Document Classification

We propose an automated, bi-level document classification pipeline for regulatory text that first filters irrelevant documents using an ensemble of classical classifiers and then routes relevant documents to the appropriate departments using a transformer-based DocBERT model, significantly improving efficiency and accuracy over single-model approaches when evaluated on a large real-world corpus from Con Edison.

This work proposes a two-stage document classification system that filters and routes regulatory documents using an ensemble of classical classifiers and a DocBERT-based multi-label model, achieving strong performance on real-world industrial data from Con Edison.