AraDIC: Arabic Document Classification using Image-Based Character Embeddings and Class-Balanced Loss

TL;DR

AraDIC is a novel end-to-end Arabic document classification framework that uses image-based character embeddings and class-balanced loss, eliminating the need for complex preprocessing and achieving significant performance improvements.

Contribution

This work introduces the first image-based character embedding framework for Arabic text classification, trained end-to-end with class-balanced loss on newly created datasets.

Findings

Achieves 12.29% and 23.05% improvements in F-score over baselines.

First to evaluate deep learning on modern, colloquial, and classical Arabic.

Demonstrates effectiveness of image-based character embeddings for Arabic classification.

Abstract

Classical and some deep learning techniques for Arabic text classification often depend on complex morphological analysis, word segmentation, and hand-crafted feature engineering. These could be eliminated by using character-level features. We propose a novel end-to-end Arabic document classification framework, Arabic document image-based classifier (AraDIC), inspired by the work on image-based character embeddings. AraDIC consists of an image-based character encoder and a classifier. They are trained in an end-to-end fashion using the class balanced loss to deal with the long-tailed data distribution problem. To evaluate the effectiveness of AraDIC, we created and published two datasets, the Arabic Wikipedia title (AWT) dataset and the Arabic poetry (AraP) dataset. To the best of our knowledge, this is the first image-based character embedding framework addressing the problem of Arabic text classification. We also present the first deep learning-based text classifier widely evaluated on modern standard Arabic, colloquial Arabic and classical Arabic. AraDIC shows performance improvement over classical and deep learning baselines by 12.29% and 23.05% for the micro and macro F-score, respectively.