Noriko Sugimoto

Constructive Learning of Translations Based on Dictionaries

Learning a translation based on a dictionary is to extract a binary relation over strings from given examples based on information supplied by the dictionary. In this paper, we introduce a restricted elementary formal system called a regular TEFS to formalize translations and dictionaries. Then, we propose a learning algorithm that identifies a translation defined by a regular TEFS from positive and negative examples. The main advantage of the learning algorithm is constructive, that is, the produced hypothesis reflects the examples directly. The learning algorithm generates the most specific clauses from examples by referring to a dictionary, generalizes these clauses, and then removes too strong clauses from them. As a result, the algorithm can learn translations over context-free languages.

Constructive Learning of Context-Free Languages with a Subpansive Tree

A subpansive tree is a rooted tree that gives a partial order of nonterminal symbols of a context-free grammar. We formalize subpansive trees as background knowledge of CFGs, and investigate query learning of CFGs with the help of subpansive trees. We show a restricted class of CFGs, which we call hierarchical CFGs, is efficiently learnable, while it is unlikely to be polynomial-time predictable.

Learnability of Translations from Positive Examples

One of the most important issues in machine translations is deducing unknown rules from pairs of input-output sentences. Since the translations are expressed by elementary formal systems (EFSs, for short), we formalize learning translations as the process of guessing an unknown EFS from pairs of input-output sentences. In this paper, we propose a class of EFSs called linearly-moded EFSs by introducing local variables and linear predicate inequalities based on mode information, which can express translations of context-sensitive languages. We show that, for a given input sentence, the set of all output sentences is finite and computable in a translation defined by a linearly-moded EFS. Finally, we show that the class of translations defined by linearly-moded EFSs is learnable under the condition that the number of clauses in an EFS and the length of the clause are bounded by some constant.

An Efficient Derivation for Elementary Formal Systems Based on Partial Unification

An EFS is a kind of logic programs expressing various formal languages. We propose an efficient derivation for EFSs called an S-derivation, where every possible unifiers are evaluated at one step of the derivation. In the S-derivation, each unifier is partially applied to each goal clause by assigning variables whose values are uniquely determined from the set of all possible unifiers. This contributes to reduce the number of backtracking, and thus the S-derivation works efficiently. In this paper, the S-derivation is shown to be complete for the class of regular EFSs.We implement an EFS interpreter based on the S-derivation in Prolog programming language, and compare the parsing time with that of DCG provided by the Prolog interpreter. As the results of experiments, we verify the efficiency of the S-derivation for accepting context-free languages.