由于现在使用拼音输入法较多,很可能输入同音字而无法搜索到结果,或者由于普通话不标准(前后鼻音不分)输入了错误的拼音。例如用户想搜索”牛奶“却输入了刘来(liulai),那我们可以对liulai进行模糊音扩展,然后一起放到搜索引擎中去搜索。为了避免过度扩展曲解了用户的意图,对于扩展词需要进行排序,越少改动越排在前面(原文排在最前),越多改动越排在后面。在搜索时按照顺序赋予不同的权重。
类似这样:
"query": {"bool":{
"should":[
{"match" : { "t.py":{"query": "lanqiuxie","analyzer":"whitespace","operator":"or" ,"boost":3}}},
{"match" : { "t.py":{"query": "langqiuxie","analyzer":"whitespace","operator":"or" ,"boost":3}}},
{"match" : { "t.py":{"query": "nangqiuxie","analyzer":"whitespace","operator":"or" ,"boost":2}}},
{"match" : { "t.py":{"query": "nanqiuxie","analyzer":"whitespace","operator":"or" ,"boost":1}}}]}}}
好,现在show一下模糊拼音扩展的算法:
1.声母l和n互换
2.韵母有g和没g互换。
3.变换少的排在前面。
例如 ["liulai"] => ["liulai","liunai","niulai","niunai"]
//汉语拼音模糊音扩展
static public List<String> expand(String[] ss) {
int n = ss.length;
List<String>[] lists = new ArrayList[n];
List<Integer> changes = new ArrayList<>();
for (int i = 0; i < n; i++) {
lists[i] = expandOne(ss[i]);
if (lists[i].size() > 1) changes.add(i);
}
int max = changes.size();
if (max > 4) max = 4;
List<String> ans = new ArrayList<>();
for (int i = 0; i <= max; i++) {
ans.addAll(miss(lists, changes, i, 0, 0));
}
return ans;
}
static private Map<String, String> confusingMap = new HashMap<String, String>() {{
this.put("in", "ing");
this.put("an", "ang");
this.put("en", "eng");
this.put("un", "ong");
}};
static private List<String> miss(List<String>[] lists, List<Integer> changes, int c, int b1, int b2) {
List<String> list = new ArrayList<>();
if (c > changes.size() - b2 || c < 0) return list;
int n = lists.length;
if (c == 0) {
StringBuilder sb = new StringBuilder();
for (int i = b1; i < n; i++) {
sb.append(lists[i].get(0));
}
list.add(sb.toString());
} else {
int fc = changes.get(b2);
StringBuilder sb = new StringBuilder();
for (int i = b1; i < fc; i++) {
sb.append(lists[i].get(0));
}
String pre = sb.toString();
List<String> left = miss(lists, changes, c - 1, fc + 1, b2 + 1);
for (int i = 1; i < lists[fc].size(); i++) {
for (String s : left) {
list.add(pre + lists[fc].get(i) + s);
}
}
pre += lists[fc].get(0);
left = miss(lists, changes, c, fc + 1, b2 + 1);
for (String s : left) {
list.add(pre + s);
}
}
return list;
}
static private List<String> expandOne(String py) {
List<String> list = new ArrayList<>();
list.add(py);
if (py.equals("yang") || py.equals("yan")) return list;
if (py.startsWith("l") || py.startsWith("n")) {
List<String> vs = changeVowel(py.substring(1));
String[] ln = py.startsWith("l") ? new String[]{"l", "n"}
: new String[]{"n", "l"};
for (String sm : ln) {
for (String s : vs) {
String nw = sm + s;
if (!nw.equals(py)) list.add(nw);
}
}
} else {
List<String> vs = changeVowel(py);
return vs;
}
return list;
}
static private List<String> changeVowel(String s) {
List<String> list = new ArrayList<>();
list.add(s);
if (s.endsWith("iang") || s.endsWith("ian")) return list;
int len = s.length();
for (Map.Entry<String, String> en : confusingMap.entrySet()) {
String k = en.getKey(), v = en.getValue();
if (s.endsWith(k)) {
list.add(s.substring(0, len - k.length()) + v);
break;
} else if (s.endsWith(v)) {
list.add(s.substring(0, len - v.length()) + k);
break;
}
}
return list;
}
public static void main(String[] args) {
System.out.println(expand(new String[]{"niu","nai"}));
}