The Important Influencing Factors in Machine Translation

What is it about?

Factored-based machine translation deals with various important linguistics information of a word during the translation process. Morphologically rich languages like Hindi provide multiple word forms from the root or dictionary word by differing in morphological information such as part of speech (POS), affixes, number, gender, etc. So, linguistic factors can provide useful information while translating morphologically rich languages. Different factors contribute in a different manner. In this chapter, we study the significance of different linguistic factors in a phrase-based statistical machine translation (SMT) framework employed for Hindi to English translation. We performed experiments over HindEnCorp and ILCI dataset for Hindi–English. We find that POS+lemma+gender achieves the highest BLEU score (16.46) for HindEnCorp dataset, and POS+lemma+number achieves the highest BLEU score (19.11) for ILCI dataset.

Featured Image

Why is it important?

This chapter highlights the importance of factored-based machine translation, particularly in morphologically rich languages like Hindi, where linguistic factors such as part of speech, lemma, gender, and number significantly impact translation quality. Through experimentation, it demonstrates how incorporating these factors into a phrase-based statistical machine translation framework improves translation accuracy, as evidenced by the notable BLEU scores achieved for HindEnCorp and ILCI datasets.

Perspectives