Efficient Classification of Long Documents Using Transformers

TL;DR

This paper evaluates various transformer-based methods for long document classification, highlighting the importance of comprehensive baselines and datasets to fairly compare model efficacy across different tasks and data organizations.

Contribution

It provides a thorough benchmark and analysis of existing approaches, emphasizing the need for standardized evaluation in long document classification.

Findings

Complex models often do not outperform simple baselines

Performance varies significantly across datasets

Current methods lack consistency and robustness

Abstract

Several methods have been proposed for classifying long textual documents using Transformers. However, there is a lack of consensus on a benchmark to enable a fair comparison among different approaches. In this paper, we provide a comprehensive evaluation of the relative efficacy measured against various baselines and diverse datasets -- both in terms of accuracy as well as time and space overheads. Our datasets cover binary, multi-class, and multi-label classification tasks and represent various ways information is organized in a long text (e.g. information that is critical to making the classification decision is at the beginning or towards the end of the document). Our results show that more complex models often fail to outperform simple baselines and yield inconsistent performance across datasets. These findings emphasize the need for future studies to consider comprehensive baselines and datasets that better represent the task of long document classification to develop robust models.