Scholarlink

<p>Multilingual communication plays an important role in modern industrial environments. This study investigates the phenomenon of code-switching and code-mixing between English-Kokborok language sentence and evaluates its implications for multilingual communication system. Multiple machine learning techniques including Support Vector Machines, N-grams Models, Hidden Markov Models, Transformers, and Decision Trees were deployed to analyze and classify language usage patterns. The experimental results show that the Transformer- based model achieved the highest with an accuracy rate of 92.2% and an F1-score of 92.0%, in traditional machine learning approaches. A statistical comparison using ANOVA assesses that the changes are of practical record (F = 15.67, <i>p</i> &lt; 0.05). The post hoc analysis showed that the transformer model works well in using contextual embeddings for different languages. These findings demonstrate the performance of machine learning techniques in building multilingual communication systems in low-resource languages fields.</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

Linguistic Innovation and Operational Efficiency: A Study on Code-Mixing and Code-Switching Low Resource Language

  • 期刊论文

  • 发表日期：

    2026年5月15日

查看全文
  • Enjula Uchoi,
  • Koj Sambyo

摘要

Multilingual communication plays an important role in modern industrial environments. This study investigates the phenomenon of code-switching and code-mixing between English-Kokborok language sentence and evaluates its implications for multilingual communication system. Multiple machine learning techniques including Support Vector Machines, N-grams Models, Hidden Markov Models, Transformers, and Decision Trees were deployed to analyze and classify language usage patterns. The experimental results show that the Transformer- based model achieved the highest with an accuracy rate of 92.2% and an F1-score of 92.0%, in traditional machine learning approaches. A statistical comparison using ANOVA assesses that the changes are of practical record (F = 15.67, p < 0.05). The post hoc analysis showed that the transformer model works well in using contextual embeddings for different languages. These findings demonstrate the performance of machine learning techniques in building multilingual communication systems in low-resource languages fields.