The key step of semantic parsing is to learn a connection between parts of the grammar and parts of the English statement. Many approaches have been generated, but we will be focusing on the mechanism introduced in work by Zettlemoyer (Artzi and Zettlemoyer, 2011). This work attempts to learn a probabilistic grammar in a bootstrapping manner, by looking for commonalities in the domain of cricket. For example, if several of the example queries in the game of Cricket have the term "centuries" in it and there is always a corresponding part of the query generated that includes a class such as give centuries clause, it might be reasonably concluded that the term "centuries" is a strong predictor of that clause. As more of these connections are made the learner can focus on the remaining words and corresponding parts of the parse tree and attempt to make further connections. This approach is similar, though a different mechanism is used by Kate (2008). Results obtained were promising and proves the efficiency of the model against previously performed work.
University of Minnesota M.S. thesis. May 2014. Major: Computer science. Advisor: Dr. Richard F. Maclin. 1 computer file (PDF); vii, 122 pages, appendix p. 73-122.
Semantic Parsing for Automatic Generation of SQL Queries using Adaptive Boosting.
Retrieved from the University of Minnesota Digital Conservancy,
Content distributed via the University of Minnesota's Digital Conservancy may be subject to additional license and use restrictions applied by the depositor.