/**
* Builds a term-phrase matrix in the same space as the main term-document matrix. If the
* processing context contains no phrases, {@link VectorSpaceModelContext#termPhraseMatrix} will
* remain <code>null</code>.
*/
public void buildTermPhraseMatrix(VectorSpaceModelContext context) {
final PreprocessingContext preprocessingContext = context.preprocessingContext;
final IntIntHashMap stemToRowIndex = context.stemToRowIndex;
final int[] labelsFeatureIndex = preprocessingContext.allLabels.featureIndex;
final int firstPhraseIndex = preprocessingContext.allLabels.firstPhraseIndex;
if (firstPhraseIndex >= 0 && stemToRowIndex.size() > 0) {
// Build phrase matrix
int[] phraseFeatureIndices = new int[labelsFeatureIndex.length - firstPhraseIndex];
for (int featureIndex = 0; featureIndex < phraseFeatureIndices.length; featureIndex++) {
phraseFeatureIndices[featureIndex] = labelsFeatureIndex[featureIndex + firstPhraseIndex];
}
final DoubleMatrix2D phraseMatrix =
TermDocumentMatrixBuilder.buildAlignedMatrix(
context, phraseFeatureIndices, termWeighting);
MatrixUtils.normalizeColumnL2(phraseMatrix, null);
context.termPhraseMatrix = phraseMatrix.viewDice();
}
}
/**
* Performs the actual clustering with an assumption that all documents are written in one <code>
* language</code>.
*/
private void cluster(LanguageCode language) {
// Preprocessing of documents
final PreprocessingContext context =
preprocessingPipeline.preprocess(documents, query, language);
// Further processing only if there are words to process
clusters = Lists.newArrayList();
if (context.hasLabels()) {
// Term-document matrix building and reduction
final VectorSpaceModelContext vsmContext = new VectorSpaceModelContext(context);
final ReducedVectorSpaceModelContext reducedVsmContext =
new ReducedVectorSpaceModelContext(vsmContext);
LingoProcessingContext lingoContext = new LingoProcessingContext(reducedVsmContext);
matrixBuilder.buildTermDocumentMatrix(vsmContext);
matrixBuilder.buildTermPhraseMatrix(vsmContext);
matrixReducer.reduce(
reducedVsmContext, computeClusterCount(desiredClusterCountBase, documents.size()));
// Cluster label building
clusterBuilder.buildLabels(lingoContext, matrixBuilder.termWeighting);
// Document assignment
clusterBuilder.assignDocuments(lingoContext);
// Cluster merging
clusterBuilder.merge(lingoContext);
// Format final clusters
final int[] clusterLabelIndex = lingoContext.clusterLabelFeatureIndex;
final BitSet[] clusterDocuments = lingoContext.clusterDocuments;
final double[] clusterLabelScore = lingoContext.clusterLabelScore;
for (int i = 0; i < clusterLabelIndex.length; i++) {
final Cluster cluster = new Cluster();
final int labelFeature = clusterLabelIndex[i];
if (labelFeature < 0) {
// Cluster removed during merging
continue;
}
// Add label and score
cluster.addPhrases(labelFormatter.format(context, labelFeature));
cluster.setAttribute(Cluster.SCORE, clusterLabelScore[i]);
// Add documents
final BitSet bs = clusterDocuments[i];
for (int bit = bs.nextSetBit(0); bit >= 0; bit = bs.nextSetBit(bit + 1)) {
cluster.addDocuments(documents.get(bit));
}
// Add cluster
clusters.add(cluster);
}
Collections.sort(clusters, Cluster.byReversedWeightedScoreAndSizeComparator(scoreWeight));
}
Cluster.appendOtherTopics(documents, clusters);
}