敏感词检测过滤是 App 中很重要的一个功能,敏感词数量动辄几千甚至上万,设计一个好的、高效的过滤算法是非常有必要的。那么有什么比较好的敏感词过滤算法吗?
Google 一下 !
DFA 全称为:Deterministic Finite Automaton,即确定有穷自动机。剩下的理论没看懂就不 BB 了,想要了解的自己 Google 去。
- 创建模型类
class Node {
var chirldren = [String: Node]()
var word = ""
}
- 读取敏感词库到数组中
private func getSentiveWords() -> [String] {
var rootArray = [String]()
guard let filePath = Bundle.main.path(forResource: "SensitiveWords", ofType: nil) else {return [String]()}
guard let filterString = try? String(contentsOfFile: filePath, encoding: .utf8) else {return [String]()}
rootArray.append(contentsOf: filterString.components(separatedBy: "\r\n"))
return rootArray
}
- 添加敏感词到 DFA 模型中
private func add(_ word: String, toRootNode root: Node) {
var node = root
for letter in word {
let letterStr = String(letter)
if node.chirldren[letterStr] == nil {
node.chirldren[letterStr] = Node()
}
node = node.chirldren[letterStr]!
}
node.word = word
}
- 遍历敏感词数组,添加敏感词到 DFA 模型,完成模型构建
private func creatDFAModel() -> Node {
let root = Node()
let words = getSentiveWords()
words.forEach { add($0, toRootNode: root) }
return root
}
/// 敏感词过滤
/// - Parameter text: 要检测的文本
/// - Parameter replaceChar: 敏感词替换为某个字符,传入 "" 表示移除敏感词
func filter(text: String, with replaceChar: String = "*") -> FilterResult {
var filterdStr = text
var isCotain = false
for i in 0..<text.count {
var p = root
var j = i
while j < text.count, p.chirldren[text[j]] != nil {
p = p.chirldren[text[j]]!
j += 1
}
if (p.word == text.subString(from: i, to: (j - 1))) && !p.word.isEmpty {
var replaceStr = ""
isCotain = true
for _ in i..<j {
replaceStr.append(replaceChar)
}
filterdStr = filterdStr.replacingOccurrences(of: p.word, with: replaceStr)
}
}
return (isCotain, filterdStr)
}