A Character Level Based and Word Level Based Approach for Chinese-Vietnamese Machine Translation

Area of Science:

• Computational Linguistics
• Natural Language Processing
• Machine Translation

Background:

• Chinese and Vietnamese are isolated languages lacking word delimiters (spaces), posing challenges for machine translation.
• Word segmentation is a critical preprocessing step in translating between these languages and others, but its necessity is debated for space-delimited language pairs.
• The Chinese-Vietnamese language pair is considered low-resource, leading to significant data sparsity issues in translation systems.

Purpose of the Study:

• To propose a new machine translation method for the Chinese-Vietnamese language pair.
• To address the challenges of word segmentation and data sparsity in low-resource machine translation.
• To improve the performance of Chinese-Vietnamese machine translation by leveraging both character and word-level translation strategies.

Main Methods:

• A novel hybrid approach combining statistical and rule-based methods for word-level translation.
• Utilizing statistical translation at the character level.
• Integrating character-level and word-level translation components to leverage their respective advantages.

Main Results:

• The proposed method demonstrated improved performance in Chinese-Vietnamese machine translation compared to traditional character-level or word-level translation methods alone.
• The hybrid approach effectively mitigated issues related to data sparsity in this low-resource language pair.
• Experimental results validated the efficacy of the combined character and word-level translation strategy.

Conclusions:

• The proposed hybrid machine translation method offers a significant improvement for the Chinese-Vietnamese language pair.
• Combining character and word-level translation strategies is beneficial, particularly for low-resource languages with segmentation ambiguities.
• This research contributes a more effective approach to machine translation for languages lacking explicit word boundaries.