Comparing Feature Importance and Rule Extraction for Interpretability on Text Data

Gianluigi Lopardo, Damien Garreau
7/4/2022
cs.LGcs.AIcs.CL

No official code found. Best implementation: gianluigilopardo/anchors_vs_lime_text (9 stars, last commit 2024-10-26).

Abstract

Complex machine learning algorithms are used more and more often in critical tasks involving text data, leading to the development of interpretability methods. Among local methods, two families have emerged: those computing importance scores for each feature and those extracting simple logical rules. In this paper we show that using different methods can lead to unexpectedly different explanations, even when applied to simple models for which we would expect qualitative coincidence. To quantify this effect, we propose a new approach to compare explanations produced by different methods.

View on arXivView PDF

Code Implementations(1)

gianluigilopardo/anchors_vs_lime_textCommunity100%

Code for the paper "Comparing Feature Importance and Rule Extraction for Interpretability on Text Data", XAIE @ ICPR 2022

90Python, Jupyter NotebookApr 16, 20221 years ago
explainable-aiinterpretabilityinterpretability-methodssentiment-analysis

Similar Papers

Position: Rethinking Explainable Machine Learning as Applied Statistics

85% match2024

Unified Explanations in Machine Learning Models: A Perturbation Approach

85% match1 repo2024

On the Importance and Evaluation of Narrativity in Natural Language AI Explanations

85% match1 repo12026

On the Relationship Between Interpretability and Explainability in Machine Learning

84% match2023

Pyreal: A Framework for Interpretable ML Explanations

84% match2023

Explainability for Machine Learning Models: From Data Adaptability to User Perception

84% match2024

Cite this paper

@article{lopardo2022comparing,
  title  = {Comparing Feature Importance and Rule Extraction for Interpretability on Text Data},
  author = {Gianluigi Lopardo and Damien Garreau},
  year   = {2022},
  eprint = {2207.01420},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2207.01420v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2207.01420.svg)](https://codeofpaper.com/paper/2207.01420)

Discussion