Direct-Bridge Combination Scenario for Persian-Spanish Low-Resource Statistical Machine Translation (original) (raw)

—In this paper we compare two different approaches for translating from Persian to Spanish, as a language pair with scarce parallel corpus. The first approach involves direct transfer using an statistical machine translation system, which is available for this language pair. The second approach involves translation through English, as a pivot language, which has more translation resources and more advanced translation systems available. The results show that, it is possible to achieve better translation quality using English as a pivot language in either approach outperforms direct translation from Persian to Spanish. Our best result is the pivot system which scores higher than direct translation by (1.12) BLEU points.

This paper is an attempt to exclusively focus on investigating the pivot language technique in which a bridging language is utilized to increase the quality of the Persian–Spanish low-resource Statistical Machine Translation (SMT). In this case, English is used as the bridging language, and the Persian–English SMT is combined with the English–Spanish one, where the relatively large corpora of each may be used in support of the Persian–Spanish pairing. Our results indicate that the pivot language technique outperforms the direct SMT processes currently in use between Persian and Spanish. Furthermore, we investigate the sentence translation pivot strategy and the phrase translation in turn, and demonstrate that, in the context of the Persian–Spanish SMT system, the phrase-level pivoting outperforms the sentence-level pivoting. Finally we suggest a method called combination model in which the standard direct model and the best triangulation pivoting model are blended in order to reach a high-quality translation

The quality of Neural Machine Translation (NMT) systems like Statistical Machine Translation (SMT) systems, heavily depends on the size of training data set, while for some pairs of languages, high-quality parallel data are poor resources. In order to respond to this low-resourced training data bottleneck reality, we employ the pivoting approach in both neural MT and statistical MT frameworks. During our experiments on the Persian-Spanish, taken as an under-resourced translation task, we discovered that, the aforementioned method, in both frameworks, significantly improves the translation quality in comparison to the standard direct translation approach.

Abstract In this paper, we present the results of experiments to enhance the performance of a baseline statistical machine translation system with an automatically extracted parallel corpus from comparable corpora. We train new translation systems by combining the extracted corpus with different sizes of parallel corpora. We also experiment with combining the phrase tables trained separately from the two resources. These phrase tables are then used in a phrase-based SMT decoder to translate test sets.