CommonsenseVIS: Visualizing and Understanding Commonsense Reasoning Capabilities of Natural Language Models
This addresses the need for interpretability in NLP by providing tools for experts to probe and edit model reasoning, though it is incremental as it builds on existing explainability techniques.
The authors tackled the problem of understanding what commonsense knowledge language models learn and whether they rely on spurious patterns, by developing CommonsenseVIS, a visual explanatory system that uses external knowledge bases to contextualize model behavior, which helped NLP experts analyze models' relational reasoning through a user study.
Recently, large pretrained language models have achieved compelling performance on commonsense benchmarks. Nevertheless, it is unclear what commonsense knowledge the models learn and whether they solely exploit spurious patterns. Feature attributions are popular explainability techniques that identify important input concepts for model outputs. However, commonsense knowledge tends to be implicit and rarely explicitly presented in inputs. These methods cannot infer models' implicit reasoning over mentioned concepts. We present CommonsenseVIS, a visual explanatory system that utilizes external commonsense knowledge bases to contextualize model behavior for commonsense question-answering. Specifically, we extract relevant commonsense knowledge in inputs as references to align model behavior with human knowledge. Our system features multi-level visualization and interactive model probing and editing for different concepts and their underlying relations. Through a user study, we show that CommonsenseVIS helps NLP experts conduct a systematic and scalable visual analysis of models' relational reasoning over concepts in different situations.