MACHINE TRANSLATION AND CORPUS-BASED POST-EDITING IN THE FINANCIAL PRESS: A COMPARATIVE STUDY OF SMT AND NMT OUTPUTS
MACHINE TRANSLATION AND CORPUS-BASED POST-EDITING IN THE FINANCIAL PRESS: A COMPARATIVE STUDY OF SMT AND NMT OUTPUTS

Summary/Abstract: Machine translation (MT) has made huge strides in the last few decades. There are many strands of research dedicated to analysing the various types and advantages of MT, such as those related to statistical machine translation (SMT), and neural machine translation (NMT). Nonetheless, there are still a few shortcomings that the literature addresses and warns against. Also, little has been investigated regarding the quality of SMT or NMT with respect to corpus evidence. This paper aims to bridge this gap by exploring the quality and accuracy of SMT and NMT in the translation, from English into Italian, of a text in the financial field. To this aim, it compares MT outputs and corpus results by addressing an extract of a newspaper article sourced from the financial press. The paper findings report that NMT performs better in terms of word order and grammar accuracy, whereas SMT is more effective with naturally sounding words and sector-specific terms. The paper claims the high reliability of corpus data. As a matter of fact, not only does corpus evidence show language patterns and word usages in context, but it also suggests collocations and a variety of alternative terms in specific contexts.

• Details
• Contents