The synergism of passive smoking and artificial feeding on hospitalization for respiratory illness in early childhood was examined among 2,227 subjects born in the last quarter of 1983 in Chang-Ning District, Shanghai, People's Republic of China. The eligible families were visited by the trained interviewers. A loglinear model shows no interaction on a multiplicative scale between these two fac...

In this paper we show that generative models are competitive with and sometimes superior to discriminative models, when both kinds of models are allowed to learn structures that are optimal for discrimination. In particular, we compare Bayesian Networks and Conditional loglinear models on two NLP tasks. We observe that when the structure of the generative model encodes very strong independence ...

Multiple response categorical variables (MRCVs), also known as “pick any” or “choose all that apply” variables, summarize survey questions for which respondents are allowed to select more than one category response option. Traditional methods for analyzing the association between categorical variables are not appropriate with MRCVs due to the within-subject dependence among responses. We have d...

Shallow-n grammars (de Gispert et al., 2010) were introduced to reduce over-generation in the Hiero translation model (Chiang, 2005) resulting in much faster decoding and restricting reordering to a desired level for specific language pairs. However, Shallow-n grammars require parameters which cannot be directly optimized using minimum error-rate tuning by the decoder. This paper introduces som...

Little work has been done on understanding the articulation patterns of users’ touch and surface gestures, despite the importance of such knowledge to inform the design of gesture recognizers and gesture sets for different applications. We report a methodology to analyze user consistency in gesture production, both between-users and within-user, by employing articulation features such as stroke...

In this study we investigate three properties of the left vowel context which we hypothesize to underlie deletion of /r/ in postvocalic, preconsonantal position in Dutch spontaneous speech: vowel type (schwa, full vowel), vowel length (long, short) and lexical stress (+,-). For each of five categories 90 instances with possible /r/-realizations were extracted from a large speech database contai...

For a given discrete decomposable graphical model, we identify several alternative parametrizations, and construct the corresponding reference priors for suitable groupings of the parameters. Specifically, assuming that the cliques of the graph are arranged in a perfect order, the parameters we consider are conditional probabilities of clique-residuals given separators, as well as generalized l...

Most statistical machine translation (SMT) systems are modeled using a loglinear framework. Although the log-linear model achieves success in SMT, it still suffers from some limitations: (1) the features are required to be linear with respect to the model itself; (2) features cannot be further interpreted to reach their potential. A neural network is a reasonable method to address these pitfall...

Markov logic is a robust approach for probabilistic relational knowledge representation that uses a log-linear model of weighted first-order formulas for probabilistic reasoning. This loglinear model always exists but may not represent the knowledge engineer’s intentions adequately. In this paper, we develop a general framework for measuring this coherence of Markov logic networks by comparing ...

The accident at the Three Mile Island nuclear power plant in 1979 led to an unprecedented set of events with potentially life threatening implications. This paper focusses on the analysis of a longitudinal study of the psychological well-being of the mothers of young children living within 10 miles of the plant. The initial analyses of the data utilize loglinear/logit model techniques from the ...