[Cover] [Contents] [Index] Previous page Next Section

Page 66
§4.2.3 A Comprehensive, Identifiable Collection of Languages
The languages comprising 0066-001.gif and 0066-002.gif are computationally transparent, yet Corollaries 3.28 and 3.29 show that neither is identifiable. In conjunction with Proposition 4.7, these facts might encourage the belief that TxtEx contains only computationally impoverished subsets of  e . The purpose of the present section is to show that this belief is mistaken. In particular, it will be seen that there is a computable scientist of simple design that identifies "nearly all" of  e . This idea is made precise by the following definition. First, recall from Chapter 2 that two languages L and L' are finite variants just in case both L - L' and L' - L are finite.
4.12 Definition Let 0066-003.gif be given. Image-0801.gif covers  e  just in case for every 0066-004.gif there is 0066-005.gif such that L and L' are finite variants.
Languages that are finite variants have similar computational properties. For example, if one is recursive, then so is the other, and similarly for being decidable in polynomial time, etc. Hence, if Image-0802.gif covers  e , then even the most computationally complex sets in  e  have equally complex counterparts in 0066-006.gif. Nonetheless:
4.13 Proposition (Wiehagen [196]) There is 0066-007.gif such that Image-0803.gif covers  e .
Proof: By Exercise 2-3, one can construct, for each 0066-008.gif, an 0066-009.gif such that:
(a) L and L' are finite variants; and
(b) L' = Wx where x is the smallest member of L'.
It follows that 0066-010.gif covers  e . Now define scientist M by the condition that for all 0066-011.gif, M( s ) = the least 0066-012.gif It is evident that M is computable and identifies Image-0804.gif.
§4.2.4 Computability in the Context of Other Constraints
In Section 3.7 we noted that a model of language acquisition places greater constraint on theories of comparative grammar to the extent that it succeeds in locating the language acquisition mechanism of children in a narrow subset of the class of all scientists. One such subset was discussed, namely, memory-limited scientists, and it was shown that there are identifiable collections of languages that are not identified by any memory-limited scientist. The computable scientists constitute another subset of the same nature, as shown by Corollary 4.5.

 
[Cover] [Contents] [Index] Previous page Next Section