On the reliability of feature attribution methods for speech classification
Not on the map. Why?
Abstract (arXiv)
As the capabilities of large-scale pre-trained models evolve, understanding the determinants of their outputs becomes more important. Feature attribution aims to reveal which parts of the input elements contribute the most to model outputs. In speech processing, the unique characteristics of the input signal make the application of feature attribution methods challenging. We study how factors such as input type and aggregation and perturbation timespan impact the reliability of standard feature attribution methods, and how these factors interact with characteristics of each classification task. We find that standard approaches to feature attribution are generally unreliable when applied to the speech domain, with the exception of word-aligned perturbation methods when applied to word-based classification tasks.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 26 of its 31 references that are in the map. First posted on arXiv 22 May 2025.
Builds on
26 of 31 references.
References · 26
- 2017Attention Is All You Needcited by 60,361
- 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
- 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,057
- 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,905
- 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
- 2021HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Unitscited by 2,342
- 2013Visualizing and Understanding Convolutional Networkscited by 2,290
- 2019Common Voice: A Massively-Multilingual Speech Corpuscited by 1,384
- 2020Captum: A unified and generic model interpretability library for PyTorchcited by 547
- 2020Explaining by Removing: A Unified Framework for Model Explanationcited by 218
- 2019Speech Model Pre-training for End-to-End Spoken Language Understandingcited by 193
- 2022The Disagreement Problem in Explainable Machine Learning: A Practitioner's Perspectivecited by 159
14 more
- 2020The elephant in the interpretability room: Why use attention as explanation when we have saliency methods?cited by 122
- 2020Evaluating Explanations: How much do explanations from the teacher aid students?cited by 94
- 2019When Explanations Lie: Why Many Modified BP Attributions Failcited by 92
- 2022Impossibility Theorems for Feature Attributioncited by 78
- 2018AudioMNIST: Exploring Explainable Artificial Intelligence for Audio Analysis on a Simple Benchmarkcited by 72
- 2021"Will You Find These Shortcuts?" A Protocol for Evaluating the Faithfulness of Input Salience Methods for Text Classificationcited by 66
- 2021Exploring the Role of BERT Token Representations to Explain Sentence Probing Resultscited by 29
- 2022A Song of (Dis)agreement: Evaluating the Evaluation of Explainable Artificial Intelligence in Natural Language Processingcited by 13
- 2023Can We Trust Explainable AI Methods on ASR? An Evaluation on Phoneme Recognitioncited by 10
- 2023Explanations for Automatic Speech Recognitioncited by 8
- 2023Explaining Speech Classification Models via Word-Level Audio Segments and Paralinguistic Featurescited by 7
- 2024SPES: Spectrogram Perturbation for Explainable Speech-to-Text Generationcited by 6
- 2024Investigating the Effectiveness of Explainability Methods in Parkinson's Detection from Speechcited by 4
- 2024Phoneme Discretized Saliency Maps for Explainable Detection of AI-Generated Voicecited by 3
Built on by
Citing papers per year: 2026: 1.
1 paper.