Skip to content

T
Toy-MT-Introduction
Commit bae0bc94 by xiaotong
Options

new pages

parent 6b043d25
正在显示 包含 318 行增加 和 210 行删除
Section04-Phrasal-and-Syntactic-Models/section04-test.tex
...@@ -140,216 +140,6 @@ ...@@ -140,216 +140,6 @@
\section{基于语言学句法的模型} \section{基于语言学句法的模型}
%%%------------------------------------------------------------------------------------------------------------ %%%------------------------------------------------------------------------------------------------------------
%%% 句法树(层次短语)
\begin{frame}{层次短语规则推导的句法树表示}
\begin{itemize}
\item 层次短语规则的使用本质上是对(源语)句子建立树状结构，这反应了文法对于字符串分析得到的推导\\
\vspace{-0.4em}
\begin{displaymath}
\visible<2->{d = r_3}\visible<3->{\circ r_1}\visible<4->{ \circ r_4}\visible<5->{ \circ r_2}\visible<6->{ \circ r_5}\visible<7->{ \circ r_2}\visible<8->{ \circ r_7}\visible<9->{ \circ r_6}\visible<10->{ \circ r_2}
\end{displaymath}
\end{itemize}
\vspace{-1em}
\begin{center}
\begin{tikzpicture}
{\scriptsize
\begin{scope}[sibling distance=0pt, level distance = 27pt]
{\scriptsize
\Tree[.\node(n1){\textbf{S}};
[.\node(n2){\textbf{S}};
[.\node(n3){\textbf{S}};
[.\node(n4){\textbf{S}};
[.\node(n5){\textbf{X}}; \node(cw1){但}; ]
]
[.\node(n6){\textbf{X}}; \node(cw2){美国}; ]
]
[.\node(n7){\textbf{X}};
[. \node(cw3){并没有}; ]
[. \node(cw4){执行}; ]
]
]
[.\node(n8){\textbf{X}};
[. \node(cw5){世贸}; ]
[.\node(n9){\textbf{X}};
[. \node(cw6){组织}; ]
[. \node(cw7){的}; ]
]
[. \node(cw8){裁决}; ]
]
]
\node<10->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = west] (l1) at ([xshift=-0.3em]n1.north east) {{\color{white} \tiny{$r_2$}}};
\node<7->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = east] (l2) at ([xshift=0.3em]n2.north west) {{\color{white} \tiny{$r_2$}}};
\node<5->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = east] (l2) at ([xshift=0.3em]n3.north west) {{\color{white} \tiny{$r_2$}}};
\node<3->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = east] (l2) at ([xshift=0.3em]n4.north west) {{\color{white} \tiny{$r_1$}}};
\visible<2->{\node[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = east] (l2) at ([xshift=0.3em]n5.north west) {{\color{white} \tiny{$r_3$}}};}
\node<4->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = west] (l1) at ([xshift=-0.3em]n6.south east) {{\color{white} \tiny{$r_4$}}};
\node<6->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = west] (l1) at ([xshift=-0.3em]n7.north east) {{\color{white} \tiny{$r_5$}}};
\node<9->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = west] (l1) at ([xshift=-0.3em]n8.north east) {{\color{white} \tiny{$r_6$}}};
\node<8->[circle, inner sep = 0.5pt, fill=blue!90!white, anchor = west] (l1) at ([xshift=-0.3em]n9.north east) {{\color{white} \tiny{$r_7$}}};
}
\end{scope}
\begin{scope}[xshift = 1.8in, yshift = 0.1in]
\node (rules) {\textbf{层次短语翻译规则:}};
\draw[-] (rules.south west)--([xshift=1.8in]rules.south west);
\node[anchor=north west] (r1) at ([yshift=-0.2em]rules.south west) {$r_1$};
\node[anchor=west] (rc1) at ([xshift=0.0em]r1.east) {$\textrm{S} \; \to \; \langle\ \textrm{X}_1, \; \; \textrm{X}_1\ \rangle$};
\node[anchor=north west] (r2) at ([yshift=-0.4em]r1.south west) {$r_2$};
\node[anchor=west] (rc2) at ([xshift=0em]r2.east) {$\textrm{S} \; \to \; \langle\ \textrm{S}_1 \; \textrm{X}_2, \; \; \textrm{S}_1 \; \textrm{X}_2\ \rangle$};
\node[anchor=north west] (r3) at ([yshift=-0.4em]r2.south west) {$r_3$};
\node[anchor=west] (rc3) at ([xshift=0em]r3.east) {$\textrm{X} \; \to \; \langle\ \text{但}, \; \; \text{but}\ \rangle$};
\node[anchor=north west] (r4) at ([yshift=-0.4em]r3.south west) {$r_4$};
\node[anchor=west] (rc4) at ([xshift=0em]r4.east) {$\textrm{X} \; \to \; \langle\ \text{美国}, \; \; \text{the U.S.}\ \rangle$};
\node[anchor=north west] (r5) at ([yshift=-0.4em]r4.south west) {$r_5$};
\node[anchor=west] (rc5) at ([xshift=0em]r5.east) {$\textrm{X} \; \to \; \langle\ \text{并没有} \; \text{执行}, \; \; \text{}$};
\node[anchor=north west] (r52) at ([yshift=-0.4em]r5.south west) {{\color{white} $r_5$}};
\node[anchor=west] (rc52) at ([xshift=2.9em]r52.east) {$\text{has not implemented}\ \rangle$};
\node[anchor=north west] (r6) at ([yshift=-0.4em]r52.south west) {$r_6$};
\node[anchor=west] (rc6) at ([xshift=0em]r6.east) {$\textrm{X} \; \to \; \langle\ \text{世贸} \; \textrm{X}_1 \; \text{裁决}, $};
\node[anchor=north west] (r61) at ([yshift=-0.4em]r6.south west) {{\color{white} $r_6$}};
\node[anchor=west] (rc61) at ([xshift=2.9em]r61.east) {$\text{the decision} \; \textrm{X}_1 \; \text{the WTO}\ \rangle$};
\node[anchor=north west] (r7) at ([yshift=-0.4em]r61.south west) {$r_7$};
\node[anchor=west] (rc7) at ([xshift=0em]r7.east) {$\textrm{X} \; \to \; \langle\ \text{组织 的}, \; \; \text{of}\ \rangle$};
\end{scope}
}
\end{tikzpicture}
\end{center}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 句法树(语言学)
\begin{frame}{语言学句法树}
\begin{itemize}
\item x
\end{itemize}
\begin{center}
\begin{tikzpicture}
\begin{scope}[xshift = -0.3in, sibling distance=18pt, level distance = 27pt]
\Tree[.IP
[.ADVP
[.AD \node(cw1){但}; ]
]
[.NP
[.NR \node(cw2){美国}; ]
]
[.VP
[.ADVP
[.AD \node(cw3){并没有}; ]
]
[.VP
[.VP
[.VV \node(cw4){执行}; ]
]
[.NP
[.NP
[.NN \node(cw5){世贸}; ]
[.NN \node(cw6){组织}; ]
]
[.DEC \node(cw7){的}; ]
[.NP
[.NN \node(cw8){裁决}; ]
]
]
]
]
]
\node (labela) at (1in,-2.5in) {(a) The CTB-style parse tree of the input sentence};
\end{scope}
\end{tikzpicture}
\end{center}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 使用句法信息的一些思考
\begin{frame}{使用句法信息的一些思考}
\begin{itemize}
\item \textbf{句法树形式选择} - 句法的表示形式有很多，侧重点也不同，机器翻译需要何种句法信息？
\item \textbf{如何获取句法知识} - 句法信息由谁提供？
\begin{itemize}
\item 自动句法分析器可以提供句法信息，分析器可以从树库中自动学习(回忆第二章)，因此得到的句法分析结果和树库标注是一致的
\item 不使用句法分析器，而是让机器翻译自动学习适合翻译任务的句法结构
\end{itemize}
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 在MT中使用句法带来的好处
\begin{frame}{使用句法的好处}
\begin{itemize}
\item 这里我们假设：
\begin{itemize}
\item 使用短语结构树表示句法信息
\item 句法树由句法分析器自动生成
\end{itemize}
\item 短语结构树可以仍容易进行远距离调序的建模，同时丰富的句法功能标记有助于翻译中的消歧和生成
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 句法树中的一些基本概念
\begin{frame}{基本概念}
\begin{itemize}
\item 一些基本概念
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 句法模型的分类
\begin{frame}{句法模型的分类}
\begin{itemize}
\item 句法模型分类
\item 各种变形
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 对比
\begin{frame}{句法模型对比}
\begin{tabular}{l | l | l | l | l}
& 形式句法 & \multicolumn{3}{c}{语言学句法} \\ \cline{3-5}
& & 树到串 & 串到树 & 树到树 \\ \hline
源语句法 & No & Yes & No & Yes \\
目标语句法 & No & No & Yes & Yes \\
基于串的解码 & Yes & No & Yes & Yes \\
基于树的解码 & No & Yes & No & Yes \\
健壮性 & High & Mid & Mid & Low
\end{tabular}
\begin{itemize}
\item 下面，以树到串/串到树为例进行介绍，一些代表性论文\\
\begin{itemize}
\item \textbf{What's in a translation rule?}\\
\textbf{Galley et al., 2004, In Proc. of HLT-NAACL}
\item \textbf{Scalable Inference and Training of Context-Rich Syntactic Translation Models}\\
\textbf{Galley et al., 2006, In Proc. of ACL}
\item \textbf{Tree-to-String Alignment Template for Statistical Machine Translation}\\
\textbf{Liu et al., 2006, In Proc. of ACL}
\end{itemize}
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
\subsection{基于树结构的文法} \subsection{基于树结构的文法}
%%%------------------------------------------------------------------------------------------------------------ %%%------------------------------------------------------------------------------------------------------------
......
Section04-Phrasal-and-Syntactic-Models/section04.tex
...@@ -1987,6 +1987,324 @@ d = r_1 \circ r_2 \circ r_3 \circ r_4 ...@@ -1987,6 +1987,324 @@ d = r_1 \circ r_2 \circ r_3 \circ r_4
\end{frame} \end{frame}
%%%------------------------------------------------------------------------------------------------------------ %%%------------------------------------------------------------------------------------------------------------
%%% 句法树(语言学)
\begin{frame}{语言学句法树}
\begin{itemize}
\item 但是，层次短语的分析过程还是一种形式文法的推导，另一种自然的想法是把句子表示成语言学上的句法树，这种表示更符合人类对语言的认知
\begin{itemize}
\item 比如，``执行''后面接名词短语NP
\item 还有，``并没有''作用于后面的动词短语``执行...''
\end{itemize}
\end{itemize}
\begin{center}
\begin{tikzpicture}
\begin{scope}[xshift = -0.3in, sibling distance=3pt, level distance = 24pt]
\Tree[.IP
[.ADVP
[.AD \node(cw1){但}; ]
]
[.NP
[.NR \node(cw2){美国}; ]
]
[.VP
[.ADVP
[.AD \node(cw3){并没有}; ]
]
[.VP
[.VP
[.VV \node(cw4){执行}; ]
]
[.NP
[.NP
[.NN \node(cw5){世贸}; ]
[.NN \node(cw6){组织}; ]
]
[.DEC \node(cw7){的}; ]
[.NP
[.NN \node(cw8){裁决}; ]
]
]
]
]
]
\end{scope}
\end{tikzpicture}
\end{center}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 在MT中使用句法带来的好处
\begin{frame}{使用句法的好处}
\begin{itemize}
\item 短语结构树可以仍容易进行远距离调序的建模，同时丰富的句法功能标记有助于翻译中的消歧和生成
\end{itemize}
\vspace{-1.0em}
\begin{center}
\begin{tikzpicture}
\begin{scope}[yshift=-0in]
{\tiny
\node[anchor=west] (ref) at (0,0) {\textbf{参考答案:} The Chinese star performance troupe presented a wonderful Peking opera as well as singing and dancing };
\node[anchor=north west] (ref2) at (ref.south west) {{\color{white} \textbf{Reference:}} performance to Hong Kong audience .};
\node[anchor=north west] (hifst) at (ref2.south west) {\textbf{层次短语系统:} Star troupe of China, highlights of Peking opera and dance show to the audience of Hong Kong .};
\node[anchor=north west] (synhifst) at (hifst.south west) {\textbf{句法系统:} Chinese star troupe};
\node[anchor=west, fill=green!20!white, inner sep=0.25em] (synhifstpart2) at (synhifst.east) {presented};
\node[anchor=west, fill=blue!20!white, inner sep=0.25em] (synhifstpart3) at ([xshift=0.2em]synhifstpart2.east) {a wonderful Peking opera singing and dancing};
\node[anchor=west, fill=red!20!white, inner sep=0.40em] (synhifstpart4) at ([xshift=0.2em]synhifstpart3.east) {to};
\node[anchor=west, fill=purple!20!white, inner sep=0.25em] (synhifstpart5) at ([xshift=0.2em]synhifstpart4.east) {Hong Kong audience};
\node[anchor=west] (synhifstpart6) at (synhifstpart5.east) {.};
\node[anchor=north west] (input) at ([yshift=-10em]synhifst.south west) {\textbf{源语句法树:}};
\node<2->[anchor=north west] (rule) at ([yshift=-2em]synhifst.south west) {句法翻译规则:};
\node<2->[anchor=north west] (rule2) at ([yshift=0.2em]rule.south west) {(VP BA(将) $x_1$:NP $x_2$:VP PP(P(给) $x_3$:NP))};
\node<2->[anchor=north west] (rule3) at ([yshift=0.2em]rule2.south west) {$\to$ $x_2$ $x_1$ to $x_3$ };
\begin{scope}[scale = 0.9, grow'=up, sibling distance=15pt, level distance=23pt, xshift=2.8in, yshift=-2.8in]
\Tree[.\node(tn1){IP};
[.\node(tn2){NP}; \edge[roof]; \node[](seg1){中国$_1$ 明星$_2$ 艺术团$_3$}; ]
[.\node(tn3){VP};
[.\node(tn4){BA}; \node[fill=red!20!white](seg2){将$_4$}; ]
[.\node(tn5){NP}; \edge[roof]; \node[fill=blue!20!white](seg3){一$_5$ 台$_6$ 精彩$_7$ 的$_8$ 京剧$_9$ 歌舞$_{10}$}; ]
[.\node(tn6){VP};
[.\node(tn7){VV}; \node[fill=green!20!white](seg4){呈现$_{11}$}; ]
]
[.\node(tn8){PP};
[.\node(tn9){P}; \node[fill=red!20!white](seg5){给$_{12}$}; ]
[.\node(tn10){NP}; \edge[roof]; \node[fill=purple!20!white](seg6){香港$_{13}$ 观众$_{14}$}; ]
]
]
[.\node(tn11){.}; ]
]
\end{scope}
\path<2-> [draw,dotted,thick,blue] (tn3.south west) -- (tn4.south west) -- (seg2.north west) -- (seg2.north east) -- ([xshift=0.1em]tn4.north east) -- ([xshift=2.8em]tn6.north east) -- (seg5.north west) -- (seg5.north east) -- ([yshift=-1.7em]seg5.south east) -- (tn10.north east) -- (tn10.south east) -- (tn8.south east) -- (tn3.south east) -- (tn3.south west);
\path [draw,thick,->,dashed] (seg2.north) .. controls +(north:1.0) and +(south:1.5) .. (synhifstpart4.south);
\path [draw,thick,->,dashed] (seg3.north) -- (synhifstpart3.south);
\path [draw,thick,->,dashed] (seg4.north) -- (synhifstpart2.south);
\path [draw,thick,->,dashed] (seg5.north) .. controls +(north:0.5) .. (synhifstpart4.south);
\path [draw,thick,->,dashed] (seg6.north) -- (synhifstpart5.south);
\path<2-> [draw,<->,dotted,blue,thick] ([yshift=-0.5em]seg2.south west) .. controls +(west:1.0) and +(south:0.5) .. ([xshift=-1.5em]rule2.south);
}
\end{scope}
\end{tikzpicture}
\end{center}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 使用句法信息的一些思考
\begin{frame}{有很多句法知识可以使用}
\begin{itemize}
\item \textbf{句法树形式选择} - 句法的表示形式有很多，侧重点也不同，机器翻译需要何种句法信息？
\begin{center}
\begin{tikzpicture}
{\footnotesize
\begin{scope}[xshift = -0.3in, sibling distance=3pt, level distance = 18pt]
\Tree[.S
[.NP
[.NNP \node(w1){Messi}; ]
]
[.VP
[.VBD \node(w2){hit}; ]
[.NP
[.DT \node(w3){the}; ]
[.NN \node(w4){ball}; ]
]
]
]
\node [anchor=north west] (cap1) at ([yshift=-2.0em,xshift=2.0em]w1.south west) {\footnotesize{(a) 短语结构树}};
\node [anchor=west] (t1) at ([xshift=3em,]w4.east) {Messi};
\node [anchor=west] (t2) at ([xshift=0.5em,]t1.east) {hit};
\node [anchor=west] (t3) at ([xshift=0.5em,]t2.east) {the};
\node [anchor=west] (t4) at ([xshift=0.5em,]t3.east) {ball};
\draw [->] ([xshift=0em]t3.north) .. controls +(north:1em) and +(north:1em) .. ([xshift=-0.2em]t4.north);
\draw [->] ([xshift=0.2em]t4.north) .. controls +(north:2.5em) and +(north:2.5em) .. ([xshift=0.2em]t2.north);
\draw [->] ([xshift=0.0em]t1.north) .. controls +(north:2.5em) and +(north:2.5em) .. ([xshift=-0.2em]t2.north);
\node [anchor=north west] (cap2) at ([yshift=-0.2em,xshift=-0.5em]t2.south west) {\footnotesize{(b) 依存树}};
\end{scope}
}
\end{tikzpicture}
\end{center}
\item<2-> \textbf{如何获取句法知识} - 句法信息由谁提供？
\begin{itemize}
\item 自动句法分析器可以提供句法信息，分析器可以从树库中自动学习(回忆第二章)，因此得到的句法分析结果和树库标注是一致的
\item 不使用句法分析器，而是让机器翻译自动学习适合翻译任务的句法结构
\end{itemize}
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 句法树中的一些基本概念
\begin{frame}{术语}
\begin{itemize}
\item 这里我们\alert{假设}：
\begin{itemize}
\item 使用短语结构树表示句法信息
\item 句法树由句法分析器自动生成
\end{itemize}
\item 涉及的术语
\end{itemize}
\vspace{-0.8em}
{\small
\begin{center}
\begin{tabular}{l | l}
术语 & 说明 \\ \hline
翻译规则 & 翻译的最小单元(或步骤) \\ \hline
推导 & 由一系列规则组成的分析或翻译过程，推导可以 \\
& 被看做是规则的序列 \\ \hline
规则表 & 翻译规则的存储表示形式，可以高效进行查询\\ \hline
层次短语模型 & 基于同步上下文无关文法的翻译模型，非终结符\\
& 只有S和X两种，规则和文法并不需要符合语言学\\
& 句法约束\\ \hline
树到串模型 & 一类翻译模型，它使用源语语言学句法树，因此\\
& 翻译可以被看做从一棵句法树到词串的转换\\ \hline
串到树模型 & 一类翻译模型，它使用目标语语言学句法树，因\\
& 此翻译可以被看做从词串到句法树的转换\\
\end{tabular}
\end{center}
}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 句法树中的一些基本概念(续)
\begin{frame}{术语(续)}
\vspace{-0.5em}
{\small
\begin{center}
\begin{tabular}{l | l}
术语 & 说明 \\ \hline
树到树模型 & 一类翻译模型，它同时使用源语和目标语语言学\\
& 句法树，因此此翻译可以被看做从句法树到句法\\
& 树的转换 \\ \hline
基于句法 & 使用语言学句法 \\ \hline
基于树 & (源语言)使用树结构(大多指句法树)\\ \hline
基于串 & (源语言)使用词串，比如串到树的翻译系统的\\
& 解码器一般都是基于串的解码方法 \\ \hline
基于森林 & (源语言)使用句法森林，这里森林只是对多个\\
& 句法树的一种压缩表示 \\ \hline
分析 & 给定一个词串，使用文法规则得到推导 \\ \hline
树分析 & 给定一棵句法树，使用文法规则得到推导 \\ \hline
句法约软束 & 不强制规则推导匹配语言学句法树，通常把 \\
& 句法信息作为特征使用 \\ \hline
句法硬约束 & 强制推导必须符合语言学句法树，不符合的 \\
& 推导会被过滤掉
\end{tabular}
\end{center}
}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 句法模型的分类
\begin{frame}{句法模型的分类}
\begin{center}
\begin{tikzpicture}
\begin{scope}
\tikzstyle{cnode} = [minimum width=7.0em,minimum height=2.5em,rounded corners=0.2em];
\tikzstyle{xnode} = [minimum width=4.5em,minimum height=2.5em,rounded corners=0.2em];
\node[xnode,anchor=west,fill=red!30,align=left] (itg) at (0,0) {\footnotesize{反向转录}\\\footnotesize{文法}};
\node[xnode,anchor=west,fill=red!30,align=left] (hiero) at ([xshift=0.5em]itg.east) {\footnotesize{层次短语}\\\footnotesize{模型}};
\node[xnode,anchor=west,fill=blue!30,align=left] (s2t) at ([xshift=0.5em]hiero.east) {\footnotesize{串到树}\\\footnotesize{模型}};
\node[xnode,anchor=west,fill=blue!30,align=left] (t2s) at ([xshift=0.5em]s2t.east) {\footnotesize{树到串}\\\footnotesize{模型}};
\node[xnode,anchor=west,fill=blue!30,align=left] (t2t) at ([xshift=0.5em]t2s.east) {\footnotesize{树到树}\\\footnotesize{模型}};
\node[cnode,anchor=south,fill=red!30,align=left] (cat1) at ([xshift=-0.2em,yshift=2em]hiero.north west) {\footnotesize{基于形式文法}\\\footnotesize{的模型}};
\node[cnode,anchor=south,fill=blue!30,align=left] (cat2) at ([xshift=-0.0em,yshift=2em]t2s.north) {\footnotesize{基于语言学}\\\footnotesize{句法的模型}};
\node[cnode,anchor=south,minimum width=10.0em,fill=green!30,align=left] (cat0) at ([xshift=-3em,yshift=2em]cat2.north west) {\footnotesize{广义上基于}\\\footnotesize{句法的模型}};
\draw [-,thick] ([yshift=0.1em,xshift=1em]cat1.north) -- ([xshift=-1.5em,yshift=-0.1em]cat0.south);
\draw [-,thick] ([yshift=0.1em,xshift=-1em]cat2.north) -- ([xshift=1.5em,yshift=-0.1em]cat0.south);
\draw [-,thick] ([yshift=0.1em]itg.north) -- ([xshift=-0.5em,yshift=-0.1em]cat1.south);
\draw [-,thick] ([yshift=0.1em]hiero.north) -- ([xshift=0.5em,yshift=-0.1em]cat1.south);
\draw [-,thick] ([yshift=0.1em]s2t.north) -- ([xshift=-0.8em,yshift=-0.1em]cat2.south);
\draw [-,thick] ([yshift=0.1em]t2s.north) -- ([xshift=-0.0em,yshift=-0.1em]cat2.south);
\draw [-,thick] ([yshift=0.1em]t2t.north) -- ([xshift=0.8em,yshift=-0.1em]cat2.south);
\node [anchor=north] (itglabel) at (itg.south) {\scriptsize{(Wu, 1995)}};
\node [anchor=north] (hierolabel) at (hiero.south) {\scriptsize{(Chiang, 2005)}};
\node [anchor=north,align=left] (s2tlabel) at (s2t.south) {\scriptsize{(Galley et al.,}\\\scriptsize{\ 2004; 2006)}};
\node [anchor=north,align=left] (t2slabel) at (t2s.south) {\scriptsize{(Liu et al.,}\\\scriptsize{\ 2006)}};
\node [anchor=north,align=left] (t2tlabel) at (t2t.south) {\scriptsize{(Eisner, 2003)}};
\end{scope}
\end{tikzpicture}
\end{center}
\vspace{-0.5em}
\begin{itemize}
\item 实际上，上面仅仅只是一种分类方法，还有很多其它分类标准，比如：句法软约束 vs 句法硬约束，基于树 vs. 基于串，等等
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 对比
\begin{frame}{句法模型对比}
\begin{tabular}{l | l | l | l | l}
& 形式句法 & \multicolumn{3}{c}{语言学句法} \\ \cline{3-5}
& & 树到串 & 串到树 & 树到树 \\ \hline
源语句法 & No & Yes & No & Yes \\
目标语句法 & No & No & Yes & Yes \\
基于串的解码 & Yes & No & Yes & Yes \\
基于树的解码 & No & Yes & No & Yes \\
健壮性 & High & Mid & Mid & Low
\end{tabular}
\vspace{0.3em}
\begin{itemize}
\item<2-> 下面，以树到串/串到树为例进行介绍，一些代表性论文\\
\begin{itemize}
\item \textbf{What's in a translation rule?}\\
\textbf{Galley et al., 2004, In Proc. of HLT-NAACL}
\item \textbf{Scalable Inference and Training of Context-Rich Syntactic Translation Models}\\
\textbf{Galley et al., 2006, In Proc. of ACL}
\item \textbf{Tree-to-String Alignment Template for Statistical Machine Translation}\\
\textbf{Liu et al., 2006, In Proc. of ACL}
\end{itemize}
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
\subsection{基于树结构的文法} \subsection{基于树结构的文法}
%%%------------------------------------------------------------------------------------------------------------ %%%------------------------------------------------------------------------------------------------------------
......
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