Abstract
We introduce a model and an algorithm for semantic parsing in λ-calculus formalism which is based on Constrained Conditional Model framework (Roth and Yih, 2005). We also introduce an approach for learning the constraints of our model directly from knowledge graph which scales up for new domains easily.
In this ongoing project we optimize our model on a dataset using integer linear programming. Our project includes two parts and our results in the first part show a meaningful improvement over previous models.