Scholarlink

In the era of advancing digital technologies, a tremendous amount of online content is continually being generated, including product evaluations expressed across different social media platforms. This analysis gives highly advantageous information to different companies serving on online platforms. The technique of deciphering and analyzing the tone and emotions conveyed within textual content is known as sentiment analysis (SA) and determines the overall sentiment of the text. However, this approach lacks the granularity to recognize a specific entity to the sentiment attributed or some crucial details such as the aspects present within the text. Aspect-based sentiment analysis (ABSA) performs a more complicated task by giving fine-grained opinion toward a specific aspect to its associated sentiment. It gives a comprehensive understanding of the sentiments expressed in relation to distinct elements within the text. Bidirectional encoder representations from transformers (BERT) is a language representing model that identifies patterns and relationships through unlabeled data autonomously, which adopts a fine-tuning approach considering the previous and the next phrases for the context of the word simultaneously. This paper explores the potential use of different approaches on pre-trained (PT) language model BERT and fine-tuning for its performance enhancement for ABSA task which outperforms the other results for Laptop reviews. SemEval2015 (task 12, subtask 2) and SemEval2016 (task 5).

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

Aspect-Based Sentiment Analysis on Social Media Platforms Using a Dynamic Pre-trained BERT Language Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • T. Sai Suvarna,
  • Sunil Kumar Mohapatra,
  • Xuyang Hu,
  • Lu Wang,
  • Bharati Rathore

摘要

In the era of advancing digital technologies, a tremendous amount of online content is continually being generated, including product evaluations expressed across different social media platforms. This analysis gives highly advantageous information to different companies serving on online platforms. The technique of deciphering and analyzing the tone and emotions conveyed within textual content is known as sentiment analysis (SA) and determines the overall sentiment of the text. However, this approach lacks the granularity to recognize a specific entity to the sentiment attributed or some crucial details such as the aspects present within the text. Aspect-based sentiment analysis (ABSA) performs a more complicated task by giving fine-grained opinion toward a specific aspect to its associated sentiment. It gives a comprehensive understanding of the sentiments expressed in relation to distinct elements within the text. Bidirectional encoder representations from transformers (BERT) is a language representing model that identifies patterns and relationships through unlabeled data autonomously, which adopts a fine-tuning approach considering the previous and the next phrases for the context of the word simultaneously. This paper explores the potential use of different approaches on pre-trained (PT) language model BERT and fine-tuning for its performance enhancement for ABSA task which outperforms the other results for Laptop reviews. SemEval2015 (task 12, subtask 2) and SemEval2016 (task 5).