Evaluation procedure
An evaluation procedure is a mechanical procedure for comparing two generative grammars of the same format and generating the same language to determine, for example, by a measure of simplicity, which Chomsky better proposes in the 1950s as a goal for linguistic theory weaker than that of a discovery procedure. An evaluation measure is a metric employed in such a procedure [Matthews, p. 129].
An evaluation procedure provides criteria for choosing the better of two analyses of a set of data, for example, when it is argued that one analysis is more straightforward, more plausible, or more elegant than another. In generative linguistics, a few (controversial) procedures have been suggested (see simplicity) that attempt to formalize the properties of alternative descriptions so that precise evaluations can be made [Crystal, p. 388].
An evaluation procedure for grammars, that is, entire grammars – is a rather tall order if the procedure is to be interpreted operationally. It implies that two grammars (that is, two complete descriptions from phoneme – or morpheme – to sentence) are to be compared to each other and a corpus to ascertain which is to be preferred [Garvin, p. 48].
⠀ David Crystal. (2008). A Dictionary of Linguistics and Phonetics, 6th Edition. New-Jersey: Wiley-Blackwell.
⠀ Matthews P.H. (2007). Oxford Concise Dictionary of Linguistics. Oxford: Oxford University Press.
⠀ Paul L. Garvin. (1972). Evaluation procedure in linguistic analysis. On Linguistic Method. Berlin: Walter de Gruyter.