Python实现敏感词过滤的五种方法_Python

1、replace替换

replace就是最简单的字符串替换，当一串字符串中有可能会出现的敏感词时，我们直接使用相应的replace方法用*替换出敏感词即可。

缺点：

文本和敏感词少的时候还可以，多的时候效率就比较差了。

示例代码：

text = '我是一个来自星星的超人，具有超人本领！'
text = text.replace("超人", '*' * len("超人")).replace("星星", '*' * len("星星"))
print(text)  # 我是一个来自***的***，具有***本领！

运行结果：

如果是多个敏感词可以用列表进行逐一替换。

示例代码：

text = '我是一个来自星星的超人，具有超人本领！'
words = ['超人', '星星']
 
for word in words:
    text = text.replace(word, '*' * len(word))
print(text)  # 我是一个来自***的***，具有***本领！

运行效果：

2、正则表达式

使用正则表达式是一种简单而有效的方法，可以快速地匹配敏感词并进行过滤。在这里我们主要是使用“|”来进行匹配，“|”的意思是从多个目标字符串中选择一个进行匹配。

示例代码：

import re
 
 
def filter_words(text, words):
    pattern = '|'.join(words)
    return re.sub(pattern, '***', text)
 
 
if __name__ == '__main__':
    text = '我是一个来自星星的超人，具有超人本领！'
    words = ['超人', '星星']
    res = filter_words(text, words)
    print(res)  # 我是一个来自***的***，具有***本领！

运行结果：

3、使用ahocorasick第三方库

ahocorasick库安装：

pip install pyahocorasick

示例代码：

import ahocorasick
 
 
def filter_words(text, words):
    a = ahocorasick.automaton()
    for index, word in enumerate(words):
        a.add_word(word, (index, word))
    a.make_automaton()
 
    result = []
    for end_index, (insert_order, original_value) in a.iter(text):
        start_index = end_index - len(original_value) + 1
        result.append((start_index, end_index))
 
    for start_index, end_index in result[::-1]:
        text = text[:start_index] + '*' * (end_index - start_index + 1) + text[end_index + 1:]
    return text
 
 
if __name__ == '__main__':
    text = '我是一个来自星星的超人，具有超人本领！'
    words = ['超人', '星星']
    res = filter_words(text, words)
    print(res)  # 我是一个来自***的***，具有***本领！

运行结果：

4、字典树

使用字典树是一种高效的方法，可以快速地匹配敏感词并进行过滤。

示例代码：

class treenode:
    def __init__(self):
        self.children = {}
        self.is_end = false
 
 
class tree:
    def __init__(self):
        self.root = treenode()
 
    def insert(self, word):
        node = self.root
        for char in word:
            if char not in node.children:
                node.children[char] = treenode()
            node = node.children[char]
        node.is_end = true
 
    def search(self, word):
        node = self.root
        for char in word:
            if char not in node.children:
                return false
            node = node.children[char]
        return node.is_end
 
 
def filter_words(text, words):
    tree = tree()
    for word in words:
        tree.insert(word)
 
    result = []
    for i in range(len(text)):
        node = tree.root
        for j in range(i, len(text)):
            if text[j] not in node.children:
                break
            node = node.children[text[j]]
            if node.is_end:
                result.append((i, j))
    for start_index, end_index in result[::-1]:
        text = text[:start_index] + '*' * (end_index - start_index + 1) + text[end_index + 1:]
    return text
 
 
if __name__ == '__main__':
    text = '我是一个来自星星的超人，具有超人本领！'
    words = ['超人', '星星']
    res = filter_words(text, words)
    print(res)  # 我是一个来自***的***，具有***本领！

运行结果：

5、dfa算法

使用dfa算法是一种高效的方法，可以快速地匹配敏感词并进行过滤。dfa的算法，即deterministic finite automaton算法，翻译成中文就是确定有穷自动机算法。它的基本思想是基于状态转移来检索敏感词，只需要扫描一次待检测文本，就能对所有敏感词进行检测。

示例代码：

class dfa:
    def __init__(self, words):
        self.words = words
        self.build()
 
    def build(self):
        self.transitions = {}
        self.fails = {}
        self.outputs = {}
        state = 0
        for word in self.words:
            current_state = 0
            for char in word:
                next_state = self.transitions.get((current_state, char), none)
                if next_state is none:
                    state += 1
                    self.transitions[(current_state, char)] = state
                    current_state = state
                else:
                    current_state = next_state
            self.outputs[current_state] = word
        queue = []
        for (start_state, char), next_state in self.transitions.items():
            if start_state == 0:
                queue.append(next_state)
                self.fails[next_state] = 0
        while queue:
            r_state = queue.pop(0)
            for (state, char), next_state in self.transitions.items():
                if state == r_state:
                    queue.append(next_state)
                    fail_state = self.fails[state]
                    while (fail_state, char) not in self.transitions and fail_state != 0:
                        fail_state = self.fails[fail_state]
                    self.fails[next_state] = self.transitions.get((fail_state, char), 0)
                    if self.fails[next_state] in self.outputs:
                        self.outputs[next_state] += ', ' + self.outputs[self.fails[next_state]]
 
    def search(self, text):
        state = 0
        result = []
        for i, char in enumerate(text):
            while (state, char) not in self.transitions and state != 0:
                state = self.fails[state]
            state = self.transitions.get((state, char), 0)
            if state in self.outputs:
                result.append((i - len(self.outputs[state]) + 1, i))
        return result
 
 
def filter_words(text, words):
    dfa = dfa(words)
    result = []
    for start_index, end_index in dfa.search(text):
        result.append((start_index, end_index))
    for start_index, end_index in result[::-1]:
        text = text[:start_index] + '*' * (end_index - start_index + 1) + text[end_index + 1:]
    return text
 
 
if __name__ == '__main__':
    text = '我是一个来自星星的超人，具有超人本领！'
    words = ['超人', '星星']
    res = filter_words(text, words)
    print(res)  # 我是一个来自***的***，具有***本领！

运行结果：