@Override
public Tuple parse(
IndexedWord gov, IndexedWord dep, SemanticGraph depGraph, Tuple t, Set<IndexedWord> visited) {
getPOSString(gov, dep);
Tuple t1;
/*
* Check for LeafNode
*/
if (depGraph.getChildren(dep).size() > 0) {
t1 = parse(dep, depGraph, visited);
} else {
Entity e = new Entity(dep.word(), EntityType.Notion);
t1 = new Tuple(e);
}
String s = depGraph.getEdge(gov, dep).getRelation().getSpecific();
Relation r = new Relation(s, RelationType.One2One);
if (t == null) {
Entity e1 = new Entity(gov.word(), EntityType.Object);
Tuple t2 = new Tuple(e1);
t = new Tuple(t1, r, t2);
} else {
t = new Tuple(t1, r, t);
}
logger.info(t.toString());
return t;
}
public void testHasChildren() {
SemanticGraph gr = SemanticGraph.valueOf("[ate subj>Bill dobj>[muffins compound>blueberry]]");
List<IndexedWord> vertices = gr.vertexListSorted();
for (IndexedWord word : vertices) {
if (word.word().equals("ate") || word.word().equals("muffins")) {
assertTrue(gr.hasChildren(word));
} else {
assertFalse(gr.hasChildren(word));
}
}
}
public Entity(IndexedWord... wrd) {
this.name = "";
for (IndexedWord w : wrd) {
this.name = this.name + " " + w.word();
}
this.name = this.name.trim();
this.type = EntityType.Unknown;
}
/**
* This method attempts to resolve noun phrases which consist of more than one word. More
* precisely, it looks for nn dependencies below {@code head} and creates an entity.
*
* @param head The head of the noun phrase
* @param graph The sentence to look in.
* @param words The words which make up the noun phrase
* @return A distinct word
*/
public static String resolveNN(
IndexedWord head, SemanticGraph graph, ArrayList<IndexedWord> words) {
List<IndexedWord> nns =
graph.getChildrenWithReln(head, EnglishGrammaticalRelations.NOUN_COMPOUND_MODIFIER);
String name = "";
// check for nulls. if there is nothing here, we have nothing to do.
if (nns != null) {
for (IndexedWord part : nns) {
name += part.word();
name += " ";
words.add(part); // save this word as a part of the results
}
// append the head word ("starting" word)
name += head.word();
words.add(head); // save this word as a part of the results
return name;
} else {
return null;
}
}
@Override
public void handle(HttpExchange httpExchange) throws IOException {
// Set common response headers
httpExchange.getResponseHeaders().add("Access-Control-Allow-Origin", "*");
Future<String> json =
corenlpExecutor.submit(
() -> {
try {
// Get the document
Properties props =
new Properties() {
{
setProperty("annotators", "tokenize,ssplit,pos,lemma,ner,depparse");
}
};
Annotation doc = getDocument(props, httpExchange);
if (!doc.containsKey(CoreAnnotations.SentencesAnnotation.class)) {
StanfordCoreNLP pipeline = mkStanfordCoreNLP(props);
pipeline.annotate(doc);
}
// Construct the matcher
Map<String, String> params = getURLParams(httpExchange.getRequestURI());
// (get the pattern)
if (!params.containsKey("pattern")) {
respondError("Missing required parameter 'pattern'", httpExchange);
return "";
}
String pattern = params.get("pattern");
// (get whether to filter / find)
String filterStr = params.getOrDefault("filter", "false");
final boolean filter =
filterStr.trim().isEmpty()
|| "true".equalsIgnoreCase(filterStr.toLowerCase());
// (create the matcher)
final SemgrexPattern regex = SemgrexPattern.compile(pattern);
// Run TokensRegex
return JSONOutputter.JSONWriter.objectToJSON(
(docWriter) -> {
if (filter) {
// Case: just filter sentences
docWriter.set(
"sentences",
doc.get(CoreAnnotations.SentencesAnnotation.class)
.stream()
.map(
sentence ->
regex
.matcher(
sentence.get(
SemanticGraphCoreAnnotations
.CollapsedCCProcessedDependenciesAnnotation
.class))
.matches())
.collect(Collectors.toList()));
} else {
// Case: find matches
docWriter.set(
"sentences",
doc.get(CoreAnnotations.SentencesAnnotation.class)
.stream()
.map(
sentence ->
(Consumer<JSONOutputter.Writer>)
(JSONOutputter.Writer sentWriter) -> {
SemgrexMatcher matcher =
regex.matcher(
sentence.get(
SemanticGraphCoreAnnotations
.CollapsedCCProcessedDependenciesAnnotation
.class));
int i = 0;
while (matcher.find()) {
sentWriter.set(
Integer.toString(i),
(Consumer<JSONOutputter.Writer>)
(JSONOutputter.Writer matchWriter) -> {
IndexedWord match = matcher.getMatch();
matchWriter.set("text", match.word());
matchWriter.set(
"begin", match.index() - 1);
matchWriter.set("end", match.index());
for (String capture :
matcher.getNodeNames()) {
matchWriter.set(
"$" + capture,
(Consumer<JSONOutputter.Writer>)
groupWriter -> {
IndexedWord node =
matcher.getNode(
capture);
groupWriter.set(
"text", node.word());
groupWriter.set(
"begin",
node.index() - 1);
groupWriter.set(
"end", node.index());
});
}
});
i += 1;
}
sentWriter.set("length", i);
}));
}
});
} catch (Exception e) {
e.printStackTrace();
try {
respondError(e.getClass().getName() + ": " + e.getMessage(), httpExchange);
} catch (IOException ignored) {
}
}
return "";
});
// Send response
byte[] response = new byte[0];
try {
response = json.get(5, TimeUnit.SECONDS).getBytes();
} catch (InterruptedException | ExecutionException | TimeoutException e) {
respondError("Timeout when executing Semgrex query", httpExchange);
}
if (response.length > 0) {
httpExchange.getResponseHeaders().add("Content-Type", "text/json");
httpExchange.getResponseHeaders().add("Content-Length", Integer.toString(response.length));
httpExchange.sendResponseHeaders(HTTP_OK, response.length);
httpExchange.getResponseBody().write(response);
httpExchange.close();
}
}
