This paper continues the investigation of the motion of solitary grains in a water stream, reported by Francis (1973)- The trajectories of solid grains are photographed by a multi-exposure technique as they are propelled by water streams along the bed of a laboratory channel. Many thousands of photographs were taken and analysed to determine the positions, velocities and accelerations of the grain. The technique does not take into account the possible effect, in multi-grain transport, of intergranular collisions. The three different modes of transport of grains were all observed — rolling, saltation and suspension, and the proportion of each found for a variety of transport stage w*/w*0. The development of suspension is much less rapid than the development of saltation from rolling, but even at the highest stage used, about 3.0, there is still a small amount of rolling. The trajectory dimensions and geometry are shown in relation to the stage which uniquely determines the geometry. Experiments where the grain is suddenly entrained from a stationary position show that several features of the subsequent trajectory are the same as those of a trajectory with a prior history of movement: thus it is inferred that the start of a trajectory is by way of hydrodynamic forces rather than by the conservation of momentum of previous trajectories. Impacts and trajectories were analysed for the coefficient of friction tan cc and for the height of the effective thrust. While tan a is shown to be rather larger than has been suspected in the past, the variation of yn throws light upon predominance of slow fluid near the bed rather than high speed inrushes of fast fluid. Better information is now available for finding the mean forward speed of grains compared to that presented in the earlier paper. There are grounds for believing the existence of a * shear-drift ’ force on grains when they are in a velocity gradient, giving a force opposing gravity: but there is no evidence of a proximity effect of the bed independent of the velocity gradient.

Non-extractive commonsense QA remains a challenging AI task, as it requires systems to reason about, synthesize, and gather disparate pieces of information, in order to generate responses to queries. Recent approaches on such tasks show increased performance, only when models are either pre-trained with additional information or when domain-specific heuristics are used, without any special consideration regarding the knowledge resource type. In this paper, we perform a survey of recent commonsense QA methods and we provide a systematic analysis of popular knowledge resources and knowledge-integration methods, across benchmarks from multiple commonsense datasets. Our results and analysis show that attention-based injection seems to be a preferable choice for knowledge integration and that the degree of domain overlap, between knowledge bases and datasets, plays a crucial role in determining model success.

No abstract

No abstract