Complexities of Order-Related Formal Language Extensions

The work presented in this thesis discusses various formal language formalisms that extend classical formalisms like regular expressions and context-free grammars with additional abilities, most relating to order. This is done while focusing on the impact these extensions have on the efficiency of parsing the languages generated. That is, rather than taking a step up on the Chomsky hierarchy to the context-sensitive languages, which makes parsing very difficult, a smaller step is taken, adding some mechanisms which permit interesting spatial (in)dependencies to be modeled. The most immediate example is shuffle formalisms, where existing language formalisms are extended by introducing operators which generate arbitrary interleavings of argument languages. For example, introducing a shuffle operator to the regular expressions does not make it possible to recognize context-free languages like anbn, but it does capture some non-context-free languages like the language of all strings containing the same number of as, bs and cs. The impact these additions have on parsing has many facets. Other than shuffle operators we also consider formalisms enforcing repeating substrings, formalisms moving substrings around, and formalisms that restrict which substrings may be concatenated. The formalisms studied here all have a number of properties in common. 1. They are closely related to existing regular and context-free formalisms. They operate in a step-wise fashion, deriving strings by sequences of rule applications of individually limited power. 2. Each step generates a constant number of symbols and does not modify parts that have already been generated. That is, strings are built in an additive fashion that does not explode in size (in contrast to e.g. Lindenmayer systems). All languages here will have a semi-linear Parikh image. 3. They feature some interesting characteristic involving order or other spatial constraints. In the example of the shuffle multiple derivations are in a sense interspersed in a way that each is unaware of. 4. All of the formalisms are intended to be limited enough to make an efficient parsing algorithm at least for some cases a reasonable goal. This thesis will give intuitive explanations of a number of formalisms fulfilling these requirements, and will sketch some results relating to the parsing problem for them. This should all be viewed as preparation for the more complete results and explanations featured in the papers given in the appendices.