Repository logo
 

Source sentence simplification for statistical machine translation

Published version
Peer-reviewed

Change log

Authors

Hasler, EC 
de Gispert, A 
Waite, A 
Byrne, WJ 

Abstract

Long sentences with complex syntax and long-distance dependencies pose difficulties for machine translation systems. Short sentences, on the other hand, are usually easier to translate. We study the potential of addressing this mismatch using text simplifi- cation: given a simplified version of the full input sentence, can we use it in addition to the full input to improve translation? We show that the spaces of original and simplified translations can be effectively combined using translation lattices and compare two decoding approaches to process both inputs at different levels of integration. We demonstrate on source-annotated portions of WMT test sets and on top of strong baseline systems combining hierarchical and neural translation for two language pairs that source simplification can help to improve translation quality.

Description

Keywords

hierarchical machine translation, text simplification, neural machine translation

Journal Title

Computer Speech & Language

Conference Name

Journal ISSN

0885-2308
1095-8363

Volume Title

45

Publisher

Elsevier
Sponsorship
Engineering and Physical Sciences Research Council (EP/L027623/1)
This work was supported by the EPSRC grant Improving Target Language Fluency in Statistical Machine Translation, grant number EP/L027623/1.
Relationships
Is supplemented by: