LIME: Weakly-Supervised Text Classification Without Seeds

TL;DR

LIME introduces a weakly-supervised text classification method that replaces seed-word reliance with entailment-based pseudo-labeling, achieving state-of-the-art results across multiple benchmarks.

Contribution

The paper presents LIME, a novel framework that eliminates seed-word dependency by using textual entailment for pseudo-labeling in weakly-supervised text classification.

Findings

LIME outperforms recent baselines in weakly-supervised classification.

Achieves state-of-the-art results on 4 benchmark datasets.

Simplifies the classification pipeline by removing seed-word generation.

Abstract

In weakly-supervised text classification, only label names act as sources of supervision. Predominant approaches to weakly-supervised text classification utilize a two-phase framework, where test samples are first assigned pseudo-labels and are then used to train a neural text classifier. In most previous work, the pseudo-labeling step is dependent on obtaining seed words that best capture the relevance of each class label. We present LIME, a framework for weakly-supervised text classification that entirely replaces the brittle seed-word generation process with entailment-based pseudo-classification. We find that combining weakly-supervised classification and textual entailment mitigates shortcomings of both, resulting in a more streamlined and effective classification pipeline. With just an off-the-shelf textual entailment model, LIME outperforms recent baselines in weakly-supervised text classification and achieves state-of-the-art in 4 benchmarks. We open source our code at https://github.com/seongminp/LIME.