Smart Email Sorting: A Machine Learning Approach for Enhanced Productivity

Abstract

Smart Email Sorting System uses Natural Language Processing (NLP) and a fine-tuned DistilBERT model to classify emails, by analyzing the contents of an email message and determining if they fall into categories (e.g., Invoices, Meetings, Project Update) or not. It also pulls deadlines and key info from unstructured text. Using OAuth integration with Gmail, the system sorts emails according to urgency and stresses time-sensitive engagements, allowing users to swiftly locate significant communications in an increasingly digital environment where they can manage their inboxes more efficiently. The system eliminates the manual overhead on a massive scale by automating email sorting and prioritization based on predictive capabilities. It prevents essential emails from getting lost in the shuffle and reminds users to respond quickly to significant activities. This results in increased productivity, efficient time management and more streamlined workflow; offering a solution to managing copious amounts of correspondence both professionally and personally.

Introduction

The text presents the Smart Email Sorting System, an AI-powered solution designed to improve email management in modern workplaces. Traditional email clients rely on chronological sorting or rule-based filters, which often fail to capture context, urgency, or deadlines, leading to information overload, missed tasks, and reduced productivity.

The proposed system leverages machine learning and NLP, specifically a fine-tuned DistilBERT model combined with a deadline extraction engine, to automatically classify and prioritize emails based on content, urgency, and key dates. Emails are categorized into 17 business-specific types (e.g., invoices, legal, customer support, urgent), and deadlines are highlighted, helping users focus on time-critical tasks.

Key Features and Workflow:

• Integration: Secure Gmail OAuth login fetches emails into the system.
• Processing: ML engine classifies emails; NLP extracts deadlines and urgency.
• Storage & Analytics: Data is stored in PostgreSQL; dashboards provide real-time insights and priorities.
• User Interface: Interactive React frontend with simplified, non-technical dashboards; admin interface allows feedback to retrain the model.
• Architecture: Microservice-based asynchronous backend with FastAPI, Celery, Redis, and PostgreSQL for scalability and efficiency.

Compared to traditional or semi-automated systems, this approach reduces manual effort, improves task tracking, and boosts workplace productivity by providing context-aware, deadline-sensitive email management.

Conclusion

This innovation illustrates a significant gain in productivity by embedding an intelligent machine learning filter, rather than traditional rule-based filters such as spam or data loss filters. The model is fine-tuned DistilBERT and refined using NLP for correct classification of emails and deadline extraction. This system does more than store information; it actively flags urgent tasks to minimize missed communications and streamline operations. This leads to privacy along with usability through role-based dashboards, reducing excess cost and improving general decision making throughout email processing via secure FastAPI architecture, new GMAIL OAuth. Future developments involve adding sentiment analysis to help prioritize findings, asynchronous processing with Celery and Redis for better performance, as well as automatic OAuth token refresh. Features like real-time notifications using WebSockets and webhook systems would enable a better response. The Lightning Web Compoents will provide enterprise-grade analytics with Power BI integration and docking-based deployment for scalability, insights, and management in large-scale environments.

References

[1] Vaswani et al., “Attention Is All You Need,” in Advances in Neural Information Processing Systems (NeurIPS), 2017. [2] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” in Proc. NAACL-HLT, 2019. [3] V. Sanh, L. Debut, J. Chaumond, and T. Wolf, “DistilBERT, a Distilled Version of BERT,” arXiv preprint arXiv:1910.01108, 2019. [4] M. J. Eppler and J. Mengis, “The Concept of Information Overload: A Review of Literature,” The Information Society, vol. 20, no. 5, pp. 325–344, 2004. [5] T. Chen et al., “MLOps Best Practices, Challenges and Maturity Models,” Journal of Software Engineering, 2025. [6] R. Kumar et al., “A Comparative Evaluation of a Multimodal Approach for Unwanted Email Detection,” Electronics, MDPI, 2025. [7] P. Singh et al., “BERT-Based Phishing Email Classification,” in Proc. COMPSAC Workshop, 2023. [8] A. Alshamrani et al., “EmailNet: Efficient Email Classification Based on Graph Similarity,” in Proc. INASS Conference, 2023. [9] J. Li et al., “A Survey on Named Entity Recognition,” Journal of Natural Language Processing, 2023. [10] M. Honnibal and I. Montani, “spaCy 2: Natural Language Understanding,” 2017.

Copyright

Copyright © 2026 Sesetti Yaswini Priya, Bala Indra Sena Reddy, Rajulapati Uday Malleswar , Bulla Vijay Kumar , CH. Jeevana Priya. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.