Polynomial-Time Learning of Very Simple Grammars from Positive Data

COLT ◽  
1991 ◽  
pp. 213-227 ◽  
Author(s):  
Takashi YOKOMORI
1999 ◽  
Vol 10 (04) ◽  
pp. 483-501
Author(s):  
OKADOME TAKESI

The class of simple flat languages defined here is shonw to be learnable in the limit from positive data. In particular, its subclass named k-consecutive, which covers a part of the class of context-sensitive languages not belonging to the class of context-free languages, is polynomial-time learnable in the limit from positive data. The class of "disjunct" unions of simple flat languages is a nontrivial example which is learnable in the limit from positive data, but does not have Wright's finite elasticity. The learning algorithm presented here for identifying the subclasses of flat languages consists essentially of identifying an arithmetic progression in the limit from positive examples using Euclidean algorithm of mutual division.


2004 ◽  
Vol 329 (1-3) ◽  
pp. 203-221 ◽  
Author(s):  
Yasuhiro Tajima ◽  
Etsuji Tomita ◽  
Mitsuo Wakatsuki ◽  
Matsuaki Terada

Author(s):  
Yuta Yoshimura ◽  
Takayoshi Shoudai ◽  
Yusuke Suzuki ◽  
Tomoyuki Uchida ◽  
Tetsuhiro Miyahara

1996 ◽  
Vol 07 (04) ◽  
pp. 309-327 ◽  
Author(s):  
SATOSHI KOBAYASHI ◽  
TAKASHI YOKOMORI

This paper introduces some subclasses of noncounting languages and presents some results on the learnability of the classes from positive data. We first establish several relationships among the language classes introduced and the class of reversible languages. Especially, we introduce the notion of local parsability, and define a class (k, l)-CLTS, which is a subclass of the class of concatenations of strictly locally testable languages. We show its close relation to the class of reversible languages. We then study on the relationship between the closure of the Boolean operations and the learnability in the limit from positive data only. Further, we explore the learnability question of some subclasses of noncounting languages in the model of identification in the limit from positive data. In particular, we show that, for each k, l≥1, (k, l)-CLTS is identifiable in the limit from positive data using reversible automata with the conjectures updated in polynomial time. Some possible applications of the result are also briefly discussed.


Sign in / Sign up

Export Citation Format

Share Document