diff --git a/others/preface/preface-en.tex b/others/preface/preface-en.tex
index de28ce7f6..b5b785b3f 100644
--- a/others/preface/preface-en.tex
+++ b/others/preface/preface-en.tex
@@ -434,7 +434,7 @@ \section*{Summary}
 \vspace{3mm}
 For example $X$ = [3, 1, 3, 5, 4], the missing number $x = 2$, the duplicated one $y = 3$. We give 4 methods: (1) divide and conquer; (2) pigeon hole sort; (3) sign encoding; and (4) equations.
 
-\textbf{Divide and conquer}: Partition the numbers with the middle point $m = \lfloor \dfrac{1 + n}{2} \rfloor$: the left $as = [a \leq m, a \gets X]$, and the right $bs = [b > m, b \gets X]$. If the length of $|as| < m$, then the missing number is on the left, let $s = 1 + 2 + ... + m = \dfrac{m (m + 1)}{2}$, then $x = s - sum(as)$. We can also calculate the missing one on the right. Let $s' = (m + 1) + (m + 2) + ... + n = \dfrac{(n + m + 1)(n - m)}{2}$, then $y = sum(bs) - s'$. If the length of $|as| > m$, then the duplicated number is on the left. Use the similar method, we calculate the missing number $x = s' - sum(bs)$, and the duplicated number $y = sum(as) - s$. Otherwise if the length $|as| = m$, then there are $m$ numbers not greater than $m$. But we don't know whether they are some permutation of 1, 2, ..., $m$. We can calculate and compare $sum(as)$ and $s$. If equal, then we can drop all numbers on the left, then recursively find $x$ and $y$ on the right; otherwise, we drop the right, recursively find on the left. In recursive finding, we need replace the lower bound of 1 with $l$. Because we halve the list every time, the overall performance is $O(n)$ according to the master theorem.
+\textbf{Divide and conquer}: Partition the numbers with the middle point $m = \lfloor \frac{1 + n}{2} \rfloor$: the left is $as = [a \leq m, a \gets X]$; the right is $bs = [b > m, b \gets X]$. If the length of $|as| < m$, then the missing number is on the left. Let $s = 1 + 2 + \dotsb + m = \frac{m (m + 1)}{2}$, then $x = s - sum(as)$. The duplicated one is on the right. Let $s' = (m + 1) + (m + 2) + \dotsb + n = \frac{(n + m + 1)(n - m)}{2}$, then $y = sum(bs) - s'$. If the length of $|as| > m$, then the duplicated number is on the left. In the same way, calculate the missing number $x = s' - sum(bs)$ and the duplicated number $y = sum(as) - s$. Otherwise if the length $|as| = m$, then there are $m$ numbers not greater than $m$. In order to know whether they are some permutation of 1, 2, ..., $m$, we calculate and compare $sum(as)$ and $s$. If equal, then drop all numbers on the left, and recursively find $x$ and $y$ on the right; otherwise, drop the right, recursively find on the left. In recursive finding, we need update the lower bound of 1 with $l$. Because we halve the list every time, the overall performance is $O(n)$ by the master theorem.
 
 \begin{Haskell}
 missDup xs = solve xs 1 (length xs) where
@@ -442,16 +442,16 @@ \section*{Summary}
                        | k > m - l + 1 = (sr - sr', sl' - sl)
                        | sl == sl' = solve bs (m + 1) u
                        | otherwise = solve as l m
-      where
-          m = (l + u) `div` 2
-          (as, bs) = partition (<=m) xs
-          k = length as
-          sl = (l + m) * (m - l + 1) `div` 2
-          sr = (m + 1 + u) * (u - m) `div` 2
-          (sl', sr') = (sum as, sum bs)
+    where
+      m = (l + u) `div` 2
+      (as, bs) = partition (<=m) xs
+      k = length as
+      sl = (l + m) * (m - l + 1) `div` 2
+      sr = (m + 1 + u) * (u - m) `div` 2
+      (sl', sr') = (sum as, sum bs)
 \end{Haskell}
 
-\textbf{Pigeon hole sort}. Since all numbers are within the range from 1 to $n$, we can do pigeon hole sort. Scan from left to right, for every number $x$ at position $i$, if $x \neq i$, we swap it with number $y$ at position $x$. We find the duplicated number if $x = y$, besides, we find the missing number $i$. Repeat this till $x = i$ or meet the duplicated number. Because every number is swapped to its right position a time, the total performance is $O(n)$.
+\textbf{Pigeon hole sort}. We apply pigeon hole sort since all numbers are within the range from 1 to $n$. Scan from left to right, for every number $x$ at position $i$, if $x \neq i$, swap it with number $y$ at position $x$. We find the duplicated number if $x = y$, besides, we find the missing number $i$. Repeat this till $x = i$ or meet the duplicated number. Because every number is swapped to its right position a time, the total performance is $O(n)$.
 
 \begin{Bourbaki}
 (Int, Int) missDup([Int] xs) {
@@ -470,9 +470,10 @@ \section*{Summary}
         }
     }
     return (miss, dup)
+}
 \end{Bourbaki}
 
-\textbf{Sign encoding}. Setup an array of $n$ flags. For every number $x$, mark the $x$-th flag in the array true. When meet the duplicated number, the corresponding flag was marked before. Let the duplicated number be $d$, we know $s = 1 + 2 + ... + n = \dfrac{n (n + 1)}{2}$, and the sum $s'$ of all numbers. We can calculate the missing number $m = d + s - s'$. However, this method need additional $n$ flags. The existence of a number is a type of binary information (yes/no), we can encode it as the positive/negative sign, hence re-use the space. For every $x$, flip the number at position $|x|$ to negative, where $|x|$ is the absolute value. If a number of some position is already negative, it's the duplicated one, and we can next calculate the missing one.
+\textbf{Sign encoding}. Setup an array of $n$ flags. For every number $x$, mark the $x$-th flag in the array true. The flag will have been marked before when meet the duplicated number. Let the duplicated number be $d$, the sum $s = 1 + 2 + \dotsb + n = \frac{n (n + 1)}{2}$, and $s'$ be the sum of all numbers. The missing number $m = d + s - s'$. However, this method need additional $n$ flags. The existence of a number is a type of binary information (yes/no), we encode it as the positive/negative sign, hence re-use the space. For every $x$, flip the number at position $|x|$ to negative, where $|x|$ is the absolute value. If a number at some position is already negative, it's the duplicated one; from where we then calculate the missing one.
 
 \begin{Bourbaki}
 (Int, Int) missDup([Int] xs) {
@@ -489,9 +490,10 @@ \section*{Summary}
         xs[j] = -abs(xs[j])
     }
     return (miss, dup)
+}
 \end{Bourbaki}
 
-\textbf{Equation}. Consider a simplified problem: random drop a number after shuffle 1 to $n$, how to find it? We sum all the numbers, then subtract it from $\dfrac{n (n + 1)}{2}$:
+\textbf{Equation}. Consider a simplified problem: random drop a number after shuffle 1 to $n$, how to find it? We sum all the numbers, then subtract it from $\frac{n (n + 1)}{2}$:
 
 \[
 m = s - s'
@@ -504,27 +506,27 @@ \section*{Summary}
 \label{eq:miss-dup-1}
 \ee
 
-Where the left hand is the sum of the $i$-th number minus $i$. Can we figure out a second equation? We can use square: sum the difference between the square of the $i$-th number and the square of $i$:
+Where the left hand is the sum of the difference between the $i$-th number and $i$. We use square to figure out the second equation: sum the difference between the square of the $i$-th number and the square of $i$:
 
 \be
 \sum (x[i]^2 - i^2) = d^2 - m^2 = (d + m)(d - m)
 \label{eq:miss-dup-2}
 \ee
 
-Since $d - m \neq 0$, we can divide \cref{eq:miss-dup-1} by \cref{eq:miss-dup-2} on both sides to get another equation:
+Since $d - m \neq 0$, divide \cref{eq:miss-dup-1} by \cref{eq:miss-dup-2} on both sides to get another equation:
 
 \be
 \sum (x[i]^2 - i^2) / \sum (x[i] - i) = d + m
 \label{eq:miss-dup-3}
 \ee
 
-Compare equation \cref{eq:miss-dup-1} and \cref{eq:miss-dup-3}, there are two equations with two unknowns. We can solve them:
+Compare \cref{eq:miss-dup-1} and \cref{eq:miss-dup-3}, there are two equations with two unknowns. Solve this system:
 
 \[
-\begin{cases}
-m = \dfrac{1}{2} (\dfrac{\sum (x[i]^2 - i^2)}{\sum (x[i] - i)} - \sum (x[i] - i)) \\
-d = \dfrac{1}{2} (\dfrac{\sum (x[i]^2 - i^2)}{\sum (x[i] - i)} + \sum (x[i] - i)) \\
-\end{cases}
+\begin{dcases}
+m & = \dfrac{1}{2} [\dfrac{\sum (x[i]^2 - i^2)}{\sum (x[i] - i)} - \sum (x[i] - i)] \\
+d & = \dfrac{1}{2} [\dfrac{\sum (x[i]^2 - i^2)}{\sum (x[i] - i)} + \sum (x[i] - i)]
+\end{dcases}
 \]
 
 \begin{Haskell}
diff --git a/others/preface/preface-zh-cn.tex b/others/preface/preface-zh-cn.tex
index 1f1464b8e..b6422bd85 100644
--- a/others/preface/preface-zh-cn.tex
+++ b/others/preface/preface-zh-cn.tex
@@ -432,6 +432,7 @@ \section*{小结}
         }
     }
     return (miss, dup)
+}
 \end{Bourbaki}
 
 \textbf{符号编码}。假设存在一个长度为$n$的标记数组，对于序列中的每个数字$x$，我们都将标记数组中的第$x$个位置做上标记。当我们遇到重复元素时，我们会发现这个位置上的标记已经做过了。记重复的数字为$d$，我们知道$s = 1 + 2 + ... + n = \dfrac{n (n + 1)}{2}$，以及序列中所有的数字和$s'$。我们可以计算出丢失的数字$m = d + s - s'$。 但是这一方法需要额外长度为$n$的空间用作标记数组。由于数字的存在与否是一种二值化的信息（有、无），我们可以将其编码为数字的正负号，从而复用待查找的数字序列。对于序列中的每个数字$x$，我们将序列中第$|x|$位置上的元素标记为负数，其中$|x|$表示绝对值。如果发现某一位置已经为负了，我们就找到了重复的元素，接下来我们就可以计算出丢失的数字。
@@ -451,6 +452,7 @@ \section*{小结}
         xs[j] = -abs(xs[j])
     }
     return (miss, dup)
+}
 \end{Bourbaki}
 
 \textbf{解方程}。考虑一个简化的问题：给定一个从1到$n$的列表，去掉一个元素，然后打乱序列的顺序，怎样能够快速找出去掉的元素呢？我们可以将列表中所有的元素相加，然后从$\dfrac{n (n + 1)}{2}$减去这一结果就得出了答案。这一思路可以表示为如下方程：