Complexity of recognition in intermediate Level languages

Complexity of sentence recognition is studied for one-way stack languages, indexed languages, and tree transducer languages. The problem is shown to be polynomial-complete in each case. A class of naturallanguage grammars is formalized and the sentence-recognition problem is shown to be polynomial-hard although the languages are context-sensitive. The proofs give new language-theoretic characterizations of the set of satisfiable propositional formulas and the set of prepositional tautologies.