MusicLIME: Explainable Multimodal Music Understanding

2024-09-16

arXiv.org

AI Summary

To address the lack of interpretability in multimodal music understanding models, this paper proposes MusicLIME—the first model-agnostic explanation framework specifically designed for cross-modal interaction between music audio and lyrics. Unlike existing methods that perform unimodal attribution in isolation, MusicLIME introduces a perturbation-based LIME variant tailored to musical signals, jointly models cross-modal feature coupling, and employs a local-to-global explanation aggregation algorithm to enable fine-grained, joint attribution and generate coherent, multimodal interpretability visualizations. Additionally, it incorporates a multimodal attention consistency verification mechanism to ensure explanation reliability. Extensive evaluation on state-of-the-art music classification and sentiment analysis models demonstrates that MusicLIME’s explanations achieve 87% agreement among domain-expert musicians, significantly enhancing decision trustworthiness and bias detection capability.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Multimodal models are critical for music understanding tasks, as they capture the complex interplay between audio and lyrics. However, as these models become more prevalent, the need for explainability grows-understanding how these systems make decisions is vital for ensuring fairness, reducing bias, and fostering trust. In this paper, we introduce MusicLIME, a model-agnostic feature importance explanation method designed for multimodal music models. Unlike traditional unimodal methods, which analyze each modality separately without considering the interaction between them, often leading to incomplete or misleading explanations, MusicLIME reveals how audio and lyrical features interact and contribute to predictions, providing a holistic view of the model's decision-making. Additionally, we enhance local explanations by aggregating them into global explanations, giving users a broader perspective of model behavior. Through this work, we contribute to improving the interpretability of multimodal music models, empowering users to make informed choices, and fostering more equitable, fair, and transparent music understanding systems.

Problem

• Music Understanding
• Multimodal Music Models
• Explainable AI

Innovation

• MusicLIME
• Multimodal Music Models
• Interpretability