Andreas Kathol (kathol@violet.berkeley.edu)
Wed, 16 Oct 1996 10:18:17 -0700 (PDT)
>From Daniele.Godard@linguist.jussieu.fr Wed Oct 16 06:56:14 1996 Subject: critique of the minimalist program The final version of "A Critique of the Minimalist Program" by David Johnson and Shalom Lappin is available via anonymous ftp from ling.ohio-state.edu at: ftp/pub/HPSG/Papers/mpcrit.ps.gz This is a gzipped postscript version, thanks to Carl Pollard, who took the time and effort to help us out. =================================================================== Overview of Revisions to Johnson/Lappin ms. The paper is now divided into 3 parts to make the overall structure of the paper, and the flow of our arguments, clearer. We trust that our original intent is now completely clear. The Parts are: Part I Global Economy Conditions Part II Feature Checking in a Local Minimalist Theory Part III Optimal Design in Natural Language In Part I, we criticize Chomsky's specific presentation of the MP and its global economy conditions both for computational complexity-theortic and ordinary (conceptual and factual) linguistic reasons. We have emphasized several key points: (1) In Section 2.1 (Numeration, Merge and Move) we provide in the first paragraph a clear, but very high level definition of an MP derivation. (2) In Section 1 (Introduction) and Section 2.3. (Global Economy Condtions and Complexity Issues), we state clearly we are not attempting to analyze the complexity of the decision problem for the MP. Indeed, given its presentation, we observe this would be impossible, (3) We present a clear but high level statement of the sentence recognition problem in Section 2.3. (4) We also address the claim that more efficient algorithms computing the same results might exist. Our position is simply that the mere assertion of such a logical possibility adds nothing of substance to the discussion. (Section 2.3.) (5) In any event, the discussion of more efficient realizations of MP global conditions, is irrelevant, since we argue that these conditions (Procrastniate, the Shortest Derivation Principle and the "Have an Effect on Priciple (HEOC)") can be either reformulated as a local constraints or discarded. On to the discussion of the Global Economy Conditions (6) Specifically, in the case of Procrastinate we offer a Local version as a replacement (Section 3.2). We observe that this entails in a sense a re-emergence of a level of structure corresponding to S-strufture. This secton ahs been improved and strengthened. In the new Section 3.3. we discuss a very serious technicla problem with the distinction between Overt and Covert movement. We believe this discovery substantially strenthens our general psotion and criticisms. Section 3.4. deals with Chomsky's main "empricial" argument for Procrastinate. This has to do with "There-Insertion" sentences like "there seems to be someone in the room" but not "there seems someone to be in the room". Chomsky invokes Procrastnate to block the latter. We offer an alternative Case-based constraint that has the same outcome but has the virtue of not requiring cmoparison of derivations. (We use thissame constraint latter in the discussion of the Shortest Derivation Principle (SDP). Since there is no known let alone compelling evidence for Procrastinate, we declare victory here and move on.Next in section 4 is the Shortest Derivation Principle: this is basically as before modulo correcting obvious errors. The point is similar: one can account for the only argument Chomsky gives for t his global principle with the previous Case-based constraint. We also lifted the footnote about Strict Cycllcity to the text as a new Section 4.3. The upshot is that the SDP can be discarded in favor of local constraints. In the case of the HEOC,the last remaining global condition, we (i) provide a much more precise statement of the problem, (ii) point out that (a) it can have no bearing on the sentence recognition problem or the fundamental problem of pairing sentences with semantic interpretations(as opposed to Lf's) - that is, the kinds of problems it can conceivably address are contrived/theory-internal, and we argue (b) on a naive interpretation, it leads to complexity, there is no factual evidence in favor of it, and so it should be discarded. (We improve this technically.) and (c) conclude it is useless. (7) These arguments collectively lead to a Local Minimalist Theory, which clearly helps alleviate any complexity problems associated with global constraints and is obviously conceptually simpler in that we have discarded global constraints altogether (see Section 5.3. SUmmingUp). On to Part II, Part II Feature Checking in a Local Minimalist Theory (8) This is as before but the HPSG anslysis has been improved. The main technical difference is fffour defeasible version of the HPSG NonLocal Feature Percolation Principle in (40) but the basic analysis, discussion and conclusions are the same. Part II Optimal Design In Natural Language (9) Sections 7.1 (Efficiency as a Property of Biological Organisms) and and 7.2 (Economy in Physics) are the same in content and conclusions but we have tightened the wording and shortened it a bit. (10) Section 8. (Conclusions) is the same save minor stylistic revisons. =========================================================================
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:29 PST