Why AGFL under a GNU licence

parser generator system for natural languages

AGFL is a typical solution in search of a problem: when given a grammar and its associated lexicon covering some natural language like english, it is easy to generate an efficient parser for that language, but very few applications demand such a parser because such parsers are a rarity:

• linguists are traditionally more concerned with giving an abstract "account" of language than with doing the painstaking work to describe a particular language in a more or less complete and consistent way
• most NLP projects set out to produce a parser intended for some particular purpose, rather than a grammar to be used for many different purposes
• most parsers are written in some programming language for efficiency, so that the knowledge contained in them is buried in code and there is no collaborative improvement of existing parsers
• others are written in some proprietary formalism and jealously hidden from the eyes of the non-initiated
• commercial users are deterred from using parsers already developed elsewhere by unpredictable financial demands and the absence of maintenance
• since there is no demand for such parsers, there is little or no supply.

In order to break this stalemate, the AGFL project has brought the AGFL system in the public domain:

• AGFL is provided as a common formalism to be used freely by any linguist, guaranteeing the automatic generation of efficient parsers
• the system comes with a number of grammars and lexica for free, in particular the EP4IR (English Phrases for Information Retrieval) grammar of English
• anybody is invited to use the AGFL system and the accompanying EP4IR grammar and lexicon of English for whatever purpose he likes, including commercial purposes, as long as the GPL is adhered to
• anybody can make and share his improvements to the free grammars and lexica, or add new ones in the same spirit.