Efficient Parsing Using Filtered-Popping Recursive Transition Networks
We present here filtered-popping recursive transition networks (FPRTNs), a special breed of RTNs, and an efficient parsing algorithm based on recursive transition network with string output (RTNSO) which constructs the set of parses of a potentially ambiguous sentence as a FPRTN in polynomial time. By constructing a FPRTN rather than a parse enumeration, we avoid the exponential explosion due to cases where the number of parses increases exponentially w.r.t. the input length. The algorithm is compatible with the grammars that can be manually developed with the Intex and Unitex systems.
Unable to display preview. Download preview PDF.