Scholarlink

<p>Efficient email management is vital to organizations and businesses that process vast volumes of messages every day. In this paper, an intelligent email processing system for automated categorization, summarization, and sentiment analysis is proposed using state-of-the-art transformer-based models. Emails are accessed securely using the Gmail API, which enables real-time processing. In categorization, Sentence-BERT embeddings are used to compute semantic similarity between emails and predefined categories allowing flexible classification while using an uncategorized option for ambiguous emails. Summarization is done through a pre-trained BART-based model on conversational data to create concise and consistent summaries of lengthy emails. Summarization is only applied when explicitly requested by the user, with clear disclosure of AI involvement to maintain transparency and user agency. Sentiment analysis is performed by a RoBERTa model trained on social data, labelling emails as positive, negative, or neutral providing emotional context for prioritizing responses. The system boosts productivity with less manual triage, readable text, and rapid insight extraction. It is applicable to various enterprises such as customer service, HR, and project coordination, providing scalable automation and actionable analytics for heavy communication-intensive environments.</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

Automated email processing via a transformer based framework for open-set email categorization, abstractive summarization and sentiment analysis

  • 期刊论文

  • 发表日期：

    2026年5月29日

查看全文
  • Soundarya Suresh,
  • M. Amrutha Sridhanya,
  • Balika J. Chelliah,
  • B. Aarthi

摘要

Efficient email management is vital to organizations and businesses that process vast volumes of messages every day. In this paper, an intelligent email processing system for automated categorization, summarization, and sentiment analysis is proposed using state-of-the-art transformer-based models. Emails are accessed securely using the Gmail API, which enables real-time processing. In categorization, Sentence-BERT embeddings are used to compute semantic similarity between emails and predefined categories allowing flexible classification while using an uncategorized option for ambiguous emails. Summarization is done through a pre-trained BART-based model on conversational data to create concise and consistent summaries of lengthy emails. Summarization is only applied when explicitly requested by the user, with clear disclosure of AI involvement to maintain transparency and user agency. Sentiment analysis is performed by a RoBERTa model trained on social data, labelling emails as positive, negative, or neutral providing emotional context for prioritizing responses. The system boosts productivity with less manual triage, readable text, and rapid insight extraction. It is applicable to various enterprises such as customer service, HR, and project coordination, providing scalable automation and actionable analytics for heavy communication-intensive environments.