Constraints as Prior Knowledge

Making complex decisions in real world problems often involves assigning values to sets of interdependent variables where an expressive dependency structure among these can influence, or even dictate, what assignments are possible. Commonly used models typically ignore expressive dependencies since the traditional way of incorporating non-local dependencies is inefficient and hence lead to expensive training and inference. This paper presents Constrained Conditional Models (CCMs), a framework that augments probabilistic models with declarative constraints as a way to support decisions in an expressive output space while maintaining modularity and tractability of training. We develop, analyze and compare novel algorithms for training and inference with CCMs. Our main experimental study exhibits the advantage our framework provides when declarative constraints are used in the context of supervised and semi-supervised training of a probabilistic model.