Semi-Automatic Word-Aligned Tool for Thai-Vietnamese Parallel Corpus Construction

Abstract

A corpus, especially a parallel corpus, which contains both source and target language, is an important resource in Natural Language Processing (NLP) research, particularly in machine translation. A quality corpus can improve the accuracy of the translation results significantly; however, corpus construction is very time consuming, and requires the expertise of linguistic experts. In this paper, we present Thai-Vietnamese parallel corpus construction and the process of building a Thai-Vietnamese parallel corpus. This work focuses on the construction of a semi-automatic word-alignment tool, capable of assisting researchers in the construction of a parallel corpus. The collection and validation within this study was achieved through the use of our development tool. In the first stage, the Vietnamese -Thai parallel corpus, containing 14,771 sentence pairs; was collected, aligned at word level, and validated by linguistic experts. This parallel corpus can be used as a reliable resource for statistical machine translation and other applications.