版权声明:本文虽为博主原创文章,但你转了博主也不知道,所以放心的转吧。 https://blog.csdn.net/javaer_lee/article/details/88029376
总的来说,敏感词过滤就是词库匹配,你定义一个词库,里面有很多敏感词,匹配到了就说明这个词是敏感词。
所以最简单的办法就是建立一个list,先把所有的敏感词读进这个list,然后再利用list的contains方法,就可以判断某一句话中是否有敏感词,如果有就弹个提示,告诉用户语句中有敏感词,禁止用户发送,但是如果须要把把敏感词屏蔽掉(比如用” * “号代替)这个时候contains方法就不行了,得自己写算法判断敏感词所在的位置并屏蔽掉,实现起来并不那么简单。
下面有个简单的办法,大体的思路就是构造词库的时候改变一下思路,不再是一个敏感词一组了,而是将敏感词变成一棵树,用Map集合来存放,举个例子:有个敏感词叫翻墙,那么key就是翻字(具体会转成char类型),value是一个节点对象,它有一个指向下一个字的引用,用代码描述大概是这样{翻=对象类{key=墙,下一个对象{},是否是结尾}}
,类似的结构很常见,值得一提的是,用一个标志位来表示这棵树是否到头了,这样做有个很明显的好处,比如有两个敏感词,一个叫法轮功,一个叫法轮大法,那么这两个敏感词会存放在同一课树中,key都是法,后面是轮,完了有两个分支,一个叫功,一个叫大法,每一个字前面都有一个标志位来标记结束,功和法字的标志位都是true,因为是两个词的结尾位置。主要就在于构建这个敏感词库,有了它查询敏感词就很简单了,下面是实现代码,词库添加和搜索敏感词都是毫秒级别的,生产上完全可用。
敏感词树的节点
public class SensitiveWordNode {
// 节点所代表的字符
private char key;
// 节点的子节点
private Map<Character,SensitiveWordNode> nextNodes;
// 该节点是否为 End 节点
private boolean end;
public SensitiveWordNode (char key) {
this.key = key;
nextNodes = new HashMap<>();
end = false;
}
public SensitiveWordNode getNextNode (char key) {
return nextNodes.get(key);
}
public void putNextNode (SensitiveWordNode node) {
nextNodes.put (node.getKey (), node);
}
public char getKey () {
return key;
}
public void setKey (char key) {
this.key = key;
}
public Map getNextNodes () {
return nextNodes;
}
public void setNextNodes (Map nextNodes) {
this.nextNodes = nextNodes;
}
public boolean isEnd () {
return end;
}
public void setEnd (boolean end) {
this.end = end;
}
@Override
public String toString() {
return "SensitiveWordNode{" +
"key=" + key +
", nextNodes=" + nextNodes +
", end=" + end +
'}';
}
}
敏感词树
public class SensitiveWordTree {
// 日志
private static final Logger LOG = Logger.getLogger("SensitiveWordTree");
// 根节点
private static SensitiveWordNode root = null;
// 敏感词库编码
// private static final String ENCODING = "utf-8";
private static final String ENCODING = "gbk";
// 敏感词库位置 (这个词库的位置自己要换掉)
// private static final String filePath = "/Users/lirenren/intellijProjects/effective-java-3e-source-code/resource/1.txt";
private static final String filePath = "D:/1.txt";
/**
* 读取敏感词库
*
* @return
*/
private static Set readSensitiveWords() {
Set keyWords = new HashSet();
BufferedReader reader = null;
try {
reader = new BufferedReader(new InputStreamReader(new FileInputStream(filePath), ENCODING));
String line;
while ((line = reader.readLine()) != null) {
keyWords.add(line.trim());
}
} catch (UnsupportedEncodingException e) {
LOG.info("敏感词库编码错误!");
} catch (FileNotFoundException e) {
LOG.info("敏感词库不存在!");
} catch (IOException e) {
LOG.info("读取敏感词库失败!");
} finally {
if (reader != null) {
try {
reader.close();
} catch (IOException e) {
LOG.info("读取敏感词库失败!");
}
}
}
return keyWords;
}
/**
* 初始化敏感词库
*/
private static void init() {
// 读取敏感词库
Set<String> keyWords = readSensitiveWords();
// 初始化根节点
root = new SensitiveWordNode(' ');
// 创建敏感词
for (String keyWord : keyWords) {
createSensitiveWordNode(keyWord);
}
}
/**
* 构建敏感词
*
* @param keyWord
*/
private static void createSensitiveWordNode(String keyWord) {
if (root == null) {
LOG.info("根节点不存在!");
return;
}
SensitiveWordNode nowNode = root;
for (Character c : keyWord.toCharArray()) {
SensitiveWordNode nextNode = nowNode.getNextNode(c);
if (nextNode == null) {
nextNode = new SensitiveWordNode(c);
nowNode.putNextNode(nextNode);
}
nowNode = nextNode;
}
nowNode.setEnd(true);
}
/**
* 检查敏感词
*
* @return 所有查出的敏感词
*/
private static String censorWords(String text) {
if (root == null) {
init();
}
StringBuilder sensitiveWords = new StringBuilder();
StringBuilder temp_sensitiveWords = new StringBuilder();
char[] text_to_char = text.toCharArray();
SensitiveWordNode sensitiveWordNode = root;
SensitiveWordNode this_sensitiveWordNode = null;
boolean flag;
for (int start = 0; start < text_to_char.length; start++) {
SensitiveWordNode temp_sensitiveWordNode = sensitiveWordNode.getNextNode(text_to_char[start]);
if (temp_sensitiveWordNode != null || this_sensitiveWordNode != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]) != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]).getKey() == text_to_char[start]) {
flag = true;
} else {
flag = false;
temp_sensitiveWords = new StringBuilder();
this_sensitiveWordNode = null;
}
if (flag) {
if (this_sensitiveWordNode != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]) == null) {
if (this_sensitiveWordNode.isEnd()) {
sensitiveWords.append(temp_sensitiveWords + ",");
this_sensitiveWordNode = null;
}
temp_sensitiveWords = new StringBuilder();
}
if (temp_sensitiveWordNode != null && temp_sensitiveWordNode.isEnd() || this_sensitiveWordNode != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]) != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]).getNextNodes() != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]).isEnd()
&& (start == text_to_char.length - 1
|| this_sensitiveWordNode.getNextNode(text_to_char[start]).getNextNodes().size() == 0)) {
temp_sensitiveWords.append(text_to_char[start] + ",");
sensitiveWords.append(temp_sensitiveWords);
temp_sensitiveWords = new StringBuilder();
this_sensitiveWordNode = null;
} else {
temp_sensitiveWords.append(text_to_char[start]);
if (this_sensitiveWordNode == null) {
this_sensitiveWordNode = temp_sensitiveWordNode;
}
if (this_sensitiveWordNode != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]) != null
&& this_sensitiveWordNode.getNextNode(text_to_char[start]).getNextNodes().size() != 0) {
this_sensitiveWordNode = this_sensitiveWordNode.getNextNode(text_to_char[start]);
}
}
}
}
return sensitiveWords.length() > 0 ?
sensitiveWords.subSequence(0, sensitiveWords.length() - 1).toString()
: "未匹配到敏感词";
}
public static void main(String[] args) {
long start_time = System.currentTimeMillis();
init();
System.out.println(System.currentTimeMillis() - start_time);
System.out.println(root);
start_time = System.currentTimeMillis();
String list = censorWords("语文数,学数学,学英英语语文文");
System.out.println(System.currentTimeMillis() - start_time);
System.out.println(list);
}
}
下面这种算法的核心跟上面类似,只不过加入了符号识别和大小写以及全角半角识别,识别率更好一些。
敏感词节点
public class WordNode {
private int value; // 节点名称
private List<WordNode> subNodes; // 子节点
private boolean isLast;// 默认false
public WordNode(int value) {
this.value = value;
}
public WordNode(int value, boolean isLast) {
this.value = value;
this.isLast = isLast;
}
/**
*
* @param subNode
* @return 就是传入的subNode
*/
private WordNode addSubNode(final WordNode subNode) {
if (subNodes == null)
subNodes = new LinkedList<WordNode>();
subNodes.add(subNode);
return subNode;
}
/**
* 有就直接返回该子节点, 没有就创建添加并返回该子节点
*
* @param value
* @return
*/
public WordNode addIfNoExist(final int value, final boolean isLast) {
if (subNodes == null) {
return addSubNode(new WordNode(value, isLast));
}
for (WordNode subNode : subNodes) {
if (subNode.value == value) {
if (!subNode.isLast && isLast)
subNode.isLast = true;
return subNode;
}
}
return addSubNode(new WordNode(value, isLast));
}
public WordNode querySub(final int value) {
if (subNodes == null) {
return null;
}
for (WordNode subNode : subNodes) {
if (subNode.value == value)
return subNode;
}
return null;
}
public boolean isLast() {
return isLast;
}
public void setLast(boolean isLast) {
this.isLast = isLast;
}
@Override
public int hashCode() {
return value;
}
}
敏感词树
public class WordFilter {
private static final FilterSet set = new FilterSet(); // 存储首字
private static final Map<Integer, WordNode> nodes = new HashMap<Integer, WordNode>(1024, 1); // 存储节点
private static final Set<Integer> stopwdSet = new HashSet<>(); // 停顿词
private static final char SIGN = '*'; // 敏感词过滤替换
static {
try {
long a = System.nanoTime();
init();
a = System.nanoTime() - a;
System.out.println("加载时间 : " + a + "ns");
System.out.println("加载时间 : " + a / 1000000 + "ms");
} catch (Exception e) {
throw new RuntimeException("初始化过滤器失败");
}
}
private static void init() {
// 获取敏感词
addSensitiveWord(readWordFromFile("wd.txt"));
addStopWord(readWordFromFile("stopwd.txt"));
}
/**
* 增加敏感词
* @param path
* @return
*/
private static List<String> readWordFromFile(String path) {
List<String> words;
BufferedReader br = null;
try {
br = new BufferedReader(new InputStreamReader(WordFilter.class.getClassLoader().getResourceAsStream(path),"UTF-8"));
words = new ArrayList<String>(1200);
for (String buf = ""; (buf = br.readLine()) != null;) {
if (buf == null || buf.trim().equals(""))
continue;
words.add(buf);
}
} catch (Exception e) {
throw new RuntimeException(e);
} finally {
try {
if (br != null)
br.close();
} catch (IOException e) {
}
}
return words;
}
/**
* 增加停顿词
*
* @param words
*/
private static void addStopWord(final List<String> words) {
if (words != null && words.size() > 0) {
char[] chs;
for (String curr : words) {
chs = curr.toCharArray();
for (char c : chs) {
stopwdSet.add(charConvert(c));
}
}
}
}
/**
* 添加DFA节点
* @param words
*/
private static void addSensitiveWord(final List<String> words) {
if (words != null && words.size() > 0) {
char[] chs;
int fchar;
int lastIndex;
WordNode fnode; // 首字母节点
for (String curr : words) {
chs = curr.toCharArray();
fchar = charConvert(chs[0]);
if (!set.contains(fchar)) {// 没有首字定义
set.add(fchar);// 首字标志位 可重复add,反正判断了,不重复了
fnode = new WordNode(fchar, chs.length == 1);
nodes.put(fchar, fnode);
} else {
fnode = nodes.get(fchar);
if (!fnode.isLast() && chs.length == 1)
fnode.setLast(true);
}
lastIndex = chs.length - 1;
for (int i = 1; i < chs.length; i++) {
fnode = fnode.addIfNoExist(charConvert(chs[i]), i == lastIndex);
}
}
}
}
/**
* 过滤判断 将敏感词转化为成屏蔽词
* @param src
* @return
*/
public static final String doFilter(final String src) {
char[] chs = src.toCharArray();
int length = chs.length;
int currc;
int k;
WordNode node;
for (int i = 0; i < length; i++) {
currc = charConvert(chs[i]);
if (!set.contains(currc)) {
continue;
}
node = nodes.get(currc);// 日 2
if (node == null)// 其实不会发生,习惯性写上了
continue;
boolean couldMark = false;
int markNum = -1;
if (node.isLast()) {// 单字匹配(日)
couldMark = true;
markNum = 0;
}
// 继续匹配(日你/日你妹),以长的优先
// 你-3 妹-4 夫-5
k = i;
for (; ++k < length;) {
int temp = charConvert(chs[k]);
if (stopwdSet.contains(temp))
continue;
node = node.querySub(temp);
if (node == null)// 没有了
break;
if (node.isLast()) {
couldMark = true;
markNum = k - i;// 3-2
}
}
if (couldMark) {
for (k = 0; k <= markNum; k++) {
chs[k + i] = SIGN;
}
i = i + markNum;
}
}
return new String(chs);
}
/**
* 是否包含敏感词
* @param src
* @return
*/
public static final boolean isContains(final String src) {
char[] chs = src.toCharArray();
int length = chs.length;
int currc;
int k;
WordNode node;
for (int i = 0; i < length; i++) {
currc = charConvert(chs[i]);
if (!set.contains(currc)) {
continue;
}
node = nodes.get(currc);// 日 2
if (node == null)// 其实不会发生,习惯性写上了
continue;
boolean couldMark = false;
if (node.isLast()) {// 单字匹配(日)
couldMark = true;
}
// 继续匹配(日你/日你妹),以长的优先
// 你-3 妹-4 夫-5
k = i;
for (; ++k < length;) {
int temp = charConvert(chs[k]);
if (stopwdSet.contains(temp))
continue;
node = node.querySub(temp);
if (node == null)// 没有了
break;
if (node.isLast()) {
couldMark = true;
}
}
if (couldMark) {
return true;
}
}
return false;
}
/**
* 大写转化为小写 全角转化为半角
*
* @param src
* @return
*/
private static int charConvert(char src) {
int r = BCConvert.qj2bj(src);
return (r >= 'A' && r <= 'Z') ? r + 32 : r;
}
public static void main(String args[]) {
System.out.println(doFilter("日%%##你你妈"));
}
}
FilterSet,枚举了0~65535的所有char是否是某个敏感词开头的状态
public class FilterSet{
private final long[] elements;
public FilterSet() {
elements = new long[1 + (65535 >>> 6)];
}
public void add(final int no) {
elements[no >>> 6] |= (1L << (no & 63));
}
public void add(final int... no) {
for(int currNo : no)
elements[currNo >>> 6] |= (1L << (currNo & 63));
}
public void remove(final int no) {
elements[no >>> 6] &= ~(1L << (no & 63));
}
/**
*
* @param no
* @return true:添加成功 false:原已包含
*/
public boolean addAndNotify(final int no) {
int eWordNum = no >>> 6;
long oldElements = elements[eWordNum];
elements[eWordNum] |= (1L << (no & 63));
boolean result = elements[eWordNum] != oldElements;
// if (result)
// size++;
return result;
}
/**
*
* @param no
* @return true:移除成功 false:原本就不包含
*/
public boolean removeAndNotify(final int no) {
int eWordNum = no >>> 6;
long oldElements = elements[eWordNum];
elements[eWordNum] &= ~(1L << (no & 63));
boolean result = elements[eWordNum] != oldElements;
return result;
}
public boolean contains(final int no) {
return (elements[no >>> 6] & (1L << (no & 63))) != 0;
}
public boolean containsAll(final int... no) {
if(no.length==0)
return true;
for(int currNo : no)
if((elements[currNo >>> 6] & (1L << (currNo & 63))) == 0)
return false;
return true;
}
/**
* 不如直接循环调用contains
* @param no
* @return
*/
public boolean containsAll_ueslessWay(final int... no) {
long[] elements = new long[this.elements.length];
for(int currNo : no){
elements[currNo >>> 6] |= (1L << (currNo & 63));
}//这一步执行完跟循环调用contains差不多了
for (int i = 0; i < elements.length; i++)
if ((elements[i] & ~this.elements[i]) != 0)
return false;
return true;
}
/**
* 目前没有去维护size,每次都是去计算size
* @return
*/
public int size() {
int size = 0;
for (long element : elements)
size += Long.bitCount(element);
return size;
}
public static void main(String[] args) {
FilterSet oi = new FilterSet();
System.out.println(oi.elements.length);
}
}
全角/半角转换
public class BCConvert {
/**
* ASCII表中可见字符从!开始,偏移位值为33(Decimal)
*/
static final char DBC_CHAR_START = 33; // 半角!
/**
* ASCII表中可见字符到~结束,偏移位值为126(Decimal)
*/
static final char DBC_CHAR_END = 126; // 半角~
/**
* 全角对应于ASCII表的可见字符从!开始,偏移值为65281
*/
static final char SBC_CHAR_START = 65281; // 全角!
/**
* 全角对应于ASCII表的可见字符到~结束,偏移值为65374
*/
static final char SBC_CHAR_END = 65374; // 全角~
/**
* ASCII表中除空格外的可见字符与对应的全角字符的相对偏移
*/
static final int CONVERT_STEP = 65248; // 全角半角转换间隔
/**
* 全角空格的值,它没有遵从与ASCII的相对偏移,必须单独处理
*/
static final char SBC_SPACE = 12288; // 全角空格 12288
/**
* 半角空格的值,在ASCII中为32(Decimal)
*/
static final char DBC_SPACE = ' '; // 半角空格
/**
* <PRE>
* 半角字符->全角字符转换
* 只处理空格,!到˜之间的字符,忽略其他
* </PRE>
*/
public static String bj2qj(String src) {
if (src == null) {
return src;
}
StringBuilder buf = new StringBuilder(src.length());
char[] ca = src.toCharArray();
for (int i = 0; i < ca.length; i++) {
if (ca[i] == DBC_SPACE) { // 如果是半角空格,直接用全角空格替代
buf.append(SBC_SPACE);
} else if ((ca[i] >= DBC_CHAR_START) && (ca[i] <= DBC_CHAR_END)) { // 字符是!到~之间的可见字符
buf.append((char) (ca[i] + CONVERT_STEP));
} else { // 不对空格以及ascii表中其他可见字符之外的字符做任何处理
buf.append(ca[i]);
}
}
return buf.toString();
}
/**
* 半角转换全角
*
* @param src
* @return
*/
public static int bj2qj(char src) {
int r = src;
if (src == DBC_SPACE) { // 如果是半角空格,直接用全角空格替代
src = SBC_SPACE;
} else if ((src >= DBC_CHAR_START) && (src <= DBC_CHAR_END)) { // 字符是!到~之间的可见字符
r = src + CONVERT_STEP;
}
return r;
}
/**
* <PRE>
* 全角字符->半角字符转换
* 只处理全角的空格,全角!到全角~之间的字符,忽略其他
* </PRE>
*/
public static String qj2bj(String src) {
if (src == null) {
return src;
}
StringBuilder buf = new StringBuilder(src.length());
char[] ca = src.toCharArray();
for (int i = 0; i < src.length(); i++) {
if (ca[i] >= SBC_CHAR_START && ca[i] <= SBC_CHAR_END) { // 如果位于全角!到全角~区间内
buf.append((char) (ca[i] - CONVERT_STEP));
} else if (ca[i] == SBC_SPACE) { // 如果是全角空格
buf.append(DBC_SPACE);
} else { // 不处理全角空格,全角!到全角~区间外的字符
buf.append(ca[i]);
}
}
return buf.toString();
}
/**
* 全角转换半角
*
* @param src
* @return
*/
public static int qj2bj(char src) {
int r = src;
if (src >= SBC_CHAR_START && src <= SBC_CHAR_END) { // 如果位于全角!到全角~区间内
r = src - CONVERT_STEP;
} else if (src == SBC_SPACE) { // 如果是全角空格
r = DBC_SPACE;
}
return r;
}
}
以上两种写法的代码我都托管在了这个地址上,当然最好还是自己动手写一写,这样更便于理解。