mazdac/Szpitale-graph
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

T2: NameNormalizer testy + transliteratePolish, token-strip titles [checkpoint]

Browse Source
This commit is contained in:
Artur Kruszewski
2026-07-07 23:49:04 +02:00
parent 3914669c14
commit afe107cea0
19 changed files with 1649 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files
szpitale-graph/src/main/java/com/developx/szpitale/ingest/normalize/NameNormalizer.java
+40 -8
View File
@@ -20,12 +20,31 @@ public class NameNormalizer {
public String makeSlug(String fullName) {
String normalized = Normalizer.normalize(fullName, Normalizer.Form.NFKD);
normalized = normalized.replaceAll("\\p{M}", "");
normalized = normalized.replaceAll("[^a-zA-Z0-9\\s-]", " ");
normalized = normalized.toLowerCase().replaceAll("\\s+", "-").replaceAll("-+", "-");
normalized = normalized.replaceAll("[^\\p{L}\\s-]", "");
normalized = transliteratePolish(normalized.toLowerCase());
normalized = normalized.replaceAll("\\s+", "-").replaceAll("-+", "-");
normalized = normalized.replaceAll("^-|-$", "");
return normalized;
}
private String transliteratePolish(String input) {
Map<String, String> map = new HashMap<>();
map.put("ł", "l"); map.put("Ł", "l");
map.put("ę", "e"); map.put("Ĕ", "e");
map.put("ó", "o"); map.put("Ō", "o");
map.put("ą", "a"); map.put("Ą", "a");
map.put("ś", "s"); map.put("Ś", "s");
map.put("ź", "z"); map.put("Ź", "z");
map.put("ż", "z"); map.put("Ż", "z");
map.put("ć", "c"); map.put("Ć", "c");
map.put("ń", "n"); map.put("Ń", "n");
String result = input;
for (Map.Entry<String, String> entry : map.entrySet()) {
result = result.replace(entry.getKey(), entry.getValue());
}
return result;
}
public String extractFullName(String rawName) {
return extractGivenAndFamilyNames(rawName);
}
@@ -62,12 +81,25 @@ public class NameNormalizer {
.collect(Collectors.toList());
}
static final List<String> TITLE_PREFIXES = List.of(
"prof. dr hab.", "prof.dr hab.", "prof. dr.", "prof. dr",
"dr hab.", "dr.hab.", "dr. hab.", "dr.hab",
"mgr inż.", "mgr inż", "mgr.inż.", "mgr.inż",
"prof.", "prof", "dr.", "dr",
"mgr.", "mgr", "inż.", "inż",
"lek.", "lek.med.", "n. med.", "n. o zdr.",
"licencjat inż.", "licencjat inż"
);
private String extractGivenAndFamilyNames(String rawName) {
Pattern pattern = Pattern.compile(
"(?i)(prof\\.\\s*|dr\\s*hab\\.\\s*|dr\\.\\s*|mgr\\s*|in\u017c\\.\\s*)",
Pattern.MULTILINE | Pattern.DOTALL
);
Matcher matcher = pattern.matcher(rawName);
return matcher.replaceAll("").trim();
String name = rawName.trim();
String lower = name.toLowerCase();
for (String prefix : TITLE_PREFIXES) {
if (lower.startsWith(prefix.toLowerCase())) {
name = name.substring(prefix.length()).trim();
lower = name.toLowerCase();
}
}
return name;
}
}