WEBVTT

1
00:00:00.360 --> 00:00:05.460
When you open an English Wikipedia article, it can feel like the
complete account of a topic.
打开英文维基百科条目时，它可能让人觉得这就是一个主题的完整介绍。

2
00:00:06.140 --> 00:00:10.060
Switching languages may seem like asking for the same account in
different words.
切换语言似乎只是让同一篇介绍换一种文字表达。

3
00:00:10.880 --> 00:00:15.420
But the editions are not simple copies. They can contain
complementary facts.
然而，不同语言版本并非简单副本，它们可能包含互补的事实。

4
00:00:16.640 --> 00:00:22.320
WikiGap asks how to make those differences visible while someone
is already reading the English article.
WikiGap 探索怎样在读者阅读英文条目时，让这些差异变得可见。

5
00:00:23.120 --> 00:00:27.200
The goal is to make other editions easier to discover as sources
of knowledge.
它的目标是让其他语言版本更容易被发现，并成为知识来源。

6
00:00:28.100 --> 00:00:35.020
The visual example here is schematic: different cards stand for
different facts, rather than real Wikipedia quotations.
这里是示意图：不同卡片代表不同事实，而非真实的维基百科引文。

7
00:00:35.400 --> 00:00:38.500
The interface places subtle markers inside the English article.
界面在英文条目中放置不显眼的标记。

8
00:00:39.180 --> 00:00:43.180
Selecting a marker reveals a related fact card in a panel beside
the text.
选择标记后，文字旁的面板会显示相关的事实卡片。

9
00:00:43.900 --> 00:00:48.120
The card presents the additional fact in English, alongside its
source-language label.
卡片用英文呈现补充事实，同时标明其来源语言。

10
00:00:48.940 --> 00:00:53.060
That keeps the new information close to the passage that
prompted the reader's curiosity.
这让新信息紧邻引发读者好奇的那段文字。

11
00:00:53.820 --> 00:00:59.300
The source link opens the original sentence in context, allowing
the reader to inspect where the fact came from.
来源链接会在上下文中打开原始句子，让读者检查事实出自哪里。

12
00:01:00.060 --> 00:01:04.300
Discovery and checking therefore become connected parts of the
same reading interaction.
这样，发现信息与核查信息就成为同一阅读交互中相互连接的部分。

13
00:01:04.827 --> 00:01:10.227
Behind the interface, WikiGap uses and extends a prior pipeline
called InfoGap.
在界面背后，WikiGap 使用并扩展了已有的 InfoGap 流水线。

14
00:01:11.047 --> 00:01:15.567
The pipeline breaks paragraphs into atomic facts before
comparing language editions.
流水线先把段落拆成独立事实，再比较不同语言版本。

15
00:01:16.527 --> 00:01:22.347
Multilingual representations retrieve likely matches, so the
verifier can examine a smaller candidate set.
多语言表示检索可能匹配的事实，让验证器检查更小的候选集合。

16
00:01:23.427 --> 00:01:27.627
The verifier then asks whether a fact is inferable from the
matched material.
验证器随后判断某个事实能否从匹配材料中推断出来。

17
00:01:28.607 --> 00:01:33.787
A fact without an aligned English counterpart becomes a
candidate knowledge gap for that English article.
没有对应英文事实的内容，会成为该英文条目的候选知识缺口。

18
00:01:34.907 --> 00:01:40.867
WikiGap turns selected gaps from French, Russian and Chinese
editions into the reader-facing cards.
WikiGap 把法语、俄语和中文版本中选出的缺口，变成面向读者的卡片。

19
00:01:41.527 --> 00:01:45.687
The interaction can start with a marker encountered during
ordinary reading.
交互可以从日常阅读中遇到的一个标记开始。

20
00:01:46.367 --> 00:01:52.767
It can also begin with a particular question. Readers can filter
by source language and search the fact cards.
它也可以从具体问题开始；读者可以按来源语言筛选，并搜索事实卡片。

21
00:01:53.507 --> 00:01:59.287
Both paths keep the source edition visible, instead of blending
all the information into an anonymous summary.
两条路径都会保留来源版本，而非把信息混成没有来源区分的摘要。

22
00:02:00.867 --> 00:02:04.907
The paper evaluates this reading design through a study with
twenty-one participants.
论文通过一项包含二十一名参与者的研究，评估这种阅读设计。

23
00:02:05.807 --> 00:02:11.327
Each participant uses both conditions for different food topics,
with treatment assignment randomized.
每位参与者针对不同食物主题使用两种条件，并随机分配实验条件。

24
00:02:12.187 --> 00:02:15.967
They answer open-book fact-finding quizzes using the supplied
articles.
他们使用所提供的条目，回答可以查阅资料的事实检索测验。

25
00:02:16.553 --> 00:02:21.053
In the reported quiz, average accuracy is seventy-three percent
with default navigation.
在报告的测验中，使用默认导航时，平均正确率为百分之七十三。

26
00:02:21.593 --> 00:02:23.793
With WikiGap, it is ninety-one percent.
使用 WikiGap 时，这一数值为百分之九十一。

27
00:02:24.713 --> 00:02:30.473
The same study reports average completion times of twenty point
eight four and twelve point three nine minutes.
同一研究报告的平均完成时间分别为二十点八四和十二点三九分钟。

28
00:02:31.553 --> 00:02:35.333
These measurements concern quizzes built around the selected
cross-language facts,
这些测量针对围绕所选跨语言事实设计的测验，

29
00:02:35.713 --> 00:02:37.453
rather than unrestricted web browsing.
而非不受限制的网页浏览。

30
00:02:38.513 --> 00:02:41.193
The detailed usability results also favor WikiGap,
详细的可用性结果也支持 WikiGap，

31
00:02:41.473 --> 00:02:45.193
with mean scores of eighty-two point five versus forty-eight
point two.
平均得分分别为八十二点五与四十八点二。

32
00:02:46.113 --> 00:02:52.013
Together, the results connect this interface design to easier
fact-finding in the study's specific setting.
这些结果将这一界面设计与研究特定设置中更方便的事实检索联系起来。

33
00:02:52.660 --> 00:02:57.880
The idea to remember is that language editions can contribute
different knowledge about the same topic.
值得记住的是：不同语言版本可以为同一个主题提供不同知识。

34
00:02:58.720 --> 00:03:04.160
WikiGap makes those differences discoverable without requiring a
reader to leave the English article first.
WikiGap 让这些差异更容易被发现，而读者无需先离开英文条目。

35
00:03:05.080 --> 00:03:11.080
Fact cards supply an English presentation, a language label, and
a path back to the original context.
事实卡片提供英文呈现、语言标签，以及返回原始上下文的路径。

36
00:03:11.700 --> 00:03:16.220
A useful interface can therefore change which knowledge becomes
visible in everyday reading.
因此，一个有用的界面能够改变日常阅读中哪些知识会被看见。