Gracy Theresa W, Pabitha C, Revathi K, Chawengsaksopark Pornpimol, Sathyanarayanan Mithilesh
Department of Artificial Intelligence and Data Science, Panimalar Engineering College, Chennai, India.
Department of Computer Science and Engineering, SRM Valliammai Engineering College, Chennai, India.
Sci Rep. 2025 Jul 21;15(1):26437. doi: 10.1038/s41598-025-12478-6.
An important part of customer relationship management (CRM) is being able to read emails for emotional cues; this helps with both communication and keeping customers satisfied. This study aims to improve email emotion identification by creating a system combining visual clues, aural signals, and textual information. To analyze text and emoji, the system uses advanced affective computing techniques such as Robustly Optimized Bidirectional Encoder Representations from Transformers Approach (RoBERTa), Convolutional Neural Networks (CNN) for images, Bidirectional Convolutional Long Short-Term Memory (BiConvLSTM) for video, and Cross-Modal BERT for audio. Together, they enable a wider variety of emotional signals to be extracted and understood from email content, yielding more insightful results than text-based analysis could on its own. By facilitating better two-way communication and customer satisfaction, the technology intends to improve CRM by providing actionable information that firms can use to personalize responses. This research delves into the possible uses of multi-modal emotional analysis across different businesses dealing with customers and builds a strong foundation.
客户关系管理(CRM)的一个重要部分是能够通过电子邮件识别情感线索;这有助于沟通并让客户保持满意。本研究旨在通过创建一个结合视觉线索、听觉信号和文本信息的系统来改进电子邮件情感识别。为了分析文本和表情符号,该系统使用了先进的情感计算技术,如基于变换器的稳健优化双向编码器表示法(RoBERTa)、用于图像的卷积神经网络(CNN)、用于视频的双向卷积长短期记忆网络(BiConvLSTM)以及用于音频的跨模态BERT。这些技术共同作用,能够从电子邮件内容中提取和理解更广泛的情感信号,产生比单纯基于文本的分析更有深度的结果。通过促进更好的双向沟通和提高客户满意度,该技术旨在通过提供企业可用于个性化回复的可操作信息来改进客户关系管理。本研究深入探讨了跨不同客户业务的多模态情感分析的可能用途,并奠定了坚实的基础。