可视化BERTScore的词元匹配，帮你看懂简化文本哪里出问题。
BERTScoreVisualizer: A Web Tool for Understanding Simplified Text Evaluation with BERTScore
- 通过词元匹配可视化揭示简化文本与参考文本的差异
- 不仅报告准确率、召回率和F1，还展示逐词对应关系
- 适合评估自动文本简化系统的研究人员和开发者
BERTScore 被广泛用于评估自动文本简化系统。然而，当前实现未能充分展现该指标所能提供的全部信息，特别是词元匹配细节对于理解简化文本质量具有重要价值。为此，我们提出 BERTScoreVisualizer——一个网页工具，不仅报告精确率、召回率和F1分数，还提供词元间的匹配可视化。该工具能帮助分析简化文本在句法层级上的偏差，提升对生成结果的可解释性。代码与演示已开源于 GitHub。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。