Thursday, February 21, 2013

Discovery Procedures and Evaluation Procedures

Discovery Procedures
American structural linguistics (pre-Chomskiyan era) tradition studied exotic languages to preserve their culture and traditions. The linguists like Franz Boas and his followers had found that the notions and concepts of traditional grammar (of Latin, Greek and Sanskrit) were useless against totally new languages with no historical relation between them and the classical languages and/or Indo-European family of languages. The study of language to study indigenous cultures made them aware of the diversity in human languages. According to unlimited diversity principle, their beliefs emerged as almost opposite to traditional grammar’s somewhat universalist beliefs. The languages were different; each language had its own way of organizing concepts of grammar hence the grammar of each language was different. There was no guarantee that techniques, methods or tools developed for language will work on very next language. Thus to study the grammar of a language a linguist had to have a number of tools (for description and analysis). As Sampson (1980) notes, the theory of language for Descriptivists was that there was no theory at all. Instead they believed that they were developing ‘techniques’ of analysis without going into details of creating a theory of language. But, as Samson (1980) argues, “any analytical technique in any domain must depend on some assumptions about the nature of the things analysed”.
Apart from this contradiction of having a theory of language or not having a theory of language, they didn’t have any mechanism to select or reject the best technique or ‘discovery procedure’ to analyse a given chunk of language. “They approached alternative techniques of description in a more catholic way, seeing them as alternative tools to be pulled out of the toolbag when needed” (Sampson 1980, p.74). Thus there was no way to measure or weigh for a method to select the best method, often it was a matter of personal choice. As for Charles Hocket (1954 quoted in Sampson 1980) two different methods of analysis of same linguistic phenomenon were equal, none better than the other one, moreover he described a third method of analysis as well which, in his view, deserved equal attention.
In later years some linguists were compelled to devise explicit ‘discovery procedures’ which, when written as a computer program, would enable a computer to analyse linguistic data and derive the grammar of that language without human intervention. Harris (1951, quoted in Sampson 1980) wrote a complete account of discovery procedures to collect utterance and analyse them at phonemic, morphemic and (to a rather less extent) at syntactic level. There was a tension between the ‘unlimited diversity principle’ and ‘view that linguistics should consist of mechanical rules for processing data into grammars’. The methods of latter view couldn’t be successful without assuming some ‘universal features’ in all languages.
The apparent lack of a theory of language, the inability to select most efficient method of analysis (or ‘grammar’ of a language) and the over-emphasis on diversity led to the criticism from Chomsky (2002) and the development of ‘evaluation procedures’ in contrast of ‘discovery procedures’.

Evaluation Procedures
For Chomsky (2002), the grammar must meet certain external criteria of adequacy e.g. generated sentences should be acceptable by native speaker, and condition of generality by which a grammar of a language must be constructed out of a theory which defines terms like ‘phoneme’ and ‘phrase’ independent of a given language, thus opposite to the idea of ‘unlimited diversity principle’.
Chomsky (2002) outlines three requirements from a theory of language. The theory must provide “a practical and mechanical method for actually constructing the grammar, given a corpus of utterances” (p.57). He calls this requirement discovery procedure, which is the strongest demand from a theory of language.
A weaker demand is to require a decision procedure through which a theory of language may be able to decide whether the provided grammar is best grammar for the language from which the corpus was collected.
A third and weakest demand would be that theory must be able to tell us which grammar is better, if a corpus and two grammars G1 and G2 are given. He calls this requirement evaluation procedure. In Chosmky’s view, the descriptivists require the most strongest of these three procedures, which in his view is unreasonable to require from a theory of language i.e. to ask it more than a practical evaluation procedure for grammars.
In his view – by demanding ‘evaluation procedures’ instead of ‘discovery procedures’ from a theory – “a number certain problems that have been the subject of intense methodological controversy simply do not arise” (p.56), e.g. the problem of inter-dependence of levels. While working to create discovery procedures, the descriptivists assume that levels are interdependent and thus they cannot go beyond phonemic and morphemic levels due to unsolved problems which their discovery procedure cannot deal with at this current level. But by abandoning this higher level requirement, the inter-dependency problem can be ignored and syntactic theory can be advanced which was otherwise being ignored. As grammar of a language is a complex system of interconnections, and other linguists’ (descriptivists) beliefs about it “has been nurtured by a faulty analogy between the order of development of linguistic theory and the presumed order of operations in discovery of grammatical structure” (p.59).

Concluding the discussion, it can be safely said that Chomsky’s proposal was a major methodological advancement which led to a rather unified and universal theory of language. His ideas provided an independent status to syntax which was previously ignored by the descriptivists. Thus the advancement of a syntactic theory in turn could provide some ideas to solve the un-resolved problems at the phonemic and morphemic level.

Sampson, Geoffrey. 1980. Schools of Linguistics: Competition and evolution. Hutchinson: London, Melbourne, Sydney, Auckland, Johannesburg.
Chomsky, Naom. 2002. Syntactic Structures. 2nd Edition. Mouton de Gruyter: Berlin, New York.

No comments:

Post a Comment