An experiment for determining semantic relations between main and subordinate clauses in complex sentences

Authors:
Sawa Takakura;Dongli Han;Teiji Furugori
Affiliations:
The Univ. of Electro-Communications, Tokyo, Japan;The Univ. of Electro-Communications, Tokyo, Japan;The Univ. of Electro-Communications, Tokyo, Japan
Venue:
WISICT '04 Proceedings of the winter international synposium on Information and communication technologies
Year:
2004

Citing 2
Cited 0

Connectionist learning procedures

Artificial Intelligence
Mechanisms of sentence processing: assigning roles to constituents

Parallel distributed processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Complex sentences are a cause for difficulties in natural language processing. A machine translation system often produces incorrect translations unless it grasps the semantic relation between main and subordinate clauses in a complex sentence. This paper describes a way for determining the semantic relation that holds between the clauses. It extracts grammatical and semantic features crucial for finding 7 semantic relations assumed to hold between them. Then, it tries to find a most likely semantic relation by taking the features as input to a neural network learning procedure. An experimental result for 100 complex sentences whose main and subordinate clauses are connected by the particle (te) shows that we are successful to find the semantic relations correctly at a rate of 73%.