Commit c0472484 by 曹润柘

合并分支 'caorunzhe' 到 'master'

Caorunzhe

查看合并请求 !218
parents fd26a251 bc2be173
...@@ -747,11 +747,11 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll} ...@@ -747,11 +747,11 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll}
\parinterval 在使用语言模型时,往往需要知道模型的质量。{\small\sffamily\bfseries{困惑度}}\index{困惑度}(Perplexity\index{Perplexity},PPL)是一种衡量语言模型的好坏的指标。对于一个真实的词序列$ w_1\dots w_m $,困惑度被定义为: \parinterval 在使用语言模型时,往往需要知道模型的质量。{\small\sffamily\bfseries{困惑度}}\index{困惑度}(Perplexity\index{Perplexity},PPL)是一种衡量语言模型的好坏的指标。对于一个真实的词序列$ w_1\dots w_m $,困惑度被定义为:
\begin{eqnarray} \begin{eqnarray}
{\rm{PPL}}&=&{\rm P}{(w_1\dots w_m)}^{- \frac{1}{m}} {\rm{PPL}}&=&\funp{P}{(w_1\dots w_m)}^{- \frac{1}{m}}
\label{eq:5-65} \label{eq:5-65}
\end{eqnarray} \end{eqnarray}
\parinterval 本质上,PPL反映了语言模型对序列可能性预测能力的一种评估。如果$ w_1\dots w_m $\\是真实的自然语言,``完美''的模型会得到$ {\rm P} (w_1\dots w_m)=1 $,它对应了最低的困惑度$ {\rm{PPL}}=1$,这说明模型可以完美地对词序列出现的可能性进行预测。当然,真实的语言模型是无法达到$ {\rm{PPL}}=1$的,比如,在著名的Penn Treebank(PTB)数据上最好的语言模型的PPL值也只能到达35左右。可见自然语言处理任务的困难程度。 \parinterval 本质上,PPL反映了语言模型对序列可能性预测能力的一种评估。如果$ w_1\dots w_m $\\是真实的自然语言,``完美''的模型会得到$ \funp{P}(w_1\dots w_m)=1 $,它对应了最低的困惑度PPL=1,这说明模型可以完美地对词序列出现的可能性进行预测。当然,真实的语言模型是无法达到PPL=1的,比如,在著名的Penn Treebank(PTB)数据上最好的语言模型的PPL值也只能到达35左右。可见自然语言处理任务的困难程度。
%---------------------------------------------------------------------------------------- %----------------------------------------------------------------------------------------
% NEW SECTION % NEW SECTION
...@@ -800,7 +800,7 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll} ...@@ -800,7 +800,7 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll}
\parinterval 从词序列建模的角度看,这两类预测问题本质上是一样的。因为,它们都在使用语言模型对词序列进行概率评估。但是,从实现上看,词序列的生成问题更难。因为,它不仅要对所有可能的词序列进行打分,同时要“找到”最好的词序列。由于潜在的词序列不计其数,因此这个“找”最优词序列的过程并不简单。 \parinterval 从词序列建模的角度看,这两类预测问题本质上是一样的。因为,它们都在使用语言模型对词序列进行概率评估。但是,从实现上看,词序列的生成问题更难。因为,它不仅要对所有可能的词序列进行打分,同时要“找到”最好的词序列。由于潜在的词序列不计其数,因此这个“找”最优词序列的过程并不简单。
\parinterval 实际上,生成最优词序列的问题也对应着自然语言处理中的一大类问题\ \dash\ {\small\bfnew{序列生成}}\index{序列生成}(Sequence Generation)\index{Sequence Generation}。机器翻译就是一个非常典型的序列生成问题:在机器翻译任务中,需要根据源语言词序列生成与之相对应的目标语言词序列。但是语言模型本身并不能“制造”单词序列的。因此,严格地说,序列生成问题的本质并非让语言模型凭空“生成”序列,而是使用语言模型在所有候选的单词序列中“找出”最佳序列。这个过程对应着经典的{\small\bfnew{搜索问题}}\index{搜索问题}(Search Problem)\index{Search Problem}。下面将着重介绍序列生成背后的建模方法,以及在序列生成里常用的搜索技术。 \parinterval 实际上,生成最优词序列的问题也自然语言处理中的一大类问题\ \dash\ {\small\bfnew{序列生成}}\index{序列生成}(Sequence Generation)\index{Sequence Generation}。机器翻译就是一个非常典型的序列生成问题:在机器翻译任务中,需要根据源语言词序列生成与之相对应的目标语言词序列。但是语言模型本身并不能“制造”单词序列的。因此,严格地说,序列生成问题的本质并非让语言模型凭空“生成”序列,而是使用语言模型在所有候选的单词序列中“找出”最佳序列。这个过程对应着经典的{\small\bfnew{搜索问题}}\index{搜索问题}(Search Problem)\index{Search Problem}。下面将着重介绍序列生成背后的建模方法,以及在序列生成里常用的搜索技术。
%---------------------------------------------------------------------------------------- %----------------------------------------------------------------------------------------
% NEW SUB-SECTION % NEW SUB-SECTION
...@@ -878,7 +878,7 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll} ...@@ -878,7 +878,7 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll}
}\end{table} }\end{table}
%------------------------------------------------------ %------------------------------------------------------
\parinterval 那么是否有比枚举策略更高效的方法呢?答案是肯定的。一种直观的方法是将搜索的过程表示成树型结构,称为解空间树。它包含了搜索过程中可生成的全部序列。该树的根节点恒为<sos>,代表序列均从<sos> 开始。该树结构中非叶子节点的兄弟节点有$|V|$个,由词表和结束符号<eos>构成。从图\ref{fig:2-14}可以看到,对于一个最大长度为4的序列的搜索过程,生成某个单词序列的过程实际上就是访问解空间树中从根节点<sos> 开始一直到叶子节点<eos>结束的某条路径,而这条的路径上节点按顺序组成了一段独特的单词序列。此时对所有可能单词序列的枚举就变成了对解空间树的遍历。并且枚举的过程与语言模型打分的过程也是一致的,每枚举一个词$i$也就是在上图选择$w_i$一列的一个节点,语言模型就可以为当前的树节点$w_i$给出一个分值,即$\funp{P}(w_i | w_1 w_2 \ldots w_{i-1})$。对于$n$-gram语言模型,这个分值$\funp{P}(w_i | w_1 w_2 \ldots w_{i-1})=\funp{P}(w_i | w_{i-n+1} \ldots w_{i-1})$ \parinterval 那么是否有比枚举策略更高效的方法呢?答案是肯定的。一种直观的方法是将搜索的过程表示成树型结构,称为解空间树。它包含了搜索过程中可生成的全部序列。该树的根节点恒为<sos>,代表序列均从<sos> 开始。该树结构中非叶子节点的兄弟节点有$|V|+1$个,由词表和结束符号<eos>构成。从图\ref{fig:2-14}可以看到,对于一个最大长度为4的序列的搜索过程,生成某个单词序列的过程实际上就是访问解空间树中从根节点<sos> 开始一直到叶子节点<eos>结束的某条路径,而这条的路径上节点按顺序组成了一段独特的单词序列。此时对所有可能单词序列的枚举就变成了对解空间树的遍历。并且枚举的过程与语言模型打分的过程也是一致的,每枚举一个词$i$也就是在上图选择$w_i$一列的一个节点,语言模型就可以为当前的树节点$w_i$给出一个分值,即$\funp{P}(w_i | w_1 w_2 \ldots w_{i-1})$。对于$n$-gram语言模型,这个分值可以表示为$\funp{P}(w_i | w_1 w_2 \ldots w_{i-1})=\funp{P}(w_i | w_{i-n+1} \ldots w_{i-1})$
%---------------------------------------------- %----------------------------------------------
\begin{figure}[htp] \begin{figure}[htp]
...@@ -1040,13 +1040,13 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll} ...@@ -1040,13 +1040,13 @@ c_{\textrm{KN}}(\cdot) = \left\{\begin{array}{ll}
\begin{adjustwidth}{1em}{} \begin{adjustwidth}{1em}{}
\begin{itemize} \begin{itemize}
\vspace{0.5em} \vspace{0.5em}
\item$n$-gram语言模型中,由于语料中往往存在大量的低频词以及未登录词,模型会产生不合理的概率预测结果。因此本章介绍了三种平滑方法,以解决上述问题。实际上,平滑方法是语言建模中的重要研究方向。除了上述三种方法之外,还有Jelinek–Mercer平滑\upcite{jelinek1980interpolated}、Katz 平滑\upcite{katz1987estimation}以及Witten–Bell平滑等等\upcite{bell1990text,witten1991the} 相关工作也对这些平滑方法进行了详细对比\upcite{chen1999empirical,goodman2001a} \item$n$-gram语言模型中,由于语料中往往存在大量的低频词以及未登录词,模型会产生不合理的概率预测结果。因此本章介绍了三种平滑方法,以解决上述问题。实际上,平滑方法是语言建模中的重要研究方向。除了上述三种方法之外,还有Jelinek–Mercer平滑\upcite{jelinek1980interpolated}、Katz 平滑\upcite{katz1987estimation}以及Witten–Bell平滑等等\upcite{bell1990text,witten1991the}。相关工作也对这些平滑方法进行了详细对比\upcite{chen1999empirical,goodman2001a}
\vspace{0.5em} \vspace{0.5em}
\item 除了平滑方法,也有很多工作对$n$-gram语言模型进行改进。比如,对于形态学丰富的语言,可以考虑对单词的形态学变化进行建模。这类语言模型在一些机器翻译系统中也体现出了很好的潜力\upcite{kirchhoff2005improved,sarikaya2007joint,koehn2007factored}。此外,如何使用超大规模数据进行语言模型训练也是备受关注的研究方向。比如,有研究者探索了对超大语言模型进行压缩和存储的方法\upcite{federico2007efficient,federico2006how,heafield2011kenlm}。另一个有趣的方向是,利用随机存储算法对大规模语言模型进行有效存储\upcite{talbot2007smoothed,talbot2007randomised},比如,在语言模型中使用Bloom\ Filter等随机存储的数据结构。 \item 除了平滑方法,也有很多工作对$n$-gram语言模型进行改进。比如,对于形态学丰富的语言,可以考虑对单词的形态学变化进行建模。这类语言模型在一些机器翻译系统中也体现出了很好的潜力\upcite{kirchhoff2005improved,sarikaya2007joint,koehn2007factored}。此外,如何使用超大规模数据进行语言模型训练也是备受关注的研究方向。比如,有研究者探索了对超大语言模型进行压缩和存储的方法\upcite{federico2007efficient,federico2006how,heafield2011kenlm}。另一个有趣的方向是,利用随机存储算法对大规模语言模型进行有效存储\upcite{talbot2007smoothed,talbot2007randomised},比如,在语言模型中使用Bloom\ Filter等随机存储的数据结构。
\vspace{0.5em} \vspace{0.5em}
\item 本章更多地关注了语言模型的基本问题和求解思路,但是基于$n$-gram的方法并不是语言建模的唯一方法。从现在自然语言处理的前沿看,端到端的深度学习方法在很多任务中都取得了领先的性能。语言模型同样可以使用这些方法\upcite{jing2019a},而且在近些年取得了巨大成功。例如,最早提出的前馈神经语言模型\upcite{bengio2003a}和后来的基于循环单元的语言模型\upcite{mikolov2010recurrent}、基于长短期记忆单元的语言模型\upcite{sundermeyer2012lstm}以及现在非常流行的Transformer\upcite{vaswani2017attention}。 关于神经语言模型的内容,会在{\chapternine}进行进一步介绍。 \item 本章更多地关注了语言模型的基本问题和求解思路,但是基于$n$-gram的方法并不是语言建模的唯一方法。从现在自然语言处理的前沿看,端到端的深度学习方法在很多任务中都取得了领先的性能。语言模型同样可以使用这些方法\upcite{jing2019a},而且在近些年取得了巨大成功。例如,最早提出的前馈神经语言模型\upcite{bengio2003a}和后来的基于循环单元的语言模型\upcite{mikolov2010recurrent}、基于长短期记忆单元的语言模型\upcite{sundermeyer2012lstm}以及现在非常流行的Transformer\upcite{vaswani2017attention}。 关于神经语言模型的内容,会在{\chapternine}进行进一步介绍。
\vspace{0.5em} \vspace{0.5em}
\item 最后,本章结合语言模型的序列生成任务对搜索技术进行了介绍。类似地,机器翻译任务也需要从大量的翻译候选中快速寻找最优译文。因此在机器翻译任务中也使用了搜索方法,这个过程通常被称作{\small\bfnew{解码}}\index{解码}(Decoding)\index{Decoding}。例如,有研究者在基于词的翻译模型中尝试使用启发式搜索\upcite{DBLP:conf/acl/OchUN01,DBLP:conf/acl/WangW97,tillmann1997a}以及贪婪搜索方法\upcite{germann2001fast}\upcite{germann2003greedy},也有研究者研究基于短语的栈解码方法\upcite{Koehn2007Moses,DBLP:conf/amta/Koehn04}。此外,解码方法还包括有限状态机解码\upcite{bangalore2001a}\upcite{DBLP:journals/mt/BangaloreR02}以及基于语言学约束的解码\upcite{venugopal2007an,zollmann2007the,liu2006tree,galley2006scalable,chiang2005a}。相关内容将在{\chaptereight}{\chapterfourteen} 进行介绍。 \item 最后,本章结合语言模型的序列生成任务对搜索技术进行了介绍。类似地,机器翻译任务也需要从大量的翻译候选中快速寻找最优译文。因此在机器翻译任务中也使用了搜索方法,这个过程通常被称作{\small\bfnew{解码}}\index{解码}(Decoding)\index{Decoding}。例如,有研究者在基于词的翻译模型中尝试使用启发式搜索\upcite{DBLP:conf/acl/OchUN01,DBLP:conf/acl/WangW97,tillmann1997a}以及贪婪搜索方法\upcite{germann2001fast}\upcite{germann2003greedy},也有研究者探索基于短语的栈解码方法\upcite{Koehn2007Moses,DBLP:conf/amta/Koehn04}。此外,解码方法还包括有限状态机解码\upcite{bangalore2001a}\upcite{DBLP:journals/mt/BangaloreR02}以及基于语言学约束的解码\upcite{venugopal2007an,zollmann2007the,liu2006tree,galley2006scalable,chiang2005a}。相关内容将在{\chaptereight}{\chapterfourteen}进行介绍。
\vspace{0.5em} \vspace{0.5em}
\end{itemize} \end{itemize}
\end{adjustwidth} \end{adjustwidth}
...@@ -213,7 +213,7 @@ $计算这种切分的概率值。 ...@@ -213,7 +213,7 @@ $计算这种切分的概率值。
\parinterval 以“确实现在数据很多”这个实例来说,如果把这句话按照“确实/现在/数据/很/多”这样的方式进行切分,这个句子切分的概率$\funp{P}$(确实/现在/数据/很/多) 可以通过每个词出现概率相乘的方式进行计算。 \parinterval 以“确实现在数据很多”这个实例来说,如果把这句话按照“确实/现在/数据/很/多”这样的方式进行切分,这个句子切分的概率$\funp{P}$(确实/现在/数据/很/多) 可以通过每个词出现概率相乘的方式进行计算。
\begin{eqnarray} \begin{eqnarray}
&\funp{P}&\textrm{(确实/现在/数据/很/多)} \nonumber \\ &\funp&{P}\textrm{(确实/现在/数据/很/多)} \nonumber \\
& = &\funp{P}\textrm{(确实)} \cdot \funp{P}\textrm{(现在)} \cdot \funp{P}\textrm{(数据)} \cdot \funp{P}\textrm{(很)} \cdot \funp{P}\textrm{(多)} & = &\funp{P}\textrm{(确实)} \cdot \funp{P}\textrm{(现在)} \cdot \funp{P}\textrm{(数据)} \cdot \funp{P}\textrm{(很)} \cdot \funp{P}\textrm{(多)}
\label{eq:3.2-1} \label{eq:3.2-1}
\end{eqnarray} \end{eqnarray}
......
\usetikzlibrary{shapes.geometric}
\begin{tikzpicture}
\node[font=\footnotesize] (overall) at (0,0){\small\bfnew{\ \ \ \ }};
\node[anchor=north,font=\footnotesize] (hypo) at ([yshift=-3em]overall.south){某种假设};
\coordinate (A) at ([yshift=-3.5em,xshift=-1.6em]overall);
\coordinate (B) at ([yshift=0.3em,xshift=0.3em]A);
\draw[] (B) .. controls +(east:1.3em) and +(west:0.3em) .. ([xshift=1.5em,yshift=2em]B) .. controls +(east:0.3em) and +(west:1.3em) .. ([xshift=3em]B);
\draw[<->] ([yshift=2.4em]A) -- (A) -- ([xshift=3.7em]A);
\begin{pgfonlayer}{background}
\node [draw,thick,rectangle,inner sep=0.5em,rounded corners=2pt,fill=red!15,drop shadow] [fit = (overall)(hypo)] (box1) {};
\end{pgfonlayer}
\node[draw,fill=yellow!15,thick,anchor=west,font=\footnotesize,align=center,drop shadow](sample) at ([xshift=4em]box1.east){样本\\观察结果};
\node[anchor=west,draw,diamond,fill=ugreen!15,drop shadow,aspect=2,font=\scriptsize,align=center,inner sep=1pt,thick] (judge) at ([xshift=3em]sample.east){小概率事件\\发生?};
\node[draw,fill=blue!10,thick,drop shadow,anchor=west,font=\footnotesize,align=center,thick](refuse) at ([xshift=6em]judge.north){拒绝原假设};
\node[draw,fill=blue!10,thick,drop shadow,anchor=west,font=\footnotesize,align=center,thick](accept) at ([xshift=6em]judge.south){接受原假设};
\draw[->,thick] (box1.east) -- node[above,font=\scriptsize]{抽样}(sample.west);
\draw[->,thick] (sample.east) -- node[above,font=\scriptsize]{检验}(judge.west);
\draw[->,thick] (judge.north) -- node[above,font=\scriptsize]{}(refuse.west);
\draw[->,thick] (judge.south) -- node[below,font=\scriptsize]{}(accept.west);
\end{tikzpicture}
\begin{tikzpicture}[scale=0.5] \begin{tikzpicture}[scale=0.5]
\tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!30] \tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!15]
\tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!30] \tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!15]
\node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small{机器译文:}};
\node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can}; \node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can};
\node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I}; \node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I};
\node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have}; \node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have};
\node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){this}; \node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){it};
\node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like}; \node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like};
\node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he}; \node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he};
\node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){do}; \node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){?};
\node[cand,anchor=west] (n19) at ([xshift=1em]n18.east){it};
\node[cand,anchor=west] (n20) at ([xshift=1em]n19.east){?};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small\bfnew{Reference :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small{参考答案:}};
\node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can}; \node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can};
\node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I}; \node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I};
\node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat}; \node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat};
\node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this}; \node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this};
\node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can}; \node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can};
\node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like}; \node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like};
\node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){he}; \node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){him};
\node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){did}; \node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){?};
\node[ref,anchor=west] (n60) at ([xshift=1em]n29.east){?};
\draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north); \draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north);
\draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north); \draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north);
\draw[line width=1.6pt,blue!40] (n15.south) -- (n25.north);
\draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north); \draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north);
\draw[line width=1.6pt,blue!40] (n17.south) -- (n28.north); \draw[line width=1.6pt,blue!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,blue!40] (n20.south) -- (n60.north);
\end{tikzpicture} \end{tikzpicture}
\begin{tikzpicture}[scale=0.5] \begin{tikzpicture}[scale=0.5]
\tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!30] \tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!15]
\tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!30] \tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!15]
\node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small{机器译文:}};
\node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can}; \node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can};
\node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I}; \node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I};
\node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have}; \node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have};
\node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){this}; \node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){it};
\node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like}; \node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like};
\node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he}; \node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he};
\node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){do}; \node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){?};
\node[cand,anchor=west] (n19) at ([xshift=1em]n18.east){it};
\node[cand,anchor=west] (n20) at ([xshift=1em]n19.east){?};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small\bfnew{Reference :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small{参考答案:}};
\node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can}; \node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can};
\node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I}; \node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I};
\node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat}; \node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat};
\node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this}; \node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this};
\node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can}; \node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can};
\node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like}; \node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like};
\node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){he}; \node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){him};
\node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){did}; \node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){?};
\node[ref,anchor=west] (n30) at ([xshift=1em]n29.east){?};
\draw[line width=1.6pt,blue!40] (n12.south) -- (n26.north); \draw[line width=1.6pt,blue!40] (n12.south) -- (n26.north);
\draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north); \draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north);
\draw[line width=1.6pt,blue!40] (n15.south) -- (n25.north); \draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north);
\draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north); \draw[line width=1.6pt,blue!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,blue!40] (n17.south) -- (n28.north);
\draw[line width=1.6pt,blue!40] (n20.south) -- (n30.north);
\end{tikzpicture} \end{tikzpicture}
\definecolor{ugreen}{rgb}{0,0.5,0} \definecolor{ugreen}{rgb}{0,0.5,0}
\begin{tikzpicture}[scale=0.5] \begin{tikzpicture}[scale=0.5]
\tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!30] \tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!15]
\tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!30] \tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!15]
\node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small{机器译文:}};
\node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can}; \node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can};
\node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I}; \node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I};
\node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have}; \node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have};
\node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){this}; \node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){it};
\node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like}; \node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like};
\node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he}; \node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he};
\node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){do}; \node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){?};
\node[cand,anchor=west] (n19) at ([xshift=1em]n18.east){it};
\node[cand,anchor=west] (n20) at ([xshift=1em]n19.east){?}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small{参考答案:}};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small\bfnew{Reference :}};
\node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can}; \node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can};
\node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I}; \node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I};
\node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat}; \node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat};
\node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this}; \node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this};
\node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can}; \node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can};
\node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like}; \node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like};
\node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){he}; \node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){him};
\node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){did}; \node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){?};
\node[ref,anchor=west] (n30) at ([xshift=1em]n29.east){?};
\draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north); \draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north);
\draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north); \draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north);
\draw[line width=1.6pt,blue!40] (n15.south) -- (n25.north);
\draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north); \draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north);
\draw[line width=1.6pt,blue!40] (n17.south) -- (n28.north); \draw[line width=1.6pt,blue!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,blue!40] (n20.south) -- (n30.north); \draw[line width=1.6pt,orange!40] (n17.south) -- (n28.north);
\draw[line width=1.6pt,orange!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,ugreen!40](n14.south) -- (n24.north); \draw[line width=1.6pt,ugreen!40](n14.south) -- (n24.north);
......
\begin{tikzpicture}[scale=0.5] \begin{tikzpicture}[scale=0.5]
\tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!30] \tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!15]
\tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!30] \tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!15]
\node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small{机器译文:}};
\node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can}; \node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can};
\node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I}; \node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I};
\node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have}; \node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have};
\node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){this}; \node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){it};
\node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like}; \node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like};
\node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he}; \node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he};
\node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){do}; \node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){?};
\node[cand,anchor=west] (n19) at ([xshift=1em]n18.east){it};
\node[cand,anchor=west] (n20) at ([xshift=1em]n19.east){?};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small\bfnew{Reference :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small{参考答案:}};
\node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can}; \node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can};
\node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I}; \node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I};
\node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat}; \node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat};
\node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this}; \node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this};
\node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can}; \node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can};
\node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like}; \node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like};
\node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){he}; \node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){him};
\node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){did}; \node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){?};
\node[ref,anchor=west] (n30) at ([xshift=1em]n29.east){?};
\node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n31) at ([yshift=-5em]n21.south){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n31) at ([yshift=-5em]n21.south){\small{机器译文:}};
\node[cand,anchor=west] (n32) at ([xshift=0.0em]n31.east){Can}; \node[cand,anchor=west] (n32) at ([xshift=0.0em]n31.east){Can};
\node[cand,anchor=west] (n33) at ([xshift=1em]n32.east){I}; \node[cand,anchor=west] (n33) at ([xshift=1em]n32.east){I};
\node[cand,anchor=west] (n34) at ([xshift=1em]n33.east){have}; \node[cand,anchor=west] (n34) at ([xshift=1em]n33.east){have};
\node[cand,anchor=west] (n35) at ([xshift=1em]n34.east){this}; \node[cand,anchor=west] (n35) at ([xshift=1em]n34.east){it};
\node[cand,anchor=west] (n36) at ([xshift=1em]n35.east){like}; \node[cand,anchor=west] (n36) at ([xshift=1em]n35.east){like};
\node[cand,anchor=west] (n37) at ([xshift=1em]n36.east){he}; \node[cand,anchor=west] (n37) at ([xshift=1em]n36.east){he};
\node[cand,anchor=west] (n38) at ([xshift=1em]n37.east){do}; \node[cand,anchor=west] (n38) at ([xshift=1em]n37.east){?};
\node[cand,anchor=west] (n39) at ([xshift=1em]n38.east){it};
\node[cand,anchor=west] (n40) at ([xshift=1em]n39.east){?}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n41) at ([yshift=-6em]n31.south){\small{参考答案:}};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n41) at ([yshift=-6em]n31.south){\small\bfnew{Reference :}};
\node[ref,anchor=west] (n42) at ([xshift=0.0em]n41.east){Can}; \node[ref,anchor=west] (n42) at ([xshift=0.0em]n41.east){Can};
\node[ref,anchor=west] (n43) at ([xshift=1em]n42.east){I}; \node[ref,anchor=west] (n43) at ([xshift=1em]n42.east){I};
\node[ref,anchor=west] (n44) at ([xshift=1em]n43.east){eat}; \node[ref,anchor=west] (n44) at ([xshift=1em]n43.east){eat};
\node[ref,anchor=west] (n45) at ([xshift=1em]n44.east){this}; \node[ref,anchor=west] (n45) at ([xshift=1em]n44.east){this};
\node[ref,anchor=west] (n46) at ([xshift=1em]n45.east){Can}; \node[ref,anchor=west] (n46) at ([xshift=1em]n45.east){Can};
\node[ref,anchor=west] (n47) at ([xshift=1em]n46.east){like}; \node[ref,anchor=west] (n47) at ([xshift=1em]n46.east){like};
\node[ref,anchor=west] (n48) at ([xshift=1em]n47.east){he}; \node[ref,anchor=west] (n48) at ([xshift=1em]n47.east){him};
\node[ref,anchor=west] (n49) at ([xshift=1em]n48.east){did}; \node[ref,anchor=west] (n49) at ([xshift=1em]n48.east){?};
\node[ref,anchor=west] (n50) at ([xshift=1em]n49.east){?};
\draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north); \draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north);
\draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north); \draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north);
\draw[line width=1.6pt,blue!40] (n15.south) -- (n25.north);
\draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north); \draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north);
\draw[line width=1.6pt,blue!40] (n17.south) -- (n28.north); \draw[line width=1.6pt,blue!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,blue!40] (n20.south) -- (n30.north); \draw[line width=2pt,orange!40] (n17.south) -- (n28.north);
\draw[line width=2pt,orange!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,blue!40] (n32.south) -- (n46.north); \draw[line width=1.6pt,blue!40] (n32.south) -- (n46.north);
\draw[line width=1.6pt,blue!40] (n33.south) -- (n43.north); \draw[line width=1.6pt,blue!40] (n33.south) -- (n43.north);
\draw[line width=1.6pt,blue!40] (n35.south) -- (n45.north);
\draw[line width=1.6pt,blue!40] (n36.south) -- (n47.north); \draw[line width=1.6pt,blue!40] (n36.south) -- (n47.north);
\draw[line width=1.6pt,blue!40] (n37.south) -- (n48.north); \draw[line width=1.6pt,blue!40] (n38.south) -- (n49.north);
\draw[line width=1.6pt,blue!40] (n40.south) -- (n50.north); \draw[line width=2pt,orange!40] (n37.south) -- (n48.north);
\draw[line width=2pt,orange!40] (n38.south) -- (n49.north);
\end{tikzpicture} \end{tikzpicture}
\ No newline at end of file
\definecolor{ugreen}{rgb}{0,0.5,0} \definecolor{ugreen}{rgb}{0,0.5,0}
\begin{tikzpicture}[scale=0.6] \begin{tikzpicture}[scale=0.6]
\tikzstyle{unit} = [draw,inner sep=3pt,font=\tiny,minimum height=1.2em] \tikzstyle{unit} = [draw,inner sep=3pt,font=\tiny,minimum height=1.2em,drop shadow={shadow xshift=0.1em,shadow yshift=-0.15em}]
\tikzstyle{bad_tag} = [fill=red!15,inner sep=1pt,align=center,font=\tiny,text=red] \tikzstyle{bad_tag} = [fill=red!15,inner sep=1pt,align=center,font=\tiny,text=red!80]
\tikzstyle{ok_tag} = [fill=ugreen!15,inner sep=1pt,align=center,font=\tiny,text=ugreen] \tikzstyle{ok_tag} = [fill=ugreen!15,inner sep=1pt,align=center,font=\tiny,text=ugreen!80]
\coordinate (o) at (0, 0); \coordinate (o) at (0, 0);
\node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (n1_1) at ([yshift=5.5em]o.east){\textbf{Source}}; \node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (n1_1) at ([yshift=5.5em]o.east){\textbf{Source}};
\node[unit,anchor=west,fill=green!20](n1_2) at ([xshift=7.6em]n1_1.east){The}; \node[unit,anchor=west,fill=green!20](n1_2) at ([xshift=7.6em]n1_1.east){Draw};
\node[unit,anchor=west,fill=green!20](n1_3) at ([xshift=0.8em]n1_2.east){Sharpen}; \node[unit,anchor=west,fill=green!20](n1_3) at ([xshift=0.8em]n1_2.east){or};
\node[unit,anchor=west,fill=green!20](n1_4) at ([xshift=0.8em]n1_3.east){tool}; \node[unit,anchor=west,fill=green!20](n1_4) at ([xshift=0.8em]n1_3.east){select};
\node[unit,anchor=west,fill=green!20](n1_5) at ([xshift=0.8em]n1_4.east){sharpens}; \node[unit,anchor=west,fill=green!20](n1_5) at ([xshift=0.8em]n1_4.east){a};
\node[unit,anchor=west,fill=green!20](n1_6) at ([xshift=0.8em]n1_5.east){areas}; \node[unit,anchor=west,fill=green!20](n1_6) at ([xshift=0.8em]n1_5.east){line};
\node[unit,anchor=west,fill=green!20](n1_7) at ([xshift=0.8em]n1_6.east){in}; \node[unit,anchor=west,fill=green!20](n1_7) at ([xshift=0.8em]n1_6.east){.};
\node[unit,anchor=west,fill=green!20](n1_8) at ([xshift=0.8em]n1_7.east){an};
\node[unit,anchor=west,fill=green!20](n1_9) at ([xshift=0.8em]n1_8.east){image};
\node[unit,anchor=west,fill=green!20](n1_10) at ([xshift=0.8em]n1_9.east){.};
\node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (n2_1) at (o.east){\textbf{PE}}; \node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (n2_1) at (o.east){\textbf{PE}};
\node[unit,anchor=west,fill=red!20](n2_2) at ([xshift=1em]n2_1.east){Mit}; \node[unit,anchor=west,fill=red!20](n2_2) at ([xshift=1em]n2_1.east){Zeichnen};
\node[unit,anchor=west,fill=red!20](n2_3) at ([xshift=0.8em]n2_2.east){dem}; \node[unit,anchor=west,fill=red!20](n2_3) at ([xshift=0.8em]n2_2.east){oder};
\node[unit,anchor=west,fill=red!20](n2_4) at ([xshift=0.8em]n2_3.east){Scharfzeichner}; \node[unit,anchor=west,fill=red!20](n2_4) at ([xshift=0.8em]n2_3.east){Sie};
\node[unit,anchor=west,fill=red!20](n2_5) at ([xshift=0.8em]n2_4.east){können}; \node[unit,anchor=west,fill=red!20](n2_5) at ([xshift=0.8em]n2_4.east){eine};
\node[unit,anchor=west,fill=red!20](n2_6) at ([xshift=0.8em]n2_5.east){Sie}; \node[unit,anchor=west,fill=red!20](n2_6) at ([xshift=0.8em]n2_5.east){linie};
\node[unit,anchor=west,fill=red!20](n2_7) at ([xshift=0.8em]n2_6.east){einzelne}; \node[unit,anchor=west,fill=red!20](n2_7) at ([xshift=0.8em]n2_6.east){,};
\node[unit,anchor=west,fill=red!20](n2_8) at ([xshift=0.8em]n2_7.east){Bereiche}; \node[unit,anchor=west,fill=red!20](n2_8) at ([xshift=0.8em]n2_7.east){order};
\node[unit,anchor=west,fill=red!20](n2_9) at ([xshift=0.8em]n2_8.east){in}; \node[unit,anchor=west,fill=red!20](n2_9) at ([xshift=0.8em]n2_8.east){wählen};
\node[unit,anchor=west,fill=red!20](n2_10) at ([xshift=0.8em]n2_9.east){einem}; \node[unit,anchor=west,fill=red!20](n2_10) at ([xshift=0.8em]n2_9.east){Sie};
\node[unit,anchor=west,fill=red!20](n2_11) at ([xshift=0.8em]n2_10.east){Bild}; \node[unit,anchor=west,fill=red!20](n2_11) at ([xshift=0.8em]n2_10.east){eine};
\node[unit,anchor=west,fill=red!20](n2_12) at ([xshift=0.8em]n2_11.east){scharfzeichnen}; \node[unit,anchor=west,fill=red!20](n2_12) at ([xshift=0.8em]n2_11.east){aus};
\node[unit,anchor=west,fill=red!20](n2_13) at ([xshift=0.8em]n2_12.east){.}; \node[unit,anchor=west,fill=red!20](n2_13) at ([xshift=0.8em]n2_12.east){.};
\node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (n3_1) at ([yshift=-5.5em]o.east){\textbf{MT}}; \node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (n3_1) at ([yshift=-5.5em]o.east){\textbf{MT}};
\node[unit,anchor=west,fill=blue!20](n3_2) at ([xshift=4.7em]n3_1.east){Der}; \node[unit,anchor=west,fill=blue!20](n3_2) at ([xshift=4.7em]n3_1.east){Zeichnen};
\node[unit,anchor=west,fill=blue!20](n3_3) at ([xshift=0.8em]n3_2.east){Schärfen-Werkezug}; \node[unit,anchor=west,fill=blue!20](n3_3) at ([xshift=0.8em]n3_2.east){oder};
\node[unit,anchor=west,fill=blue!20](n3_4) at ([xshift=0.8em]n3_3.east){Bereiche}; \node[unit,anchor=west,fill=blue!20](n3_4) at ([xshift=0.8em]n3_3.east){wählen};
\node[unit,anchor=west,fill=blue!20](n3_5) at ([xshift=0.8em]n3_4.east){in}; \node[unit,anchor=west,fill=blue!20](n3_5) at ([xshift=0.8em]n3_4.east){Sie};
\node[unit,anchor=west,fill=blue!20](n3_6) at ([xshift=0.8em]n3_5.east){einem}; \node[unit,anchor=west,fill=blue!20](n3_6) at ([xshift=0.8em]n3_5.east){eine};
\node[unit,anchor=west,fill=blue!20](n3_7) at ([xshift=0.8em]n3_6.east){Bild}; \node[unit,anchor=west,fill=blue!20](n3_7) at ([xshift=0.8em]n3_6.east){Linie};
\node[unit,anchor=west,fill=blue!20](n3_8) at ([xshift=0.8em]n3_7.east){Schärfer}; \node[unit,anchor=west,fill=blue!20](n3_8) at ([xshift=0.8em]n3_7.east){aus};
\node[unit,anchor=west,fill=blue!20](n3_9) at ([xshift=0.8em]n3_8.east){erscheint}; \node[unit,anchor=west,fill=blue!20](n3_9) at ([xshift=0.8em]n3_8.east){.};
\node[unit,anchor=west,fill=blue!20](n3_10) at ([xshift=0.8em]n3_9.east){.};
\node[bad_tag,anchor=south] at ([yshift=2pt]n1_2.north){BAD}; \node[bad_tag,anchor=south] at ([yshift=2pt]n1_2.north){BAD};
\node[bad_tag,anchor=south] at ([yshift=2pt]n1_3.north){BAD}; \node[bad_tag,anchor=south] at ([yshift=2pt]n1_3.north){BAD};
\node[bad_tag,anchor=south] at ([yshift=2pt]n1_4.north){BAD}; \node[ok_tag,anchor=south] at ([yshift=2pt]n1_4.north){OK};
\node[bad_tag,anchor=south] at ([yshift=2pt]n1_5.north){BAD}; \node[bad_tag,anchor=south] at ([yshift=2pt]n1_5.north){BAD};
\node[ok_tag,anchor=south] at ([yshift=2pt]n1_6.north){OK}; \node[bad_tag,anchor=south] at ([yshift=2pt]n1_6.north){BAD};
\node[ok_tag,anchor=south] at ([yshift=2pt]n1_7.north){OK}; \node[ok_tag,anchor=south] (tag1) at ([yshift=2pt]n1_7.north){OK};
\node[ok_tag,anchor=south] at ([yshift=2pt]n1_8.north){OK};
\node[ok_tag,anchor=south] at ([yshift=2pt]n1_9.north){OK};
\node[ok_tag,anchor=south] (tag1) at ([yshift=2pt]n1_10.north){OK};
\node[bad_tag,anchor=north] at ([yshift=-2pt]n3_2.south){BAD}; \node[ok_tag,anchor=north] at ([yshift=-3pt]n3_2.south){OK};
\node[bad_tag,anchor=north] at ([yshift=-2pt]n3_3.south){BAD}; \node[ok_tag,anchor=north] at ([yshift=-3pt]n3_3.south){OK};
\node[ok_tag,anchor=north] at ([yshift=-2pt]n3_4.south){OK}; \node[ok_tag,anchor=north] at ([yshift=-3pt]n3_4.south){OK};
\node[ok_tag,anchor=north] at ([yshift=-2pt]n3_5.south){OK}; \node[ok_tag,anchor=north] at ([yshift=-3pt]n3_5.south){OK};
\node[ok_tag,anchor=north] at ([yshift=-2pt]n3_6.south){OK}; \node[ok_tag,anchor=north] at ([yshift=-3pt]n3_6.south){OK};
\node[ok_tag,anchor=north] at ([yshift=-2pt]n3_7.south){OK}; \node[bad_tag,anchor=north] at ([yshift=-3pt]n3_7.south){BAD};
\node[bad_tag,anchor=north] at ([yshift=-2pt]n3_8.south){BAD}; \node[ok_tag,anchor=north] at ([yshift=-3pt]n3_8.south){OK};
\node[bad_tag,anchor=north] at ([yshift=-2pt]n3_9.south){BAD}; \node[ok_tag,anchor=north] (tag2) at ([yshift=-3pt]n3_9.south){OK};
\node[ok_tag,anchor=north] (tag2) at ([yshift=-2pt]n3_10.south){OK};
\node[bad_tag,anchor=north] (gap_1)at ([xshift=-2em,yshift=-2em]n3_2.south){BAD}; \node[ok_tag,anchor=north] (gap_1)at ([xshift=-2.6em,yshift=-2em]n3_2.south){OK};
\node[ok_tag,anchor=north] (gap_2)at ([xshift=1.6em,yshift=-2em]n3_2.south){OK}; \node[bad_tag,anchor=north] (gap_2)at ([xshift=2.55em,yshift=-2em]n3_2.south){BAD};
\node[bad_tag,anchor=north] (gap_3)at ([xshift=4.4em,yshift=-2em]n3_3.south){BAD}; \node[ok_tag,anchor=north] (gap_3)at ([xshift=1.85em,yshift=-2em]n3_3.south){OK};
\node[ok_tag,anchor=north] (gap_4)at ([xshift=2.5em,yshift=-2em]n3_4.south){OK}; \node[ok_tag,anchor=north] (gap_4)at ([xshift=2.3em,yshift=-2em]n3_4.south){OK};
\node[ok_tag,anchor=north] (gap_5)at ([xshift=1.3em,yshift=-2em]n3_5.south){OK}; \node[ok_tag,anchor=north] (gap_5)at ([xshift=1.5em,yshift=-2em]n3_5.south){OK};
\node[ok_tag,anchor=north] (gap_6)at ([xshift=2em,yshift=-2em]n3_6.south){OK}; \node[ok_tag,anchor=north] (gap_6)at ([xshift=1.8em,yshift=-2em]n3_6.south){OK};
\node[ok_tag,anchor=north] (gap_7)at ([xshift=1.7em,yshift=-2em]n3_7.south){OK}; \node[ok_tag,anchor=north] (gap_7)at ([xshift=2.0em,yshift=-2em]n3_7.south){OK};
\node[ok_tag,anchor=north] (gap_8)at ([xshift=2.4em,yshift=-2em]n3_8.south){OK}; \node[ok_tag,anchor=north] (gap_8)at ([xshift=1.60em,yshift=-2em]n3_8.south){OK};
\node[ok_tag,anchor=north] (gap_9)at ([xshift=2.5em,yshift=-2em]n3_9.south){OK}; \node[ok_tag,anchor=north] (tag3) at ([xshift=1.7em,yshift=-2em]n3_9.south){OK};
\node[ok_tag,anchor=north](tag3) at ([xshift=1.3em,yshift=-2em]n3_10.south){OK};
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_1.north) -- ([yshift=2em]gap_1.north); \draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_1.north) -- ([yshift=2em]gap_1.north);
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_2.north) -- ([yshift=2em]gap_2.north); \draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_2.north) -- ([yshift=2em]gap_2.north);
...@@ -83,29 +74,26 @@ ...@@ -83,29 +74,26 @@
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_6.north) -- ([yshift=2em]gap_6.north); \draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_6.north) -- ([yshift=2em]gap_6.north);
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_7.north) -- ([yshift=2em]gap_7.north); \draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_7.north) -- ([yshift=2em]gap_7.north);
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_8.north) -- ([yshift=2em]gap_8.north); \draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_8.north) -- ([yshift=2em]gap_8.north);
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](gap_9.north) -- ([yshift=2em]gap_9.north);
\draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](tag3.north) -- ([yshift=2em]tag3.north); \draw[dash pattern=on 2pt off 1pt,gray,line width=1pt](tag3.north) -- ([yshift=2em]tag3.north);
\draw [line width=1pt](n1_2.south) -- (n2_3.north); \draw [line width=1pt,blue!30](n1_2.south) -- (n2_2.north);
\draw [line width=1pt](n1_3.south) -- (n2_4.north); \draw [line width=1pt,blue!30](n1_3.south) -- (n2_3.north);
\draw [line width=1pt](n1_4.south) -- (n2_4.north); \draw [line width=1pt,blue!30](n1_4.south) -- (n2_4.north);
\draw [line width=1pt](n1_5.south) -- (n2_12.north); \draw [line width=1pt,blue!30](n1_4.south) -- (n2_9.north);
\draw [line width=1pt](n1_6.south) -- (n2_8.north); \draw [line width=1pt,blue!30](n1_5.south) -- (n2_5.north);
\draw [line width=1pt](n1_7.south) -- (n2_9.north); \draw [line width=1pt,blue!30](n1_6.south) -- (n2_6.north);
\draw [line width=1pt](n1_8.south) -- (n2_10.north); \draw [line width=1pt,blue!30](n1_7.south) -- (n2_13.north);
\draw [line width=1pt](n1_9.south) -- (n2_11.north);
\draw [line width=1pt](n1_10.south) -- (n2_13.north); \draw[line width=1pt,ugreen!60] (n2_2.south) -- (n3_2.north);
\draw[line width=1pt,ugreen!60] (n2_3.south) -- (n3_3.north);
\draw[line width=1pt,red!60] (n2_3.south) -- (n3_2.north); \draw[line width=1pt,ugreen!60] (n2_4.south) -- (n3_5.north);
\draw[line width=1pt,red!60] (n2_4.south) -- (n3_3.north); \draw[line width=1pt,ugreen!60] (n2_5.south) -- (n3_6.north);
\draw[line width=1pt,ugreen!60] (n2_8.south) -- (n3_4.north); \draw[line width=1pt,red!60] (n2_6.south) -- (n3_7.north);
\draw[line width=1pt,ugreen!60] (n2_9.south) -- (n3_5.north); \draw[line width=1pt,ugreen!60] (n2_9.south) -- (n3_4.north);
\draw[line width=1pt,ugreen!60] (n2_10.south) -- (n3_6.north); \draw[line width=1pt,ugreen!60] (n2_12.south) -- (n3_8.north);
\draw[line width=1pt,ugreen!60] (n2_11.south) -- (n3_7.north); \draw[line width=1pt,ugreen!60] (n2_13.south) -- (n3_9.north);
\draw[line width=1pt,red!60] (n2_12.south) -- (n3_8.north);
\draw[line width=1pt,ugreen!60] (n2_13.south) -- (n3_10.north);
\node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] at ([xshift=4.6em]tag1.east){\textbf{Source tags}}; \node[anchor=west,inner sep=0pt,align=center,font=\scriptsize](st) at ([xshift=8em]tag1.east){\textbf{Source tags}};
\node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] at ([xshift=2.6em]tag2.east){\textbf{MT tags}}; \node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] at ([xshift=3.6em]tag2.east){\textbf{MT tags}};
\node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] at ([xshift=1.1em]tag3.east){\textbf{Gap tags}}; \node[anchor=west,inner sep=0pt,align=center,font=\scriptsize] (gt) at ([xshift=1.6em]tag3.east){\textbf{Gap tags}};
\end{tikzpicture} \end{tikzpicture}
\ No newline at end of file
\definecolor{ugreen}{rgb}{0,0.5,0} \definecolor{ugreen}{rgb}{0,0.5,0}
\begin{tikzpicture}[scale=0.5] \begin{tikzpicture}[scale=0.5]
\tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!30] \tikzstyle{cand} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=green!15]
\tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!30] \tikzstyle{ref} = [draw,line width=1pt,align=center,minimum width=2.6em,minimum height=1.6em,drop shadow={shadow xshift=0.15em},fill=red!15]
\node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,minimum width=6em] (n11) at (0,0){\small{机器译文:}};
\node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can}; \node[cand,anchor=west] (n12) at ([xshift=0.0em]n11.east){Can};
\node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I}; \node[cand,anchor=west] (n13) at ([xshift=1em]n12.east){I};
\node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have}; \node[cand,anchor=west] (n14) at ([xshift=1em]n13.east){have};
\node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){this}; \node[cand,anchor=west] (n15) at ([xshift=1em]n14.east){it};
\node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like}; \node[cand,anchor=west] (n16) at ([xshift=1em]n15.east){like};
\node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he}; \node[cand,anchor=west] (n17) at ([xshift=1em]n16.east){he};
\node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){do}; \node[cand,anchor=west] (n18) at ([xshift=1em]n17.east){?};
\node[cand,anchor=west] (n19) at ([xshift=1em]n18.east){it};
\node[cand,anchor=west] (n20) at ([xshift=1em]n19.east){?}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small{参考答案:}};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n21) at ([yshift=-6em]n11.south){\small\bfnew{Reference :}};
\node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can}; \node[ref,anchor=west] (n22) at ([xshift=0.0em]n21.east){Can};
\node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I}; \node[ref,anchor=west] (n23) at ([xshift=1em]n22.east){I};
\node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat}; \node[ref,anchor=west] (n24) at ([xshift=1em]n23.east){eat};
\node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this}; \node[ref,anchor=west] (n25) at ([xshift=1em]n24.east){this};
\node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can}; \node[ref,anchor=west] (n26) at ([xshift=1em]n25.east){Can};
\node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like}; \node[ref,anchor=west] (n27) at ([xshift=1em]n26.east){like};
\node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){he}; \node[ref,anchor=west] (n28) at ([xshift=1em]n27.east){him};
\node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){did}; \node[ref,anchor=west] (n29) at ([xshift=1em]n28.east){?};
\node[ref,anchor=west] (n30) at ([xshift=1em]n29.east){?};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n31) at ([yshift=-5em]n21.south){\small\bfnew{Candidate :}}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n31) at ([yshift=-5em]n21.south){\small{机器译文:}};
\node[cand,anchor=west] (n32) at ([xshift=0.0em]n31.east){Can}; \node[cand,anchor=west] (n32) at ([xshift=0.0em]n31.east){Can};
\node[cand,anchor=west] (n33) at ([xshift=1em]n32.east){I}; \node[cand,anchor=west] (n33) at ([xshift=1em]n32.east){I};
\node[cand,anchor=west] (n34) at ([xshift=1em]n33.east){have}; \node[cand,anchor=west] (n34) at ([xshift=1em]n33.east){have};
\node[cand,anchor=west] (n35) at ([xshift=1em]n34.east){this}; \node[cand,anchor=west] (n35) at ([xshift=1em]n34.east){it};
\node[cand,anchor=west] (n36) at ([xshift=1em]n35.east){like}; \node[cand,anchor=west] (n36) at ([xshift=1em]n35.east){like};
\node[cand,anchor=west] (n37) at ([xshift=1em]n36.east){he}; \node[cand,anchor=west] (n37) at ([xshift=1em]n36.east){he};
\node[cand,anchor=west] (n38) at ([xshift=1em]n37.east){do}; \node[cand,anchor=west] (n38) at ([xshift=1em]n37.east){?};
\node[cand,anchor=west] (n39) at ([xshift=1em]n38.east){it};
\node[cand,anchor=west] (n40) at ([xshift=1em]n39.east){?}; \node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n41) at ([yshift=-6em]n31.south){\small{参考答案:}};
\node[align=center,minimum width=2.4em,minimum height=1.6em,anchor=north,minimum width=6em] (n41) at ([yshift=-6em]n31.south){\small\bfnew{Candidate :}};
\node[ref,anchor=west] (n42) at ([xshift=0.0em]n41.east){Can}; \node[ref,anchor=west] (n42) at ([xshift=0.0em]n41.east){Can};
\node[ref,anchor=west] (n43) at ([xshift=1em]n42.east){I}; \node[ref,anchor=west] (n43) at ([xshift=1em]n42.east){I};
\node[ref,anchor=west] (n44) at ([xshift=1em]n43.east){eat}; \node[ref,anchor=west] (n44) at ([xshift=1em]n43.east){eat};
\node[ref,anchor=west] (n45) at ([xshift=1em]n44.east){this}; \node[ref,anchor=west] (n45) at ([xshift=1em]n44.east){this};
\node[ref,anchor=west] (n46) at ([xshift=1em]n45.east){Can}; \node[ref,anchor=west] (n46) at ([xshift=1em]n45.east){Can};
\node[ref,anchor=west] (n47) at ([xshift=1em]n46.east){like}; \node[ref,anchor=west] (n47) at ([xshift=1em]n46.east){like};
\node[ref,anchor=west] (n48) at ([xshift=1em]n47.east){he}; \node[ref,anchor=west] (n48) at ([xshift=1em]n47.east){him};
\node[ref,anchor=west] (n49) at ([xshift=1em]n48.east){did}; \node[ref,anchor=west] (n49) at ([xshift=1em]n48.east){?};
\node[ref,anchor=west] (n50) at ([xshift=1em]n49.east){?};
\draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north); \draw[line width=1.6pt,blue!40] (n12.south) -- (n22.north);
\draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north); \draw[line width=1.6pt,blue!40] (n13.south) -- (n23.north);
\draw[line width=1.6pt,blue!40] (n15.south) -- (n25.north);
\draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north); \draw[line width=1.6pt,blue!40] (n16.south) -- (n27.north);
\draw[line width=1.6pt,blue!40] (n17.south) -- (n28.north); \draw[line width=1.6pt,blue!40] (n18.south) -- (n29.north);
\draw[line width=1.6pt,blue!40] (n20.south) -- (n30.north); \draw[line width=1.6pt,orange!40] (n17.south) -- (n28.north);
\draw[line width=1.6pt,orange!40] (n18.south) -- (n29.north);
\draw[line width=2pt,ugreen!40](n14.south) -- (n24.north); \draw[line width=2pt,ugreen!40](n14.south) -- (n24.north);
\draw[line width=1.6pt,blue!40] (n32.south) -- (n46.north); \draw[line width=1.6pt,blue!40] (n32.south) -- (n46.north);
\draw[line width=1.6pt,blue!40] (n33.south) -- (n43.north); \draw[line width=1.6pt,blue!40] (n33.south) -- (n43.north);
\draw[line width=1.6pt,blue!40] (n35.south) -- (n45.north);
\draw[line width=1.6pt,blue!40] (n36.south) -- (n47.north); \draw[line width=1.6pt,blue!40] (n36.south) -- (n47.north);
\draw[line width=1.6pt,blue!40] (n37.south) -- (n48.north); \draw[line width=1.6pt,blue!40] (n38.south) -- (n49.north);
\draw[line width=1.6pt,blue!40] (n40.south) -- (n50.north); \draw[line width=1.6pt,orange!40] (n37.south) -- (n48.north);
\draw[line width=1.6pt,orange!40] (n38.south) -- (n49.north);
\draw[line width=2pt,ugreen!40](n34.south) -- (n44.north); \draw[line width=2pt,ugreen!40](n34.south) -- (n44.north);
\end{tikzpicture} \end{tikzpicture}
This source diff could not be displayed because it is too large. You can view the blob instead.
...@@ -1206,7 +1206,6 @@ ...@@ -1206,7 +1206,6 @@
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%% chapter 4------------------------------------------------------ %%%%% chapter 4------------------------------------------------------
@inproceedings{DBLP:conf/acl/PapineniRWZ02, @inproceedings{DBLP:conf/acl/PapineniRWZ02,
author = {Kishore Papineni and author = {Kishore Papineni and
Salim Roukos and Salim Roukos and
...@@ -1931,16 +1930,355 @@ ...@@ -1931,16 +1930,355 @@
publisher = {Annual Meeting of the Association for Computational Linguistics}, publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2003}, year = {2003},
} }
@inproceedings{chen-guo-2015-representation, @inproceedings{DBLP:conf/wmt/Post18,
author = {Boxing Chen and author = {Matt Post},
Hongyu Guo}, title = {A Call for Clarity in Reporting {BLEU} Scores},
title = {Representation Based Translation Evaluation Metrics}, pages = {186--191},
pages = {150--155}, publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2018},
}
@inproceedings{he2012maximum,
title={Maximum expected bleu training of phrase and lexicon translation models},
author={He, Xiaodong and Deng, Li},
publisher={Annual Meeting of the Association for Computational Linguistics},
pages={292--301},
year={2012}
}
@inproceedings{DBLP:conf/acl/ChenG15,
author = {Boxing Chen and
Hongyu Guo},
title = {Representation Based Translation Evaluation Metrics},
pages = {150--155},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2015},
}
@inproceedings{DBLP:conf/acl/ShenCHHWSL16,
author = {Shiqi Shen and
Yong Cheng and
Zhongjun He and
Wei He and
Hua Wu and
Maosong Sun and
Yang Liu},
title = {Minimum Risk Training for Neural Machine Translation},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2016},
}
@inproceedings{kulesza2004learning,
title={A learning approach to improving sentence-level MT evaluation},
author={Kulesza, Alex and Shieber, Stuart},
publisher={Proceedings of the 10th International Conference on Theoretical and Methodological Issues in Machine Translation},
year={2004}
}
@inproceedings{corston2001machine,
title={A machine learning approach to the automatic evaluation of machine translation},
author={Corston-Oliver, Simon and Gamon, Michael and Brockett, Chris},
publisher={Annual Meeting of the Association for Computational Linguistics},
pages={148--155},
year={2001}
}
@article{albrecht2008regression,
title={Regression for machine translation evaluation at the sentence level},
author={Albrecht, Joshua S and Hwa, Rebecca},
volume={22},
number={1-2},
pages={1},
year={2008},
publisher={Springer}
}
@inproceedings{duh2008ranking,
title={Ranking vs. regression in machine translation evaluation},
author={Duh, Kevin},
publisher={Proceedings of the Third Workshop on Statistical Machine Translation},
pages={191--194},
year={2008}
}
@inproceedings{chen2015multi,
title={Multi-level evaluation for machine translation},
author={Chen, Boxing and Guo, Hongyu and Kuhn, Roland},
publisher={Proceedings of the Tenth Workshop on Statistical Machine Translation},
pages={361--365},
year={2015}
}
@book{DBLP:books/sp/EfronT93,
author = {Bradley Efron and
Robert Tibshirani},
title = {An Introduction to the Bootstrap},
publisher = {Springer},
year = {1993}
}
@inproceedings{DBLP:conf/emnlp/Koehn04,
author = {Philipp Koehn},
title = {Statistical Significance Tests for Machine Translation Evaluation},
pages = {388--395},
publisher = {{ACL}},
year = {2004}
}
@book{noreen1989computer,
title={Computer-intensive methods for testing hypotheses},
author={Noreen, Eric W},
year={1989},
publisher={Wiley New York}
}
@inproceedings{DBLP:conf/acl/RiezlerM05,
author = {Stefan Riezler and
John T. Maxwell III},
title = {On Some Pitfalls in Automatic Evaluation and Significance Testing
for {MT}},
pages = {57--64},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2005}
}
@inproceedings{DBLP:conf/emnlp/Berg-KirkpatrickBK12,
author = {Taylor Berg{-}Kirkpatrick and
David Burkett and
Dan Klein},
title = {An Empirical Investigation of Statistical Significance in {NLP}},
pages = {995--1005},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2012}
}
@inproceedings{DBLP:conf/wmt/Bicici13a,
author = {Ergun Bi{\c{c}}ici},
title = {Referential Translation Machines for Quality Estimation},
pages = {343--351},
publisher = {Annual Meeting of the Association for Computational Linguistics}, publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2015}, year = {2013}
}
@inproceedings{DBLP:conf/wmt/SouzaBTN13,
author = {Jos{\'{e}} Guilherme Camargo de Souza and
Christian Buck and
Marco Turchi and
Matteo Negri},
title = {FBK-UEdin Participation to the {WMT13} Quality Estimation Shared Task},
pages = {352--358},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2013}
}
@inproceedings{DBLP:conf/wmt/SouzaGBTN14,
author = {Jos{\'{e}} Guilherme Camargo de Souza and
Jes{\'{u}}s Gonz{\'{a}}lez{-}Rubio and
Christian Buck and
Marco Turchi and
Matteo Negri},
title = {FBK-UPV-UEdin participation in the {WMT14} Quality Estimation shared-task},
pages = {322--328},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2014}
}
@inproceedings{DBLP:conf/wmt/Espla-GomisSF15,
author = {Miquel Espl{\`{a}}{-}Gomis and
Felipe S{\'{a}}nchez{-}Mart{\'{\i}}nez and
Mikel L. Forcada},
title = {UAlacant word-level machine translation quality estimation system
at {WMT} 2015},
pages = {309--315},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2015}
}
@inproceedings{DBLP:conf/wmt/KreutzerSR15,
author = {Julia Kreutzer and
Shigehiko Schamoni and
Stefan Riezler},
title = {QUality Estimation from ScraTCH {(QUETCH):} Deep Learning for Word-level
Translation Quality Estimation},
pages = {316--322},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2015}
}
@inproceedings{DBLP:conf/wmt/MartinsAHK16,
author = {Andr{\'{e}} F. T. Martins and
Ram{\'{o}}n Fern{\'{a}}ndez Astudillo and
Chris Hokamp and
Fabio Kepler},
title = {Unbabel's Participation in the {WMT16} Word-Level Translation Quality
Estimation Shared Task},
pages = {806--811},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2016}
}
@inproceedings{DBLP:conf/wmt/ChenTZXZLW17,
author = {Zhiming Chen and
Yiming Tan and
Chenlin Zhang and
Qingyu Xiang and
Lilin Zhang and
Maoxi Li and
Mingwen Wang},
title = {Improving Machine Translation Quality Estimation with Neural Network
Features},
pages = {551--555},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2017}
}
@inproceedings{kreutzer2015quality,
title={Quality estimation from scratch (quetch): Deep learning for word-level translation quality estimation},
author={Kreutzer, Julia and Schamoni, Shigehiko and Riezler, Stefan},
publisher={Proceedings of the Tenth Workshop on Statistical Machine Translation},
pages={316--322},
year={2015}
}
@inproceedings{DBLP:conf/wmt/ScartonBSSS16,
author = {Carolina Scarton and
Daniel Beck and
Kashif Shah and
Karin Sim Smith and
Lucia Specia},
title = {Word embeddings and discourse information for Quality Estimation},
pages = {831--837},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2016}
}
@inproceedings{DBLP:conf/wmt/AbdelsalamBE16,
author = {Amal Abdelsalam and
Ondrej Bojar and
Samhaa El{-}Beltagy},
title = {Bilingual Embeddings and Word Alignments for Translation Quality Estimation},
pages = {764--771},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2016}
}
@inproceedings{DBLP:conf/wmt/BasuPN18,
author = {Prasenjit Basu and
Santanu Pal and
Sudip Kumar Naskar},
title = {Keep It or Not: Word Level Quality Estimation for Post-Editing},
pages = {759--764},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2018}
}
@inproceedings{DBLP:conf/wmt/Lo19,
author = {Chi{-}kiu Lo},
title = {YiSi - a Unified Semantic {MT} Quality Evaluation and Estimation Metric
for Languages with Different Levels of Available Resources},
pages = {507--513},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2019}
}
@inproceedings{DBLP:conf/wmt/YankovskayaTF19,
author = {Elizaveta Yankovskaya and
Andre T{\"{a}}ttar and
Mark Fishel},
title = {Quality Estimation and Translation Metrics via Pre-trained Word and
Sentence Embeddings},
pages = {101--105},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2019}
}
@inproceedings{DBLP:conf/wmt/Qi19,
author = {Hou Qi},
title = {{NJU} Submissions for the {WMT19} Quality Estimation Shared Task},
pages = {95--100},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2019}
}
@inproceedings{DBLP:conf/wmt/ZhouZH19,
author = {Junpei Zhou and
Zhisong Zhang and
Zecong Hu},
title = {{SOURCE:} SOURce-Conditional Elmo-style Model for Machine Translation
Quality Estimation},
pages = {106--111},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2019}
}
@inproceedings{DBLP:conf/wmt/Hokamp17,
author = {Chris Hokamp},
title = {Ensembling Factored Neural Machine Translation Models for Automatic
Post-Editing and Quality Estimation},
pages = {647--654},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2017}
}
@inproceedings{DBLP:conf/wmt/YankovskayaTF19,
author = {Elizaveta Yankovskaya and
Andre T{\"{a}}ttar and
Mark Fishel},
title = {Quality Estimation and Translation Metrics via Pre-trained Word and
Sentence Embeddings},
pages = {101--105},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2019}
}
@inproceedings{DBLP:conf/wmt/KimLKN19,
author = {Hyun Kim and
Joon{-}Ho Lim and
Hyun{-}Ki Kim and
Seung{-}Hoon Na},
title = {{QE} {BERT:} Bilingual {BERT} Using Multi-task Learning for Neural
Quality Estimation},
pages = {85--89},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2019}
}
@inproceedings{DBLP:conf/wmt/HildebrandV13,
author = {Silja Hildebrand and
Stephan Vogel},
title = {{MT} Quality Estimation: The {CMU} System for WMT'13},
pages = {373--379},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2013}
}
@article{kepler2019unbabel,
title={Unbabel's Participation in the WMT19 Translation Quality Estimation Shared Task},
author={Kepler, F{\'a}bio and Tr{\'e}nous, Jonay and Treviso, Marcos and Vera, Miguel and G{\'o}is, Ant{\'o}nio and Farajian, M Amin and Lopes, Ant{\'o}nio V and Martins, Andr{\'e} FT},
year={2019}
}
@inproceedings{martins2016unbabel,
title={Unbabel’s participation in the wmt16 word-level translation quality estimation shared task},
author={Martins, Andr{\'e} FT and Astudillo, Ram{\'o}n and Hokamp, Chris and Kepler, Fabio},
publisher={Proceedings of the First Conference on Machine Translation},
pages={806--811},
year={2016}
}
@inproceedings{DBLP:conf/wmt/ShahLPBBBS15,
author = {Kashif Shah and
Varvara Logacheva and
Gustavo Paetzold and
Fr{\'{e}}d{\'{e}}ric Blain and
Daniel Beck and
Fethi Bougares and
Lucia Specia},
title = {{SHEF-NN:} Translation Quality Estimation with Neural Networks},
pages = {342--347},
publisher = {Annual Meeting of the Association for Computational Linguistics},
year = {2015}
}
@book{huang2019machine,
title={Machine Translation: 15th China Conference, CCMT 2019, Nanchang, China, September 27--29, 2019, Revised Selected Papers},
author={Huang, Shujian and Knight, Kevin},
volume={1104},
year={2019},
publisher={Springer Nature}
}
@article{akaike1974new,
title={A new look at the statistical model identification},
author={Akaike, Hirotugu},
volume={19},
number={6},
pages={716--723},
year={1974},
publisher={IEEE}
}
@inproceedings{wang2019niutrans,
title={NiuTrans Submission for CCMT19 Quality Estimation Task},
author={Wang, Ziyang and Liu, Hui and Chen, Hexuan and Feng, Kai and Wang, Zeyang and Li, Bei and Xu, Chen and Xiao, Tong and Zhu, Jingbo},
pages={82--92},
year={2019},
publisher={Springer}
}
@book{jurafsky2000speech,
title={Speech \& language processing},
author={Jurafsky, Dan},
year={2000},
publisher={Pearson Education India}
}
@article{devlin2018bert,
title={Bert: Pre-training of deep bidirectional transformers for language understanding},
author={Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina},
journal={arXiv preprint arXiv:1810.04805},
year={2018}
} }
%%%%% chapter 4------------------------------------------------------ %%%%% chapter 4------------------------------------------------------
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论