From 6096f1a6b9da1104084ce89a2389a83a746c7023 Mon Sep 17 00:00:00 2001
From: "Mr.Gung" <249053957@qq.com>
Date: Wed, 13 Feb 2019 17:07:06 +0800
Subject: [PATCH] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E5=9B=BE=E7=89=87=E5=95=8A?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

还缺一张图片和文字第二次调整
---
 .idea/workspace.xml                           | 101 ++++++++++++------
 ...d Embeddings_Encoding Lexical Semantics.md |  59 +++++-----
 ...dings-0d61f8370cad1d412f80b84d143e1257.gif | Bin 0 -> 145 bytes
 ...dings-1ed346930917426bc46d41e22cc525ec.gif | Bin 0 -> 145 bytes
 ...dings-2b13c62ff7b8f4c26887bf369a4e44c8.gif | Bin 0 -> 301 bytes
 ...dings-3c3d846fb2913b4605e8d59bc5a14e6c.gif | Bin 0 -> 764 bytes
 ...dings-3f53ca9954dfc8730d4f58c937097118.gif | Bin 0 -> 368 bytes
 ...dings-3ff042e26bf4d74a5488884f76800e6c.gif | Bin 0 -> 401 bytes
 ...dings-4856f4335c4b3d6613590c6d264e3541.gif | Bin 0 -> 168 bytes
 ...dings-5206560a306a2e085a437fd258eb57ce.gif | Bin 0 -> 142 bytes
 ...dings-56cc30cd24196bea5f078cb9b1f1d0dc.gif | Bin 0 -> 1700 bytes
 ...dings-62ca51f900bf27324a2be4e6b8609f4b.gif | Bin 0 -> 2263 bytes
 ...dings-74fcb594bdd93c0f956682ae1ea013e6.gif | Bin 0 -> 178 bytes
 ...dings-8b4e6bfa073defa91d3f23cdec8f1f0e.gif | Bin 0 -> 1859 bytes
 ...dings-8d9c307cb7f3c4a32822a51922d1ceaa.gif | Bin 0 -> 155 bytes
 ...dings-a0be0498fd8216177330deffbfcb6ea2.gif | Bin 0 -> 2161 bytes
 ...dings-aa38f107289d4d73d516190581397349.gif | Bin 0 -> 161 bytes
 ...dings-cf775cf1814914c00f5bf7ada7de4369.gif | Bin 0 -> 949 bytes
 ...dings-f1290186a5d0b1ceab27f4e77c0c5d68.gif | Bin 0 -> 138 bytes
 ...dings-f623e75af30e62bbd73d6df5b50bb7b5.gif | Bin 0 -> 149 bytes
 20 files changed, 94 insertions(+), 66 deletions(-)
 create mode 100644 docs/1.0/img/WordEmbeddings-0d61f8370cad1d412f80b84d143e1257.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-1ed346930917426bc46d41e22cc525ec.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-2b13c62ff7b8f4c26887bf369a4e44c8.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-3c3d846fb2913b4605e8d59bc5a14e6c.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-3f53ca9954dfc8730d4f58c937097118.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-3ff042e26bf4d74a5488884f76800e6c.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-4856f4335c4b3d6613590c6d264e3541.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-5206560a306a2e085a437fd258eb57ce.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-56cc30cd24196bea5f078cb9b1f1d0dc.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-62ca51f900bf27324a2be4e6b8609f4b.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-74fcb594bdd93c0f956682ae1ea013e6.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-8b4e6bfa073defa91d3f23cdec8f1f0e.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-8d9c307cb7f3c4a32822a51922d1ceaa.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-a0be0498fd8216177330deffbfcb6ea2.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-aa38f107289d4d73d516190581397349.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-cf775cf1814914c00f5bf7ada7de4369.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif
 create mode 100644 docs/1.0/img/WordEmbeddings-f623e75af30e62bbd73d6df5b50bb7b5.gif

diff --git a/.idea/workspace.xml b/.idea/workspace.xml
index 59fb91d3..998c4d93 100644
--- a/.idea/workspace.xml
+++ b/.idea/workspace.xml
@@ -11,7 +11,7 @@
   </component>
   <component name="FileEditorManager">
     <leaf SIDE_TABS_SIZE_LIMIT_KEY="300">
-      <file leaf-file-name="test.py" pinned="false" current-in-tab="true">
+      <file leaf-file-name="test.py" pinned="false" current-in-tab="false">
         <entry file="file://E:/常用文件/OCR之word转换xml或json/工作文件/代码/test.py">
           <provider selected="true" editor-type-id="text-editor">
             <state relative-caret-position="19">
@@ -23,12 +23,24 @@
           </provider>
         </entry>
       </file>
-      <file leaf-file-name="Word Embeddings_Encoding Lexical Semantics.md" pinned="false" current-in-tab="false">
+      <file leaf-file-name="Word Embeddings_Encoding Lexical Semantics.md" pinned="false" current-in-tab="true">
         <entry file="file://$PROJECT_DIR$/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md">
           <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
             <state split_layout="SPLIT">
-              <first_editor relative-caret-position="285">
-                <caret line="126" column="11" selection-start-line="126" selection-start-column="5" selection-end-line="126" selection-end-column="11" />
+              <first_editor relative-caret-position="152">
+                <caret line="8" column="10" lean-forward="true" selection-start-line="8" selection-start-column="10" selection-end-line="8" selection-end-column="24" />
+              </first_editor>
+              <second_editor />
+            </state>
+          </provider>
+        </entry>
+      </file>
+      <file leaf-file-name="blitz_autograd_tutorial.md" pinned="false" current-in-tab="false">
+        <entry file="file://$PROJECT_DIR$/docs/1.0/blitz_autograd_tutorial.md">
+          <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
+            <state split_layout="SPLIT">
+              <first_editor relative-caret-position="2565">
+                <caret line="177" column="20" lean-forward="true" selection-start-line="177" selection-start-column="20" selection-end-line="177" selection-end-column="20" />
               </first_editor>
               <second_editor />
             </state>
@@ -44,6 +56,11 @@
       </list>
     </option>
   </component>
+  <component name="FindInProjectRecents">
+    <findStrings>
+      <find>根据链式法则，雅可比向量积应</find>
+    </findStrings>
+  </component>
   <component name="Git.Settings">
     <option name="RECENT_GIT_ROOT_PATH" value="$PROJECT_DIR$" />
   </component>
@@ -53,11 +70,12 @@
         <option value="E:/常用文件/OCR之word转换xml或json/工作文件/代码/aaa.py" />
         <option value="E:/常用文件/OCR之word转换xml或json/工作文件/代码/error.txt" />
         <option value="E:/常用文件/OCR之word转换xml或json/工作文件/代码/test.py" />
+        <option value="$PROJECT_DIR$/docs/1.0/blitz_autograd_tutorial.md" />
         <option value="$PROJECT_DIR$/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md" />
       </list>
     </option>
   </component>
-  <component name="ProjectFrameBounds" extendedState="7">
+  <component name="ProjectFrameBounds" extendedState="6">
     <option name="x" value="281" />
     <option name="y" value="-11" />
     <option name="width" value="853" />
@@ -68,6 +86,7 @@
       <foldersAlwaysOnTop value="true" />
     </navigator>
     <panes>
+      <pane id="Scope" />
       <pane id="ProjectPane">
         <subPane>
           <expand>
@@ -93,15 +112,10 @@
               <item name="1.0" type="462c0819:PsiDirectoryNode" />
               <item name="img" type="462c0819:PsiDirectoryNode" />
             </path>
-            <path>
-              <item name="pytorch-doc-zh" type="b2602c69:ProjectViewProjectNode" />
-              <item name="代码" type="462c0819:PsiDirectoryNode" />
-            </path>
           </expand>
           <select />
         </subPane>
       </pane>
-      <pane id="Scope" />
     </panes>
   </component>
   <component name="PropertiesComponent">
@@ -165,10 +179,10 @@
     <servers />
   </component>
   <component name="ToolWindowManager">
-    <frame x="-8" y="-8" width="1696" height="1026" extended-state="7" />
+    <frame x="-8" y="-8" width="1696" height="1026" extended-state="6" />
     <editor active="true" />
     <layout>
-      <window_info content_ui="combo" id="Project" order="0" weight="0.22916667" />
+      <window_info active="true" content_ui="combo" id="Project" order="0" visible="true" weight="0.2029762" />
       <window_info anchor="bottom" id="TODO" order="6" />
       <window_info anchor="bottom" id="Event Log" order="7" side_tool="true" />
       <window_info anchor="bottom" id="Version Control" order="7" />
@@ -196,6 +210,26 @@
     </breakpoint-manager>
   </component>
   <component name="editorHistoryManager">
+    <entry file="file://E:/常用文件/OCR之word转换xml或json/工作文件/代码/test.py">
+      <provider selected="true" editor-type-id="text-editor">
+        <state relative-caret-position="1767">
+          <caret line="93" column="20" selection-start-line="93" selection-start-column="20" selection-end-line="93" selection-end-column="20" />
+          <folding>
+            <element signature="e#0#25#0" expanded="true" />
+          </folding>
+        </state>
+      </provider>
+    </entry>
+    <entry file="file://$PROJECT_DIR$/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md">
+      <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
+        <state split_layout="SPLIT">
+          <first_editor relative-caret-position="3952">
+            <caret line="208" lean-forward="true" selection-start-line="208" selection-end-line="208" />
+          </first_editor>
+          <second_editor />
+        </state>
+      </provider>
+    </entry>
     <entry file="file://E:/常用文件/OCR之word转换xml或json/工作文件/代码/test.py">
       <provider selected="true" editor-type-id="text-editor">
         <state relative-caret-position="1767">
@@ -346,16 +380,6 @@
         </state>
       </provider>
     </entry>
-    <entry file="file://$PROJECT_DIR$/docs/1.0/blitz_autograd_tutorial.md">
-      <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
-        <state split_layout="SPLIT">
-          <first_editor relative-caret-position="19">
-            <caret line="1" lean-forward="true" selection-start-line="1" selection-end-line="1" />
-          </first_editor>
-          <second_editor />
-        </state>
-      </provider>
-    </entry>
     <entry file="file://$PROJECT_DIR$/docs/1.0/aws_distributed_training_tutorial.md">
       <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
         <state split_layout="SPLIT">
@@ -418,16 +442,6 @@
     <entry file="file://E:/常用文件/OCR之word转换xml或json/工作文件/代码/error.txt">
       <provider selected="true" editor-type-id="text-editor" />
     </entry>
-    <entry file="file://$PROJECT_DIR$/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md">
-      <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
-        <state split_layout="SPLIT">
-          <first_editor relative-caret-position="285">
-            <caret line="126" column="11" selection-start-line="126" selection-start-column="5" selection-end-line="126" selection-end-column="11" />
-          </first_editor>
-          <second_editor />
-        </state>
-      </provider>
-    </entry>
     <entry file="file://E:/常用文件/OCR之word转换xml或json/工作文件/代码/test.py">
       <provider selected="true" editor-type-id="text-editor">
         <state relative-caret-position="19">
@@ -438,5 +452,28 @@
         </state>
       </provider>
     </entry>
+    <entry file="file://$PROJECT_DIR$/docs/1.0/img/WordEmbeddings-74fcb594bdd93c0f956682ae1ea013e6.gif">
+      <provider selected="true" editor-type-id="images" />
+    </entry>
+    <entry file="file://$PROJECT_DIR$/docs/1.0/blitz_autograd_tutorial.md">
+      <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
+        <state split_layout="SPLIT">
+          <first_editor relative-caret-position="2565">
+            <caret line="177" column="20" lean-forward="true" selection-start-line="177" selection-start-column="20" selection-end-line="177" selection-end-column="20" />
+          </first_editor>
+          <second_editor />
+        </state>
+      </provider>
+    </entry>
+    <entry file="file://$PROJECT_DIR$/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md">
+      <provider selected="true" editor-type-id="split-provider[text-editor;markdown-preview-editor]">
+        <state split_layout="SPLIT">
+          <first_editor relative-caret-position="152">
+            <caret line="8" column="10" lean-forward="true" selection-start-line="8" selection-start-column="10" selection-end-line="8" selection-end-column="24" />
+          </first_editor>
+          <second_editor />
+        </state>
+      </provider>
+    </entry>
   </component>
 </project>
\ No newline at end of file
diff --git a/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md b/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md
index 475dc654..05ec0835 100644
--- a/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md	
+++ b/docs/1.0/Word Embeddings_Encoding Lexical Semantics.md	
@@ -1,12 +1,14 @@
 # 词嵌入：编码形式的词汇语义
 
-词嵌入是一种由真实数字组成的稠密向量，每个向量都代表了单词表里的一个单词. 在自然语言处理中，总会遇到这样的情况：特征全是单词！ 但是，如何在电脑上表述一个单词呢？你在电脑上存储的单词的ascii码，但是它仅仅代表单词怎么拼写，没有说明单词的内在含义(你也许能够从词缀中了解它的词性，或者从大小写中得到一些属性，但仅此而已). 更重要的是，你能把这些ascii码字符组合成什么含义？? 当V代表词汇表、输入数据是|V|维的情况下，我们往往想从神经网络中得到数据密集的结果，但是结果只有很少的几个维度（例如，预测的数据只有几个标签时）。我们如何从大的数据维度空间中得到稍小一点的维度空间？
+> 译者：[巩子惠](https://github.com/sight007)
 
-放弃使用ascii码字符的形式表示单词，换用one-hot encoding会怎么样了？好吧，W这个单词就能这样表示：
+词嵌入是一种由真实数字组成的稠密向量，每个向量都代表了单词表里的一个单词。 在自然语言处理中，总会遇到这样的情况：特征全是单词！但是，如何在电脑上表述一个单词呢？你在电脑上存储的单词的`ascii`码，但是它仅仅代表单词怎么拼写，没有说明单词的内在含义(你也许能够从词缀中了解它的词性，或者从大小写中得到一些属性，但仅此而已)。 更重要的是，你能把这些`ascii`码字符组合成什么含义？当![](./img/WordEmbeddings-5206560a306a2e085a437fd258eb57ce.gif)代表词汇表、输入数据是![](./img/WordEmbeddings-74fcb594bdd93c0f956682ae1ea013e6.gif)维的情况下，我们往往想从神经网络中得到数据密集的结果，但是结果只有很少的几个维度（例如，预测的数据只有几个标签时）。我们如何从大的数据维度空间中得到稍小一点的维度空间？
 
-cf775cf1814914c00f5bf7ada7de4369.gif
+放弃使用`ascii`码字符的形式表示单词，换用`one-hot encoding`会怎么样了？好吧，![](./img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif)这个单词就能这样表示：
 
-其中，1就是表示w的独有位置，其他位置全是0。其他的词都类似，在另外不一样的位置有一个1代表它，其他位置也都是0。
+![](./img/WordEmbeddings-cf775cf1814914c00f5bf7ada7de4369.gif)
+
+其中，1 表示![](./img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif)的独有位置，其他位置全是0。其他的词都类似，在另外不一样的位置有一个1代表它，其他位置也都是0。
 这种表达除了占用巨大的空间外，还有个很大的缺陷。 它只是简单的把词看做一个单独个体，认为它们之间毫无联系。 我们真正想要的是能够表达单词之间一些相似的含义。为什么要这样做呢？来看下面的例子：
 
 假如我们正在搭建一个语言模型，训练数据有下面一些句子：
@@ -32,29 +34,21 @@ cf775cf1814914c00f5bf7ada7de4369.gif
 
 如果每个属性都表示一个维度，那我们也许可以用一个向量表示一个单词，就像这样：
 
-```py
-\[ q_\text{mathematician} = \left[ \overbrace{2.3}^\text{can run}, \overbrace{9.4}^\text{likes coffee}, \overbrace{-5.5}^\text{majored in Physics}, \dots \right]\]
-```
+![](img/WordEmbeddings-62ca51f900bf27324a2be4e6b8609f4b.gif)
 
-```py
-\[ q_\text{physicist} = \left[ \overbrace{2.5}^\text{can run}, \overbrace{9.1}^\text{likes coffee}, \overbrace{6.4}^\text{majored in Physics}, \dots \right]\]
-```
+![](./img/WordEmbeddings-a0be0498fd8216177330deffbfcb6ea2.gif)
 
 那么，我们就这可以通过下面的方法得到这些单词之间的相似性：
 
-```py
-\[\text{Similarity}(\text{physicist}, \text{mathematician}) = q_\text{physicist} \cdot q_\text{mathematician}\]
-```
+![](./img/WordEmbeddings-8b4e6bfa073defa91d3f23cdec8f1f0e.gif)
 
 尽管通常情况下需要进行长度归一化：
 
-```py
-\[ \text{Similarity}(\text{physicist}, \text{mathematician}) = \frac{q_\text{physicist} \cdot q_\text{mathematician}} {\| q_\text{\physicist} \| \| q_\text{mathematician} \|} = \cos (\phi)\]
-```
+![](a./img/WordEmbeddings-8b4e6bfa073defa91d3f23cdec8f1f0e.gif)
 
-Φ是两个向量的夹角。 这就意味着，完全相似的单词相似度为1。完全不相似的单词相似度为-1。
+![](./img/WordEmbeddings-1ed346930917426bc46d41e22cc525ec.gif)是两个向量的夹角。 这就意味着，完全相似的单词相似度为1。完全不相似的单词相似度为-1。
 
-你可以把本章开头介绍的one-hot稀疏向量看做是我们新定义向量的一种特殊形式，那里的单词相似度为0， 现在我们给每个单词一些独特的语义属性。 这些向量数据密集，也就是说它们数字通常都非零。
+你可以把本章开头介绍的`one-hot`稀疏向量看做是我们新定义向量的一种特殊形式，那里的单词相似度为0， 现在我们给每个单词一些独特的语义属性。 这些向量数据密集，也就是说它们数字通常都非零。
 
 但是新的这些向量存在一个严重的问题： 你可以想到数千种不同的语义属性，它们可能都与决定相似性有关，而且，到底如何设置不同属性的值呢？深度学习的中心思想是用神经网络来学习特征的表示，而不是程序员去设计它们。 所以为什么不把词嵌入只当做模型参数，而是通过训练来更新呢？ 这就才是我们要确切做的事。我们将用神经网络做一些潜在语义属性，但是原则上，学习才是关键。 注意，词嵌入可能无法解释。就是说，尽管使用我们上面手动制作的向量，能够发现数学家和物理学家都喜欢喝咖啡的相似性， 如果我们允许神经网络来学习词嵌入，那么就会发现数学家和物理学家在第二维度有个较大的值，它所代表的含义很不清晰。 它们在一些潜在语义上是相似的，但是对我们来说无法解释。
 
@@ -62,11 +56,11 @@ cf775cf1814914c00f5bf7ada7de4369.gif
 
 ## Pytorch中的词嵌入
 
-在我们举例或练习之前，这里有一份关于如何在Pytorch和常见的深度学习中使用词嵌入的简要介绍。 与制作one-hot向量时对每个单词定义一个特殊的索引类似，当我们使用词向量时也需要为每个单词定义一个索引。这些索引将是查询表的关键点。意思就是，词嵌入被被存储在一个|V|*D的向量中，其中D是词嵌入的维度。词被被分配的索引i，表示在向量的第i行存储它的嵌入。在所有的代码中，从单词到索引的映射是一个叫'word_to_ix'的字典。
+在我们举例或练习之前，这里有一份关于如何在`Pytorch`和常见的深度学习中使用词嵌入的简要介绍。 与制作`one-hot`向量时对每个单词定义一个特殊的索引类似，当我们使用词向量时也需要为每个单词定义一个索引。这些索引将是查询表的关键点。意思就是，词嵌入被被存储在一个![](./img/WordEmbeddings-2b13c62ff7b8f4c26887bf369a4e44c8.gif)的向量中，其中![](./img/WordEmbeddings-f623e75af30e62bbd73d6df5b50bb7b5.gif)是词嵌入的维度。词被被分配的索引`i`，表示在向量的第`i`行存储它的嵌入。在所有的代码中，从单词到索引的映射是一个叫`word_to_ix`的字典。
 
-能使用词嵌入的模块是torch.nn.Embedding，这里面有两个参数：词汇表的大小和词嵌入的维度。
+能使用词嵌入的模块是`torch.nn.Embedding`，这里面有两个参数：词汇表的大小和词嵌入的维度。
 
-索引这张表时，你必须使用torch.LongTensor（因为索引是整数，不是浮点数）。
+索引这张表时，你必须使用`torch.LongTensor`（因为索引是整数，不是浮点数）。
 
 ```py
 # 作者: Robert Guthrie
@@ -99,12 +93,11 @@ tensor([[ 0.6614,  0.2669,  0.0617,  0.6213, -0.4519]],
 
 ## 例子： N-Gram语言模型
 
-回想一下，在n-gram语言模型中,给定一个单词序列向量w，我们要计算的是
-```py
-\[P(w_i | w_{i-1}, w_{i-2}, \dots, w_{i-n+1} )\]
-```
+回想一下，在`n-gram`语言模型中,给定一个单词序列向量![](./img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif)，我们要计算的是
 
-wi是单词序列的第i个单词。
+![](./img/WordEmbeddings-3c3d846fb2913b4605e8d59bc5a14e6c.gif)
+
+![](./img/WordEmbeddings-aa38f107289d4d73d516190581397349.gif)是单词序列的第`i`个单词。
 在本例中，我们将在训练样例上计算损失函数，并且用反向传播算法更新参数。
 ```py
 CONTEXT_SIZE = 2
@@ -192,20 +185,18 @@ print(losses)  # 用训练数据每次迭代，损失函数都会下降。
 
 ## 练习：计算连续词袋模型的词向量
 
-连续词袋模型（CBOW）在NLP深度学习中使用很频繁。它是一个模型，尝试通过目标词前后几个单词的文本，来预测目标词。这有别于语言模型，因为CBOW不是序列的，也不必是概率性的。 CBOW常用于快速地训练词向量，得到的嵌入用来初始化一些复杂模型的嵌入。通常情况下，这被称为预训练嵌入。 它几乎总能帮忙把模型性能提升几个百分点。
+连续词袋模型（`CBOW`）在`NLP`深度学习中使用很频繁。它是一个模型，尝试通过目标词前后几个单词的文本，来预测目标词。这有别于语言模型，因为`CBOW`不是序列的，也不必是概率性的。 `CBOW`常用于快速地训练词向量，得到的嵌入用来初始化一些复杂模型的嵌入。通常情况下，这被称为`预训练嵌入`。 它几乎总能帮忙把模型性能提升几个百分点。
 
-CBOW模型如下所示： 给定一个单词Wi，N代表两边的滑窗距，如Wi-1, .... ,Wi-N 和 Wi+1, ..... Wi+N，并将所有的上下文词统称为C，CBOW试图最小化
+`CBOW`模型如下所示： 给定一个单词![](./img/WordEmbeddings-aa38f107289d4d73d516190581397349.gif) ，![](./img/WordEmbeddings-8d9c307cb7f3c4a32822a51922d1ceaa.gif)代表两边的滑窗距，如![](./img/WordEmbeddings-3f53ca9954dfc8730d4f58c937097118.gif)和![](./img/WordEmbeddings-3ff042e26bf4d74a5488884f76800e6c.gif)，并将所有的上下文词统称为![](./img/WordEmbeddings-0d61f8370cad1d412f80b84d143e1257.gif) ，`CBOW`试图最小化
 
-```py
-\[-\log p(w_i | C) = -\log \text{Softmax}(A(\sum_{w \in C} q_w) + b)\]
-```
+![](./img/WordEmbeddings-56cc30cd24196bea5f078cb9b1f1d0dc.gif)
 
-其中q_w是单词W的嵌入。
+其中![](./img/WordEmbeddings-4856f4335c4b3d6613590c6d264e3541.gif)是单词![](./img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif)的嵌入。
 
-在Pytorch中，通过填充下面的类来实现这个模型，有两条需要注意：
+在`Pytorch`中，通过填充下面的类来实现这个模型，有两条需要注意：
 
 *   考虑下你需要定义哪些参数。
-*   确保你知道每步操作后的结构，如果想重构，请使用.view()
+*   确保你知道每步操作后的结构，如果想重构，请使用`.view()`
 
 ```py
 CONTEXT_SIZE = 2  # 左右各两个词
diff --git a/docs/1.0/img/WordEmbeddings-0d61f8370cad1d412f80b84d143e1257.gif b/docs/1.0/img/WordEmbeddings-0d61f8370cad1d412f80b84d143e1257.gif
new file mode 100644
index 0000000000000000000000000000000000000000..4b5ffe596011e0f313548d832edb157ae57eb27a
GIT binary patch
literal 145
zcmZ?wbhEHb<YVAr*v!E2|Nnmm28MU<-j$V=g@lBprKNRrblkafM@dQP%$YO0cI|R;
zaS;#@ShZ>uP#F*?{$ybU(K;XkWCjC^oq&fXo93~G#m^m1vecZ~qO)iUOYlLCl`TwO
s7mi%w)!~`P#xh^yLZ_@+&ms@ONg`=VjY16pYRiO}6DMT)DKS_B0DO!v$N&HU

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-1ed346930917426bc46d41e22cc525ec.gif b/docs/1.0/img/WordEmbeddings-1ed346930917426bc46d41e22cc525ec.gif
new file mode 100644
index 0000000000000000000000000000000000000000..909611705ede5e2309b9a411d8c9f22893fa3919
GIT binary patch
literal 145
zcmZ?wbhEHb<Yo|H*v!E2|Nnmm28LCuR?V0(<IbHsXU?2?_wHR;Sy@L%$F5zwLPA1Z
zTwIitl#Gmw($dm^DiwdSFajwC9S{LBgMr0P-~=bj-h>C9r>0#?lW1=UXbEHNRFdqT
uA`zsZyO4{;M1Lix+gs7B77;N<-@;ikY<-MP8!qNL&1U)9ywrt(!5RQnFfxb$

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-2b13c62ff7b8f4c26887bf369a4e44c8.gif b/docs/1.0/img/WordEmbeddings-2b13c62ff7b8f4c26887bf369a4e44c8.gif
new file mode 100644
index 0000000000000000000000000000000000000000..96af6340453bacc138140d3aec2cd45790805404
GIT binary patch
literal 301
zcmV+|0n+|QNk%w1VLSj60J8u9|Ns90006qWx<o`oR8&-Uc6JaD5N2j(%*@P~n3ys$
zGODVoh=_>p?(W>&+#(_(A^8LW00000EC2ui06YK_000EqP=HCRt@7&4_vqXQj^rc)
zg%KP`l?2K{oh4X2L7^W*Ml*p+<6xoHIs-zb@kjt@i$KG|xEPO{fD<F7Ap$LmAY8dl
z994@Tvk^!wjevj_mrWY!BBziCER-EL7Y<rRITU?mfoXpu3PfshMhbvq9ySb(BLQ|5
z1zi>d3IKQ^BLfd4Zc`QyN=7LtNFaYYBZ3w$Q${ZdWEy4^nzkJe1r`nlK9Px59GVlv
z8gXm@0uO9P4Wl5z4uGE36a@+a1Um$Y!QEKxitX#}?C|pL@uu}y;79xUBoP2RoUUxg

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-3c3d846fb2913b4605e8d59bc5a14e6c.gif b/docs/1.0/img/WordEmbeddings-3c3d846fb2913b4605e8d59bc5a14e6c.gif
new file mode 100644
index 0000000000000000000000000000000000000000..5b58c467e0c8d183fb39273a83669094ce4bf0c7
GIT binary patch
literal 764
zcmV<Y0t5X=Nk%w1VcY-{0J8u9|Ns90004G&b|NAoW@ctYL`2NY%v4lVy1KgD+}xO$
znC|ZGh=_;~5D+plGODVoA^8LW00000EC2ui0Nel*000F35XedET*~T=w*O!#j$~<`
zD5xwcK(s&s&vb3y_VQZ75C}j*!9ZaUTQZz-$!yLL3{BxcK?D*L$1LgWdcR(+C6KuU
zq5wr132wXJEykl_#ZC-?PR!1R`~QFx22>OR4HX4q0EblyJ1!Xu4J;O12#ym3k&{f9
zkD36Ilz}9knVX|<j9G*d3=X5Kp&1QFwGs|`h!Pl!wnw>3y|=!&x}+h)zs7(86#)ed
z3_!;T162SC&e_`l43rN76G<J~66UVp;vflJ9_8EU-s&I^(~oZO=iT2D;Y;l{3>anQ
zhz4ss2r1yQE))fK0u(a55JEwZ5fLRUoX{|ph6Vm;4m5m|G0j9?GYlL{b5X-Uj<PIB
zvM^u{fCFf}ZE#mp0?r9JM<5i@^JdPVAP|UApy*A}pRR5out{O41z7`HZHNgV0HvY^
zjxf;b&Z`8lVzqj(3c@S^w64@1lpp|~8@F)1$*llVfG0i@rdnW2Vby{FRCfYighi6g
zG8tY<aMz%5gU1@*dOUco>VeIbLtW0CM)QaPb`pH9pxR?=)~h{(wv4&7<PCEI5D~B@
zaN={FD-5FR0NU>gYUpTEU_ev9b4}eLSl(R0^K4?K0ok6Duz?1>15<ZCDna(`+$kK;
z^*#J|gA(2gCw2$Ch3RUlUpMaxJrFboB&%@%5jqecL;(g8kXmp>90$Sz?ufwKLnTZY
z0%87~U?B)bWJ7_2G=TWQh8{>G0at>t#$XmG2moIQ1FaDu2r*u`VumKv_`w4YERaA@
zBEk@)K)rl01ONyekiao1R(0bgDcA*q01WUk<A_xzMZg(=T}J|yi?J5Km_Gou84X@y
ul%<ip91vcU;&FL_lq3|1!j(-7xrCl5@YxBUD*&<=GK3bo!I>aL0029xp*5`l

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-3f53ca9954dfc8730d4f58c937097118.gif b/docs/1.0/img/WordEmbeddings-3f53ca9954dfc8730d4f58c937097118.gif
new file mode 100644
index 0000000000000000000000000000000000000000..420db9acdeaaa12670fd6b014ce632cae1a1d6bc
GIT binary patch
literal 368
zcmV-$0gwJiNk%w1VR8Tr0J8u9|Ns90006qWx|o=lW@cvY?(Pr}5X{WXGBPsU+}wzW
zh<0{%R8&+%L_{JYBC4vYA^8LW00000EC2ui0CE5f000F35Qss5B?A;wtp8vrj$}zr
z)s=0X=~_i*9M)_h!mIE67XvmR4S?&N0EvJl=~1*E3s<6Htcr3>q|)NFK&6bp(~trz
z7?uIm5{0=23x{<MI;YoWPwUNS0~8+)4;Bo7Ts;YLbc=$7hJQu_2L>J)9%2@a00%<~
zXi|!cbC>{{oLrx8M1uke9|o=)g8-Hp4tpetAPEe(NWHtjAH2TAA+Q<-v$dtZA_+MU
z2(%yq7?K!WOVfV~-rtY9ACs0IlSAe~>F5^ZA=U%f1KLR33g6(B9y8@S3;nr5eV8Q*
z1i>BuUX=3$(8aBWA;cLh7~-Ighb|r!2*B>e0tFtfN>~d3@&i4Jf1J7a7;*$J1nA13
OG+~iq&6y=40027)0g(Iv

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-3ff042e26bf4d74a5488884f76800e6c.gif b/docs/1.0/img/WordEmbeddings-3ff042e26bf4d74a5488884f76800e6c.gif
new file mode 100644
index 0000000000000000000000000000000000000000..f3c91bcf203f124ad16d59a99f40e070fa588b49
GIT binary patch
literal 401
zcmV;C0dD?BNk%w1VR8Ts0J8u9|Ns90006qWx|o=lW@cvY?(Pr}5X{WXGBPsU+}wzW
zh<0{%R8&+%L_{JYBC4vYA^8LW00000EC2ui0CE5g000F35Qss5B?A;wtp8vrj$}zr
z)s=0X=~_i*9M)_h!mIE67XvmR4S?&N0EvJl=~1*E3s<6Htcr3>q|)NFK&6bp(~trz
z7?uIm5`}>dxzAwXu-ZZA^xEucdEg=g3~c}k9}W)|42E1Mi69Aecaw;UjD?PY2L>J)
z9%2@i00%<~XdeQPACZ%jpa7zzT&EtXfkcS{2_FW#8;_nD4t^4<9;~Yj#7N4J&B@CN
z&%C|{z-h3=wvagw2mqe77y}0wT}$5u1nUhA>jb+<l79x_7Mnx#`S(He_7(!hl^bvl
zNxFCK-o*<y<DZKK|F*4~_rX{z4;EiQ1R=3g69zvHO$@WwaS26?7ei!pF%nhAO9cY#
v)2JXGK#Ktb)cjB|Wuav>Z5B(0AYYj~B}uGowbV#bpCEujooYfVi3k8Y<GP>h

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-4856f4335c4b3d6613590c6d264e3541.gif b/docs/1.0/img/WordEmbeddings-4856f4335c4b3d6613590c6d264e3541.gif
new file mode 100644
index 0000000000000000000000000000000000000000..0d668f1de55b0cd9307178220014269d91dc9df9
GIT binary patch
literal 168
zcmZ?wbhEHb6lCCG*v!E2|Nnmm28ND~ju|s%tXj3|&Ye3UAt5d<E@fq90s;cNcI`TI
z=8Tb%(Ytr=l$4ax($W-vvM>TQFzA2?kQodtp#mO;9)?EQl^rI>w{a@AcSISbG&Xs;
z26%XXxb*Fak3@n=tFv;r$DFV}Hid(X0s>Q~EqTbG<Hh9b^RkA4ZTT@R;}w2k9XgHL
Tl@*EoUUyje4Q#e^FjxZsFEKl%

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-5206560a306a2e085a437fd258eb57ce.gif b/docs/1.0/img/WordEmbeddings-5206560a306a2e085a437fd258eb57ce.gif
new file mode 100644
index 0000000000000000000000000000000000000000..f53085db02cdbe3db6a85a731115c6308ecb498b
GIT binary patch
literal 142
zcmZ?wbhEHb<YVAr*v!E2|Nnmm28Ob-G64aBw6wG{XU@!+F~i8nXw|A!9UUF--o4wk
zYgb4}2v7|KDE?$&1hI5L1jq~q7HfeHK@UcW83h7<j?*+-I#{|mRI&tPZZh>Zwrj58
lVsV=Gnw`yU*<k}dh8RKPi8_YThu7@j_%n^`i4hZnH2|OTEDZnv

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-56cc30cd24196bea5f078cb9b1f1d0dc.gif b/docs/1.0/img/WordEmbeddings-56cc30cd24196bea5f078cb9b1f1d0dc.gif
new file mode 100644
index 0000000000000000000000000000000000000000..72ff3b67b05ab23bf0cb05f287679c24f546ff52
GIT binary patch
literal 1700
zcma)#`CF0+0)Rg*<&wVesC3pLM8*3oO<e~SA<7b!yerf&GjlN&Ev-d7@&XYb^Vmii
zO*}>?)G}i?=OZmIRK^@_%6810(zf!zHl1y3xBp<@AKstd=XsedhUcNU1Q-<<0)Wq-
zKLY@u(P-#&dU0{_`}gnX=jS;bj!LEK=;)}etrZ9aiHV71GTGJD6$HV?#>W5k|3CY;
zk+3hd{TF?C0;nAlNpj{~LsG{NDgPSa+fQX&xD^cg<gV&ea>q0~(~6Os7e`aP%COJn
zJoZ-^l2*a^yURavU(=Y`#wIYo?B@>DBy?R9^ux2GsZm!sBQ0?lJR?22zzwArD{>EW
zL6xNxXJtyAtI2RL0~cI|N5v6ja&C<yJfS4G>3}#Jr;kVS^mt_Sow!z5`#~(O2Med$
z>2-IV{@xa!Mj5qCg6W^X%hXG-nsNJkbX2Dm&Y+pRpdmj`d3GOX0XO)I9U<-8qLZ(w
zXxOU7zby#?=Rc*8Jo_pbDl6>f1j$oK5(ZhK@iFPrM~GlDu?H<&2Oc??o}jByq@@yH
zubu2ml_UOs_l$6}bzGa~MH$SN_f-Oa9q9*~*8bMHeX#;7*;gr0+Yo8_M}oo@a3CM2
zz3vn;<ehfM>J}19)EcR2d7ZtML&UZwU^<_Oek!5j<WtPuU2DgJ&M$UpcXJ=UcgP7h
z!8B|hMk;IogM1XY7g@}r3xBlDBlffS^4L5icVBksLU29I+!eMJxStdfK>R-JFoBKL
z%Twx!`ijL5ek+B;XcfllXMtYu_7%7%$<0J~qh43pW$ZnbA2K(=Lur#~>&=-G>%R`?
z{;nfNz}=VK;oGrSR`66~*r$_<EOfZrvSQ(hFwk7Bc;Jjraea_w<-Q{y6gc9YH84U8
z2!z{Fq{o9Tf|gI5)QvK)d6Pj<3ZT%+r*>2OlE%|b$L1QWF(Wfa+@0O|d!qE0+FnZ9
ze^DCn_g-3_@g_I1AHsYn@FlcA9Q-wxjD+y~u;j=HTgQ!cTI|>R<V_gE-OBnJ&a&{k
z3(r6Fi^O(YilF2e8TX})!O!QcSs_;OahOmzjmAb$XUbDQw`?6_{~?x|GTatNmoO=&
zey!hT)hj0&D4W~Xr$qH^A!3?azuztRpVb6@yFKf)KQi*%(Hj)$C4Uze2fi5H(?==9
z?}CO&*h=%hW`z7rn<DH)>B||h*~&^sRT@b?dux72{gLrC#<#evdHzo$4)8Jay$!v&
z4DWZC>{t52wh&pVW$5A!Tc>gYMP^xrSj&Hh$$3#_E3%I5j;L#=ZNedo{t{8g+YW{J
zzCO=?xZ9dqEi8R#^E2cOFYmEAm41i=+$A(Zkc3(jF?4Wao$LY_Rjq}YIP^Bp{KJx0
zYdd*dvKW7e=P7@cg*ZzlJorP}DN<h*=(?G+n}KoIR<Sdr-BnQ)>oAyBEkP{OGv%^f
zc-^E~rTJrk1((hn2euEn@0mmBSfOX5r@GRMlA*i0$qc}wv@i$xWcX<OGcuKLXxZQ|
zxpsRC@=ALAVlG(KB6n8yt?Bw~b|UaldDK7<<upjhNDlPJwVjbU@K6Df(%0bTYo|#K
z$72r-MY_zm>%dik?cb#9T^yIvRBN>br*LQ0K1N>sk3U`&nrba*<E}_*TT#jFmiW+A
zPFw+GOu9a^;&I|0m=Jko_Rgrye#Y$_%-<S?kkhc>3X<KP%nT<e2UPUKD4cc}7Ns3v
zxjP9B<kt~VY#~Y^7hS-H8`K*W%Q^?Cuaf+BxLU|8LyFAKKh}R(DKzeY7^_M-2kWUe
zl#mJyZ}{PO%E25TF4c`KEKNo9x+WDN)6T(qW2Hk@L(2z8M#t{6*=Dt!s56qbc*5KV
zT(3P%6cf9XXHp$UiM%HVj-CN1rMs{981c#(0Llzp*!5Gaw8&2S0z))><ZBBnIQy0Z
zoBk+zYLsvLaqw4SCB*z@mXt0jze$N9-QF7+C#%)mwDPD!j@%dKa+hvWnXECED4{Zf
z_z3=~J3mylQLTL2Tv$LGkDhi-?AEk6X5B+B&Ah3B@QFd}(eZdwUF}Ga!nI~pljx<Z
z)0QjTTUisy(YpE=p9cG+youDyk1H`?%=!Rp@}fXp^XC*MU6?c}P^F`S1Lzj!Y0ctl
zsBl%&>g(3iDLWvFhkj|L>8q662bxSG{?lR#u37#LZhp7=R_^p>gAlOssu-N0S|2Le
rD$Bgol<8}84e)RON#<20&U-Td;kfqUa@}Ujc1oG)I%^CB0Pp_-O2N{|

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-62ca51f900bf27324a2be4e6b8609f4b.gif b/docs/1.0/img/WordEmbeddings-62ca51f900bf27324a2be4e6b8609f4b.gif
new file mode 100644
index 0000000000000000000000000000000000000000..68c7267f371b95b1cda34f21f8ba61caef62acac
GIT binary patch
literal 2263
zcmV;|2q^bQNk%w1VT%Dn0K@<Q|Ns90004-Hh?tm|s;a8o+}u=DR76BXc6N3U5D>b$
zy3EYXGBPsm?(QNYB4%c0u&}T_Jv|f@6dfHM000000000000000000000000000000
z000000000000000A^8LW00000EC2ui0E+=b000I4ARvxpX`X1Ru59bRV<69PZQppV
z@1`K|z#z^slP!eEq%su{5T1mkbV@}D9=e6&#5k@NhNb}VXaE$6D6CsWDWHUaw@t6B
z<aU_Cp=P7!NDT-BWC8|t1%NgUTP6f%EDH%E0}LV$YA=s@D0rNocAlUsm?E1128uxl
z3IGL@Kch8_S(73P0hgJwH=?}1LcYKxr6s9L#|@S^djo_61Oo<;2n5w44F}5#0s}v2
z0}nsmKDUDcjU)>Ut<BFSk|Si7-Gcye2mq1_1x~E%t<DU>($88zD#C^-?4?M-KoQlh
zxwBUSoH#U59!#1zKt(BlB91vYP=MTzf~@{3K0MRyn*(tI5(p>&z@!M9C?4bk5P)Vk
z5eG!1D_{wXhy_>-2xzdtD5?cSMHEO!OpQ(ev`Q=p;LOR02UxNq_|YJWfK~-sm5Pe0
z!C6-arzBiR*M%>((*Oho(CBJbV@r<|v*nkniDJc4E$HYp)!Vcr@G?E5FX@r2jhj+F
zcA{9WnmZdfO8_bA*chnPATaO*!GbH;4C|T=z?M4ORYlmETJvav104aBf;~bKa5{7A
zDrW+S_z2MhNDEhkdj#vR)M-~Qfmn?J1dUd=Zu&d;2<HG!i^!EMJZozU#KX^-9keS3
zW*n5Zq@L$fh&zceMr@r~Ro`s|>i#xD7r_-!)^WpOF@l2x99Knh%VkJ{FX$1KpBaOl
zz}a*Pl31HZ28Lu!J_rQp%7f+sfL?#>T^62;EwT`zec|X<Kon6eu$KadaJJ4-1+b+>
zf+lVk&tV5v;zb!)VX;;tNmAK>gb1mk;g{<H!ka54NtnQoI~p^`PiSa|B5fiywG32R
za_I_|i40k!jYrLyh@U5DL6QJm5!FQl26XiRO>9{+MgwRVK+`=BIAh8n2ymqp1JelF
zkEX?iv6MO^eHxrH=hgFPUIbLykq?AS+9jB@Zm3ABmLjSs1f>!h;~5D^DwV5Ec<L((
z>vYN~Q!k;Cz#Fd0sA>tg{s!kL0?ba6skPBkYH1K`)!N&a7>Y|*x#p%@oB&RNInyZ`
zqO0yN^3E%60bt4cr8m1|Sg*aH{0s2F1k>wf!LkU9@WKqw+OWY7M=UXN2T$CF!YD)_
z!Iv6atU|{gcT#c4QefOe025HA9HA+bTms7~yBwy>b(lQi0Ra%Ol>;MxQUK062f(w-
zKj*x2&kx&7vkeL-LVy4a$Pk<)LRSNo&kH;qHK<fOV70yg7chg?Qu|hQ&@x!8^wKcY
zTVWA0FrY7@a0kG^s8qZSx7;Mqy*J+G>RmVA&xKL9-GDE=HrsafJQdt8#T6NNa6c=>
z<3&V{LIIQ)VR^0om_z=6<P&yYIq3s0&NvI@4j=<-3qbNhS{emlXArIj0K4oi(4K(p
zwC28h4!#4gIMS=<fMFa5@Sej10l3Hm^UmV{z4UcapTzaYCtvcx4J3fQ4!9uSLipq7
zP=1E!OR;{zs(UXo3au(ZND%cWasP7pzfk=Nkk}u=;wL{7D9jrVH~{Ky;*CUc!ch-E
zT6`uZwi(=RLPfYBzzA4CTfJZarODX|DAE8>(P034;S1PI&@73Sq$w(Rq4~tc!4A&M
z1w#voY=o1-V1*(R^GJa-G>EVtvA~ECtDy}AmH_i)Y6T_y)fB{bkm#g90Z1WY!H^ik
zesxbk&?x>u%xw4qr5Rv{BtSq4+dxH1e8vQ#5snrK#yM!D<OFpz2*W&a0Pn1kjE%cN
z3B5RiR^`PG4lob&Rw$!O%yD3G6bkAh$(IGH3l9QF;Q`X8f(A?qBY}_=8@MpT!x-R{
zs7xg%N05M0j^F^XP=O>|nT5S@E&$czha+94Mi%rEVXdr<c>-X`6cFGb#lYV|;6<(f
z)q!v@C`StnupKIt<d}qlqaq)v93dUxo8SazEJ={f6_m4JaueI6wz3FzU`GhbD5gR%
z_B0iYC{v7VLjoPqq82%T0j`Ny8qQPAfeNUa^O6HQ1!M`1MB|VVi2?y$5({y>qb5FZ
ziT*PN@la}5L!uPbhdTw*Bao!$qi0CuMNOH?mag)KE>M7Z#wisMhUATH7{Ebtn1Tav
z(_SR8&KQ9rt5Y7Ygi;aTC3eyi83kuAjW7T*!?K0U2#2XYd8$xCf=nPSBQM2(>Qb?S
zMji+ZDGEqy{=`TPNQ953cs!&IEaV)vn!u;TlPg_7nAAzxwI-!G6CM6KA-%f7u!J?>
z5v&0XV$$X_`-6^jr25Vi$tR$mxk+e7V+P9320bQ{EQk)21SZ0bGP69-SgsR^Ch~xD
zV~9~|ivzs`Eb0c5kfK7oK?YgP7NfnbLvFPa%-bm9w~CDnF5P4w5Amh5XlYV<{-n2~
z`#D!XsEBTaxZ~9iNp}fK)Bz!BHC{j(Q!aoQKsu_Mj~>+qO$11UbdRJDny|MfW`eJL
zuOt=wM!~&jY87db5Z)Y=_e%#huOrY)3<(*lPz*R!q7ayft30?P)zU|=AS_`;RX88X
zGDC(#tSbcaYEUL5FfEsmIB~UD!6|rgFuZ^-yCoRL9t6gX`G?~c>{xj#?lDh-EaW1G
zg2*^7u#uMx1SNaV$xMziL!{gSC{y{$tV1%4t&HU^pMcA31GAUMJY+H_*UV^Ef|zNx
z<}^#W%`Stpo0lx-kkxt4LcX)ww2bFD>v>{-?z4FVokTth8qI`GnxPL|P3A;5%+ZW~
z@}kKI=}51x(g?$}rI)7Z0Zw|;RPJ=@J`L(9i+Y2kE;XG~Ex%K%8qlnk^s8U(>V(EO
lN%@@hm~owmT3-@wy2kVGiV*AMHQ?CD4)&apo$L?*06SA=$}Ru^

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-74fcb594bdd93c0f956682ae1ea013e6.gif b/docs/1.0/img/WordEmbeddings-74fcb594bdd93c0f956682ae1ea013e6.gif
new file mode 100644
index 0000000000000000000000000000000000000000..612bf26e8df31e714bc30b7029edc4d6b3e10560
GIT binary patch
literal 178
zcmZ?wbhEHb6lD-**v!E2|Nnmm28La`cDcB?goK2Ym6Zty2&AQ@ojG%6#*7(8Mn<bv
zt?KCLc=zrdP#F*?{$ybU(K;XkWCjCEd_aTe>a|ICYTM_v3d(p)WjW<kDJmRuHgV$>
zArrC7wZcMEL>M<X%{F1uTAY{8DWc(>=IS_u!}2JnBV&fk(kTKDPBNx2E<aLGEXJ}g
aXWqo<d2?R7SlaFH|85;xFUlgwU=09AK0VL?

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-8b4e6bfa073defa91d3f23cdec8f1f0e.gif b/docs/1.0/img/WordEmbeddings-8b4e6bfa073defa91d3f23cdec8f1f0e.gif
new file mode 100644
index 0000000000000000000000000000000000000000..58b5390b2450343914cfa41cdb4935da750c81ac
GIT binary patch
literal 1859
zcma)#`#;kQ1AxD~+Su6SvPRxn<CsP><XXNm_d_lt;$-M$s(DA~xUH`xM!6=p!VpQ$
zv7}rM@0n|ICYQ5>*ZowVUZHfNI`#6t{Ri*!!}HVg`8=-fF4i{T(=ZKS5CHb}_5c7V
zFE39@N)n63a=E;@xtU6(5(tFN&CTxa?x?6J4u>;8KhI*Z*le~`D*a#o|D*pI3Hw$v
zzUkW&K<!c}`DbFPdCfnsmt**h@dW<TkFhE_n|m^ez^%l9>^xnU_`PlR*d2?~An(?h
zmOvVbfDwr@Qk$@&Q2#;bc2!&^+;i+{Z@|Zmw1w&GCzGv|k-{rk*(N_6L@6<20?ko{
zK%zx4&7P&oxnWgiTtlz$ucu<GPBs)As<pl;mfp|C96bon2b-*L4;dY-Zeh>K0b1XQ
zp%7$b+t^com@@6~@yF(HSnnaVbjEb>%$)m*!4msL3JN#AzHwXy(;TCQaYLz|hoe={
zhtasHSI=~O=@~EtTub4%EU5;R@Cw|#Hdh8@erh3Lq~1ulI<9_UHM#{l1_8Tz=!p@t
zmyR$s<c=8=tg*=9f6%|?K1qGB@8kQ8Hy2qH;4(%t_jPiS<jEx$805}V3qz!T{aS0W
z?MDY1BL(@A7=Od(w;zg5Upc#4Phg_S2?^QopF_KyNJxx;3%2_6y={yMvLbkgt6jjU
zcgnBCXxK2L5D|oY@oN<(4?N%XPV9oT)E0yUBWvLZmjuR!wO7?>Q@};7i^rvy{w$*T
zEA<OWHj~|!R+Np|#7*Ka{yS(v@~MXD%%T$SMcXazQz{&jEW6eh-$U!X`cC!qNB|mx
zj(WRqjPqr1ct$A}G3yN2Y^C3OH1*bu{G}HI&7|H@^aGcU8tn=|olm~hg;anPkS)Tk
zb-*)7-C#|UR+P3+hCDGI1WCe?6<?iWrCv5_WP_Eos{N65)=Cp-865Yb>5u%o6`(9$
z?L$z1xWmvgc-cTVIK-;s0Q3UmkWV;+_a)HT8h}x<sFY*NEQ`p{Y+p@5Zf}VI9oR@j
zcIGyua$&FQxyqSOITh2noeTm1$Bk&Ol%HGdjIKG*NQo-Hv``#Xn<Uhnb^KCDGORN-
zN{P7enGW!TjX|SyF-?JC+asNw;qK-iE{d8BNZhkzOph{fS8saRINtd$!5hCL!e?)=
zm3e}oia|pcdc_DC@D=Y=__$xd+Y(%@u)AEP=5u^qTTwg8=gWBczSvXizDqm6#+B=%
zU5d8t&(o4e=vyRSjKugV`xpJ4cc$u<b3r-1a;=cLC=eW$^cYCi*l^X6Tv1<#zMMVK
ziSCmQ*rb&WA5vAGH)=cJFA7Vay0yw^=p;d)6ltfu4E@{mVVg<1<S%ljyoe5tMM9rm
zVdo43e@izjy>{pE$GT#^T%tbKG8wC$>cR1hwrWc%?tV;58rgVli8yW=PleqydVNKk
z$^cxv@*9s-!to4q2>uay8}#y2jNO`7CJCT9GXnbWvv~=%1;9|KqGSnL>U?yNCnSci
zSn8KpdqpyZ<RRi}ZydOrTTr`fa^AoHeV~5!_lG51!>q6#MW_mE@*mW~L%_WC$KPQN
zsEx@J*gPg!#b<YE-0!=}?0H5HF2zl)RliR;l0Tt(6hO!=Jtim~ziMpD|M1J%3UlAE
zSbz37?O&m`<#Jv4QI_o5k1C8JfXvgO-LSyBIr$-4NoaSB%ppdSkJ{AnF3_5&w}l*$
zy^e^rwTmUJrbzpEE~P?GY)5pPFxpe$s`+U4wKu#Hj(W;wwFrbn<T4B;SV6pQEZg5*
z2s3L5R}C}DyMzH^k=$CGQvE;!lY*mRZkjA6D7)Pkq}!L&`u1lPmSr}G!-8cd>$L;M
zfpHAi!QFA*B5jDZhRP1Jp%QA!U1*g>*`ZB2I_KO5jH-QAkA+5pX+}!>|LJ|z&eMtb
zC&@~G$~V`i`+(|*dMk7u6=xNZ6+-sOdSfpmB5tv1;fYVSxaR|Iq!x^=&d_1oTv730
z^;cf8W0b&+=-P!c0U(dNJoSL+N`vOkK)rom<r+&xRXE{Yl=IHi0bTR)a=otnA7X!_
zm72MIXI>#Svj8iZAoh9?M2~KySpw@BP3>Jzg2PYbEOOY^FsAn|pCx{FCLb|q)6da+
zfJD0GncaU*32&y<ytErU_I4JBBK)qDY)wV}Cwjbu=S~lpp&6K5Mizq!*so&ou+HxF
zrwUrDp3`KL`P;agxpV%+zwtr05bSrkv8)ECEsUeP4|Ld`9`3`AR5?q|b|NgFG{54^
zz6$nj@rwA@mWN$*`<O;QnNUx!^&Jz&shv)%RoR)K;BzY#otkLK)k8*{Y=^`=={?Ue
zYJi$V(P~R+iNx0hFK7ik?3XlS3HR!>0NA6HZHePgcy6lAgiP<ao&yi6)}5$25dkOW
p;(<uK=j`aIowhM+ij<?wk*DlDkl^`TwLk5{Or6?aVqt*Ye*ykD2rd8s

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-8d9c307cb7f3c4a32822a51922d1ceaa.gif b/docs/1.0/img/WordEmbeddings-8d9c307cb7f3c4a32822a51922d1ceaa.gif
new file mode 100644
index 0000000000000000000000000000000000000000..900275c2edd12b4c2bbf39da3493cd45eeeae885
GIT binary patch
literal 155
zcmZ?wbhEHb6ky<C*v!E2|Nnmm28Ob-G9@LYGiT1cd-pCRB*ewVW!0)x9UUEMX=yWN
z%-FSS*PT0e1Ox<(jEoe2vM>TQFzA2?kQodt9s(W;ZVH=c6kM2cGNGoy$;(?vWTv~3
zub_(Xp-Ck#eJu3^8QEThMEJ<DJ2?Nyy28Z!*wd<zqe_02%Yy^|wlGK*FDl?Uu*{!@
G!5RP=wKrn`

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-a0be0498fd8216177330deffbfcb6ea2.gif b/docs/1.0/img/WordEmbeddings-a0be0498fd8216177330deffbfcb6ea2.gif
new file mode 100644
index 0000000000000000000000000000000000000000..70f5fba201de51cf1497cd0a3eb53e8a8e226bcf
GIT binary patch
literal 2161
zcmait=|7Z>0*0S?W-*2tV{Aj?ZH%R{Wfxg`WGq=?G@(-A#jEmW$#F(nhR_HTjy>BX
z5z<)Nn0O2-vP%hB3YDcMvefD5oc@Ayf4Dwf*Y9>Z;AmkP5Q?w|IsxF<uU`NFaJk(4
z{QT<b>dDDT7K_DTFk)h2h(uy@b8~NRud%W5=g*&MG+IbVh(I8)v9XbtmseF){a^Th
zr}iC%_$}%E#_vx6y-lU^f=-o$R@c?42qSa{qW_*|m*5gU|AIKkjiadR&ksRRC@E{O
zDqD;FQIRLt@#rWy7wjxYcxse1t*|O0NZ{msA*OU%p3pQ-aWW?&13^p-fWBysKca;S
z(vy-@M@rLVLQT?>c5@4hHU3dzoJ%VJFhvj+mz-(C-*e3;lZHd&<lVC@d(3YU(#tg~
znH~6S3pO$yi9%v{Skw>=FYycqB5gvDGB)GZO&l;)LrS`)`-N79dSK2Q4iJt7GH__O
z>JK%+9PjV9){nhk^5G}owo|y~?!`jVq_7}%m73Lo&~f}KXVlvs56+xNvCzzbg0P8g
z3R^A%f)Ie3Lnq2x2IAzsVNi(7IzZEars!)#Eo%qMZYyrdGs`F`{xXYQI`veJv6lKD
zz6ICHoUZHwy%l`TPT#qM3Bcovyas0|CqqgRW7@Z}Se5Er6*Mu-n*|(-=S#+%MmXN8
z9Oh}tzNV|i*{8cXT=@1?AlDw&Ui$_*0y??2@C3mCK4ExRk;PBm%sm4%_I6#m7UF#<
z-%gGKKUb>Fk<EyiIJK1PUplI>S7vMSSm&}Z??>yU6QGU?602BzHesgf?38+jL3FT*
zw*N|1OSp&r*yG;ww%rzh`++9xupjZIylk@BLi_!C`{U;X=>l{(`ogy6B>kQ8O)5K#
z?^Pt<YDz98208yF3P$}<@W;K1)Nr@?VYES&#E<)&FdAQxfKqfr$ym2GO;1?PRUgU&
z6a8WidZP1_x-O4MItRrY-)UD;#``i~(d>ALvGAwvY|ot!l|ircJ*Vkz6w3mp1{a75
zN<=!TmS<<X8)N3xtTyOxX#2*`aSILV-!pB!zc)D`F@Ec?YsdjPM`Q>&V;4e(ZjYm4
zeH|2#1O8t#fh=&-fw{1pq8<J;QE8w!%4m@;6F+yPA}3reP{)yj!^Kw?@G?S6(58D~
zQtqlG-!fZ8K~K?Mb1=CMD7(g~zD2e2@WGPjVUx`5)s54QJWPOMjn2=WyY8e*J5Z7D
za^sQ!=Ykj0>AZZ`y5_w?sV_pkxxMr!<1=1$!f(T}XWIY?y$H#l1mXU6%&uC)8mxhv
zxQ*bX7imEBus+(Q$DqbHs~oeofA+`2L2;*gIp-7V`o*}Y9<>`3phtgaT78#ho+pez
z5Iq$NYp06V`b`X$8~YJrN+Ze6)o)b|g@s`$p`BN>sY{ZmX~gq``ThQ0+!JWT<1d~2
zNY+19qxuNyZa_Z+1p|$o2P@{4XKM8~C%P{VnIog6!ZxQOmfTU(E5k+6oj=V-AGn+`
z$~`{u)P-*$n^x`#zh^KA;yFXsCDHu0ebAJ}yQ`J|`BI0-egMttryXVjtUuzL7i(YF
zz)STtRxSPa@6DVC1#xK`qA#tH&=+%SA}4hp`T7NXkBLFu8pH_S8lq=RK;ZC_r|Tzr
z*z`rhbd9bpX<irKDw;E{w*O)R&R!yY&9mA^TED@w88!`FgAwI_;$)-tNZ8xVoE*ja
ziwbY!x%&PW7AHz29nFEkx#e?63l};DQc}zBBXY1DP+|ek6D*^2+NZ>0N~YxnH_eg<
z^R#w9B7&C!6A_1Cqo$Yo$gzx|CD*a%B%s9k;=$w?ItQz1R5mn2^q1yP8HuYM*ww4b
zXp7nPv;F}21W#E*dpnUN(V(kjU4PP>`P5W&RrYRGhLpskgAxOnR?MIPRjeG3>}R7S
zI8L!Ug)%#5D;PRmLqR!Jv`<;1?a1hGnwe*oq((szWD5iNy|=U~K2pT09BUm2rp%YA
z-)1bvx<>mRh8v6IX4EssTsgAbkDVKXW%6ksRS}p|NZ}}zq7SonK@F|}x2ZW&jqf$5
z;AU&lYH0YOz>@0yT9)K1dE3=?BdmyIv1ugXmr40B+|f%)bKXbb1`<cn5~o+MlGvPF
zU12@p{8)#tIumyMIfRcZ>mdFJz2X8DD7dT0Tj|jYK-k{1xC-~LLNW~bRiVNPrT9PN
z)wU%GWzHjSGPNh=G7r)l6js=1y+Ac79#2gPQ}I$>vAF$t;!RFrzd<Umit3E)CO_VS
zEV^%!4oWJHl-PR69RcBe_}P*4m}tDcS#@7wk9@W5@{wNW3dE1uG=h<)6!YJ_Gb%5O
zX2c;i@;DVSS4?*9wA&Flb0<s~eC0(FV*lBSWcu4zsuU=pj!$$^z-0N%&<zlqI6eK~
zD_jxil>4mH3!ub%=MzP)5`jG_<m{G)BJAC$u67)s(;1uNozRy`l*D-Qpihj7&j(0K
zV{;mmCwDBHR{GS+497K!ypOjz{7p#ManJTqrh0mo(>+^<bXA>HR?wqYmg^;7{hmj7
zH7!~1p9UTrp9T3X&VWPRWK6LgxN+=>4H#MZ;rMjjA}Z%CqGd4u?Z$W>o8<D6xrP>e
zo7!wcu7$%W=^1b!75-)>NSF>ii89z~mx~g|FG<bfl$=CjksxW0wLNb0bq9qId3A?$
zpzMvUC!$etijQ?2$s@nZa3S~F=KMSI1i3|{wVoGi^k?Kc`x&%wGVU^Ejy7HSbeYkR
znRj81V)DIv_ONj}-z}I*EC}$bYvA1r)7*U~w(CS|L~a}9y<tmNZ_t8a;pNOmDq6cQ
jA|&lbjd6(bL;ZgC30>*M-fGO&_W?F<trVdO0#^S8-%+Ar

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-aa38f107289d4d73d516190581397349.gif b/docs/1.0/img/WordEmbeddings-aa38f107289d4d73d516190581397349.gif
new file mode 100644
index 0000000000000000000000000000000000000000..545f83b7eccb802c2f0c9e29a8217a783a24ab1c
GIT binary patch
literal 161
zcmZ?wbhEHb6lCCL*v!E2|Nnmm28La`cFmYEBP}iM-Me=J0s?2woG~&ox^w4FM@L6l
zSy@O(h>MGhl9JM@RjU+#vM>TQFzA2?kQodtegYki4XVr{!E1D97Dg;;<(S8!(s;DW
zi`}MstHHGk6<?oQ@^d>gdRH<TEqH3cq*1V><(L3V${L1tp9V=814DnNM4jbg;l+`&
M!#<02axho}0D<Q?E&u=k

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-cf775cf1814914c00f5bf7ada7de4369.gif b/docs/1.0/img/WordEmbeddings-cf775cf1814914c00f5bf7ada7de4369.gif
new file mode 100644
index 0000000000000000000000000000000000000000..ab8206e72c6d3c1c66a2b3f31a76c7c66150ba8e
GIT binary patch
literal 949
zcmZ?wbhEHboWY>OaD;*3|Ns9C3=A`7%viN*m6DRuu3fv{y?bY5WaQ%Fa^}n#0Re%s
zva*nnkUMwobaZs2rKQ>0+KPyX$jZtBbrAy;f3h%w)aif-kev*y{}dGZQZnadtXh}z
zdS8Lg`R5J-tr;>NPW`Nh9xOL9={WJ^w40CyWB<d+CMSH?N3@DQ2yxrka97OdpRUBB
zmVC8>9j@`~ZZmRlv2Y)BIAG9mE5EFwiiN*~myek{Iki`lm$5F^h%<wM$1Z-V(u@FZ
zB}R8XzuD7g&Sztp9>!<M+tcS+zIn^$AO?27?OXR$siZS^&2&C+C?JTje*RJJ!%AGM
zGr2e>Z@i|)%+6eA%spQv{+?>#g_zU#m7iSX<z-{Oa{Z$!<BKWmneMZ`GjRS4dRp=0
z?d$S4H#f3z%Kb3>;4sgb;cN#_MS**QJcF>dnnD9ZAOj;WpUj7e6Ghznnf?jY6*WIU
zad^5x@~JtVo0pxPZBhK{&dJTo&l?FS{gH_}=E%S~twYQ85`$v{BPWwU#AXq;B@7Xp
zv+kbSy88P1gw1La4<0aF)8iEJh-ENcb9;Ni{#`L2_14|pUCw`o_b|iy`}^bgEgpm|
zVq`qf!6@FOy69lj;ZEh-SA(1w!q@^h!|%rIIKaG>*OiB(!NK_XWevFt+=~~;y}GW>
z#Uddswe77iWB%*7MuyMtuGdT|T4c0JlzUUI#Bt*Who7cTj=dC<SN<a;z@qH7N~0mR
z<y(87u66%WTkESg?iL2FohNc|-7D2Q{STCl4^+5+`hP%u;(YI*Z^~!bTr4#>G@KY&
z_WetV_hGv>aRDc<k_A%(gM(3c$g<vqha3_{74eeN%LF;O6h9d+DpiRPy4t#M!3L#v
z-9uX*c<XoV5a%?V^+KrC@>zsazoFI-MM;Ny9Ln9+(-JsULQYOgOctD{q2lAiq?no*
zSmdn8V4U>jflV~ehov^Lc2}0p(3<1<VA_X{2M6kuw{RYq{gv0VZBEIuhUfKOant8|
z^TxP1Ro-LNTG7^SHZ7yIo9&m%^kj>uh{aQ8sl8nA^5xPMnJtfuu1}Nb{d7PqGgNuj
ziX~bJSy9hFE_PVG=HH4ePBxtfR~Q9%?s~O))1gz^eVeB>+*u*R<~@huh~JHL?(GKN
zZkvzWax-tyGHqDJwDr2z?v`zO3EfRwHF&-??TDCgyJ=_eg56EKuIhPT-*U%_yKTG1
Ug=(kLy*Gc)_CMGuz{Fq;0GpP9Z2$lO

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif b/docs/1.0/img/WordEmbeddings-f1290186a5d0b1ceab27f4e77c0c5d68.gif
new file mode 100644
index 0000000000000000000000000000000000000000..160e397556d9e00cd7bdd0c3d0521660a4b08ed7
GIT binary patch
literal 138
zcmZ?wbhEHb<YnMs*v!E2|Nnmm28La`cFmYEBP}iM-Me=J0s?2woG~&ox^w4FM@L6l
zSy@O(h>MGhl9JM@RjU+#vM>TQFzA2?kQodt<^mm#4XVr{!Cq$VtWg~b44w%~#TJNM
oI~3u3S#Ry<h@MNorZMm=*ev6;sO8uO=9DO=GY=XnBb6Ae0lca*egFUf

literal 0
HcmV?d00001

diff --git a/docs/1.0/img/WordEmbeddings-f623e75af30e62bbd73d6df5b50bb7b5.gif b/docs/1.0/img/WordEmbeddings-f623e75af30e62bbd73d6df5b50bb7b5.gif
new file mode 100644
index 0000000000000000000000000000000000000000..5d8937d82f44cdb35a5b423f90f6096df0677ef5
GIT binary patch
literal 149
zcmZ?wbhEHb<Y(Yv*v!E2|Nnmm28Ob-G9@LYw6ruA7nfbTc7=q5ynFX<#*7(v?%Xjl
zGFr82m4JXiM@I)xrQ%N(Mj*wY10q0XFt9iYcqq6D%n&iIwv;tenJY43uAsupC?!Vo
yP}K!<?mAsrm1AVu=)EXuD%*ma3#LhUu{BLv^pN?$sy4^a4IDB??CzRO4AuahL@|^A

literal 0
HcmV?d00001

-- 
GitLab