Multi-modal emotional analysis in customer relation management and enhancing communication through integrated...

Area of Science:

• Affective computing
• Natural Language Processing (NLP)
• Customer Relationship Management (CRM)

Background:

• Effective customer relationship management (CRM) relies on understanding customer emotions.
• Current methods often focus solely on textual analysis, limiting emotional cue detection.

Purpose of the Study:

• To develop an advanced system for identifying emotional cues in customer emails.
• To enhance CRM by integrating visual, aural, and textual data for comprehensive emotion analysis.

Main Methods:

• Utilized Robustly Optimized Bidirectional Encoder Representations from Transformers Approach (RoBERTa) for text and emoji analysis.
• Employed Convolutional Neural Networks (CNN) for image data and Bidirectional Convolutional Long Short-Term Memory (BiConvLSTM) for video.
• Integrated Cross-Modal BERT for audio signal processing to capture a wider range of emotional signals.

Main Results:

• The multi-modal system extracts a broader spectrum of emotional signals compared to text-only analysis.
• Demonstrated the potential for more insightful analysis of customer email content.

Conclusions:

• The developed technology offers actionable insights for personalizing customer responses.
• This research lays the groundwork for multi-modal emotional analysis applications across various customer-facing industries.