Modified Earley parsing and MPM Method for Attributed Grammar and Seismic Pattern Recognition
Kou-Yuan Huang and Dar-Ren Leu* Department of Computer and Information Science National Chiao Tung University Hsinchu, Taiwan, R.O.C. *Department of Computer Science University of Houston Houston, TEXAS 77004, U.S.A.
Two methods of parsing attributed string are proposed. One is the modified error-correcting Earley parsing, and the other is a parsing using the match primitive measure (MPM). The computation of the match primitive measure between two attributed strings using dynamic programming is proposed. The modified minimum distance errorcorrecting Earley parsing algorithm and the MPM parsing algorithm for an attributed string can handle three type of error. The MPM parsing algorithm is obtained from the computation between the input string and the string generated by the attributed grammar. The MPM parsing is more efficient than the modified Earley parsing. The recognition criterion of the modified Earley algorithm is "minimum-distance," and the recognition criterion of the MPM parsing algorithm is "maximum-matching." The parsing methods are applied to the recognition of seismic Ricker wavelets and the recognition of wavelets in real seismic data.