Evaluating Text Classification Robustness to Part-of-Speech Adversarial Examples

TL;DR

This paper investigates the robustness of text classification models against adversarial examples, revealing biases in CNNs related to parts of speech that expose vulnerabilities in linguistic understanding.

Contribution

It identifies specific parts of speech that significantly impact CNN-based classifiers, highlighting a critical vulnerability in their linguistic processing.

Findings

CNNs show bias against certain parts of speech in review datasets

Part-of-speech tokens significantly influence classifier decisions

Vulnerabilities in CNN linguistic processing are uncovered

Abstract

As machine learning systems become more widely used, especially for safety critical applications, there is a growing need to ensure that these systems behave as intended, even in the face of adversarial examples. Adversarial examples are inputs that are designed to trick the decision making process, and are intended to be imperceptible to humans. However, for text-based classification systems, changes to the input, a string of text, are always perceptible. Therefore, text-based adversarial examples instead focus on trying to preserve semantics. Unfortunately, recent work has shown this goal is often not met. To improve the quality of text-based adversarial examples, we need to know what elements of the input text are worth focusing on. To address this, in this paper, we explore what parts of speech have the highest impact of text-based classifiers. Our experiments highlight a distinct bias in CNN algorithms against certain parts of speech tokens within review datasets. This finding underscores a critical vulnerability in the linguistic processing capabilities of CNNs.