/**
* 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;
}
/**
* Computes the similarity of the two edges as the Jaccard index of the neighbors of two impost
* nodes. The impost nodes are the two nodes the edges do not have in common. Subclasses may
* override this method to define a new method for computing edge similarity.
*
* <p><i>Implementation Note</i>: Subclasses that wish to override this behavior should be aware
* that this method is likely to be called by multiple threads and therefor should make provisions
* to be thread safe. In addition, this method may be called more than once per edge pair if the
* similarity matrix is being computed on-the-fly.
*
* @param sm a matrix containing the connections between edges. A non-zero value in location (i,j)
* indicates a node <i>i</i> is connected to node <i>j</i> by an edge.
* @param e1 an edge to be compared with {@code e2}
* @param e2 an edge to be compared with {@code e1}
* @return the similarity of the edges.a
*/
protected double getEdgeSimilarity(SparseMatrix sm, Edge e1, Edge e2) {
// Determing the keystone (shared) node by the edges and the other two
// impost (unshared) nodes.
int keystone = -1;
int impost1 = -1;
int impost2 = -1;
if (e1.from == e2.from) {
keystone = e1.from;
impost1 = e1.to;
impost2 = e2.to;
} else if (e1.from == e2.to) {
keystone = e1.from;
impost1 = e1.to;
impost2 = e2.from;
} else if (e2.to == e1.from) {
keystone = e1.from;
impost1 = e1.to;
impost2 = e2.from;
} else if (e1.to == e2.to) {
keystone = e1.to;
impost1 = e1.from;
impost2 = e2.from;
} else return 0d;
// Determine the overlap between the neighbors of the impost nodes
int[] impost1edges = getImpostNeighbors(sm, impost1);
int[] impost2edges = getImpostNeighbors(sm, impost2);
double similarity = Similarity.jaccardIndex(impost1edges, impost2edges);
return similarity;
}
/**
* Compares the specified row to all other rows, returning the k-nearest rows according to the
* similarity metric.
*
* @param m The {@link Matrix} containing data points to be compared
* @param row The current row in {@code m} to be compared against all other rows
* @param kNearestRows The number of most similar rows to retain
* @param similarityType The similarity method to use when comparing rows
* @return a mapping from the similarity to the {@code kNearestRows} most similar rows
*/
public SortedMultiMap<Double, Integer> getMostSimilar(
Matrix m, int row, int kNearestRows, Similarity.SimType similarityType) {
return getMostSimilar(m, row, kNearestRows, Similarity.getSimilarityFunction(similarityType));
}
