2014/04/09 by Mircea Marin, Marin, Mircea, Gabriel Istrate +1
Computer Science · #FOS: Computer and information sciences #Formal Languages and Automata Theory (cs.FL) #cs.FL
paper · pdf · doi:10.48550/arxiv.1404.2409
arxiv created 2014/04/09 · arxiv updated 2014/04/10
We consider the problem of learning an unknown context-free grammar when the only knowledge available and of interest to the learner is about its structural descriptions with depth at most ℓ. The goal is to learn a cover context-free grammar (CCFG) with respect to ℓ, that is, a CFG whose structural descriptions with depth at most ℓ agree with those of the unknown CFG. We propose an algorithm, called LA^ℓ, that efficiently learns a CCFG using two types of queries: structural equivalence and structural membership. We show that LA^ℓ runs in time polynomial in the number of states of a minimal deterministic finite cover tree automaton (DCTA) with respect to ℓ. This number is often much smaller than the number of states of a minimum deterministic finite tree automaton for the structural descriptions of the unknown grammar.