private static List<LinkedHashMap<String, Object>> getFeaturesForClass(
GraphDatabaseService db, Node classNode) {
List<LinkedHashMap<String, Object>> patternIds = new ArrayList<>();
for (Path p :
db.traversalDescription()
.depthFirst()
.relationships(withName("HAS_CLASS"), Direction.INCOMING)
.evaluator(Evaluators.fromDepth(1))
.evaluator(Evaluators.toDepth(1))
.traverse(classNode)) {
LinkedHashMap<String, Object> featureMap = new LinkedHashMap<>();
if (p.relationships().iterator().hasNext()) {
featureMap.put("frequency", p.relationships().iterator().next().getProperty("matches"));
} else {
featureMap.put("frequency", 0);
}
featureMap.put("feature", ((Long) p.endNode().getId()).intValue());
patternIds.add(featureMap);
}
return patternIds;
}
public static List<LinkedHashMap<String, Object>> getPhrases(
GraphDatabaseService db, String text, GraphManager graphManager) {
// This method trains a model on a supplied label and text content
Map<Long, Integer> patternMatchers =
PatternMatcher.match(GraphManager.ROOT_TEMPLATE, text, db, graphManager);
// Translate map to phrases
List<LinkedHashMap<String, Object>> results =
patternMatchers
.keySet()
.stream()
.map(
a -> {
LinkedHashMap<String, Object> linkHashMap = new LinkedHashMap<>();
linkHashMap.put("feature", NodeManager.getNodeFromGlobalCache(a).get("phrase"));
linkHashMap.put("frequency", patternMatchers.get(a));
return linkHashMap;
})
.collect(Collectors.toList());
results.sort(
(a, b) -> {
Integer diff = ((Integer) a.get("frequency")) - ((Integer) b.get("frequency"));
return diff > 0 ? -1 : diff.equals(0) ? 0 : 1;
});
return results;
}
private List<Map<String, Object>> createJson() {
final List<Map<String, Object>> rows = new ArrayList<>();
for (Map<String, Object> row : this) {
final LinkedHashMap<String, Object> newRow = new LinkedHashMap<>();
for (String column : columns) {
final Object value = row.get(column);
newRow.put(column, toJsonCompatible(value));
}
rows.add(newRow);
}
return rows;
}
public static Map<String, List<LinkedHashMap<String, Object>>> similarDocumentMapForClass(
GraphDatabaseService db, String className) {
Map<String, List<LinkedHashMap<String, Object>>> documents;
Map<String, List<LinkedHashMap<String, Object>>> results = new HashMap<>();
List<Integer> featureIndexList;
VsmCacheModel vsmCacheModel = new VsmCacheModel(db).invoke();
featureIndexList = vsmCacheModel.getFeatureIndexList();
documents = vsmCacheModel.getDocuments();
final String key = className;
List<LinkedHashMap<String, Object>> resultList = new ArrayList<>();
LinkedHashMap<String, Double> classMap = new LinkedHashMap<>();
List<Double> v1 = getFeatureVectorForDocumentClass(documents, featureIndexList, key);
documents
.keySet()
.stream()
.filter(otherKey -> !key.equals(otherKey))
.forEach(
otherKey -> {
List<Double> v2 =
getBinaryFeatureVectorForDocumentClass(documents, featureIndexList, otherKey);
classMap.put(otherKey, cosineSimilarity(v1, v2));
});
classMap
.keySet()
.forEach(
ks -> {
if (!ks.equals(key) && classMap.get(ks) > 0.0) {
LinkedHashMap<String, Object> localMap = new LinkedHashMap<>();
localMap.put("class", ks);
localMap.put("similarity", classMap.get(ks));
resultList.add(localMap);
}
});
resultList.sort(
(a, b) -> {
Double diff = (((double) a.get("similarity")) - ((double) b.get("similarity")));
return diff > 0 ? -1 : diff.equals(0.0) ? 0 : 1;
});
results.put("classes", resultList);
return results;
}
public static Map<String, List<LinkedHashMap<String, Object>>> similarDocumentMapForVector(
GraphDatabaseService db, GraphManager graphManager, String input) {
Map<String, List<LinkedHashMap<String, Object>>> documents;
Map<String, List<LinkedHashMap<String, Object>>> results = new HashMap<>();
List<Integer> featureIndexList;
VsmCacheModel vsmCacheModel = new VsmCacheModel(db).invoke();
featureIndexList = vsmCacheModel.getFeatureIndexList();
documents = vsmCacheModel.getDocuments();
List<Double> features = getFeatureVector(db, graphManager, input, featureIndexList);
List<LinkedHashMap<String, Object>> resultList = new ArrayList<>();
LinkedHashMap<String, Double> classMap = new LinkedHashMap<>();
documents
.keySet()
.stream()
.forEach(
otherKey -> {
List<Double> v2 = getWeightVectorForClass(documents, otherKey, featureIndexList, db);
classMap.put(otherKey, cosineSimilarity(v2, features));
});
classMap
.keySet()
.stream()
.forEach(
ks -> {
if (classMap.get(ks) > 0.0) {
LinkedHashMap<String, Object> localMap = new LinkedHashMap<>();
localMap.put("class", ks);
localMap.put("similarity", classMap.get(ks));
resultList.add(localMap);
}
});
try {
resultList.sort(
(a, b) -> {
Double diff = (((double) a.get("similarity")) - ((double) b.get("similarity")));
return diff > 0 ? -1 : diff.equals(0.0) ? 0 : 1;
});
} catch (NullPointerException ex) {
// resultList is empty or null
}
results.put("classes", resultList);
return results;
}
public static Map<String, Object> getCosineSimilarityVector(GraphDatabaseService db) {
Map<String, List<LinkedHashMap<String, Object>>> documents = getFeaturesForAllClasses(db);
Map<String, List<LinkedHashMap<String, Object>>> results = new HashMap<>();
List<Integer> featureIndexList = getFeatureIndexList(db);
List<String> documentList = documents.keySet().stream().collect(Collectors.toList());
Collections.sort(documentList, (a, b) -> a.compareToIgnoreCase(b));
for (String key : documentList) {
List<LinkedHashMap<String, Object>> resultList = new ArrayList<>();
LinkedHashMap<String, Double> classMap = new LinkedHashMap<>();
List<Double> v1 =
featureIndexList
.stream()
.map(i -> documents.get(key).contains(i) ? featureIndexList.indexOf(i) : 0.0)
.collect(Collectors.toList());
documents
.keySet()
.stream()
.forEach(
otherKey -> {
List<Double> v2 =
featureIndexList
.stream()
.map(
i ->
documents.get(otherKey).contains(i)
? featureIndexList.indexOf(i)
: 0.0)
.collect(Collectors.toList());
classMap.put(otherKey, cosineSimilarity(v1, v2));
});
final List<LinkedHashMap<String, Object>> finalResultList = resultList;
classMap
.keySet()
.forEach(
ks -> {
LinkedHashMap<String, Object> localMap = new LinkedHashMap<>();
localMap.put("class", ks);
localMap.put("similarity", classMap.get(ks));
finalResultList.add(localMap);
});
Collections.sort(
finalResultList,
(a, b) -> ((String) a.get("class")).compareToIgnoreCase((String) b.get("class")));
results.put(key, finalResultList);
}
List<LinkedHashMap<String, Object>> similarityVector = new ArrayList<>();
for (String key : results.keySet()) {
List<Double> cosineVector;
cosineVector =
results
.get(key)
.stream()
.map(a -> Convert.toDouble(Math.round(100000 * (Double) a.get("similarity"))))
.collect(Collectors.toList());
LinkedHashMap<String, Object> row = new LinkedHashMap<>();
row.put("class", key);
row.put("vector", cosineVector);
similarityVector.add(row);
}
Collections.sort(
similarityVector,
(a, b) -> ((String) a.get("class")).compareToIgnoreCase((String) b.get("class")));
Map<String, Object> vectorMap = new LinkedHashMap<>();
List<ArrayList<Double>> vectors = new ArrayList<>();
List<String> classNames = new ArrayList<>();
for (LinkedHashMap<String, Object> val : similarityVector) {
vectors.add((ArrayList<Double>) val.get("vector"));
classNames.add((String) val.get("class"));
}
vectorMap.put("classes", classNames);
vectorMap.put("vectors", vectors);
return vectorMap;
}