A formal theory for optimal and information theoretic syntactic pattern recognition

摘要: Abstract In this paper we present a foundational basis for optimal and information theoretic syntactic pattern recognition. We do by developing rigorous model, m ∗ , channels which permit arbitrarily distributed substitution, deletion insertion errors. More explicitly, if A is any finite alphabet the set of words over A, specify stochastically consistent scheme string U∈A can be transformed into Y∈A means operations. The shown to functionally complete consistent. Apart from synthesis aspects, also deal with analysis such model derive technique Pr[Y∣U], probability receiving Y given that U was transmitted, computed in cubic time using dynamic programming. One salient features it demonstrates how programming applied evaluate quantities involving complex combinatorial expressions maintain rigid consistency constraints. Experimental results involve dictionaries strings lengths between 7 14 an overall average noise 39.75% demonstrate superiority our system existing methods. its straightforward applications generation recognition, believe has extensive potential speech, uni-dimensional signal processing, computational molecular biology.