BERTScoreVisualizer: A Web Tool for Understanding Simplified Text Evaluation with BERTScore
This tool addresses the need for better interpretability in text simplification evaluation for researchers and developers, though it is incremental as it builds on existing BERTScore methodology.
The researchers tackled the problem of limited visibility in BERTScore evaluations for text simplification by developing BERTScoreVisualizer, a web tool that visualizes token matchings to provide clause-level insights, resulting in improved analysis capabilities without reporting specific numerical improvements.
The BERTScore metric is commonly used to evaluate automatic text simplification systems. However, current implementations of the metric fail to provide complete visibility into all information the metric can produce. Notably, the specific token matchings can be incredibly useful in generating clause-level insight into the quality of simplified text. We address this by introducing BERTScoreVisualizer, a web application that goes beyond reporting precision, recall, and F1 score and provides a visualization of the matching between tokens. We believe that our software can help improve the analysis of text simplification systems by specifically showing where generated, simplified text deviates from reference text. We host our code and demo on GitHub.