Bidirectional chart generation of natural language texts

Authors:
Masahiko Haruno;Yasuharu Den;Yuji Matsumoto;Makoto Nagao
Affiliations:
NTT Corporation and Department of Electrical Engineering, Kyoto University;ATR Interpreting Telecommunication Research Laboratories;Advanced Institute of Science and Technology, Nara;Department of Electrical Engineering, Kyoto University
Venue:
AAAI'93 Proceedings of the eleventh national conference on Artificial intelligence
Year:
1993

Citing 7
Cited 1

An assumption-based TMS

Artificial Intelligence
Guarded horn clauses

Proceedings of the 4th conference on Logic programming '85
A Parallel Parsing System for Natural Language Analysis

Proceedings of the Third International Conference on Logic Programming
A semantic-head-driven generation algorithm for unification-based formalisms

ACL '89 Proceedings of the 27th annual meeting on Association for Computational Linguistics
Some chart-based techniques for parsing ill-formed input

ACL '89 Proceedings of the 27th annual meeting on Association for Computational Linguistics
Planning text for advisory dialogues

ACL '89 Proceedings of the 27th annual meeting on Association for Computational Linguistics
Generation as structure driven derivation

COLING '88 Proceedings of the 12th conference on Computational linguistics - Volume 2

A modular toolkit for machine translation based on layered charts

Proceedings of the COLING-2000 Workshop on Using Toolsets and Architectures To Build NLP Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents Bidirectional Chart Generation (BCG) algorithm as an uniform control mechanism for sentence generation and text planning. It is an extension of Semantic Head Driven Generation algorithm [Shieber et al., 1989] in that recomputation of partial structures and backtracking are avoided by using a chart table. These properties enable to handle a large scale grammar including text planning and t.o implement the algorithm in parallel programming languages. Other merits of the algorithm are to deal with multiple contexts and to keep every partial structure in the chart. It becomes easier for the generator to find a recovery strategy when user cannot understand the generated text.