GSOC 2009: Statistical Learning and Refinement of RelEx Graph Transformation Rules
When given a sentence to RelEx, it is first parsed using Link Parser and from then RelEx engine takes over the parsed sentence in the form of a graph. Many a times the parse may contain unknown words or the link information may not be sufficient enough for RelEx to proceed forward confidently. So our aim is to learn these new transformation (link grammar) rules which increase the coverage of parser as well as refine word disjuncts into classes so as to encode more information in the disjunct.
Some of the links useful to my work are