Document Classification using File Names

2024-10-02

AI Summary

For time-sensitive applications such as digital forensics and large-scale media classification, this paper proposes a lightweight document classification method relying solely on filenames—bypassing full-content parsing to drastically reduce inference latency. We systematically demonstrate, for the first time, that filenames alone contain sufficient semantic information to support high-accuracy classification. Our approach innovatively combines TF-IDF–based feature extraction with lightweight supervised models (e.g., SVM, Random Forest), achieving an unprecedented trade-off between accuracy and efficiency. Evaluated on two real-world datasets, our method attains 99.63% and 96.57% classification accuracy, respectively, while accelerating inference by 442× over DiT. It covers over 90% of target documents, establishing a novel paradigm for real-time, resource-constrained classification.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Natural Language Processing: Text Classification & Sentiment Analysis
• Computer Vision: Image and Video Retrieval

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Rapid document classification is critical in several time-sensitive applications like digital forensics and large-scale media classification. Traditional approaches that rely on heavy-duty deep learning models fall short due to high inference times over vast input datasets and computational resources associated with analyzing whole documents. In this paper, we present a method using lightweight supervised learning models, combined with a TF-IDF feature extraction-based tokenization method, to accurately and efficiently classify documents based solely on file names, that substantially reduces inference time. Our results indicate that file name classifiers can process more than 90% of in-scope documents with 99.63% and 96.57% accuracy when tested on two datasets, while being 442x faster than more complex models such as DiT. Our method offers a crucial solution to efficiently process vast document datasets in critical scenarios, enabling fast and more reliable document classification.

Problem

• Rapid document classification for time-sensitive applications.
• Reducing inference time using lightweight supervised learning models.
• Accurate classification based solely on file names.

Innovation

• Lightweight supervised learning models for classification
• TF-IDF feature extraction for tokenization
• File name-based document classification for speed