Computational Linguistics: A Dissenter's Voice
Indian Statistical Institute
January 5, 2003
Indian Journal of Linguistics, Vol. XXI, No. 1, pp. 1-18, 2003
The Bangla version of this paper can be found at http://ssrn.com/abstract=2381320
The status of Computational linguistics is questioned here from four perspectives: from the standpoints of philosophy of science, natural science (mismatch between human Cognitive domain and machine algorithms), Social Science Problem (Plurilingual condition and the advent of the simulated hyper-real), Algocentricism (the discourse, which is only controlled or appropriated by the algorithm) in contrast with post-formal subjective and substantive task of Linguistics.
These problems are summarized as follows: Philosophy of science problem there is nothing called 'pen-paper-card linguistics', when these tools were used to taxonomize corpus. If computer manipulates linguistic data through the 'pen-paper-card' method, is it justified to label it as a separate discipline? Natural science problem, this problem deals with the matching condition between 'human cognitive domain' and 'machine algorithm' (identity and difference between computer and human being) on the basis of Russel's paradox and Goedel's theorem Problem raised by Penrose (1990,1994) and Seairle's Chinese room puzzle. Computer's halting problem. Fuzziness of natral language (The paper deals with some Bangla usages of numerals , where the status of number one is not always equal to one. The value of this fuzzy one is determined by the context, speakers socio-economic status etc.) Post - formal approach that denies analytical procedures proposed by structralists (speaking subject perceives language as a whole. Fragmenting language-object by deploying grammatical rules implies understanding symbolic order by means of another (meta-)symbolic order. A binary machine appropriates language-object according its own algorithmic program. It leads to a metonymic transformation of speaking subject as subjects’ non-algorithmic capability is ignored.) Social Science problem: Hyper-real communication through binary machines is a capital-intensive area, but the pluriligual communication system (as it is found in India) needs no such investment as such communication falls within the shadow area of economics. Furthermore, the non-bio-degradability of machines would create a future problem as semi-conductors used in electronic industry contains spurious elements. Problem of algocentricism: The equation 'man=machine' leads to a discursive formation that totally depends on the use of technical metaphors (e.g., The terms like 'Computation’, 'array' 'interface', 'parser' etc or operations like 'command', 'satisfy', 'spell out'). All these reflect the metonymic transformation of creative speaking subject as all these functions in uppercase letters make us remember Schank’s (1975) language-free representation (Propel, move, ingest or control, part etc.) which combines primitive conceptual roles and conceptual categories employed to process so-called natural language. All these are not at all metaphors (a case of displacement of algorithmic/non- algorithmic human mind with an algorithmic machine), but these are metonyms, by which the potentialities of human mind are condensed and telescoped.
Number of Pages in PDF File: 10
Keywords: algocentric discourse, artificial intelligence, fuzzy logic, Russell's paradox, Goedel's theorem, turing machineAccepted Paper Series
Date posted: March 5, 2012
© 2014 Social Science Electronic Publishing, Inc. All Rights Reserved.
This page was processed by apollo3 in 0.344 seconds