US 6,983,239 B1
Method and apparatus for embedding grammars in a natural language understanding (NLU) statistical parser
Mark E. Epstein, Katonah, N.Y. (US)
Assigned to International Business Machines Corporation, Armonk, N.Y. (US)
Filed on Oct. 25, 2000, as Appl. No. 9/696,764.
Int. Cl. G06F 17/27 (2006.01)
U.S. Cl. 704—9 51 Claims
OG exemplary drawing
 
1. In a natural language understanding (NLU) system, a method for including grammars in a statistical parser comprising:
receiving a text input; and,
applying a first context free grammar (CFG) to said text input to determine substrings and corresponding parse trees, wherein said substrings and said corresponding parse trees further correspond to said first CFG; and,
examining each said possible substring and corresponding parse tree using an inventory of ordered grammar-related queries corresponding to said CFG, wherein said queries are ordered training of the NLU system to form a decision tree, the ordering of the queries being a statistical ordering based on a minimization of conditional entropy or maximization of a likelihood that a resulting model predicts data used to train the NLU system.