@Test
public void testOnUnknownSynsets() {
assertEquals(
LeacockChodorow.min, rc.calcRelatednessOfSynset(null, n1Synsets.get(0)).getScore(), 0.0001);
assertEquals(LeacockChodorow.min, rc.calcRelatednessOfWords(null, n1), 0.0001);
assertEquals(LeacockChodorow.min, rc.calcRelatednessOfWords("", n1), 0.0001);
}
/**
* Test method for {@link
* edu.cmu.lti.similarity.RelatednessCalculator#calcRelatednessOfWords(java.lang.String,
* java.lang.String)}.
*/
@Test
public void testHappyPathOnWordsWithPOS() {
assertEquals(2.5902, rc.calcRelatednessOfWords(nv1 + "#n", nv2 + "#n"), 0.0001D);
assertEquals(0.0000, rc.calcRelatednessOfWords(nv1 + "#n", nv2 + "#v"), 0.0001D);
assertEquals(0.0000, rc.calcRelatednessOfWords(nv1 + "#v", nv2 + "#n"), 0.0001D);
assertEquals(2.2335, rc.calcRelatednessOfWords(nv1 + "#v", nv2 + "#v"), 0.0001D);
assertEquals(0.0000, rc.calcRelatednessOfWords(nv1 + "#other", nv2 + "#other"), 0.0001D);
}
/**
* Test method for {@link
* edu.cmu.lti.similarity.impl.LeacockChodorow#calcRelatedness(java.lang.String,
* java.lang.String)}.
*/
@Test
public void testHappyPathOnSynsets() {
// both japanese and english!, n and v
// English pair
assertEquals(
2.9957, rc.calcRelatednessOfSynset(n1Synsets.get(1), n2Synsets.get(0)).getScore(), 0.0001);
assertEquals(
1.6094, rc.calcRelatednessOfSynset(n1Synsets.get(0), n2Synsets.get(0)).getScore(), 0.0001);
assertEquals(
2.6391, rc.calcRelatednessOfSynset(v1Synsets.get(0), v2Synsets.get(0)).getScore(), 0.0001);
assertEquals(
1.9459, rc.calcRelatednessOfSynset(v1Synsets.get(1), v2Synsets.get(0)).getScore(), 0.0001);
}
public static double run(String word1, String word2) {
WS4JConfiguration.getInstance().setMFS(true);
String word1_v = word1.substring(0, word1.indexOf('_'));
String word1_n = word1.substring(word1.indexOf('_') + 1);
String word2_v = word2.substring(0, word2.indexOf('_'));
String word2_n = word2.substring(word2.indexOf('_') + 1);
double s1 = rc.calcRelatednessOfWords(word1_v, word2_v);
double s2 = rc.calcRelatednessOfWords(word1_n, word2_n);
return s1 + s2;
}
public static double similarity(String w1, String w2) {
WS4JConfiguration.getInstance().setMFS(true);
if (w1 != null) w1 = w1.toLowerCase().replaceAll("-", "").replaceAll(" ", "");
if (w2 != null) w2 = w2.toLowerCase().replaceAll("-", "").replaceAll(" ", "");
// double res = 0.0d;
// double res1 = simCalc.calcRelatednessOfWords(w1, w2);
// res = res1;
// double res2 = simCalc.calcRelatednessOfWords(w1.replace('-', ' '),
// w2.replace('-', ' '));
// if (res2 > res)
// res = res2;
// double res3 = simCalc.calcRelatednessOfWords(w1.replaceAll("-", ""),
// w2.replaceAll("-", ""));
// if (res3 > res)
// res = res3;
return simCalc.calcRelatednessOfWords(w1, w2);
}
/**
* Test method for {@link
* edu.cmu.lti.similarity.RelatednessCalculator#calcRelatednessOfWords(java.lang.String,
* java.lang.String)}.
*/
@Test
public void testHappyPathOnWords() {
assertEquals(2.9957, rc.calcRelatednessOfWords(n1, n2), 0.0001D);
assertEquals(2.6391, rc.calcRelatednessOfWords(v1, v2), 0.0001D);
}
public static void main(String[] args) throws ClassNotFoundException, IOException {
int nbPCA = 250;
// String sourcePath = "/Vrac/3000693/RI_Image/bows_" + nbPCA + ".ser";
String sourcePath = "/Users/remicadene/Dropbox/_Docs/UPMC/RI/bows_" + nbPCA + ".ser";
int dimpsi = 0;
Set<String> classes;
// Modèle MultiClass
// --------------------------------------------------------------------------------
// Hyperparamètres
double eps = 1e-2;
double lambda = 1e-6;
int maxIter = 100;
// Définition du modèle
DataSet<double[], String> dataSet = VisualIndexes.load(sourcePath);
classes = dataSet.outputs();
double min = Double.MAX_VALUE;
double max = -Double.MAX_VALUE;
int n = classes.size();
double[][] distances = new double[n][n];
double[][] new_distances = new double[n][n];
ILexicalDatabase db = new NictWordNet();
RelatednessCalculator calculator = new WuPalmer(db);
HashMap<String, Integer> hm_classes = new HashMap<String, Integer>();
int id = 0;
for (String classe : classes) {
hm_classes.put(classe, id);
id++;
}
for (String word1 : classes) {
int i = hm_classes.get(word1);
for (String word2 : classes) {
int j = hm_classes.get(word2);
// this.distances[i][j] = calculator.calcRelatednessOfWords(word1, word2);
if (i == j) {
distances[i][j] = 0;
} else {
distances[i][j] = 1 - calculator.calcRelatednessOfWords(word1, word2);
if (distances[i][j] < min) {
min = distances[i][j];
}
if (distances[i][j] > max) {
max = distances[i][j];
}
}
}
}
// [a; b] -> [0.1 ; 2]
// (1.9 * (x - a) / (b - a)) + .1
for (int i = 0; i < n; i++) {
for (int j = 0; j < n; j++) {
if (i != j) {
new_distances[i][j] = (distances[i][j] - min) * (2 - 0.1) / (max - min) + 0.1;
}
}
}
System.out.println("--------------");
for (int i = 0; i < classes.size(); i++) {
for (int j = i + 1; j < classes.size(); j++) {
System.out.println("[" + i + "," + j + "]");
System.out.println("avant: " + distances[i][j]);
System.out.println("apres: " + new_distances[i][j]);
}
}
System.out.println(max);
System.out.println(min);
double a =
0.1 + (2 - 0.1) * (0.6 - 0.0714285714285714) / (0.15384615384615385 - 0.0714285714285714);
}