This is the Introduction of my Ph.D. research project. Soon I’ll post the whole project in pdf.
New approach to formal grammars: Pāṇini’s grammar as an answer.
The form is the possibility of the structure.
L.Wittgenstein (T.L.Ph., 2.033)
Since Noam Chomsky said the Pāṇini’s grammar (PG) is a “generative grammar, in essentially the contemporary sense of this term” (1965: v), a new perspective in the studies of Aṣṭādhyāyī has appeared. Some linguists trained in PG have been asserting the link between PG and generative grammar at least since 1965, even assuming that PG is within a specific hierarchy of the Chomsky Hierarchy. But none of them ever questioned the Chomsky’s theory, on the contrary, just accepted it and have applied it to PG its whole framework, however, “the generative perspective has misled linguists concerning the properties of natural language” (Pullum: 2003).
The formalization of grammars has been very restricted to what we call generative frameworks, and these frameworks have been considered theoretically important for any description of natural language without proving empirically its efficacy. It is like saying that for the formal grammar have its potential proved, it has to pass through an analysis of a generative framework, known as mathematical foundations. PG has been a victim of Chomsky’s theory since his declaration. When some linguists trying to apply these new models of framework to PG to prove that PG can be one kind of generative grammar, we see two mistakes: first, to turn PG’s framework to generative framework (something that contributes to PG to lose its linguistic potential); second, to lose the oportunity to challenge Chomsky’s theory with a grammar that describes a language (Sanskrit, in this case) with total efficacy, and empirically.
My research goes in this new perspective, not trying to demonstrate that PG is one kind of generative grammar, but to prove that PG’s method is efficient for describing any natural language. PG will be proved as a model grammar, meaning that it can be applied at least to every natural language in the Indo-european family. This approuch implies two important points: outdo the Chomsky’s theory of grammar (note that I am discussing grammars, not languages); provide a framework completely linguistic for description of languages (avoiding mathematical foundations). With the second point is possible to affirm that PG can even be a model of grammar for Natural Language Processing (NLP).
If a generative grammar is a “system of rules that in some explicit and well-defined way assigns structural descriptions to sentences” (Chomsky, 1965: 8), the grammar of Pāṇini could be considered a generative grammar, but the main problem to consider it as this kind of grammar is that for Pāṇini the objects that linguists investigate (words, phrases, clauses, sentences, lexemes, syllables) are real and have a real structure, than “merely structures imposed on them or posited for them by linguists” (Pullum:2007). The PG is known as saṃjñaka vyākaraṇa, a grammar of categories of technical terms (Kapoor, 2005:70). For Pāṇini, to generate a language (Sanskrit), one grammar firstly must describe it, and if its description is not real, the generation is not effectual. In the end, PG exposes the mistakes that have been investigated by generative linguists.
 See J.F.Staal, “Context-Sensitive Rules in Pāṇini” (1965); and M.D.Hyman “From Pāṇinian sandhi to finite state calculus” (2007).
 My research has the support of all the examination made by Geoffrey K. Pullum and András Kornai regarding the mistakes of generative grammars and Chomsky’s theory, as well as parts of the Cognitive Linguistics, Geoffrey Sampson for example.