/**
* Returns the ranking <i>k</i> such that {@code other} is the k<sup>th</th> most similar word to
* {@code target} in the semantic space.
*/
private int findRank(SemanticSpace sspace, String target, String other) {
Vector v1 = sspace.getVector(target);
Vector v2 = sspace.getVector(other);
// Compute the base similarity between the two words
double baseSim = Similarity.cosineSimilarity(v1, v2);
int rank = 0;
// Next, count how many words are more similar to the target than the
// other word is
for (String word : sspace.getWords()) {
Vector v = sspace.getVector(word);
double sim = Similarity.cosineSimilarity(v1, v);
if (sim > baseSim) rank++;
}
return rank;
}
/** Saves the {@link BasisMapping} created from the {@link OccurrenceCounter}. */
protected void saveSSpace(SemanticSpace sspace, File outputFile) throws IOException {
BasisMapping<String, String> savedTerms = new StringBasisMapping();
for (String term : sspace.getWords()) savedTerms.getDimension(term);
ObjectOutputStream ouStream = new ObjectOutputStream(new FileOutputStream(outputFile));
ouStream.writeObject(savedTerms);
ouStream.close();
}
/**
* Returns the association of the two words on a scale of 0 to 1.
*
* @return the assocation or {@code null} if either {@code word1} or {@code word2} are not in the
* semantic space
*/
protected Double computeAssociation(SemanticSpace sspace, String word1, String word2) {
Vector v1 = sspace.getVector(word1);
Vector v2 = sspace.getVector(word2);
if (v1 == null || v2 == null) return null;
// Find the ranks of each of the two words to each other
double rank1 = findRank(sspace, word1, word2);
double rank2 = findRank(sspace, word2, word1);
return 2d / (rank1 + rank2);
}