“牢记”
用英语怎么说？

Now, let's talk about QMSum. It's a query-based summarization dataset consisting of 232 meeting transcripts. It covers academic group meetings, industrial meetings, and parliament committees. Right. The annotators had to write queries about the broad contents, ensuring the relevant text spans at least 200 words. This will keep track of how models handle specific context retrieval. Then there is SQuALITY. SQuALITY is question-focused summarization over stories from Project Gutenberg. Experienced writers designed questions that require reading significant parts of the story. We should bear in mind that these questions and summaries are crowdsourced and original. It is not just a simple copy-paste task; it requires deep understanding. Definitely. Models must comprehend the narrative flow to make sense of the story aspects based on the guiding questions.

现在，我们来谈谈 QMSum。它是一个基于查询的摘要数据集，由 232 个会议记录组成。它涵盖了学术小组会议、工业会议和议会委员会。对。标注者必须针对宽泛的内容撰写查询，确保相关文本跨度至少为 200 字。这将追踪模型如何处理特定的上下文检索。然后是 SQuALITY。SQuALITY 是针对古登堡计划故事的以问题为核心的摘要。经验丰富的作者设计了需要阅读故事很大一部分才能正确回答的问题。我们应该牢记，这些问题和摘要都是众包且原创的。这不仅仅是一个简单的复制粘贴任务；它需要深入的理解。确实如此。模型必须理解叙事流程，才能根据引导性问题理解故事的各个方面。