In this paper, we address the (to the best of our knowledge) new problem of ex tracting a structured description of real estate properties from their natural lan guage descriptions in classifieds. We sur vey and present several models to (a) iden tify important entities of a property (e.g., rooms) from classifieds and (b) structure them into a tree format, with the entities as nodes and edges representing a part-of relation. Experiments show that a graph-based system deriving the tree from an ini tially fully connected entity graph, outper forms a transition-based system starting from only the entity nodes, since it better reconstructs the tree.
展开▼