A Hybrid Method for Word Segmentation with English-Vietnamese Bilingual Text

This paper proposes a hybrid approach for Vietnamese word segmentation. The approach combines a dictionary-based method and a machine learning method to detect word boundaries in Vietnamese text by comparing English-Vietnamese pairs. We also point out several characteristics of Vietnamese which affect the Vietnamese word segmentation task and word alignment of English-Vietnamese text. Moreover, we built an English-Vietnamese bilingual corpus with nearly 10 million words, namely EVBCorpus, while a part of EVBNews has been manually segmented at the word level. We evaluate the performance of our approach by comparing its word segmentation results on this corpus. Our hybrid approach achieves 97% accuracy on the EVBNews corpus.