public virtual void TestEnglishDependenciesByTree()
{
// the trees to test
string[] testTrees = new string[] { "((S (NP (NNP Sam)) (VP (VBD died) (NP-TMP (NN today)))))", "(ROOT (S (NP (PRP I)) (VP (VBD saw) (NP (NP (DT the) (NN book)) (SBAR (WHNP (WDT which)) (S (NP (PRP you)) (VP (VBD bought)))))) (. .)))" };
// the expected dependency answers (basic)
string[] testAnswers = new string[] { "root(ROOT-0, died-2) nsubj(died-2, Sam-1) tmod(died-2, today-3)", "nsubj(saw-2, I-1) root(ROOT-0, saw-2) det(book-4, the-3) dobj(saw-2, book-4) dobj(bought-7, which-5) ref(book-4, which-5) dobj(bought-7, which-5) nsubj(bought-7, you-6) rcmod(book-4, bought-7)" };
// the expected dependency answers (collapsed dependencies)
string[] testAnswersCollapsed = new string[] { "root(ROOT-0, died-2) nsubj(died-2, Sam-1) tmod(died-2, today-3)", "nsubj(saw-2, I-1) root(ROOT-0, saw-2) det(book-4, the-3) dobj(saw-2, book-4) dobj(bought-7, book-4) nsubj(bought-7, you-6) rcmod(book-4, bought-7)" };
// the expected dependency answers (conjunctions processed)
string[] testAnswersCCProcessed = new string[] { "root(ROOT-0, died-2) nsubj(died-2, Sam-1) tmod(died-2, today-3)", "nsubj(saw-2, I-1) root(ROOT-0, saw-2) det(book-4, the-3) dobj(saw-2, book-4) dobj(bought-7, book-4) nsubj(bought-7, you-6) rcmod(book-4, bought-7)" };
for (int i = 0; i < testTrees.Length; i++)
{
string testTree = testTrees[i];
string testAnswer = testAnswers[i];
string testAnswerCollapsed = testAnswersCollapsed[i];
string testAnswerCCProcessed = testAnswersCCProcessed[i];
HashSet <string> testAnswerTokens = TokenSet(testAnswer);
HashSet <string> testAnswerCollapsedTokens = TokenSet(testAnswerCollapsed);
HashSet <string> testAnswerCCProcessedTokens = TokenSet(testAnswerCCProcessed);
Tree tree;
try
{
tree = new PennTreeReader(new StringReader(testTree), new LabeledScoredTreeFactory()).ReadTree();
}
catch (IOException e)
{
// these trees should all parse correctly
throw new Exception(e);
}
GrammaticalStructure gs = new EnglishGrammaticalStructure(tree);
NUnit.Framework.Assert.AreEqual("Unexpected basic dependencies for tree " + testTree, testAnswerTokens, TokenSet(gs.TypedDependencies(GrammaticalStructure.Extras.Maximal)));
NUnit.Framework.Assert.AreEqual("Unexpected collapsed dependencies for tree " + testTree, testAnswerCollapsedTokens, TokenSet(gs.TypedDependenciesCollapsed(GrammaticalStructure.Extras.Maximal)));
NUnit.Framework.Assert.AreEqual("Unexpected cc-processed dependencies for tree " + testTree, testAnswerCCProcessedTokens, TokenSet(gs.TypedDependenciesCCprocessed(GrammaticalStructure.Extras.Maximal)));
}
}
/// <summary>
/// Returns a new
/// <c>SemanticGraph</c>
/// constructed from a given
/// <see cref="Edu.Stanford.Nlp.Trees.Tree"/>
/// with given options.
/// This factory method is intended to replace a profusion of highly similar
/// factory methods, such as
/// <c>typedDependencies()</c>
/// ,
/// <c>typedDependenciesCollapsed()</c>
/// ,
/// <c>allTypedDependencies()</c>
/// ,
/// <c>allTypedDependenciesCollapsed()</c>
/// , etc.
/// For a fuller explanation of the meaning of the boolean arguments, see
/// <see cref="Edu.Stanford.Nlp.Trees.GrammaticalStructure"/>
/// .
/// </summary>
/// <param name="tree">A tree representing a phrase structure parse</param>
/// <param name="includeExtras">
/// Whether to include extra dependencies, which may
/// result in a non-tree
/// </param>
/// <param name="filter">A filter to exclude certain dependencies; ignored if null</param>
/// <param name="originalDependencies">
/// generate original Stanford dependencies instead of new
/// Universal Dependencies
/// </param>
/// <returns>A SemanticGraph</returns>
public static SemanticGraph MakeFromTree(Tree tree, SemanticGraphFactory.Mode mode, GrammaticalStructure.Extras includeExtras, IPredicate <TypedDependency> filter, bool originalDependencies, bool includePunctuationDependencies)
{
GrammaticalStructure gs;
if (originalDependencies)
{
IPredicate <string> wordFilt;
if (includePunctuationDependencies)
{
wordFilt = Filters.AcceptFilter();
}
else
{
wordFilt = new PennTreebankLanguagePack().PunctuationWordRejectFilter();
}
gs = new EnglishGrammaticalStructure(tree, wordFilt, new SemanticHeadFinder(true));
}
else
{
IPredicate <string> tagFilt;
if (includePunctuationDependencies)
{
tagFilt = Filters.AcceptFilter();
}
else
{
tagFilt = new PennTreebankLanguagePack().PunctuationTagRejectFilter();
}
gs = new UniversalEnglishGrammaticalStructure(tree, tagFilt, new UniversalSemanticHeadFinder(true));
}
return(MakeFromTree(gs, mode, includeExtras, filter));
}
/// <summary>Just for testing.</summary>
public static void Main(string[] args)
{
try
{
ITreeReader tr = new PennTreeReader(new StringReader("(S (NP (NNP Sam)) (VP (VBD died) (NP (NN today))))"), new LabeledScoredTreeFactory());
Tree t = tr.ReadTree();
System.Console.Out.WriteLine(t);
TreeGraphNode tgn = new TreeGraphNode(t, (TreeGraphNode)null);
System.Console.Out.WriteLine(tgn.ToPrettyString(0));
EnglishGrammaticalStructure gs = new EnglishGrammaticalStructure(tgn);
System.Console.Out.WriteLine(tgn.ToPrettyString(0));
tgn.PercolateHeads(new SemanticHeadFinder());
System.Console.Out.WriteLine(tgn.ToPrettyString(0));
}
catch (Exception e)
{
log.Error("Horrible error: " + e);
log.Error(e);
}
}