统计机器翻译中非线性参数学习的研究

lthough the log-linear model achieves great success in SMT, it still suffers from some drawbacks: first, the features which are used in the model must be linear with respect to the model itself; then, the model restricts its features must be linear with each other. This constraint makes the model unable to describe the complex combination between features so that it cannot further interpret the features to reach their potential information, which makes feature designation hard and often needs human knowledge and understanding to help. A non-linear model is a reasonable method to relax these linear constraints. This paper focuses on the effects of using non-linear model in SMT and conducts experiments to verify which learning framework, object function and sample method are suitable for this kind of situation.