In this paper we address the task of summarizing television shows, which
touches key areas in AI research: complex reasoning, multiple modalities, and
long narratives. We present a modular approach where separate components
perform specialized sub-tasks which we argue affords greater flexibility
compared to end-to-end methods. Our modules involve detecting scene boundaries,
reordering scenes so as to minimize the number of cuts between different
events, converting visual information to text, summarizing the dialogue in each
scene, and fusing the scene summaries into a final summary for the entire
episode. We also present a new metric, PREFS (\textbf{P}recision and
\textbf{R}ecall \textbf{E}valuation of Summary \textbf{F}act\textbf{s}), to
measure both precision and recall of generated summaries, which we decompose
into atomic facts. Tested on the recently released SummScreen3D dataset
Papalampidi and Lapata (2023), our method produces higher quality summaries
than comparison models, as measured with ROUGE and our new fact-based metric.

这篇论文主要研究了如何用模块化方法总结电视节目，涉及人工智能研究中的复杂推理、多模态和长篇叙述等关键领域。作者提出了一种模块化方法，通过不同组件执行特定子任务，可以比端到端方法更灵活地实现。论文还引入了一个新的指标 —— 基于 PREFS（Summary Facts 的精确度与召回率评估）来衡量生成摘要的精确度与召回率，并将其分解为原子事实。在最近发布的 SummScreen3D 数据集上测试，该方法比其他模型生成的摘要具有更高的质量，这是通过 ROUGE 评估和新的基于事实的指标得出的。