We consider a generalization of the "dual" approach to distributional learning of contextfree grammars, where each nonterminal A is associated with a string set X A characterized by a finite set C of contexts. Rather than letting X A be the set of all strings accepted by all contexts in C as in previous works, we allow more flexible uses of the contexts in C, using some of them positively (contexts that accept the strings in X A ) and others negatively (contexts that do not accept any strings

more »

... X A ). The resulting more general algorithm works in essentially the same way as before, but on a larger class of context-free languages.