Scholarlink

<p>Machine Translation (MT) technology makes documents available in any one of the languages to another language, functioning as a bridge to access cross-lingual information. Compared to human translation, such technologies are economical, instantaneous, and multiplicative. However, coming up with such a system along with a pair of languages is non-trivial. The automatic translation of an arbitrary text from one language to another is too hard to automate completely. Depending on the similarities and contrasts between the language pairings selected for translation, a fully automatic, high-quality translation can be built to varying degrees of complexity. However, the dream or goal of building an MT system is bit by bit turning into a reality. Providing translations that are almost as accurate as those generated by human translators while removing linguistic barriers is a challenge. Contrary to the general belief, there is a substantial lack of human translators even for technical translations. There is a growing demand worldwide for MT systems to fill this vacancy. Machine Translation systems for Indian languages must consider the region’s diverse linguistic and cultural landscape. Different languages require different algorithms, with Rule-Based MT and Neural-Based MT focusing on structural patterns and cultural nuances. The Sampark MT system has been developed by DeitY for translating the Indian languages. Sampark System for Malayalam–Tamil doesn’t include the verb agreement, affecting the output quality. This paper emphasizes the need for grammatical transfer in the current Malayalam–Tamil Sampark MT system. The paper also explores how the morphological analyzer’s improvisation and transfer grammar incorporation improves the translated text’s output quality. The accuracy of the MT system increased to 71.75% with the enhancement of the modules and the incorporation of subject-object and semantic role identification into the Sampark MT system. This study aims to develop an improved system for Malayalam–Tamil.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Challenges and improvisation in machine translation: the case of malayalam–tamil machine translation

  • 期刊论文

  • 发表日期：

    2025年10月30日

查看全文
  • Jisha P. Jayan,
  • J. Satheesh Kumar,
  • T. Amudha

摘要

Machine Translation (MT) technology makes documents available in any one of the languages to another language, functioning as a bridge to access cross-lingual information. Compared to human translation, such technologies are economical, instantaneous, and multiplicative. However, coming up with such a system along with a pair of languages is non-trivial. The automatic translation of an arbitrary text from one language to another is too hard to automate completely. Depending on the similarities and contrasts between the language pairings selected for translation, a fully automatic, high-quality translation can be built to varying degrees of complexity. However, the dream or goal of building an MT system is bit by bit turning into a reality. Providing translations that are almost as accurate as those generated by human translators while removing linguistic barriers is a challenge. Contrary to the general belief, there is a substantial lack of human translators even for technical translations. There is a growing demand worldwide for MT systems to fill this vacancy. Machine Translation systems for Indian languages must consider the region’s diverse linguistic and cultural landscape. Different languages require different algorithms, with Rule-Based MT and Neural-Based MT focusing on structural patterns and cultural nuances. The Sampark MT system has been developed by DeitY for translating the Indian languages. Sampark System for Malayalam–Tamil doesn’t include the verb agreement, affecting the output quality. This paper emphasizes the need for grammatical transfer in the current Malayalam–Tamil Sampark MT system. The paper also explores how the morphological analyzer’s improvisation and transfer grammar incorporation improves the translated text’s output quality. The accuracy of the MT system increased to 71.75% with the enhancement of the modules and the incorporation of subject-object and semantic role identification into the Sampark MT system. This study aims to develop an improved system for Malayalam–Tamil.