Scholarlink

<p>Code-mixing is a prevalent occurrence in cross-linguistic environments, where conversations frequently incorporate multiple languages. Users of online social media platforms, discussion forums, and subject experts who maintain different community channels and subscribers to these community channels typically express their opinions in a code-mixed language. Since code-mixed scenarios do not impose restrictions on language usage, analyzing code-mixed data poses a considerable challenge. This research focuses on sentiment analysis of Telugu–English code-mixed text utilizing transformer-based models. The transformer models demonstrated enhanced performance over other state-of-the-art baseline models on a Telugu–English code-mixed dataset, achieving a 6% higher accuracy and a 9% improvement in F1-score on the CMTE-IIITH dataset, and a 5% enhancement in accuracy and a 6% boost in F1-score on the CMTE-NITANP dataset.</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

Sentiment analysis of code-mixed Telugu–English text using transformers

  • 期刊论文

  • 发表日期：

    2025年11月4日

查看全文
  • Upendar Rao Rayala,
  • Karthick Seshadri

摘要

Code-mixing is a prevalent occurrence in cross-linguistic environments, where conversations frequently incorporate multiple languages. Users of online social media platforms, discussion forums, and subject experts who maintain different community channels and subscribers to these community channels typically express their opinions in a code-mixed language. Since code-mixed scenarios do not impose restrictions on language usage, analyzing code-mixed data poses a considerable challenge. This research focuses on sentiment analysis of Telugu–English code-mixed text utilizing transformer-based models. The transformer models demonstrated enhanced performance over other state-of-the-art baseline models on a Telugu–English code-mixed dataset, achieving a 6% higher accuracy and a 9% improvement in F1-score on the CMTE-IIITH dataset, and a 5% enhancement in accuracy and a 6% boost in F1-score on the CMTE-NITANP dataset.