id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_ocrpsnsynrh2nb7fy3scvt4ltu	Matthew R. Gormley	Approximation-Aware Dependency Parsing by Belief Propagation	2015	14	.pdf	application/pdf	8679	1097	71	Approximation-Aware Dependency Parsing by Belief Propagation Recent improvements to dependency parsing accuracy have been driven by higher-order features. parsers depend on approximate inference and decoding procedures, which may prevent them from predicting the best parse. In contrast, we train the parser such that the approximate system performs well on the final evaluation function. We treat the entire parsing computation as a differentiable circuit, and backpropagate the evaluation function through our approximate inference and decoding methods to improve models, Stoyanov and Eisner (2012) call this approach ERMA, for "empirical risk minimization under approximations." For objectives besides empirical risk, Domke (2011) refers to it as "learning with approximation-aware learning method in the parsing setting, for which the graphical model involves to those probabilities.3 When our inference algorithm is approximate, we replace the exact marginal and Eisner (2008)—is exact for first-order dependency parsing. regularized objective function over the training sample of (sentence, parse) pairs {(x(d),y(d))}Dd=1.	./cache/work_ocrpsnsynrh2nb7fy3scvt4ltu.pdf	./txt/work_ocrpsnsynrh2nb7fy3scvt4ltu.txt