Scholarlink

<p>With the rapid expansion and growing complexity of social networks, developing effective approaches for extracting meaningful knowledge from the vast volumes of user-generated data has become an essential challenge. Among these approaches, sentiment extraction and analysis<b>-</b> one of the core tasks in natural language processing (NLP)—plays a critical role in uncovering users’ attitudes, opinions, and emotions. Insights gained from such analysis can inform strategic decision-making across a wide range of domains, including marketing, public policy, healthcare, and service management. Over the past decade, sentiment analysis has attracted increasing attention within interdisciplinary research, with many studies leveraging artificial intelligence-based techniques. In particular, transformer architectures have emerged as leading methods in NLP, primarily due to their capacity to model long-range textual dependencies and to generate richer, more precise semantic representations. This paper presents a systematic literature review of research on transformer-based sentiment analysis. Specifically, we categorize and analyze different models, widely adopted benchmark datasets, evaluation metrics, as well as the challenges and diverse applications associated with these methods. A qualitative assessment of the reviewed studies, based on established systematic review standards, highlights the progress and maturity of the field from technical, linguistic, and applied perspectives. Furthermore, we analyze current research trends and identify promising directions for future work. The findings of this review underscore both the substantial growth in research output and the increasing significance of transformer-based sentiment analysis, reflecting its dynamism, innovation, and impact on interdisciplinary studies.</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

Systematic literature review on sentiment analysis using transformers

  • 期刊论文

  • 发表日期：

    2026年1月31日

查看全文
  • Sepehr Mostafavi,
  • Yeganeh Yahyavi,
  • Reza Ravanmehr

摘要

With the rapid expansion and growing complexity of social networks, developing effective approaches for extracting meaningful knowledge from the vast volumes of user-generated data has become an essential challenge. Among these approaches, sentiment extraction and analysis- one of the core tasks in natural language processing (NLP)—plays a critical role in uncovering users’ attitudes, opinions, and emotions. Insights gained from such analysis can inform strategic decision-making across a wide range of domains, including marketing, public policy, healthcare, and service management. Over the past decade, sentiment analysis has attracted increasing attention within interdisciplinary research, with many studies leveraging artificial intelligence-based techniques. In particular, transformer architectures have emerged as leading methods in NLP, primarily due to their capacity to model long-range textual dependencies and to generate richer, more precise semantic representations. This paper presents a systematic literature review of research on transformer-based sentiment analysis. Specifically, we categorize and analyze different models, widely adopted benchmark datasets, evaluation metrics, as well as the challenges and diverse applications associated with these methods. A qualitative assessment of the reviewed studies, based on established systematic review standards, highlights the progress and maturity of the field from technical, linguistic, and applied perspectives. Furthermore, we analyze current research trends and identify promising directions for future work. The findings of this review underscore both the substantial growth in research output and the increasing significance of transformer-based sentiment analysis, reflecting its dynamism, innovation, and impact on interdisciplinary studies.