addition of doi normalization in PidMatch comparator, addition of keywordsclustering (clustering based on terms in the translation maps for the organizations), minor changes

dnet-pace-core/src/main/java/eu/dnetlib/pace/clustering/KeywordsClustering.java

@@ -0,0 +1,39 @@
+package eu.dnetlib.pace.clustering;
+
+import com.google.common.base.Joiner;
+import eu.dnetlib.pace.common.AbstractPaceFunctions;
+
+import java.util.*;
+
+@ClusteringClass("keywordsclustering")
+public class KeywordsClustering extends AbstractClusteringFunction {
+
+    private static Map<String,String> translationMap = AbstractPaceFunctions.loadMapFromClasspath("/eu/dnetlib/pace/config/translation_map.csv");
+
+    private static Map<String,String> cityMap = AbstractPaceFunctions.loadMapFromClasspath("/eu/dnetlib/pace/config/city_map.csv");
+
+    public KeywordsClustering(Map<String, Integer> params) {
+        super(params);
+    }
+
+    @Override
+    protected Collection<String> doApply(String s) {
+
+        List<String> keywords = getCodes(s, translationMap, params.getOrDefault("windowSize", 4));
+        List<String> cities = getCodes(s, cityMap, params.getOrDefault("windowSize", 4));
+
+        final Collection<String> combinations = new LinkedHashSet<String>();
+
+        int size = 0;
+        for (String keyword: keywords){
+            for (String city: cities) {
+                combinations.add(keyword+"-"+city);
+                if (++size>params.getOrDefault("max", 2)) {
+                    return combinations;
+                }
+            }
+        }
+
+        return combinations;
+    }
+}

dnet-pace-core/src/main/java/eu/dnetlib/pace/common/AbstractPaceFunctions.java

@@ -40,6 +40,8 @@ public abstract class AbstractPaceFunctions {
 	private static final String aliases_from = "⁰¹²³⁴⁵⁶⁷⁸⁹⁺⁻⁼⁽⁾ⁿ₀₁₂₃₄₅₆₇₈₉₊₋₌₍₎àáâäæãåāèéêëēėęîïíīįìôöòóœøōõûüùúūßśšłžźżçćčñń";
 	private static final String aliases_to = "0123456789+-=()n0123456789+-=()aaaaaaaaeeeeeeeiiiiiioooooooouuuuussslzzzcccnn";
 
+	public final String DOI_PREFIX = "(https?:\\/\\/dx\\.doi\\.org\\/)|(doi:)";
+
 	protected final static FieldList EMPTY_FIELD = new FieldListImpl();
 
 	protected String concat(final List<String> l) {
@@ -313,5 +315,41 @@ public abstract class AbstractPaceFunctions {
 		return Iterables.limit(Splitter.on(" ").omitEmptyStrings().trimResults().split(s), maxTokens);
 	}
 
+	public String normalizePid(String pid) {
+		return pid.toLowerCase().replaceAll(DOI_PREFIX, "");
+	}
+
+	//get the list of codes into the input string
+	public List<String> getCodes(String s1, Map<String, String> translationMap, int windowSize){
+
+		String s = cleanup(s1);
+
+		s = filterAllStopWords(s);
+
+		List<String> tokens = Arrays.asList(s.toLowerCase().split(" "));
+
+		List<String> codes = new ArrayList<>();
+
+		if (tokens.size()<windowSize)
+			windowSize = tokens.size();
+
+		int length = windowSize;
+
+		while (length != 0) {
+
+			for (int i = 0; i<=tokens.size()-length; i++){
+				String candidate = Joiner.on(" ").join(tokens.subList(i, i + length));
+				if (translationMap.containsKey(candidate)) {
+					codes.add(translationMap.get(candidate));
+					s = s.replace(candidate, "");
+				}
+			}
+
+			tokens = Arrays.asList(s.split(" "));
+			length-=1;
+		}
+
+		return codes;
+	}
 
 }

dnet-pace-core/src/main/java/eu/dnetlib/pace/condition/PidMatch.java

@@ -52,9 +52,11 @@ public class PidMatch extends AbstractCondition {
 		return new ConditionEval(cond, a, b, result);
 	}
 
+	//lowercase + normalization of the pid before adding it to the set
 	private Set<String> toHashSet(List<Pid> pbl) {
+
 		return pbl.stream()
-					.map(pid -> pid.getType() + pid.getValue())
+					.map(pid -> pid.getType() + normalizePid(pid.getValue()))
 					.collect(Collectors.toCollection(HashSet::new));
 	}
 

dnet-pace-core/src/main/java/eu/dnetlib/pace/distance/algo/JaroWinklerNormalizedName.java

@@ -43,8 +43,11 @@ public class JaroWinklerNormalizedName extends SecondStringDistanceAlgo {
         String ca = cleanup(a);
         String cb = cleanup(b);
 
-        ca = removeStopwords(ca);
-        cb = removeStopwords(cb);
+        ca = normalize(ca);
+        cb = normalize(cb);
+
+        ca = filterAllStopWords(ca);
+        cb = filterAllStopWords(cb);
 
         //replace keywords with codes
         String codesA = keywordsToCode(ca, translationMap, params.getOrDefault("windowSize", 4).intValue());
@@ -80,16 +83,4 @@ public class JaroWinklerNormalizedName extends SecondStringDistanceAlgo {
         return d;
     }
 
-    public String removeStopwords(String s) {
-        String normString = normalize(s);
-
-        normString = filterStopWords(normString, stopwordsIt);
-        normString = filterStopWords(normString, stopwordsEn);
-        normString = filterStopWords(normString, stopwordsDe);
-        normString = filterStopWords(normString, stopwordsFr);
-        normString = filterStopWords(normString, stopwordsPt);
-        normString = filterStopWords(normString, stopwordsEs);
-
-        return normString;
-    }
 }

dnet-pace-core/src/main/resources/eu/dnetlib/pace/config/city_map.csv

@@ -7915,7 +7915,7 @@ city::5134086;Rochester;Ga-sko-sa-ga;Gaskosago;ROC;Rocestera;Rocesteris;Rocestro
 city::5136454;Schenectady;SCH;Schenectady;Shinetadi;Skanehtati;Skanéhtati;Skenektadi;Skenektadis;Skunektadi;seukinegteodi;si ke nei ke ta di;sknktdy  nywywrk;sknyktady;sqnqtdy;sukenekutadi;Скенектади;Скенектаді;Скънектади;Шинетади;סקנקטדי;سكنيكتادي;سکنکتدی، نیویورک;سکینکٹڈی ، نیویارک;سکینیکٹیڈی، نیو یارک;スケネクタディ;斯克内克塔迪;스키넥터디;
 city::5137849;Sheepshead Bay;;
 city::5139568;Staten Island;Borough of Staten Island;Staten Island;
-city::5140405;Syracuse;Bogardus Corners;Cossitts Corners;Gorad Sirakjus;Kah-ya-hoo-neh;Ken-tue-ho-ne;Milan;Na-ta-dunk;SYR;Siracusa;Sirak'jus;Sirakjus;Sirakjuz;Sirakuso;Sirakuz;South Salina;Sy-kuse;Syracusae;Syracuse;Syracuse i New York;Tu-na-ten-tonk;sayrakywz;shirakyusu;shirakyuzu;sileokyuseu;sirekyuja;syrakywz;syrakywz  nyw yark;syrakywz  nywywrk;syrqywz;xi la qiu ci;Горад Сіракюс;Сиракуз;Сиракьюс;Сиракюз;Сиракјус;Сіракюс;סירקיוז;سائراکیوز;سيراكيوز;سیراکیوز، نیو یارک;سیراکیوز، نیویورک;सिरॅक्युज;სირაკიუსი;シラキュース;シラキューズ;锡拉丘兹;시러큐스;
+city::5140405;Syracuse;Bogardus Corners;Cossitts Corners;Gorad Sirakjus;Kah-ya-hoo-neh;Ken-tue-ho-ne;Na-ta-dunk;SYR;Siracusa;Sirak'jus;Sirakjus;Sirakjuz;Sirakuso;Sirakuz;South Salina;Sy-kuse;Syracusae;Syracuse;Syracuse i New York;Tu-na-ten-tonk;sayrakywz;shirakyusu;shirakyuzu;sileokyuseu;sirekyuja;syrakywz;syrakywz  nyw yark;syrakywz  nywywrk;syrqywz;xi la qiu ci;Горад Сіракюс;Сиракуз;Сиракьюс;Сиракюз;Сиракјус;Сіракюс;סירקיוז;سائراکیوز;سيراكيوز;سیراکیوز، نیو یارک;سیراکیوز، نیویورک;सिरॅक्युज;სირაკიუსი;シラキュース;シラキューズ;锡拉丘兹;시러큐스;
 city::5142056;Utica;Fort Schuyler;Gorad Jutyka;Iotekha;Iotékha;Jutika;Nundadasis;Old Fort Schuyler;Tevadahahtodague;Twa-dah-ah-lo-dah-que;UCA;Unungagages;Utica;Utika;Yanundadasis;ZUA;awtyka;you ti ka;yutika;ywtyka  nywywrk;ywtyqh;Јутика;Горад Ютыка;Утика;Ютика;Ютіка;יוטיקה;أوتيكا;یوتیکا، نیویورک;یوٹیکا، نیو یارک;ユーティカ;由提卡;
 city::5143056;Wakefield;;
 city::5143307;Washington Heights;Harlem Heights;Pen-a-bick;Washington Heights;washintonhaitsu;wosingteonhaicheu;ワシントンハイツ;워싱턴하이츠;

dnet-pace-core/src/main/resources/eu/dnetlib/pace/config/stopwords_it.txt

@@ -332,7 +332,6 @@ male
 malgrado
 malissimo
 mancanza
-marche
 me
 medesimo
 mediante

dnet-pace-core/src/main/resources/eu/dnetlib/pace/config/translation_map.csv

@@ -38,7 +38,7 @@ key::37;federation;federazione;fédération;федерация;federatie;ομο
 key::38;observatory;osservatorio;observatoire;обсерватория;observatorium;αστεροσκοπείο
 key::39;bureau;ufficio;bureau;офис;bureau;γραφείο
 key::40;company;impresa;compagnie;société;компания;bedrijf;εταιρία
-key::41;polytechnic;politecnico;polytechnique;политехника;polytechnisch;πολυτεχνείο
+key::41;polytechnic;politecnico;polytechnique;политехника;polytechnisch;πολυτεχνείο;universita politecnica;polytechnic university
 key::42;coalition;coalizione;coalition;коалиция;coalitie;συνασπισμός
 key::43;initiative;iniziativa;initiative;инициатива;initiatief;πρωτοβουλία
 key::44;academic;accademico;académique;universitaire;акадеческий academisch;ακαδημαϊκός;ακαδημαϊκή;ακαδημαϊκό;ακαδημαϊκές;ακαδημαϊκοί

dnet-pace-core/src/test/java/eu/dnetlib/pace/clustering/ClusteringFunctionTest.java

@@ -117,4 +117,19 @@ public class ClusteringFunctionTest extends AbstractPaceTest {
 		System.out.println(cf.apply(Lists.newArrayList(person(s))));
 	}
 
+	@Test
+	public void testKeywordsClustering() {
+
+		final ClusteringFunction cf = new KeywordsClustering(params);
+		final String s = "Polytechnic University of Turin";
+		System.out.println(s);
+		System.out.println(cf.apply(Lists.newArrayList(title(s))));
+
+		final String s1 = "POLITECNICO DI TORINO";
+		System.out.println(s1);
+		System.out.println(cf.apply(Lists.newArrayList(title(s1))));
+
+
+	}
+
 }

dnet-pace-core/src/test/java/eu/dnetlib/pace/common/PaceFunctionTest.java

@@ -0,0 +1,25 @@
+package eu.dnetlib.pace.common;
+
+import org.junit.Test;
+
+import static junit.framework.Assert.assertEquals;
+import static junit.framework.Assert.assertTrue;
+
+public class PaceFunctionTest extends AbstractPaceFunctions {
+
+    @Test
+    public void normalizePidTest(){
+
+        assertEquals("identifier", normalizePid("IdentifIer"));
+        assertEquals("10.1109/tns.2015.2493347", normalizePid("10.1109/TNS.2015.2493347"));
+        assertEquals("10.0001/testdoi", normalizePid("http://dx.doi.org/10.0001/testDOI"));
+        assertEquals("10.0001/testdoi", normalizePid("https://dx.doi.org/10.0001/testDOI"));
+
+    }
+
+    @Test
+    public void filterAllStopwordsTest(){
+
+        assertEquals("universita politecnica marche", filterAllStopWords("universita politecnica delle marche"));
+    }
+}

dnet-pace-core/src/test/java/eu/dnetlib/pace/condition/ConditionTest.java

@@ -0,0 +1,7 @@
+package eu.dnetlib.pace.condition;
+
+import eu.dnetlib.pace.AbstractPaceTest;
+
+public class ConditionTest extends AbstractPaceTest {
+
+}

dnet-pace-core/src/test/java/eu/dnetlib/pace/distance/DistanceAlgoTest.java

@@ -104,4 +104,14 @@ public class DistanceAlgoTest extends AbstractPaceFunctions {
 
     }
 
+    @Test
+	public void testJaroWinklerNormalizedName7() {
+
+		final JaroWinklerNormalizedName jaroWinklerNormalizedName = new JaroWinklerNormalizedName(params);
+		double result = jaroWinklerNormalizedName.distance("Polytechnic University of Turin", "POLITECNICO DI TORINO");
+
+		System.out.println("result = " + result);
+		assertTrue(result> 0.9);
+	}
+
 }