Integration of the Dual Approaches in the Distributional Learning of Context-Free Grammars

• Models and exploits the distribution of strings in contexts • (x,z)  y = xyz ∈ L ? (ε,ε) (a,ε) (ε,b) (a,b) (ab,ε) (ε,abb) (aab,ε) ε a b ab aab abb aabb ε a b ab ab abb aab a aa ab aab aba aabb aaba b ab bb abb abb babb aabb ab aab abb aabb abab ababb aabab aab aaab aabb aaabb abaab aababb aabaab abb aabb abbb aabbb ababb abbabb aababb aabb aaabb aabbb aaabbb abaabb aabbabb aabaabb • ex. for L = { a n b n | n≧0 }, L⊘ab = { (a n ,b n) | n≧0 }. • ex. for L = { a n b n | n≧0 }, L⊘ab = { (a n ,b n) | n≧0 }. • ex. for L = { a n b n | n≧0 }, L⊘ab = { (a n ,b n) | n≧0 }. • (L⊘v)v ⊆ L and w(L⊘w) ⊆ L. • ex. for L = { a n b n | n≧0 }, L⊘ab = { (a n ,b n) | n≧0 }. • (L⊘v)v ⊆ L and w(L⊘w) ⊆ L. • ex. for L = { a n b n | n≧0 }, L⊘ab = { (a n ,b n) | n≧0 }. • (L⊘v)v ⊆ L and w(L⊘w) ⊆ L .