/// <summary>Parse a JSON formatted tree into a SemanticGraph.</summary>
/// <param name="jsonString">
/// The JSON string tree to parse, e.g:
/// "[{\"\"dependent\"\": 7, \"\"dep\"\": \"\"root\"\", \"\"governorgloss\"\": \"\"root\"\", \"\"governor\"\": 0, \"\"dependentgloss\"\": \"\"sport\"\"}, {\"\"dependent\"\": 1, \"\"dep\"\": \"\"nsubj\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"chess\"\"}, {\"\"dependent\"\": 2, \"\"dep\"\": \"\"cop\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"is\"\"}, {\"\"dependent\"\": 3, \"\"dep\"\": \"\"neg\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"not\"\"}, {\"\"dependent\"\": 4, \"\"dep\"\": \"\"det\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"a\"\"}, {\"\"dependent\"\": 5, \"\"dep\"\": \"\"advmod\"\", \"\"governorgloss\"\": \"\"physical\"\", \"\"governor\"\": 6, \"\"dependentgloss\"\": \"\"predominantly\"\"}, {\"\"dependent\"\": 6, \"\"dep\"\": \"\"amod\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"physical\"\"}, {\"\"dependent\"\": 9, \"\"dep\"\": \"\"advmod\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"yet\"\"}, {\"\"dependent\"\": 10, \"\"dep\"\": \"\"nsubj\"\", \"\"governorgloss\"\": \"\"shooting\"\", \"\"governor\"\": 12, \"\"dependentgloss\"\": \"\"neither\"\"}, {\"\"dependent\"\": 11, \"\"dep\"\": \"\"cop\"\", \"\"governorgloss\"\": \"\"shooting\"\", \"\"governor\"\": 12, \"\"dependentgloss\"\": \"\"are\"\"}, {\"\"dependent\"\": 12, \"\"dep\"\": \"\"parataxis\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"shooting\"\"}, {\"\"dependent\"\": 13, \"\"dep\"\": \"\"cc\"\", \"\"governorgloss\"\": \"\"shooting\"\", \"\"governor\"\": 12, \"\"dependentgloss\"\": \"\"and\"\"}, {\"\"dependent\"\": 14, \"\"dep\"\": \"\"parataxis\"\", \"\"governorgloss\"\": \"\"sport\"\", \"\"governor\"\": 7, \"\"dependentgloss\"\": \"\"curling\"\"}, {\"\"dependent\"\": 14, \"\"dep\"\": \"\"conj:and\"\", \"\"governorgloss\"\": \"\"shooting\"\", \"\"governor\"\": 12, \"\"dependentgloss\"\": \"\"curling\"\"}, {\"\"dependent\"\": 16, \"\"dep\"\": \"\"nsubjpass\"\", \"\"governorgloss\"\": \"\"nicknamed\"\", \"\"governor\"\": 23, \"\"dependentgloss\"\": \"\"which\"\"}, {\"\"dependent\"\": 18, \"\"dep\"\": \"\"case\"\", \"\"governorgloss\"\": \"\"fact\"\", \"\"governor\"\": 19, \"\"dependentgloss\"\": \"\"in\"\"}, {\"\"dependent\"\": 19, \"\"dep\"\": \"\"nmod:in\"\", \"\"governorgloss\"\": \"\"nicknamed\"\", \"\"governor\"\": 23, \"\"dependentgloss\"\": \"\"fact\"\"}, {\"\"dependent\"\": 21, \"\"dep\"\": \"\"aux\"\", \"\"governorgloss\"\": \"\"nicknamed\"\", \"\"governor\"\": 23, \"\"dependentgloss\"\": \"\"has\"\"}, {\"\"dependent\"\": 22, \"\"dep\"\": \"\"auxpass\"\", \"\"governorgloss\"\": \"\"nicknamed\"\", \"\"governor\"\": 23, \"\"dependentgloss\"\": \"\"been\"\"}, {\"\"dependent\"\": 23, \"\"dep\"\": \"\"dep\"\", \"\"governorgloss\"\": \"\"shooting\"\", \"\"governor\"\": 12, \"\"dependentgloss\"\": \"\"nicknamed\"\"}, {\"\"dependent\"\": 25, \"\"dep\"\": \"\"dobj\"\", \"\"governorgloss\"\": \"\"nicknamed\"\", \"\"governor\"\": 23, \"\"dependentgloss\"\": \"\"chess\"\"}, {\"\"dependent\"\": 26, \"\"dep\"\": \"\"case\"\", \"\"governorgloss\"\": \"\"ice\"\", \"\"governor\"\": 27, \"\"dependentgloss\"\": \"\"on\"\"}, {\"\"dependent\"\": 27, \"\"dep\"\": \"\"nmod:on\"\", \"\"governorgloss\"\": \"\"chess\"\", \"\"governor\"\": 25, \"\"dependentgloss\"\": \"\"ice\"\"}, {\"\"dependent\"\": 29, \"\"dep\"\": \"\"amod\"\", \"\"governorgloss\"\": \"\"chess\"\", \"\"governor\"\": 25, \"\"dependentgloss\"\": \"\"5\"\"}]");
/// </param>
/// <param name="tokens">The tokens of the sentence, to form the backing labels of the tree.</param>
/// <returns>A semantic graph of the sentence, according to the given tree.</returns>
public static SemanticGraph ParseJsonTree(string jsonString, IList <CoreLabel> tokens)
{
// Escape quoted string parts
IJsonReader json = Javax.Json.Json.CreateReader(new StringReader(jsonString));
SemanticGraph tree = new SemanticGraph();
IJsonArray array = json.ReadArray();
if (array == null || array.IsEmpty())
{
return(tree);
}
IndexedWord[] vertices = new IndexedWord[tokens.Count + 2];
// Add edges
for (int i = 0; i < array.Count; i++)
{
IJsonObject entry = array.GetJsonObject(i);
// Parse row
int dependentIndex = entry.GetInt("dependent");
if (vertices[dependentIndex] == null)
{
if (dependentIndex > tokens.Count)
{
// Bizarre mismatch in sizes; the malt parser seems to do this often
return(new SemanticGraph());
}
vertices[dependentIndex] = new IndexedWord(tokens[dependentIndex - 1]);
}
IndexedWord dependent = vertices[dependentIndex];
int governorIndex = entry.GetInt("governor");
if (governorIndex > tokens.Count)
{
// Bizarre mismatch in sizes; the malt parser seems to do this often
return(new SemanticGraph());
}
if (vertices[governorIndex] == null && governorIndex > 0)
{
vertices[governorIndex] = new IndexedWord(tokens[governorIndex - 1]);
}
IndexedWord governor = vertices[governorIndex];
string relation = entry.GetString("dep");
// Process row
if (governorIndex == 0)
{
tree.AddRoot(dependent);
}
else
{
tree.AddVertex(dependent);
if (!tree.ContainsVertex(governor))
{
tree.AddVertex(governor);
}
if (!"ref".Equals(relation))
{
tree.AddEdge(governor, dependent, GrammaticalRelation.ValueOf(Language.English, relation), double.NegativeInfinity, false);
}
}
}
return(tree);
}
/// <summary>Parse a CoNLL formatted tree into a SemanticGraph.</summary>
/// <param name="conll">The CoNLL tree to parse.</param>
/// <param name="tokens">The tokens of the sentence, to form the backing labels of the tree.</param>
/// <returns>A semantic graph of the sentence, according to the given tree.</returns>
public static SemanticGraph ParseTree(string conll, IList <CoreLabel> tokens)
{
SemanticGraph tree = new SemanticGraph();
if (conll == null || conll.IsEmpty())
{
return(tree);
}
string[] treeLines = newline.Split(conll);
IndexedWord[] vertices = new IndexedWord[tokens.Count + 2];
// Add edges
foreach (string line in treeLines)
{
// Parse row
string[] fields = tab.Split(line);
int dependentIndex = System.Convert.ToInt32(fields[0]);
if (vertices[dependentIndex] == null)
{
if (dependentIndex > tokens.Count)
{
// Bizarre mismatch in sizes; the malt parser seems to do this often
return(new SemanticGraph());
}
vertices[dependentIndex] = new IndexedWord(tokens[dependentIndex - 1]);
}
IndexedWord dependent = vertices[dependentIndex];
int governorIndex = System.Convert.ToInt32(fields[1]);
if (governorIndex > tokens.Count)
{
// Bizarre mismatch in sizes; the malt parser seems to do this often
return(new SemanticGraph());
}
if (vertices[governorIndex] == null && governorIndex > 0)
{
vertices[governorIndex] = new IndexedWord(tokens[governorIndex - 1]);
}
IndexedWord governor = vertices[governorIndex];
string relation = fields[2];
// Process row
if (governorIndex == 0)
{
tree.AddRoot(dependent);
}
else
{
tree.AddVertex(dependent);
if (!tree.ContainsVertex(governor))
{
tree.AddVertex(governor);
}
if (!"ref".Equals(relation))
{
tree.AddEdge(governor, dependent, GrammaticalRelation.ValueOf(Language.English, relation), double.NegativeInfinity, false);
}
}
}
return(tree);
}
/// <summary>Parse a CoNLL formatted tree into a SemanticGraph object (along with a list of tokens).</summary>
/// <param name="conll">The CoNLL formatted tree.</param>
/// <returns>
/// A pair of a SemanticGraph and a token list, corresponding to the parse of the sentence
/// and to tokens in the sentence.
/// </returns>
protected internal virtual Pair <SemanticGraph, IList <CoreLabel> > MkTree(string conll)
{
IList <CoreLabel> sentence = new List <CoreLabel>();
SemanticGraph tree = new SemanticGraph();
foreach (string line in conll.Split("\n"))
{
if (line.Trim().Equals(string.Empty))
{
continue;
}
string[] fields = line.Trim().Split("\\s+");
int index = System.Convert.ToInt32(fields[0]);
string word = fields[1];
CoreLabel label = IETestUtils.MkWord(word, index);
sentence.Add(label);
if (fields[2].Equals("0"))
{
tree.AddRoot(new IndexedWord(label));
}
else
{
tree.AddVertex(new IndexedWord(label));
}
if (fields.Length > 4)
{
label.SetTag(fields[4]);
}
if (fields.Length > 5)
{
label.SetNER(fields[5]);
}
if (fields.Length > 6)
{
label.SetLemma(fields[6]);
}
}
int i = 0;
foreach (string line_1 in conll.Split("\n"))
{
if (line_1.Trim().Equals(string.Empty))
{
continue;
}
string[] fields = line_1.Trim().Split("\\s+");
int parent = System.Convert.ToInt32(fields[2]);
string reln = fields[3];
if (parent > 0)
{
tree.AddEdge(new IndexedWord(sentence[parent - 1]), new IndexedWord(sentence[i]), new GrammaticalRelation(Language.UniversalEnglish, reln, null, null), 1.0, false);
}
i += 1;
}
return(Pair.MakePair(tree, sentence));
}
public override void Evaluate(SemanticGraph sg, SemgrexMatcher sm)
{
IndexedWord rootNode = this.GetNamedNode(rootName, sm);
ICollection <IndexedWord> subgraphNodeSet = sg.GetSubgraphVertices(rootNode);
if (!sg.IsDag(rootNode))
{
/* Check if there is a cycle going back to the root. */
foreach (IndexedWord child in sg.GetChildren(rootNode))
{
ICollection <IndexedWord> reachableSet = sg.GetSubgraphVertices(child);
if (reachableSet.Contains(rootNode))
{
throw new ArgumentException("Subtree cannot contain cycle leading back to root node!");
}
}
}
IList <IndexedWord> sortedSubgraphNodes = Generics.NewArrayList(subgraphNodeSet);
sortedSubgraphNodes.Sort();
IndexedWord newNode = new IndexedWord(rootNode.DocID(), rootNode.SentIndex(), rootNode.Index());
/* Copy all attributes from rootNode. */
foreach (Type key in newNode.BackingLabel().KeySet())
{
newNode.Set(key, rootNode.Get(key));
}
newNode.SetValue(StringUtils.Join(sortedSubgraphNodes.Stream().Map(null), " "));
newNode.SetWord(StringUtils.Join(sortedSubgraphNodes.Stream().Map(null), " "));
newNode.SetLemma(StringUtils.Join(sortedSubgraphNodes.Stream().Map(null), " "));
if (sg.GetRoots().Contains(rootNode))
{
sg.GetRoots().Remove(rootNode);
sg.AddRoot(rootNode);
}
foreach (SemanticGraphEdge edge in sg.IncomingEdgeIterable(rootNode))
{
sg.AddEdge(edge.GetGovernor(), newNode, edge.GetRelation(), edge.GetWeight(), edge.IsExtra());
}
foreach (IndexedWord node in sortedSubgraphNodes)
{
sg.RemoveVertex(node);
}
}
/// <summary>Returns all of the entailed shortened clauses (as per natural logic) from the given clause.</summary>
/// <remarks>
/// Returns all of the entailed shortened clauses (as per natural logic) from the given clause.
/// This runs the forward entailment component of the OpenIE system only.
/// It is usually chained together with the clause splitting component:
/// <see cref="ClausesInSentence(Edu.Stanford.Nlp.Util.ICoreMap)"/>
/// .
/// </remarks>
/// <param name="clause">The premise clause, as a sentence fragment in itself.</param>
/// <returns>A list of entailed clauses.</returns>
public virtual IList <SentenceFragment> EntailmentsFromClause(SentenceFragment clause)
{
if (clause.parseTree.IsEmpty())
{
return(Java.Util.Collections.EmptyList());
}
else
{
// Get the forward entailments
IList <SentenceFragment> list = new List <SentenceFragment>();
if (entailmentsPerSentence > 0)
{
Sharpen.Collections.AddAll(list, forwardEntailer.Apply(clause.parseTree, true).Search().Stream().Map(null).Collect(Collectors.ToList()));
}
list.Add(clause);
// A special case for adjective entailments
IList <SentenceFragment> adjFragments = new List <SentenceFragment>();
SemgrexMatcher matcher = adjectivePattern.Matcher(clause.parseTree);
while (matcher.Find())
{
// (get nodes)
IndexedWord subj = matcher.GetNode("subj");
IndexedWord be = matcher.GetNode("be");
IndexedWord adj = matcher.GetNode("adj");
IndexedWord obj = matcher.GetNode("obj");
IndexedWord pobj = matcher.GetNode("pobj");
string prep = matcher.GetRelnString("prep");
// (if the adjective, or any earlier adjective, is privative, then all bets are off)
foreach (SemanticGraphEdge edge in clause.parseTree.OutgoingEdgeIterable(obj))
{
if ("amod".Equals(edge.GetRelation().ToString()) && edge.GetDependent().Index() <= adj.Index() && Edu.Stanford.Nlp.Naturalli.Util.PrivativeAdjectives.Contains(edge.GetDependent().Word().ToLower()))
{
goto OUTER_continue;
}
}
// (create the core tree)
SemanticGraph tree = new SemanticGraph();
tree.AddRoot(adj);
tree.AddVertex(subj);
tree.AddVertex(be);
tree.AddEdge(adj, be, GrammaticalRelation.ValueOf(Language.English, "cop"), double.NegativeInfinity, false);
tree.AddEdge(adj, subj, GrammaticalRelation.ValueOf(Language.English, "nsubj"), double.NegativeInfinity, false);
// (add pp attachment, if it existed)
if (pobj != null)
{
System.Diagnostics.Debug.Assert(prep != null);
tree.AddEdge(adj, pobj, GrammaticalRelation.ValueOf(Language.English, prep), double.NegativeInfinity, false);
}
// (check for monotonicity)
if (adj.Get(typeof(NaturalLogicAnnotations.PolarityAnnotation)).IsUpwards() && be.Get(typeof(NaturalLogicAnnotations.PolarityAnnotation)).IsUpwards())
{
// (add tree)
adjFragments.Add(new SentenceFragment(tree, clause.assumedTruth, false));
}
OUTER_continue :;
}
OUTER_break :;
Sharpen.Collections.AddAll(list, adjFragments);
return(list);
}
}
public virtual SemanticGraph ConvertIntermediateGraph(IList <CoreLabel> sentence)
{
SemanticGraph graph = new SemanticGraph();
// First construct the actual nodes; keep them indexed by their index and copy count.
// Sentences such as "I went over the river and through the woods" have
// two copies for "went" in the collapsed dependencies.
TwoDimensionalMap <int, int, IndexedWord> nodeMap = TwoDimensionalMap.HashMap();
foreach (AnnotationSerializer.IntermediateNode @in in nodes)
{
CoreLabel token = sentence[@in.index - 1];
// index starts at 1!
IndexedWord word;
if (@in.copyAnnotation > 0)
{
// TODO: if we make a copy wrapper CoreLabel, use it here instead
word = new IndexedWord(new CoreLabel(token));
word.SetCopyCount(@in.copyAnnotation);
}
else
{
word = new IndexedWord(token);
}
// for backwards compatibility - new annotations should have
// these fields set, but annotations older than August 2014 might not
if (word.DocID() == null && @in.docId != null)
{
word.SetDocID(@in.docId);
}
if (word.SentIndex() < 0 && @in.sentIndex >= 0)
{
word.SetSentIndex(@in.sentIndex);
}
if (word.Index() < 0 && @in.index >= 0)
{
word.SetIndex(@in.index);
}
nodeMap.Put(word.Index(), word.CopyCount(), word);
graph.AddVertex(word);
if (@in.isRoot)
{
graph.AddRoot(word);
}
}
// add all edges to the actual graph
foreach (AnnotationSerializer.IntermediateEdge ie in edges)
{
IndexedWord source = nodeMap.Get(ie.source, ie.sourceCopy);
if (source == null)
{
throw new RuntimeIOException("Failed to find node " + ie.source + "-" + ie.sourceCopy);
}
IndexedWord target = nodeMap.Get(ie.target, ie.targetCopy);
if (target == null)
{
throw new RuntimeIOException("Failed to find node " + ie.target + "-" + ie.targetCopy);
}
// assert(target != null);
lock (Lock)
{
// this is not thread-safe: there are static fields in GrammaticalRelation
GrammaticalRelation rel = GrammaticalRelation.ValueOf(ie.dep);
graph.AddEdge(source, target, rel, 1.0, ie.isExtra);
}
}
// compute root nodes if they weren't stored in the graph
if (!graph.IsEmpty() && graph.GetRoots().Count == 0)
{
graph.ResetRoots();
}
return(graph);
}
protected internal virtual ICollection <string> Clauses(string conll)
{
IList <CoreLabel> sentence = new List <CoreLabel>();
SemanticGraph tree = new SemanticGraph();
foreach (string line in conll.Split("\n"))
{
if (line.Trim().Equals(string.Empty))
{
continue;
}
string[] fields = line.Trim().Split("\\s+");
int index = System.Convert.ToInt32(fields[0]);
string word = fields[1];
CoreLabel label = MkWord(word, index);
sentence.Add(label);
if (fields[2].Equals("0"))
{
tree.AddRoot(new IndexedWord(label));
}
else
{
tree.AddVertex(new IndexedWord(label));
}
if (fields.Length > 4)
{
label.SetTag(fields[4]);
}
if (fields.Length > 5)
{
label.SetNER(fields[5]);
}
if (fields.Length > 6)
{
label.SetLemma(fields[6]);
}
}
int i = 0;
foreach (string line_1 in conll.Split("\n"))
{
if (line_1.Trim().Equals(string.Empty))
{
continue;
}
string[] fields = line_1.Trim().Split("\\s+");
int parent = System.Convert.ToInt32(fields[2]);
string reln = fields[3];
if (parent > 0)
{
tree.AddEdge(new IndexedWord(sentence[parent - 1]), new IndexedWord(sentence[i]), new GrammaticalRelation(Language.English, reln, null, null), 1.0, false);
}
i += 1;
}
// Run extractor
ClauseSplitterSearchProblem problem = new ClauseSplitterSearchProblem(tree, true);
ICollection <string> clauses = new HashSet <string>();
problem.Search(null, new LinearClassifier <ClauseSplitter.ClauseClassifierLabel, string>(new ClassicCounter <Pair <string, ClauseSplitter.ClauseClassifierLabel> >()), ClauseSplitterSearchProblem.HardSplits, null, 100000);
return(clauses);
}