Neural information retrieval: at the end of the early years

A recent “third wave” of neural network (NN) approaches now delivers state-of-the-art performance in many machine learning tasks, spanning speech recognition, computer vision, and natural language processing. Because these modern NNs often comprise multiple interconnected layers, work in this area is often referred to as deep learning. Recent years have witnessed an explosive growth of research into NN-based approaches to information retrieval (IR). A significant body of work has now been created. In this paper, we survey the current landscape of Neural IR research, paying special attention to the use of learned distributed representations of textual units. We highlight the successes of neural IR thus far, catalog obstacles to its wider adoption, and suggest potentially promising directions for future research.     

The prediction of academic and clinical performance in medical school

Prior studies have suggested the importance of distinguishing between performance in the earlier and later years of medical education. Factor analysis of grades by year and faculty ratings of clinical and general medical competence for 661 medical students identified two uncorrelated factors clearly reflecting this distinction. The clinical performance factor, accounting for 48% of the communality of the matrix, was more or less unpredictable from aptitude and premedical academic achievement indices; it was marginally predictable from scales on the Adjective Check List. The academic performance factor, accounting for 31% of the variance among criteria, was forecast with acceptable accuracy (cross-validated R=0.43) by equations based on the Medical College Admission Test and premedical grade point average. Future study should stress the clinical performance factor; this criterion appears to be more important than academic attainment, and it is also less accurately predicted by current measures.     

Four fictional views of the future of learning

We are at a critical moment in the history of human learning, and in the history of humans. Learning is very important in our future. With 6 billion people on Earth, going to 9 billion by the middle of the century, current educational systems everywhere at all levels have major problems, probably not solvable with present approaches. In many areas, little education is available. We need new learning modes and structures, and we need them quickly and globally. Creating new learning materials and systems will require great imagination. We need approaches that go beyond current situations. Learning should be greatly improved, less expensive, and available lifelong for all in all subjects. We need a society that focuses on learning, where everyone loves to learn. Solving many of the world's problems, including population, water, violence, health, and environmental destruction, will depend on improving learning. One neglected source of future ideas is fictional accounts of the future of learning. I have long found them helpful in my own thinking and have written about them before. I keep finding new sources of this kind. Some, not mentioned here, are negative views.     

Notes on Contributors

Concepts adapted from structuration theory are employed to analyse how a disjunction has been created and sustained between the form taken by annual development plans for schools and a more flexible approach to planning that appears, from small‐scale research, to be effective at school level. The interest of LEAs and central government in a high degree of external control over school development is interpreted as contradictory to the interest at school level in maintaining a high degree of control over internal development. It is argued that action according to these contradictory interests has not led to conflict because completion of the plan has been kept largely separate from ongoing strategic planning at school level. The concepts employed in this analysis are defined, features of LEA development plans for schools are discussed, research evidence on development planning is summarized, and unintended consequences of development plans as a form of bureaucratic control are suggested.