A method for optimizing text preprocessing and text classification using multiple cycles of learning with an...

Area of Science:

• Computer Science
• Data Science
• Natural Language Processing

Background:

• Optimizing text preprocessing and classification is vital for large organizations.
• Handling unstructured, noisy business emails with abbreviations is challenging.
• Efficient email management is crucial for knowledge extraction and cost reduction.

Purpose of the Study:

• To propose a novel two-step classification approach for efficient email processing.
• To introduce a two-cycle labeling procedure to accelerate the labeling process.
• To compare various text classification and vectorization algorithms for optimal performance.

Main Methods:

• A two-step heuristic classification approach is employed.
• A two-cycle labeling procedure is implemented to speed up data annotation.
• Comparison of classification and text vectorization algorithms using F1 score and balanced accuracy.

Main Results:

• The proposed algorithm demonstrated excellent performance in a real-world application.
• Significant improvements in organization and administration were observed.
• Reduction in operational expenses was achieved through automated email handling.

Conclusions:

• The developed approach effectively addresses challenges in business email classification.
• The method offers a scalable and efficient solution for large organizations.
• The study highlights the potential for cost savings and improved knowledge management.