算法-DFA算法-敏感词过滤算法（OC、Swift、Python）「建议收藏」

大家好，又见面了，我是你们的朋友全栈君。

前言

前段时间，公司的IM SDK想做敏感词过滤，但是后端的小伙伴《比较忙》，在开产品需求会的时候想把敏感词过滤放到前端，让iOS、安卓自己搞，但是前端小伙伴写了一个方法来检测一段文本，耗时【一两秒】钟而且比较耗CPU，这样肯定不行的，最后后端小伙伴妥协了，把敏感词过滤放到后端了。

一般的思路可能是遍历敏感词库，然后把一段文字的敏感词过滤掉，但是针对比较大的词库时（比如我们的敏感词库10万），这样非常耗时和耗内存，在电脑上还能跑跑，但是在手机上分分钟钟被系统杀死掉，这样肯定是不行的，这里就用到一种DFA算法。

但是使用了DFA算法，十万的敏感词库过滤一句话只需要【0.434510秒】！

2019-10-23 14:34:08.230918+0800 DFAFilterDemo[4728:4650502] message == 小明骂小王是个王八蛋，小王骂小明是个王八羔子！
2019-10-23 14:34:08.232972+0800 DFAFilterDemo[4728:4650502] result == 小明骂小王是个***，小王骂小明是个王八羔子！
2019-10-23 14:34:08.316380+0800 DFAFilterDemo[4728:4650502] 总共耗时: 0.434510 

DFA算法

简介

何谓DFA，它的全称是Deterministic Finite Automaton，即确定有穷自动机；其特征为：有一个有限状态集合和一些从一个状态通向另一个状态的边，每条边上标记有一个符号，其中一个状态是初态，某些状态是终态。但不同于不确定的有限自动机，DFA中不会有从同一状态出发的两条边标志有相同的符号；DFA算法的核心是建立了以敏感词为基础的许多敏感词树。

描述

我们先把敏感词库拆分解析成一个”敏感词树“，我们以敏感词”王八蛋“和”王八羔子“为例：

拆成的敏感词树如下：

代码

OC代码

//
// DFAFilter.m
// DFAFilterDemo
//
// Created by 张福杰 on 2019/10/22.
// Copyright © 2019 张福杰. All rights reserved.
//
#import "DFAFilter.h"
@interface DFAFilter ()
@property (nonatomic,strong) NSMutableDictionary *keyword_chains;
@property (nonatomic,  copy) NSString *delimit;
@end
@implementation DFAFilter
- (instancetype)init{ 

if(self == [super init]){ 

_delimit = @"\x00";
}
return self;
}
/// 读取解析敏感词
- (void)parseSensitiveWords:(NSString *)path{ 

if(path == nil){ 

path = [[NSBundle mainBundle] pathForResource:@"sensitive_words" ofType:@"txt"];
}
NSString *content = [[NSString alloc] initWithContentsOfFile:path encoding:NSUTF8StringEncoding error:nil];
NSArray *keyWordList = [content componentsSeparatedByString:@","];
for (NSString *keyword in keyWordList) { 

[self addSensitiveWords:keyword];
}
NSLog(@"keyword_chains == %@",self.keyword_chains);
}
- (void)addSensitiveWords:(NSString *)keyword{ 

keyword = keyword.lowercaseString;
keyword = [keyword stringByTrimmingCharactersInSet:[NSCharacterSet whitespaceAndNewlineCharacterSet]];
NSMutableDictionary *node = self.keyword_chains;
for (int i = 0; i < keyword.length; i ++) { 

NSString *word = [keyword substringWithRange:NSMakeRange(i, 1)];
if (node[word] == nil) { 

node[word] = [NSMutableDictionary dictionary];
}
node = node[word];
}
//敏感词最后一个字符标识
[node setValue:@0 forKey:self.delimit];
}
- (NSString *)filterSensitiveWords:(NSString *)message replaceKey:(NSString *)replaceKey{ 

replaceKey = replaceKey == nil ? @"*" : replaceKey;
message = message.lowercaseString;
NSMutableArray *retArray = [[NSMutableArray alloc] init];
NSInteger start = 0;
while (start < message.length) { 

NSMutableDictionary *level = self.keyword_chains.mutableCopy;
NSInteger step_ins = 0;
NSString *message_chars = [message substringWithRange:NSMakeRange(start, message.length - start)];
for(int i = 0; i < message_chars.length; i++){ 

NSString *chars_i = [message_chars substringWithRange:NSMakeRange(i, 1)];
if([level.allKeys containsObject:chars_i]){ 

step_ins += 1;
NSDictionary *level_char_dict = level[chars_i];
if(![level_char_dict.allKeys containsObject:self.delimit]){ 

level = level_char_dict.mutableCopy;
}else{ 

NSMutableString *ret_str = [[NSMutableString alloc] init];
for(int i = 0; i < step_ins; i++){ 

[ret_str appendString:replaceKey];
}
[retArray addObject:ret_str];
start += (step_ins - 1);
break;
}
}else{ 

[retArray addObject:[NSString stringWithFormat:@"%C",[message characterAtIndex:start]]];
break;
}
}
start ++;
}
return [retArray componentsJoinedByString:@""];
}
- (NSMutableDictionary *)keyword_chains{ 

if(_keyword_chains == nil){ 

_keyword_chains = [[NSMutableDictionary alloc] initWithDictionary:@{ 
}];
}
return _keyword_chains;
}
@end

Swift代码

//
// DFAFilter.swift
// DFAFilterDemo
//
// Created by 张福杰 on 2019/10/23.
// Copyright © 2019 张福杰. All rights reserved.
//
import UIKit
class DFAFilter: NSObject { 

lazy var keyword_chains: NSMutableDictionary = { 

let dict = NSMutableDictionary()
return dict
}()
lazy var delimit: String = "\\x00";
/// 读取敏感词
func parseSensitiveWords() -> Void { 

let path = Bundle.main.path(forResource: "sensitive_words", ofType: "txt");
let url = URL(fileURLWithPath: path!)
do { 

let data = try Data(contentsOf: url)
let content: String = String(data: data, encoding: String.Encoding.utf8)!
let keyWordList = content.components(separatedBy: ",")
for keyword in keyWordList { 

addSensitiveWords(keyword)
}
} catch let error as Error? { 

print(error?.localizedDescription as Any)
}
}
/// 添加敏感词到敏感词树
func addSensitiveWords(_ keyword: String) -> Void { 

let keyword: String = keyword.lowercased().trimmingCharacters(in: .whitespacesAndNewlines)
var node = self.keyword_chains
if keyword.count <= 0{ 

return
}
for index in 0...keyword.count - 1 { 

let index0 = keyword.index(keyword.startIndex, offsetBy: index)
let index1 = keyword.index(keyword.startIndex, offsetBy: index + 1)
let word = keyword[index0..<index1]
if node[word] == nil{ 

node[word] = NSMutableDictionary()
}
node = node[word] as! NSMutableDictionary
}
node[self.delimit] = 0
}
/// 开始过滤敏感词
func filterSensitiveWords(_ message: String, replaceKey: String) -> String { 

let replaceKey = replaceKey.count > 0 ? replaceKey : "*"
let message = message.lowercased()
let retArray: NSMutableArray = NSMutableArray()
var start = 0
while start < message.count { 

var level: NSMutableDictionary = self.keyword_chains.mutableCopy() as! NSMutableDictionary
var step_ins = 0
let message_chars = getChar(message, startIndex: start, endIndex: message.count)
for index in 0...message_chars.count - 1 { 

let chars_i = getChar(message_chars, startIndex: index, endIndex: index + 1)
if level[chars_i] != nil{ 

step_ins += 1
let level_char_dict: NSDictionary = level[chars_i] as! NSDictionary
if level_char_dict[self.delimit] == nil{ 

level = level_char_dict.mutableCopy() as! NSMutableDictionary
}else{ 

var ret_str = ""
for _ in 0...step_ins - 1 { 

ret_str += replaceKey
}
retArray.add(ret_str)
start += (step_ins - 1)
break
}
}else{ 

let word = getChar(message, startIndex: start, endIndex: start + 1)
retArray.add(word)
break
}
}
start += 1
}
return retArray.componentsJoined(by: "")
}
func getChar(_ message: String, startIndex: NSInteger, endIndex: NSInteger) -> String { 

let index0 = message.index(message.startIndex, offsetBy: startIndex)
let index1 = message.index(message.startIndex, offsetBy: endIndex)
let word = message[index0..<index1]
return String(word)
}
}

Python代码

# -*- coding: utf-8 -*-
# @Author: zhangfujie
# @Date: 2019/10/22
# @Last Modified by: zhangfujie
# @Last Modified time: 2019/10/22
# @ ---------- DFA过滤算 ---------- 
import time
time1 = time.time()
class DFAFilter(object):
"""DFA过滤算法"""
def __init__(self):
super(DFAFilter, self).__init__()
self.keyword_chains = { 
}
self.delimit = '\x00'
# 读取解析敏感词
def parseSensitiveWords(self, path):
ropen = open(path,'r')
text = ropen.read()
keyWordList = text.split(',')
for keyword in keyWordList:
self.addSensitiveWords(str(keyword).strip())
# 生成敏感词树
def addSensitiveWords(self, keyword):
keyword = keyword.lower()
chars = keyword.strip()
if not chars:
return
level = self.keyword_chains
for i in range(len(chars)):
if chars[i] in level:
level = level[chars[i]]
else:
if not isinstance(level, dict):
break
for j in range(i, len(chars)):
level[chars[j]] = { 
}
last_level, last_char = level, chars[j]
level = level[chars[j]]
last_level[last_char] = { 
self.delimit: 0}
break
if i == len(chars) - 1:
level[self.delimit] = 0
# 过滤敏感词
def filterSensitiveWords(self, message, repl="*"):
message = message.lower()
ret = []
start = 0
while start < len(message):
level = self.keyword_chains
step_ins = 0
message_chars = message[start:]
for char in message_chars:
if char in level:
step_ins += 1
if self.delimit not in level[char]:
level = level[char]
else:
ret.append(repl * step_ins)
start += step_ins - 1
break
else:
ret.append(message[start])
break
start += 1
return ''.join(ret)
if __name__ == "__main__":
gfw = DFAFilter()
gfw.parseSensitiveWords('shieldwords.txt')
text = "小明骂小王是个王八蛋，小王骂小明是个王八羔子！"
result = gfw.filterSensitiveWords(text)
print(result)
time2 = time.time()
print('总共耗时:' + str(time2 - time1) + 's')	

结束语

demo下载地址: https://gitee.com/zfj1128/DFAFilterDemo
过往大佬喜欢的给个小星星吧！

欢迎各位大神提出宝贵的意见和建议，也欢迎大家进群交流365152048！