We compared spectral tilt modifications of plain speech made by Linear Predictive Coding—LPC transplantation and fractional roll-off filtering—FRF. These modifications were done at utterance-, phone-, or frame-level to equate the spectral tilt of the same utterances produced in noise (Lombard speech) by the same speakers. When mixed with speech-shaped noise at the same Signal to Noise Ratio (SNR), speech treated with LPC transplantation yielded larger objective intelligibility gains relative to those of speech treated with FRF. However, it also yielded larger spectral tilt errors for frame-based modifications. This finding suggests that the hitherto spectral tilt benefits assigned to Lombard speech may be smaller than previously thought, especially for those reports based on phone- or frame-based LPC transplantation.

A graph is a very common and powerful data structure used for modeling communication and social networks. Models that generate graphs with arbitrary features are important basic technologies in repeated simulations of networks and prediction of topology changes. Although existing generative models for graphs are useful for providing graphs similar to real-world graphs, graph generation models with tunable features have been less explored in the field. Previously, we have proposed GraphTune, a generative model for graphs that continuously tune specific graph features of generated graphs while maintaining most of the features of a given graph dataset. However, the tuning accuracy of graph features in GraphTune has not been sufficient for practical applications. In this paper, we propose a method to improve the accuracy of GraphTune by adding a new mechanism to feed back errors of graph features of generated graphs and by training them alternately and independently. Experiments on a real-world graph dataset showed that the features in the generated graphs are accurately tuned compared with conventional models.

With the development of graph applications, generative models for graphs have been more crucial. Classically, stochastic models that generate graphs with a pre-defined probability of edges and nodes have been studied. Recently, some models that reproduce the structural features of graphs by learning from actual graph data using machine learning have been studied. However, in these conventional studies based on machine learning, structural features of graphs can be learned from data, but it is not possible to tune features and generate graphs with specific features. In this paper, we propose a generative model that can tune specific features, while learning structural features of a graph from data. With a dataset of graphs with various features generated by a stochastic model, we confirm that our model can generate a graph with specific features.

In Web-based investigative learning, learners are expected to construct wider and deeper knowledge by navigating a great number of Web resources/pages. In elaborately investigating an initial question, learners are expected to decompose an initial question into related question to be further investigated. However, it is difficult for learners to conduct question decomposition in concurrence with their knowledge construction. In our previous study, we have proposed a model of Web-based investigative learning, and developed the system named interactive Learning Scenario Builder (iLSB for short). Although iLSB could promote self-directed investigative learning, learners often decompose a question into unrelated sub-questions. This suggests the necessity of promoting reflection on question decomposition by diagnosing the appropriateness of question decomposition. Toward this issue, we have proposed a method for diagnosing the appropriateness of question decomposition with Linked Open Data (LOD). In this paper, we describe an adaptive prompting with diagnosed results for reflection on question decomposition. This paper also reports a case study whose results suggest the potential for promoting reflection on improper question decomposition.

A typical implementation of scientific applications includes a large number of iterative calculations. For performance optimization, these calculations are often partitioned, grouped, and reordered for execution. Since this refactoring is repeatedly performed during development, it is one of the major source of bugs and thus tool support is necessary for debugging. This study discusses this problem and proposes tool support through a testing framework. This testing framework can help developers perform the tests we call calculation coverage testing. It investigates whether the grouped calculations cover all the calculations performed by the original (and often naively implemented) program. It also investigates whether their execution order is correct. To demonstrate this idea, we also presents HPCUnit, our prototype testing framework for Java, and then reports an empirical study applying it to the Java Grande Forum Benchmark Suite.