Extending Logic Explained Networks to Text Classification

TL;DR

This paper introduces LENp, an improved logic explanation method for text classification that offers more accurate and user-friendly local explanations than existing methods like LIME.

Contribution

We extend Logic Explained Networks to text classification with LENp, enhancing local explanations through input perturbation and demonstrating superior performance over LIME.

Findings

LENp provides better local explanations than LIME in sensitivity and faithfulness.

Logic explanations are more useful and user-friendly than feature scoring.

LENp outperforms LIME in human surveys for explanation quality.

Abstract

Recently, Logic Explained Networks (LENs) have been proposed as explainable-by-design neural models providing logic explanations for their predictions. However, these models have only been applied to vision and tabular data, and they mostly favour the generation of global explanations, while local ones tend to be noisy and verbose. For these reasons, we propose LENp, improving local explanations by perturbing input words, and we test it on text classification. Our results show that (i) LENp provides better local explanations than LIME in terms of sensitivity and faithfulness, and (ii) logic explanations are more useful and user-friendly than feature scoring provided by LIME as attested by a human survey.