From 4dfa82378566009898ae96077f51767d423159c6 Mon Sep 17 00:00:00 2001
From: Yuan <yuant614@gmail.com>
Date: Mon, 12 Dec 2022 18:11:09 +0800
Subject: [PATCH] docs(zh-cn): Reviewed 57_what-is-perplexity.srt (#401)

---
 subtitles/zh-CN/57_what-is-perplexity.srt | 68 +++++++++++------------
 1 file changed, 34 insertions(+), 34 deletions(-)

diff --git a/subtitles/zh-CN/57_what-is-perplexity.srt b/subtitles/zh-CN/57_what-is-perplexity.srt
index 8c1141ddd..b78d4fe1f 100644
--- a/subtitles/zh-CN/57_what-is-perplexity.srt
+++ b/subtitles/zh-CN/57_what-is-perplexity.srt
@@ -15,22 +15,22 @@
 
 4
 00:00:05,379 --> 00:00:06,720
-- 在这段视频中，我们来看看
+- 在这段视频中，我们来了解一下
 - In this video, we take a look
 
 5
 00:00:06,720 --> 00:00:09,483
-在称为困惑度的神秘测深指标上。
+在称为困惑度的评估指标上。
 at the mysterious sounding metric called perplexity.
 
 6
 00:00:11,070 --> 00:00:12,630
-你可能遇到过困惑
+你在研究生成模型时
 You might have encountered perplexity
 
 7
 00:00:12,630 --> 00:00:14,970
-在阅读生成模型时。
+可能遇到过困惑度。
 when reading about generative models.
 
 8
@@ -40,7 +40,7 @@ You can see two examples here,
 
 9
 00:00:16,680 --> 00:00:18,577
-一张来自原始变压器纸，
+一个来自最初的 transformer 论文，
 one from the original transformer paper,
 
 10
@@ -50,17 +50,17 @@ one from the original transformer paper,
 
 11
 00:00:19,950 --> 00:00:23,340
-另一篇来自最近的 GPT-2 论文。
+另一个来自最近的 GPT-2 论文。
 and the other one from the more recent GPT-2 paper.
 
 12
 00:00:23,340 --> 00:00:25,740
-困惑度是衡量绩效的常用指标
+困惑度是衡量语言模型的性能
 Perplexity is a common metric to measure the performance
 
 13
 00:00:25,740 --> 00:00:27,150
-的语言模型。
+的常用指标。
 of language models.
 
 14
@@ -70,7 +70,7 @@ The smaller its value, the better the performance.
 
 15
 00:00:30,000 --> 00:00:32,950
-但它究竟意味着什么，我们又该如何计算呢？
+但它究竟意味着什么，我们又该如何计算得到它呢？
 But what does it actually mean and how can we calculate it?
 
 16
@@ -80,32 +80,32 @@ A very common quantity in machine learning
 
 17
 00:00:36,180 --> 00:00:37,650
-是可能性。
+是相似度。
 is the likelihood.
 
 18
 00:00:37,650 --> 00:00:39,240
-我们可以计算可能性
+我们可以计算似然性
 We can calculate the likelihood
 
 19
 00:00:39,240 --> 00:00:42,390
-作为每个标记概率的乘积。
+作为每个词元的概率的乘积。
 as the product of each token's probability.
 
 20
 00:00:42,390 --> 00:00:44,730
-这意味着对于每个令牌，
+这意味着对于每个词元，
 What this means is that for each token,
 
 21
 00:00:44,730 --> 00:00:47,340
-我们使用语言模型来预测它的概率
+我们使用语言模型基于之前的词元
 we use the language model to predict its probability
 
 22
 00:00:47,340 --> 00:00:49,560
-基于之前的标记。
+来预测它的概率。
 based on the previous tokens.
 
 23
@@ -115,12 +115,12 @@ In the end, we multiply all probabilities
 
 24
 00:00:52,050 --> 00:00:53,253
-得到的可能性。
+从而得到似然性。
 to get the likelihood.
 
 25
 00:00:55,892 --> 00:00:57,000
-有可能，
+通过似然性，
 With the likelihood,
 
 26
@@ -135,27 +135,27 @@ the cross-entropy.
 
 28
 00:01:01,200 --> 00:01:03,450
-你可能已经听说过交叉熵
+当你接触损失函数时
 You might have already heard about cross-entropy
 
 29
 00:01:03,450 --> 00:01:05,670
-在查看损失函数时。
+可能已经听说过交叉熵。
 when looking at loss function.
 
 30
 00:01:05,670 --> 00:01:09,210
-它通常用作分类中的损失函数。
+它通常在分类中作为损失函数使用。
 It is often used as a loss function in classification.
 
 31
 00:01:09,210 --> 00:01:11,610
-在语言建模中，我们预测下一个标记
+在语言建模中，我们基于之前的词元
 In language modeling, we predict the next token
 
 32
 00:01:11,610 --> 00:01:12,930
-基于之前的令牌，
+预测下一个词元，
 based on the previous token,
 
 33
@@ -185,32 +185,32 @@ with its inputs as labels.
 
 38
 00:01:23,580 --> 00:01:26,433
-然后损失对应于交叉熵。
+其损失与交叉熵相关。
 The loss then corresponds to the cross-entropy.
 
 39
 00:01:29,130 --> 00:01:31,110
-我们现在只差一个手术了
+现在对于计算困惑度
 We are now only a single operation away
 
 40
 00:01:31,110 --> 00:01:33,510
-从计算困惑度。
+我们现在只差一个操作了。
 from calculating the perplexity.
 
 41
 00:01:33,510 --> 00:01:37,710
-通过对交叉熵取幂，我们得到了困惑。
+通过对交叉熵取幂，我们得到了困惑度。
 By exponentiating the cross-entropy, we get the perplexity.
 
 42
 00:01:37,710 --> 00:01:40,260
-所以你看到困惑是密切相关的
+所以你可以发现困惑度和损失
 So you see that the perplexity is closely related
 
 43
 00:01:40,260 --> 00:01:41,163
-到损失。
+是密切相关的。
 to the loss.
 
 44
@@ -220,27 +220,27 @@ Plugging in previous results
 
 45
 00:01:43,380 --> 00:01:47,010
-表明这相当于求幂
+表明这相当于对每个词元
 shows that this is equivalent to exponentiating
 
 46
 00:01:47,010 --> 00:01:51,033
-每个令牌的负平均锁定概率。
+的负平均锁定概率取幂值。
 the negative average lock probability of each token.
 
 47
 00:01:52,050 --> 00:01:54,630
-请记住，损失只是一个弱代理
+请记住，损失只是针对模型
 Keep in mind that the loss is only a weak proxy
 
 48
 00:01:54,630 --> 00:01:57,360
-用于模型生成高质量文本的能力
+生成高质量文本的能力的一个弱代理
 for a model's ability to generate quality text
 
 49
 00:01:57,360 --> 00:02:00,510
-困惑也是如此。
+困惑度也是如此。
 and the same is true for perplexity.
 
 50
@@ -255,7 +255,7 @@ more sophisticated metrics
 
 52
 00:02:03,840 --> 00:02:07,413
-例如生成任务上的 BLEU 或 ROUGE。
+例如生成式任务上的 BLEU 或 ROUGE。
 such as BLEU or ROUGE on generative tasks.
 
 53