Languages Between Context Free and Context Sensitive

Abstract

Most investigators supposed from the time that the Chomsky Hierarchy was first established that natural languages, considered as string sets, would fall somewhere between the context free and the context sensitive languages and, further, that they would lie in some sense “close” to the context free class. On the one hand, the context sensitive languages seemed much too inclusive, containing as they do species such as {aⁿb^n!} (n! is n factorial, i.e., 1 x 2 x 3 x ... x n ) and {aⁿ : n is prime}, which seem unlikely candidates for any sort of linguistic model. On the other hand, a large part of natural language syntax seems to be handled quite nicely by a cfg, and the aspects which seem to cause languages to fall outside the cfl class (as string sets) could be considered rather isolated and infrequent. After all, it took nearly thirty years to find one completely convincing example of a natural language which was not context free.