Supervised Large Margin Loss

We present a novel semi-supervised training algorithm for learning dependency parsers. By combining a supervised large margin loss with an unsupervised least squares loss, a dis- criminative, convex, semi-supervised learning …...