蒙古语多方言语音识别及共享识别模型探索

通过语音输入方式准确识别蒙古语主要方言 ,并为其建立共享识别模型 ,对实现方言间的信息自动交换及语音识别技术的实用化具有实际意义。针对蒙古语多方言的语音特点 ,可以采取独立的或者共享识别模型进行识别。本文介绍的抽出初步模型和建立共享模型的方法 ,简便易行 ,但识别精度有待提高。文中探索了通过借用数据量和质较完备的语言的语音数据提高识别率的可能性。     

Job turnover, wage rates, and marital stability: How are they related?

This study examines the interplay between job stability, wage rates, and marital stability. We use a Dynamic Selection Control model in which young men make sequential choices about work and family and estimate the model using an approach that takes account of self-selection, simultaneity and unobserved heterogeneity. The results quantify how job stability affects wage rates, how both affect marital status, and how marital status affects earnings and job stability. The study reveals robust evidence that job changes lower wages and the likelihood of getting married and remaining married. At the same time, marriage raises wage rates and job stability. To project the sequential effects linking job change, marital status, and earnings, we simulate the impacts of shocks that raise preferences for marriage and that increase education. Feedback effects cause the simulated wage gains from marriage to cumulate over time, indicating that long-run marriage wage premiums exceed conventional short-run estimates.     

Risk Factors for Clinically Significant Intimate Partner Violence Among Active‐Duty Members

Hypothesized risk factors for men's and women's clinically significant intimate partner violence (CS‐IPV) from four ecological levels (i.e., individual, family, workplace, community) were tested in a representative sample of active‐duty U.S. Air Force members (N = 42,744). When considered together, we expected only individual and family factors to account for unique variance in CS‐IPV perpetration. Hypothesized factors from all four ecological levels were related to men's CS‐IPV perpetration bivariately, but, as expected, only individual and family factors accounted for unique variance across ecological levels. For women, only risk factors from the individual and family levels were significantly related to CS‐IPV perpetration even bivariately. Results imply somewhat different risk profiles across gender and identify ecological risk factors of men's CS‐IPV not previously studied.     

On the relative accuracy of certain bootstrap procedures

We show the second-order relative accuracy, on bounded sets, of the Studentized bootstrap, exponentially tilted bootstrap and nonparametric likelihood tilted bootstrap, for means and smooth functions of means. We also consider the relative errors for larger deviations. Our method exploits certain connections between Edgeworth and saddlepoint approximations to simplify the computations.     

Employee exposure to coworker substance use and negative consequences: the moderating effects of work group membership.

The current study assesses: (1) whether the relationship between individual exposure to coworker substance use and negative consequences resulting from exposure depends on work group membership, and (2) whether group-level characteristics moderate the relationship between exposure and consequences. At the group-level, we assessed occupations involving safety risk or high mobility and social factors of drinking climate and group cohesiveness. We conducted Hierarchical Linear Modeling (HLM) across two samples of municipal employees (ns = 650, 878; n of groups = 50, 49). Our results revealed that groups with higher proportions of jobs involving risk (e.g., machine work) and, to a lesser extent, groups with a higher level of drinking climate were those most vulnerable to consequences under conditions of exposure. Importantly, our findings controlled for individual risk factors (e.g., personal drinking, job stress). Our discussion examines the implications of this study for theory and policy related to workplace substance abuse.     

THE PHYSICIAN LABOR MARKET IN A MANAGED CARE-DOMINATED ENVIRONMENT

The belief that doctors respond to declining demand by treating patients more aggressively has created skepticism about relying on market forces to restructure physician supply. We argue that even if the physician labor market is dysfunctional under fee-for-service incentives, it can perform better as managed care becomes dominant. Our model implies a nonlinear effect of managed care penetration on incomes. Physicians can offset most or all of initial declines in demand, but cannot insulate themselves indefinitely. This may explain the observation that, until recently, the growth of managed care has not been accompanied by large physician income changes. ( JEL III, J31)     

The coming medical apocalypse.

Is there a medical apocalypse in our future? Will it happen soon? No one can say for sure, but five ominous trends suggest that a medical meltdown could occur at any time. These trends are: (1) The practice of providing medical care becoming too complex from both a business and a legal perspective; (2) Less money being spent on medical care without any corresponding reduction in services provided, creating long-term operating deficits; (3) Investor-owned, for-profit corporations changing the focus of medicine by putting shareholder concerns ahead of patient care; (4) Employment-linked health care insurance creating a growing uninsured population, adding extra financial stress to our hospitals; and, (5) Providers losing faith in their future and becoming increasingly demoralized about practicing the healing arts. These dangerous trends are considered, along with some suggestions that physician executives and organizations might take to protect themselves.     

A Discrete Variable Chain Graph for Applicants for Credit

A bank offering unsecured personal loans may be interested in several related outcome variables, including defaulting on the repayments, early repayment or failing to take up an offered loan. Current predictive models used by banks typically consider such variables individually. However, the fact that they are related to each other, and to many interrelated potential predictor variables, suggests that graphical models may provide an attractive alternative solution. We developed such a model for a data set of 15 variables measured on a set of 14 000 applications for unsecured personal loans. The resulting global model of behaviour enabled us to identify several previously unsuspected relationships of considerable interest to the bank. For example, we discovered important but obscure relationships between taking out insurance, prior delinquency with a credit card and delinquency with the loan.