Feature Extraction Using Deep Generative Models for Bangla Text Classification on a New Comprehensive Dataset

TL;DR

This paper introduces a new comprehensive Bangla text dataset and explores deep generative models for feature extraction, demonstrating that adversarial autoencoders yield superior features for classification.

Contribution

It presents a large, publicly available Bangla dataset and applies three deep generative models for text feature extraction, comparing their effectiveness in classification.

Findings

Adversarial autoencoder outperforms other models in feature quality.

Deep generative models from computer vision can be adapted for text analysis.

The new dataset enables further research in Bangla NLP.

Abstract

The selection of features for text classification is a fundamental task in text mining and information retrieval. Despite being the sixth most widely spoken language in the world, Bangla has received little attention due to the scarcity of text datasets. In this research, we collected, annotated, and prepared a comprehensive dataset of 212,184 Bangla documents in seven different categories and made it publicly accessible. We implemented three deep learning generative models: LSTM variational autoencoder (LSTM VAE), auxiliary classifier generative adversarial network (AC-GAN), and adversarial autoencoder (AAE) to extract text features, although their applications are initially found in the field of computer vision. We utilized our dataset to train these three models and used the feature space obtained in the document classification task. We evaluated the performance of the classifiers and found that the adversarial autoencoder model produced the best feature space.