BLEU指标的计算示例

题目如下：

根据下图例子，计算下面机器翻译译文的BLEU-4值 (列出1-4元文法的统计数)。

解：由题意得

$1-4元的统计数如下表所示：$
$$\begin{array}{cccc}\text{1-gram}& \text{count}& \text{ref}& \text{count-clip}\\ \text{it}& 1& 0& 0\\ \text{is}& 1& 1& 1\\ \text{a}& 1& 1& 1\\ \text{nice}& 1& 1& 1\\ \text{day}& 1& 1& 1\\ \text{today}& 1& 1& 1\end{array}$$

$$\begin{array}{cccc}\text{2-gram}& \text{count}& \text{ref}& \text{count-clip}\\ \text{it is}& 1& 0& 0\\ \text{is a}& 1& 1& 1\\ \text{a nice}& 1& 1& 1\\ \text{nice day}& 1& 1& 1\\ \text{day today}& 1& 0& 0\end{array}$$

$$\begin{array}{cccc}\text{3-gram}& \text{count}& \text{ref}& \text{count-clip}\\ \text{it is a}& 1& 0& 0\\ \text{is a nice}& 1& 1& 1\\ \text{a nice day}& 1& 1& 1\\ \text{nice day today}& 1& 0& 0\end{array}$$

$$\begin{array}{cccc}\text{4-gram}& \text{count}& \text{ref}& \text{count-clip}\\ \text{it is a nice}& 1& 0& 0\\ \text{is a nice day}& 1& 1& 1\\ \text{a nice day today}& 1& 0& 0\end{array}$$

由表可得
$$p_1=\frac{5}{6}，p_2=\frac{3}{5}，p_3=\frac{2}{4}，p_4=\frac{1}{3}$$
易知
$$c=6,r=5$$
即
$$c>r$$
故
$$BP=1$$
画表如下：
$$\overline{\begin{array}{lllllll}& \text{Translation}& p_1& p_2& p_3& p_4& \text{BP}\\ \text{人工译文：}& \text{Today is a nice day}& & & & & \\ \text{机器译文：}& \text{It is a nice day today}& \frac{5}{6}& \frac{3}{5}& \frac{2}{4}& \frac{1}{3}& 1\end{array}}$$
因为 $N=4$ ,故 $w_n=\frac{1}{4}$

所以
$$\begin{array}{rl}\text{BLEU-4}& =\mathrm{BP}\ast \exp (\sum _{n=1}^N{w}_n\log P_n)\\ & =1\times \exp (\frac{1}{4}\times (\log (p_1)+\log (p_2)+\log (p_3)+log(p_4)))\\ & =\exp (\frac{1}{4}\times \log (p_1{p}_2{p}_3{p}_4))\\ & =\exp (\frac{1}{4}\times log(\frac{5}{6}\times \frac{3}{5}\times \frac{2}{4}\times \frac{1}{3}))\\ & =\exp (\frac{1}{4}\times log(\frac{1}{12}))\\ & \approx 0.537\end{array}$$

倘若大佬发现什么错误，敬请斧正，感谢感谢！