Title :
Fine-Tuning in Portuguese-English Statistical Machine Translation
Author :
Aziz, Wilker F. ; Pardo, Thiago A S ; Paraboni, Ivandré
Author_Institution :
NILC/ICMC, Univ. of Sao Paulo, Sao Carlos, Brazil
Abstract :
In previous work we have shown results of a first experiment in Statistical Machine Translation (SMT) for Brazilian Portuguese and American English using state-of-the-art phrase-based models. In this paper we compare a number of training and decoding parameter choices for fine-tuning the system as an attempt to obtain optimal results for this language pair.
Keywords :
Decoding; Hardware; Heuristic algorithms; Humans; NIST; Natural languages; Surface-mount technology; Testing; Time factors; Training data;
Conference_Titel :
Information and Human Language Technology (STIL), 2009 Seventh Brazilian Symposium in
Conference_Location :
Sao Carlos, TBD, Brazil
Print_ISBN :
978-1-4244-6008-3
DOI :
10.1109/STIL.2009.16
