Skip to content

T
Toy-MT-Introduction
Commit 96c091cf by xiaotong
Options

add dashes in figure names

parent 3ef421f4
正在显示 包含 53 行增加 和 51 行删除
.gitignore
......@@ -25,3 +25,5 @@ Book/mt-book.pdf
Section07-Making-A-Strong-NMT-System/section07.pdf
Section07-Towards-Strong-NMT-Systems/section07.pdf
Book/mt-book.run.xml
Book/mt-book-xelatex.bcf
Book/mt-book-xelatex.idx
Book/Chapter6/Chapter6.tex
......@@ -13,7 +13,7 @@
%公式1.7之后往后串一个
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\chapter{神经机器翻译}
\chapter{神经机器翻译模型}
\parinterval 神经机器翻译（Neural Machine Translation）是机器翻译的前沿方法。近几年，随着深度学习技术的发展和在各领域中的深入应用，基于端到端表示学习的方法正在改变着我们处理自然语言的方式，神经机器翻译在这种趋势下应运而生。一方面，神经机器翻译仍然延续着统计建模和基于数据驱动的思想，因此在基本问题的定义上与前人的研究是一致的；另一方面，神经机器翻译脱离了统计机器翻译中对隐含翻译结构的假设，同时使用分布式表示来对文字序列进行建模，这使得它可以从一个全新的视角看待翻译问题。现在，神经机器翻译已经成为了机器翻译研究及应用的热点，译文质量得到了巨大的提升。在本章中，我们将会对神经机器翻译的基础模型和方法进行介绍。
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
......@@ -80,7 +80,7 @@
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-example of mt}
\input{./Chapter6/Figures/figure-example-of-mt}
\label{fig:6-61}
\end{figure}
%----------------------------------------------
......@@ -92,7 +92,7 @@
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-numbers of WMT systems}
\input{./Chapter6/Figures/figure-numbers-of-WMT-systems}
\caption{WMT冠军系统的数量}
\label{fig:6-2}
\end{figure}
......@@ -243,7 +243,7 @@ NMT & $ 21.7^{\ast}$ & $18.7^{\ast}$ & -1
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure- encoder-decoder-process}
\input{./Chapter6/Figures/figure-encoder-decoder-process}
\caption{ encoder-decoder过程 }
\label{fig:6-4}
\end{figure}
......@@ -264,7 +264,7 @@ NMT & $ 21.7^{\ast}$ & $18.7^{\ast}$ & -1
% 图3.6
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Presentation space}
\input{./Chapter6/Figures/figure-Presentation-space}
\caption{统计机器翻译和神经机器翻译的表示空间}
\label{fig:6-5}
\end{figure}
......@@ -281,7 +281,7 @@ NMT & $ 21.7^{\ast}$ & $18.7^{\ast}$ & -1
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-a simple example for tl}
\input{./Chapter6/Figures/figure-a-simple-example-for-tl}
\label{fig:6-62}
\end{figure}
%----------------------------------------------
......@@ -292,7 +292,7 @@ NMT & $ 21.7^{\ast}$ & $18.7^{\ast}$ & -1
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-A working example of neural machine translation}
\input{./Chapter6/Figures/figure-A-working-example-of-neural-machine-translation}
\caption{神经机器翻译的运行实例}
\label{fig:6-6}
\end{figure}
......@@ -386,7 +386,7 @@ NMT & $ 21.7^{\ast}$ & $18.7^{\ast}$ & -1
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Structure of a recurrent network model}
\input{./Chapter6/Figures/figure-Structure-of-a-recurrent-network-model}
\caption{循环网络模型的结构}
\label{fig:6-8}
\end{figure}
......@@ -398,7 +398,7 @@ NMT & $ 21.7^{\ast}$ & $18.7^{\ast}$ & -1
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Model structure based on recurrent neural network translation}
\input{./Chapter6/Figures/figure-Model-structure-based-on-recurrent-neural-network-translation}
\caption{基于循环神经网络翻译的模型结构}
\label{fig:6-9}
\end{figure}
......@@ -454,7 +454,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-3 base problom of P}
\input{./Chapter6/Figures/figure-3-base-problom-of-P}
\caption{求解$\textrm{P} (y_j | \mathbf{y}_{<j},\mathbf{x})$的三个基本问题}
\label{fig:6-10}
\end{figure}
......@@ -481,7 +481,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Word embedding structure}
\input{./Chapter6/Figures/figure-Word-embedding-structure}
\caption{词嵌入层结构}
\label{fig:6-11}
\end{figure}
......@@ -493,7 +493,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Output layer structur}
\input{./Chapter6/Figures/figure-Output-layer-structur}
\caption{输出层结构}
\label{fig:6-12}
\end{figure}
......@@ -628,7 +628,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-the whole of LSTM}
\input{./Chapter6/Figures/figure-the-whole-of-LSTM}
\caption{LSTM的整体结构}
\label{fig:6-15}
\end{figure}
......@@ -703,7 +703,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Double layer RNN} \hspace{10em}
\input{./Chapter6/Figures/figure-Double-layer-RNN} \hspace{10em}
\caption{双层循环神经网络}
\label{fig:6-18}
\end{figure}
......@@ -750,7 +750,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Attention of source and target words}
\input{./Chapter6/Figures/figure-Attention-of-source-and-target-words}
\caption{源语词和目标语词的关注度}
\label{fig:6-20}
\end{figure}
......@@ -764,7 +764,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-encoder-decoder with Attention}
\input{./Chapter6/Figures/figure-encoder-decoder-with-Attention}
\caption{不使用（a）和使用（b）注意力机制的翻译模型对比}
\label{fig:6-21}
\end{figure}
......@@ -786,7 +786,7 @@ $\textrm{P}({y_j | \mathbf{s}_{j-1} ,y_{j-1},C})$由Softmax实现，Softmax的
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Calculation process of context vector C}
\input{./Chapter6/Figures/figure-Calculation-process-of-context-vector-C}
\caption{上下文向量$\mathbf{C}_j$的计算过程}
\label{fig:6-22}
\end{figure}
......@@ -828,7 +828,7 @@ $\textrm{a}(\cdot)$可以被看作是目标语表示和源语言表示的一种'
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Matrix Representation of Attention Weights Between Chinese-English Sentence Pairs}
\input{./Chapter6/Figures/figure-Matrix-Representation-of-Attention-Weights-Between-Chinese-English-Sentence-Pairs}
\caption{一个汉英句对之间的注意力权重{$\alpha_{i,j}$}的矩阵表示}
\label{fig:6-23}
\end{figure}
......@@ -841,7 +841,7 @@ $\textrm{a}(\cdot)$可以被看作是目标语表示和源语言表示的一种'
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Example of context vector calculation process}
\input{./Chapter6/Figures/figure-Example-of-context-vector-calculation-process}
\caption{上下文向量计算过程实例}
\label{fig:6-24}
\end{figure}
......@@ -882,7 +882,7 @@ $\textrm{a}(\cdot)$可以被看作是目标语表示和源语言表示的一种'
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Query model corresponding to traditional query model vs attention mechanism}
\input{./Chapter6/Figures/figure-Query-model-corresponding-to-traditional-query-model-vs-attention-mechanism}
\caption{传统查询模型（a） vs 注意力机制（b）所对应的查询模型}
\label{fig:6-25}
\end{figure}
......@@ -902,7 +902,7 @@ $\textrm{a}(\cdot)$可以被看作是目标语表示和源语言表示的一种'
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Query model corresponding to attention mechanism}
\input{./Chapter6/Figures/figure-Query-model-corresponding-to-attention-mechanism}
\caption{注意力机制所对应的查询模型}
\label{fig:6-26}
\end{figure}
......@@ -1009,7 +1009,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Relationship between learning rate and number of updates}
\input{./Chapter6/Figures/figure-Relationship-between-learning-rate-and-number-of-updates}
\caption{学习率与更新次数的变化关系}
\label{fig:6-28}
\end{figure}
......@@ -1052,7 +1052,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Data parallel process}
\input{./Chapter6/Figures/figure-Data-parallel-process}
\caption{数据并行过程}
\label{fig:6-29}
\end{figure}
......@@ -1111,7 +1111,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Decoding process based on greedy method}
\input{./Chapter6/Figures/figure-Decoding-process-based-on-greedy-method}
\caption{基于贪婪方法的解码过程}
\label{fig:6-31}
\end{figure}
......@@ -1123,7 +1123,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Decode the word probability distribution at the first position}
\input{./Chapter6/Figures/figure-Decode-the-word-probability-distribution-at-the-first-position}
\caption{解码第一个位置输出的单词概率分布（''Have''的概率最高）}
\label{fig:6-32}
\end{figure}
......@@ -1148,7 +1148,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Beam search process}
\input{./Chapter6/Figures/figure-Beam-search-process}
\caption{束搜索过程}
\label{fig:6-33}
\end{figure}
......@@ -1201,7 +1201,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-structure of GNMT}
\input{./Chapter6/Figures/figure-structure-of-GNMT}
\caption{GNMT结构}
\label{fig:6-59}
\end{figure}
......@@ -1283,7 +1283,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Dependencies between words in a recurrent neural network}
\input{./Chapter6/Figures/figure-Dependencies-between-words-in-a-recurrent-neural-network}
\caption{循环神经网络中单词之间的依赖关系}
\label{fig:6-34}
\end{figure}
......@@ -1295,7 +1295,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Dependencies between words in a recurrent neural network}
\input{./Chapter6/Figures/figure-Dependencies-between-words-in-a-recurrent-neural-network}
\caption{注意力机制中单词之间的依赖关系}
\label{fig:6-35}
\end{figure}
......@@ -1307,7 +1307,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Example of self-attention mechanism calculation}
\input{./Chapter6/Figures/figure-Example-of-self-attention-mechanism-calculation}
\caption{自注意力计算实例}
\label{fig:6-36}
\end{figure}
......@@ -1350,7 +1350,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.30
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-self-att vs enco-deco att}
\input{./Chapter6/Figures/figure-self-att-vs-enco-deco-att}
\caption{ 注意力模型的输入（自注意力子层 vs 编码-解码注意力子层）}
\label{fig:6-37}
\end{figure}
......@@ -1381,7 +1381,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Calculation of context vector C}
\input{./Chapter6/Figures/figure-Calculation-of-context-vector-C}
\caption{上下文向量C的计算}
\label{fig:6-39}
\end{figure}
......@@ -1393,7 +1393,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Transformer input and position encoding}
\input{./Chapter6/Figures/figure-Transformer-input-and-position-encoding}
\caption{Transformer输入与位置编码}
\label{fig:6-40}
\end{figure}
......@@ -1416,7 +1416,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-A combination of position encoding and word encoding}
\input{./Chapter6/Figures/figure-A-combination-of-position-encoding-and-word-encoding}
\caption{位置编码与词编码的组合}
\label{fig:6-41}
\end{figure}
......@@ -1450,7 +1450,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Position of self-attention mechanism in the model}
\input{./Chapter6/Figures/figure-Position-of-self-attention-mechanism-in-the-model}
\caption{自注意力机制在模型中的位置}
\label{fig:6-42}
\end{figure}
......@@ -1477,7 +1477,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Point product attention model}
\input{./Chapter6/Figures/figure-Point-product-attention-model}
\caption{点乘注意力力模型 }
\label{fig:6-43}
\end{figure}
......@@ -1490,7 +1490,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
\begin{figure}[htp]
\centering
%\includegraphics[scale=0.4]{./Chapter6/Figures/process of 5.png}
\input{./Chapter6/Figures/process of 5}
\input{./Chapter6/Figures/figure-process-of-5}
\caption{公式（5）的执行过程示例}
\label{fig:6-44}
\end{figure}
......@@ -1509,7 +1509,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Mask instance for future positions in Transformer}
\input{./Chapter6/Figures/figure-Mask-instance-for-future-positions-in-Transformer}
\caption{Transformer中对于未来位置进行的屏蔽的Mask实例}
\label{fig:6-45}
\end{figure}
......@@ -1533,7 +1533,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Multi-Head Attention Model}
\input{./Chapter6/Figures/figure-Multi-Head-Attention-Model}
\caption{多头注意力模型}
\label{fig:6-46}
\end{figure}
......@@ -1562,7 +1562,7 @@ L(\mathbf{Y},\hat{\mathbf{Y}}) = \sum_{j=1}^n L_{ce}(\mathbf{y}_j,\hat{\mathbf{y
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Residual network structure}
\input{./Chapter6/Figures/figure-Residual-network-structure}
\caption{残差网络结构}
\label{fig:6-47}
\end{figure}
......@@ -1581,7 +1581,7 @@ x_{l+1} = x_l + \digamma (x_l)
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Position of difference and layer regularization in the model}
\input{./Chapter6/Figures/figure-Position-of-difference-and-layer-regularization-in-the-model}
\caption{残差和层正则化在模型中的位置}
\label{fig:6-48}
\end{figure}
......@@ -1602,7 +1602,7 @@ x_{l+1} = x_l + \digamma (x_l)
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Different regularization methods}
\input{./Chapter6/Figures/figure-Different-regularization-methods}
\caption{不同正则化方式 }
\label{fig:6-49}
\end{figure}
......@@ -1615,7 +1615,7 @@ x_{l+1} = x_l + \digamma (x_l)
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Position of feedforward neural network in the model}
\input{./Chapter6/Figures/figure-Position-of-feedforward-neural-network-in-the-model}
\caption{前馈神经网络在模型中的位置}
\label{fig:6-50}
\end{figure}
......@@ -1637,7 +1637,7 @@ x_{l+1} = x_l + \digamma (x_l)
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Structure of the network during Transformer training}
\input{./Chapter6/Figures/figure-Structure-of-the-network-during-Transformer-training}
\caption{Transformer训练时网络的结构}
\label{fig:6-51}
\end{figure}
......@@ -1662,7 +1662,7 @@ lrate = d_{model}^{-0.5} \cdot \textrm{min} (step^{-0.5} , step \cdot warmup\_st
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-lrate of transformer}
\input{./Chapter6/Figures/figure-lrate-of-transformer}
\caption{Transformer模型的学习率调整曲线}
\label{fig:6-52}
\end{figure}
......@@ -1677,7 +1677,7 @@ lrate = d_{model}^{-0.5} \cdot \textrm{min} (step^{-0.5} , step \cdot warmup\_st
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Comparison of the number of padding in batch}
\input{./Chapter6/Figures/figure-Comparison-of-the-number-of-padding-in-batch}
\caption{batch中padding数量对比（白色部分为padding）}
\label{fig:6-53}
\end{figure}
......@@ -1729,7 +1729,7 @@ Transformer Deep(48层) & 30.2 & 43.1 & 194$\times 10^{6}$
% 图3.10
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-decode of transformer}
\input{./Chapter6/Figures/figure-decode-of-transformer}
\caption{Transformer推断过程示例}
\label{fig:6-54}
\end{figure}
......@@ -1752,7 +1752,7 @@ Transformer Deep(48层) & 30.2 & 43.1 & 194$\times 10^{6}$
% 图3.6.1
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Generate summary}
\input{./Chapter6/Figures/figure-Generate-summary}
\label{fig:6-64}
\end{figure}
%---------------------------
......@@ -1763,7 +1763,7 @@ Transformer Deep(48层) & 30.2 & 43.1 & 194$\times 10^{6}$
% 图3.6.1
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Example of automatic translation of classical Chinese}
\input{./Chapter6/Figures/figure-Example-of-automatic-translation-of-classical-Chinese}
\caption{文言文自动翻译实例}
\label{fig:6-56}
\end{figure}
......@@ -1779,7 +1779,7 @@ Transformer Deep(48层) & 30.2 & 43.1 & 194$\times 10^{6}$
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Automatically generate instances of couplets}
\input{./Chapter6/Figures/figure-Automatically-generate-instances-of-couplets}
\caption{对联自动生成实例（人工给定上联）}
\label{fig:6-57}
\end{figure}
......@@ -1795,7 +1795,7 @@ Transformer Deep(48层) & 30.2 & 43.1 & 194$\times 10^{6}$
\begin{figure}[htp]
\centering
\input{./Chapter6/Figures/figure-Automatic generation of ancient poems based on encoder-decoder framework}
\input{./Chapter6/Figures/figure-Automatic-generation-of-ancient-poems-based-on-encoder-decoder-framework}
\caption{基于编码器-解码器框架的古诗自动生成}
\label{fig:6-58}
\end{figure}
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论