Diagnosis of Dysarthria Severity and Explanation Generation Using XAI-Enhanced CLINIC-GENIE on Diadochokinetic Tasks

ACL ARR 2025 May Submission7514 Authors

Abstract: Deep neural network classifiers for dysarthria severity face limitations regarding interpretability and treatment guidance. To overcome these, we introduce CLINIC-GENIE, an explainable two-stage framework consisting of: (1) CLassification model using INtegrated Information from Clinically explainable acoustic features and speech representations (CLINIC), a dysarthria severity classification model combining acoustic and speech embeddings with Clinically Explainable Acoustic Features (CEAFs) for enhanced interpretability and performance; and (2) Generation of Explanations from Numerical features using Interpretability and patient Examples (GENIE), a module translating numerical data, such as CEAFs and their Shapley values, into intuitive natural language explanations via a large language model. In severity classification experiments on the DDK dataset, CLINIC achieved a balanced accuracy of 0.952, a 17.3% improvement over using CEAFs alone. In evaluation of the generated diagnosis, certified speech-language pathologists rated explanations from CLINIC-GENIE highly, with an average fidelity score of 4.94, confirming enhanced clinical utility through intuitive, human-like interpretations. These results demonstrate that CLINIC-GENIE enhances clinical utility by improving classification accuracy and providing intuitive, human-like explanations. The code will be made publicly available on GitHub.

Paper Type: Long

Research Area: Interpretability and Analysis of Models for NLP

Research Area Keywords: Generation, Language Modeling, NLP Applications

Contribution Types: Model analysis & interpretability

Languages Studied: korean

Submission Number: 7514