The Strong, Weak, and Very Weak Finite Context and Kernel Properties
Language and Automata Theory and Applications - 11th International Conference, (LATA2017),
Lecture Notes in Computer Science Vol.10168, pp.77-88
(2017), [peer-reviewed]
Event Date:
March 6-9, 2017
Abstract / 概要
We identify the properties of context-free grammars that exactly correspond to the behavior of the dual and primal versions of Clark and Yoshinaka’s distributional learning algorithm and call them the very weak finite context/kernel property. We show that the very weak finite context property does not imply Yoshinaka’s weak finite context property, which has been assumed to hold of the target language for the dual algorithm to succeed. We also show that the weak finite context property is genuinely weaker than Clark’s strong finite context property, settling a question raised by Yoshinaka.