diffraction and molecular dynamics simulation study.

The structure of a series of aqueous sodium nitrate solutions (1.9-7.6 M) was studied using a combination of experimental and theoretical methods. The results obtained from diffraction (X-ray, neutron) and molecular dynamics simulation have been compared and the capabilities and limitations of the methods in describing solution structure are discussed. For the solutions studied, diffraction methods were found to perform very well in description of hydration spheres of the sodium ion but do not yield detailed structural information on the anion's hydration structure. Molecular dynamics simulations proved to be a suitable tool in the detailed interpretation of the hydration sphere of ions, ion pair formation, and bulk structure of solutions.

Water-methanol mixtures: topology of hydrogen bonded network.

Molecular dynamics simulation has been performed to study the structure of water-methanol mixtures. Besides the evaluation of partial radial distribution functions describing the hydrogen-bonded structure of the mixtures with different composition, the statistical analysis of configurations was introduced resulting in a new insight in the clustering properties and topology of hydrogen-bonded network. The results have shown that mixtures of methanol and water exhibit extended structures in solution. At low methanol concentration the water molecules form a percolated network, the methanol molecules are incorporated as monomers or short chains and together form a percolated system. In methanol-rich mixtures short water chains and longer methanol chains build up the hydrogen-bonded clusters in the system. On the basis of the statistical analysis of configurations obtained from molecular dynamics simulation it has been found that more methanol molecules are connected to non-cyclic entities, while more water molecules form rings that might have been predicted on the basis of the stoichiometry of the mixtures. This finding can be explained by the presence of microscopic configurational inhomogeneity in water-methanol mixtures.

Complete structural characterization of metallacyclic complexes in solution-phase using simultaneously X-ray diffraction and molecular dynamics simulation.

Wide-angle X-ray diffraction and molecular dynamics simulation has been used to perform complete structural characterization of nitromethane solution of a 16-membered gold(I) ring. The joint application of these two methods was an adequate tool to describe not only the structure of the complex but also the solvation properties of the complex in nitromethane and the effect of the solvation on the bulk structure. It has been found that a relatively diffuse slightly distorted solvation shell is formed around the complex, following the shape of the molecule. Nitromethane molecules in the solvation sphere are distributed randomly; no special orientation can be detected. The interaction energy of the complex with nitromethane molecules is attractive. In bulk, besides the antiparallel orientation of the nitromethane molecules, T-shape orientation and long-range order in antidipole orientation can also be detected.

The structure of aqueous sodium hydroxide solutions: a combined solution x-ray diffraction and simulation study.

To determine the structure of aqueous sodium hydroxide solutions, results obtained from x-ray diffraction and computer simulation (molecular dynamics and Car-Parrinello) have been compared. The capabilities and limitations of the methods in describing the solution structure are discussed. For the solutions studied, diffraction methods were found to perform very well in describing the hydration spheres of the sodium ion and yield structural information on the anion's hydration structure. Classical molecular dynamics simulations were not able to correctly describe the bulk structure of these solutions. However, Car-Parrinello simulation proved to be a suitable tool in the detailed interpretation of the hydration sphere of ions and bulk structure of solutions. The results of Car-Parrinello simulations were compared with the findings of diffraction experiments.

Structure of liquid nitromethane: comparison of simulation and diffraction studies.

Simulation (molecular dynamics and Car-Parrinello [Phys. Rev. Lett. 55, 2471 (1985)]) and diffraction (x-ray and neutron) studies on nitromethane are compared aiming at the determination of the liquid structure. Beyond that, the capabilities of the methods to describe liquid structure are discussed. For the studied liquid, the diffraction methods are performing very well in the determination of intramolecular structure, but they do not give detailed structural information on the intermolecular structure. The good agreement between the diffraction experiments and the results of molecular dynamics simulations justifies the use of simulations for the more detailed description of the liquid structure using partial radial distribution functions and orientational correlation functions. Liquid nitromethane is described as a molecular liquid without strong intermolecular interactions such as hydrogen bonding, but with detectable orientational correlations resulting in preferential antiparallel order of the neighboring molecules.

X-ray diffraction and DOSY NMR characterization of self-assembled supramolecular metallocyclic species in solution.

Wide-angle X-ray scattering and diffusion NMR techniques have been used to obtain structural information on three self-assembled metallacyclic supramolecular complexes in solution: a rectangle, a triangle, and a three-diminsional cage. The low-angle region of the measured diffraction patterns and hydrodynamic radii calculations, determined from DOSY NMR experiments, suggest that the supramolecular assemblies retain their shape when dissolved in nitromethane. The experimental structure functions for the large-angle region have been analyzed, and the intramolecular contributions of the platinum-platinum interactions are discussed. These scattering measurements provide evidence that the supramolecular assemblies are not as rigid in solution as they are in the single crystal. Finally, by analysis of the radial distribution functions of the solutions, direct structural information (e.g., platinum-platinum intramolecular distances and coordination number) about the supramolecular assemblies has been obtained.