Enhancing the Comprehensibility of Text Explanations via Unsupervised Concept Discovery

2025/05/26 by Yifan Sun, Danding Wang, Sun, Yifan +7
Computer Science · #Advanced Text Analysis Techniques #Computation and Language (cs.CL) #FOS: Computer and information sciences #Topic Modeling #Web Data Mining and Analysis

paper · pdf · doi:10.48550/arxiv.2505.20293

openalex publication_date 2025/05/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Concept-based explainable approaches have emerged as a promising method in explainable AI because they can interpret models in a way that aligns with human reasoning. However, their adaption in the text domain remains limited. Most existing methods rely on predefined concept annotations and cannot discover unseen concepts, while other methods that extract concepts without supervision often produce explanations that are not intuitively comprehensible to humans, potentially diminishing user trust. These methods fall short of discovering comprehensible concepts automatically. To address this issue, we propose ECO-Concept, an intrinsically interpretable framework to discover comprehensible concepts with no concept annotations. ECO-Concept first utilizes an object-centric architecture to extract semantic concepts automatically. Then the comprehensibility of the extracted concepts is evaluated by large language models. Finally, the evaluation result guides the subsequent model fine-tuning to obtain more understandable explanations. Experiments show that our method achieves superior performance across diverse tasks. Further concept evaluations validate that the concepts learned by ECO-Concept surpassed current counterparts in comprehensibility.

Citations

• On Evaluating Explanation Utility for Human-AI Decision Making in NLP
• Sparsity-Guided Holistic Explanation for LLMs with Interpretable Inference-Time Intervention
• Concept-based Explainable Artificial Intelligence: A Survey
• Interpreting Pretrained Language Models via Concept Bottlenecks
• Concept-Centric Transformers: Enhancing Model Interpretability through Object-Centric Concept Learning within a Shared Global Workspace
• COCKATIEL: COntinuous Concept ranKed ATtribution with Interpretable ELements for explaining neural net classifiers on NLP tasks
• Learning Bottleneck Concepts in Image Classification
• Language in a Bottle: Language Model Guided Concept Bottlenecks for Interpretable Image Classification
• Causal Proxy Models for Concept-Based Model Explanations
• Concept Activation Regions: A Generalized Framework For Concept-Based Explanations
• From attribution maps to human-understandable explanations through Concept Relevance Propagation
• CEBaB: Estimating the Causal Effects of Real-World Concepts on NLP Model Behavior
• Discovering Latent Concepts Learned in BERT
• ProtoTEx: Explaining Model Decisions with Prototype Tensors
• Rationalization through Concepts
• SelfExplain: A Self-Explaining Architecture for Neural Text Classifiers
• Concept Bottleneck Models
• Invertible Concept-based Explanations for CNN Models with Non-negative Concept Activation Vectors
• Object-Centric Learning with Slot Attention
• On Completeness-aware Concept-Based Explanations in Deep Neural Networks
• RoBERTa: A Robustly Optimized BERT Pretraining Approach
• Explaining Classifiers with Causal Concept Effect (CaCE)
• Structural Scaffolds for Citation Intent Classification in Scientific\n Publications
• Deriving Machine Attention from Human Rationales
• Towards Robust Interpretability with Self-Explaining Neural Networks
• Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)
• Deep Learning for Case-Based Reasoning through Prototypes: A Neural Network that Explains Its Predictions
• Learning Attitudes and Attributes from Multi-Aspect Reviews
• A NEW MEASURE OF RANK CORRELATION
• Interpretable-by-Design Text Understanding with Iteratively Generated Concept Bottleneck
• BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Related

• SurroCBM: Concept Bottleneck Surrogate Models for Generative Post-hoc Explanation
• Towards Automatic Concept-based Explanations
• Automatic Concept Embedding Model (ACEM): No train-time concepts, No issue!
• Overlooked factors in concept-based explanations: Dataset choice, concept learnability, and human capability
• Towards Human-Understandable Multi-Dimensional Concept Discovery