public MentionExtractor(Dictionaries dict, Semantics semantics)
{
this.headFinder = new SemanticHeadFinder();
this.dictionaries = dict;
this.semantics = semantics;
this.mentionFinder = new RuleBasedCorefMentionFinder();
}
/* applies a TreeVisitor to all projections (including the node itself) of a node in a Tree.
* Does nothing if head is not in root.
* @return the maximal projection of head in root.
*/
public static Tree ApplyToProjections(ITreeVisitor v, Tree head, Tree root, IHeadFinder hf)
{
Tree projection = head;
Tree parent = projection.Parent(root);
if (parent == null && projection != root)
{
return(null);
}
v.VisitTree(projection);
if (projection == root)
{
return(root);
}
while (hf.DetermineHead(parent) == projection)
{
projection = parent;
v.VisitTree(projection);
if (projection == root)
{
return(root);
}
parent = projection.Parent(root);
}
return(projection);
}
/// <exception cref="System.TypeLoadException"/>
/// <exception cref="System.IO.IOException"/>
private CorefMentionFinder GetMentionFinder(Properties props, IHeadFinder headFinder)
{
switch (CorefProperties.MdType(props))
{
case CorefProperties.MentionDetectionType.Dependency:
{
mdName = "dependency";
return(new DependencyCorefMentionFinder(props));
}
case CorefProperties.MentionDetectionType.Hybrid:
{
mdName = "hybrid";
mentionAnnotatorRequirements.Add(typeof(TreeCoreAnnotations.TreeAnnotation));
mentionAnnotatorRequirements.Add(typeof(CoreAnnotations.BeginIndexAnnotation));
mentionAnnotatorRequirements.Add(typeof(CoreAnnotations.EndIndexAnnotation));
return(new HybridCorefMentionFinder(headFinder, props));
}
case CorefProperties.MentionDetectionType.Rule:
default:
{
mentionAnnotatorRequirements.Add(typeof(TreeCoreAnnotations.TreeAnnotation));
mentionAnnotatorRequirements.Add(typeof(CoreAnnotations.BeginIndexAnnotation));
mentionAnnotatorRequirements.Add(typeof(CoreAnnotations.EndIndexAnnotation));
mdName = "rule";
return(new RuleBasedCorefMentionFinder(headFinder, props));
}
}
}
public UnlabeledAttachmentEval(string str, bool runningAverages, IHeadFinder headFinder, IPredicate <string> punctRejectFilter)
: base(str, runningAverages)
{
this.headFinder = headFinder;
this.punctRejectWordFilter = punctRejectFilter;
this.punctRejectFilter = new _IPredicate_58(this);
}
public RuleBasedCorefMentionFinder(bool allowReparsing)
{
// protected int maxID = -1;
SieveCoreferenceSystem.logger.Fine("Using SEMANTIC HEAD FINDER!!!!!!!!!!!!!!!!!!!");
this.headFinder = new SemanticHeadFinder();
this.allowReparsing = allowReparsing;
}
public TreeAnnotator(IHeadFinder hf, ITreebankLangParserParams tlpp, Options op)
{
this.tlpParams = tlpp;
this.hf = hf;
this.tf = new LabeledScoredTreeFactory();
this.trainOptions = op.trainOptions;
}
public override TregexMatcher Matcher(Tree root, Tree tree,
IdentityDictionary<Tree, Tree> nodesToParents,
Dictionary<string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder)
{
return new CoordinationMatcher(this, root, tree, nodesToParents, namesToNodes, variableStrings, headFinder);
}
/// <exception cref="System.TypeLoadException"/>
/// <exception cref="System.IO.IOException"/>
public DocumentMaker(Properties props, Dictionaries dictionaries)
{
this.props = props;
this.dict = dictionaries;
reader = GetDocumentReader(props);
headFinder = CorefProperties.GetHeadFinder(props);
md = CorefProperties.UseGoldMentions(props) ? new RuleBasedCorefMentionFinder(headFinder, props) : null;
}
public TregexParser(TextReader stream,
Func <string, string> basicCatFunction,
IHeadFinder headFinder) :
this(stream)
{
this.basicCatFunction = basicCatFunction;
this.headFinder = headFinder;
}
public override TregexMatcher Matcher(Tree root, Tree tree,
IdentityDictionary <Tree, Tree> nodesToParents,
Dictionary <string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder)
{
return(new CoordinationMatcher(this, root, tree, nodesToParents, namesToNodes, variableStrings, headFinder));
}
/// <summary>
/// Create a TreeNormalizer that maintains some functional annotations,
/// particularly those involving temporal annotation.
/// </summary>
/// <param name="temporalAnnotation">
/// One of the constants:
/// TEMPORAL_NONE (no temporal annotation kept on trees),
/// TEMPORAL_ACL03PCFG (temporal annotation on NPs, and percolated down
/// to head of constituent until and including POS tag),
/// TEMPORAL_ANY_TMP_PERCOLATED (temporal annotation on any phrase is
/// kept and percolated via head chain to and including POS tag),
/// TEMPORAL_ALL_TERMINALS (temporal annotation is kept on NPs, and
/// is placed on all POS tag daughters of that NP (but is not
/// percolated down a head chain through phrasal categories),
/// TEMPORAL_ALL_NP (temporal annotation on NPs, and it is percolated
/// down via the head chain, but only through NPs: annotation stops
/// at either a POS tag (which is annotated) or a non-NP head
/// (which isn't annotated)),
/// TEMPORAL_ALL_NP_AND_PP (keeps temporal annotation on NPs and PPs,
/// and it is percolated down via the head chain, but only through
/// NPs: annotation stops at either a POS tag (which is annotated)
/// or a non-NP head (which isn't annotated)).
/// TEMPORAL_NP_AND_PP_WITH_NP_HEAD (like TEMPORAL_ALL_NP_AND_PP
/// except an NP is regarded as the head of a PP)
/// TEMPORAL_ALL_NP_EVEN_UNDER_PP (like TEMPORAL_ALL_NP, but a PP-TMP
/// annotation above an NP is 'passed down' to annotate that NP
/// as temporal (but the PP itself isn't marked))
/// TEMPORAL_ALL_NP_PP_ADVP (keeps temporal annotation on NPs, PPs, and
/// ADVPs
/// and it is percolated down via the head chain, but only through
/// those categories: annotation stops at either a POS tag
/// (which is annotated)
/// or a non-NP/PP/ADVP head (which isn't annotated)),
/// TEMPORAL_9 (annotates like the previous one but
/// does all NP inside node, and their children if
/// pre-pre-terminal rather than only if head).
/// </param>
/// <param name="doSGappedStuff">
/// Leave -SBJ marking on subject NP and then mark
/// S-G sentences with a gapped subject.
/// </param>
/// <param name="leaveItAll">
/// 0 means the usual stripping of functional tags and indices;
/// 1 leaves all functional tags but still strips indices;
/// 2 leaves everything
/// </param>
/// <param name="doAdverbialNP">
/// Leave -ADV functional tag on adverbial NPs and
/// maybe add it to their head
/// </param>
/// <param name="headFinder">
/// A head finder that is used with some of the
/// options for temporalAnnotation
/// </param>
public NPTmpRetainingTreeNormalizer(int temporalAnnotation, bool doSGappedStuff, int leaveItAll, bool doAdverbialNP, IHeadFinder headFinder)
{
this.temporalAnnotation = temporalAnnotation;
this.doSGappedStuff = doSGappedStuff;
this.leaveItAll = leaveItAll;
this.doAdverbialNP = doAdverbialNP;
this.headFinder = headFinder;
}
public TregexParser(TextReader stream,
Func<string, string> basicCatFunction,
IHeadFinder headFinder) :
this(stream)
{
this.basicCatFunction = basicCatFunction;
this.headFinder = headFinder;
}
/// <summary>
/// Construct a new <code>GrammaticalStructure</code> from an existing parse tree.
/// The new <code>GrammaticalStructure</code> has the same tree structure
/// and label values as the given tree (but no shared storage). As part of
/// construction, the parse tree is analyzed using definitions from
/// {@link GrammaticalRelation <code>GrammaticalRelation</code>} to populate
/// the new <code>GrammaticalStructure</code> with as many labeled grammatical
/// relations as it can.
/// </summary>
/// <param name="t">Parse tree to make grammatical structure from</param>
/// <param name="puncFilter">Filter for punctuation words</param>
/// <param name="hf">HeadFinder to use when building it</param>
/// <param name="threadSafe">Whether or not to support simultaneous instances among multiple threads</param>
public EnglishGrammaticalStructure(Tree t, Predicate<string> puncFilter, IHeadFinder hf, bool threadSafe) :
base((new CoordinationTransformer(hf)).TransformTree(t.DeepCopy()),
EnglishGrammaticalRelations.Values(threadSafe),
threadSafe ? EnglishGrammaticalRelations.valuesLock : null,
hf, puncFilter)
{
// the tree is normalized (for index and functional tag stripping) inside CoordinationTransformer
}
/// <summary>
/// Returns a lexicalized Tree whose Labels are CategoryWordTag
/// instances, all corresponds to the input tree.
/// </summary>
public static Tree Lexicalize(Tree t, IHeadFinder hf)
{
IFunction <Tree, Tree> a = TreeFunctions.GetLabeledTreeToCategoryWordTagTreeFunction();
Tree t1 = a.Apply(t);
t1.PercolateHeads(hf);
return(t1);
}
public override IHeadFinder HeadFinder()
{
if (headFinder == null)
{
headFinder = new DybroFrenchHeadFinder(TreebankLanguagePack());
}
//Superior for vanilla PCFG over Arun's headfinding rules
return(headFinder);
}
private void SetHeadFinder(IHeadFinder hf)
{
if (hf == null)
{
throw new ArgumentException();
}
headFinder = hf;
CompileAnnotations(hf);
}
/// <summary>Constructor</summary>
/// <param name="hf">the headfinder</param>
/// <param name="performMWETransformation">
/// Parameter for backwards compatibility.
/// If set to false, multi-word expressions won't be attached to a new "MWE" node
/// </param>
public CoordinationTransformer(IHeadFinder hf, bool performMWETransformation)
{
//to get rid of unwanted nodes and tag
//to flatten date patterns
//to restructure the QP constituents
// default constructor
this.headFinder = hf;
this.performMWETransformation = performMWETransformation;
qp = new QPTreeTransformer(performMWETransformation);
}
// junk tree just for testing setCategoriesToAvoid (NP never matches VBZ but shouldn't pick the punctuation marks)
// Tree.valueOf("(FOO (BAR a) (BAZ b))") // todo: If change to always do something rather than Exception (and edit hfFeads array)
// , "BAR"
private void RunTesting(IHeadFinder hf, string[] heads)
{
NUnit.Framework.Assert.AreEqual("Test arrays out of balance", testTrees.Length, heads.Length);
for (int i = 0; i < testTrees.Length; i++)
{
Tree h = hf.DetermineHead(testTrees[i]);
string headCat = h.Value();
NUnit.Framework.Assert.AreEqual("Wrong head found", heads[i], headCat);
}
}
internal TregexMatcher(Tree root, Tree tree, IdentityHashMap <Tree, Tree> nodesToParents, IDictionary <string, Tree> namesToNodes, VariableStrings variableStrings, IHeadFinder headFinder)
{
// these things are used by "find"
this.root = root;
this.tree = tree;
this.nodesToParents = nodesToParents;
this.namesToNodes = namesToNodes;
this.variableStrings = variableStrings;
this.headFinder = headFinder;
}
public TregexMatcher(Tree root, Tree tree, IdentityDictionary<Tree, Tree> nodesToParents,
Dictionary<string, Tree> namesToNodes, VariableStrings variableStrings, IHeadFinder headFinder)
{
this.root = root;
this.tree = tree;
this.nodesToParents = nodesToParents;
this.namesToNodes = namesToNodes;
this.variableStrings = variableStrings;
this.headFinder = headFinder;
}
public TregexMatcher(Tree root, Tree tree, IdentityDictionary <Tree, Tree> nodesToParents,
Dictionary <string, Tree> namesToNodes, VariableStrings variableStrings, IHeadFinder headFinder)
{
this.root = root;
this.tree = tree;
this.nodesToParents = nodesToParents;
this.namesToNodes = namesToNodes;
this.variableStrings = variableStrings;
this.headFinder = headFinder;
}
public static ShallowParseMentionFinder GetInstance(IHeadFinder headFinder)
{
if (mInstance == null)
{
mInstance = new ShallowParseMentionFinder(headFinder);
}
else if (mInstance.HeadFinder != headFinder)
{
mInstance = new ShallowParseMentionFinder(headFinder);
}
return(mInstance);
}
public DescriptionMatcher(DescriptionPattern n, Tree root, Tree tree,
IdentityDictionary <Tree, Tree> nodesToParents,
Dictionary <string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder) :
base(root, tree, nodesToParents, namesToNodes, variableStrings, headFinder)
{
myNode = n;
// no need to reset anything - everything starts out as null or false.
// lazy initialization of children to save time.
// resetChildIter();
}
public static ShallowParseMentionFinder GetInstance(IHeadFinder headFinder)
{
if (mInstance == null)
{
mInstance = new ShallowParseMentionFinder(headFinder);
}
else if (mInstance.HeadFinder != headFinder)
{
mInstance = new ShallowParseMentionFinder(headFinder);
}
return mInstance;
}
public static PennTreebankMentionFinder GetInstance(IHeadFinder headFinder)
{
if (mInstance == null)
{
mInstance = new PennTreebankMentionFinder(headFinder);
}
else if (mInstance.HeadFinder != headFinder)
{
mInstance = new PennTreebankMentionFinder(headFinder);
}
return mInstance;
}
/// <summary>
/// Returns the set of dependencies in a tree, according to some
/// <see cref="Edu.Stanford.Nlp.Trees.IDependencyTyper{T}"/>
/// .
/// </summary>
public static ICollection <E> DependencyObjectify <E>(Tree t, IHeadFinder hf, ITreeTransformer collinizer, IDependencyTyper <E> typer)
{
ICollection <E> deps = new List <E>();
Tree t1 = collinizer.TransformTree(t);
if (t1 == null)
{
return(deps);
}
DependencyObjectifyHelper(t1, t1, hf, deps, typer);
return(deps);
}
public static PennTreebankMentionFinder GetInstance(IHeadFinder headFinder)
{
if (mInstance == null)
{
mInstance = new PennTreebankMentionFinder(headFinder);
}
else if (mInstance.HeadFinder != headFinder)
{
mInstance = new PennTreebankMentionFinder(headFinder);
}
return(mInstance);
}
public NegraPennTreebankParserParams()
: base(new NegraPennLanguagePack(defaultLeaveGF, defaultGFCharacter))
{
//Features
//Grammatical function parameters
//TODO: fix this so it really works
//wsg2010: Commented out by Roger?
//return new NegraHeadFinder();
//return new LeftHeadFinder();
headFinder = new NegraHeadFinder();
// override output encoding: make it UTF-8
SetOutputEncoding("UTF-8");
}
public CollinsDepEval(string str, bool runningAverages, IHeadFinder hf, string startSymbol)
: base(str, runningAverages)
{
this.hf = hf;
this.startSymbol = startSymbol;
precisions = new ClassicCounter <CollinsRelation>();
recalls = new ClassicCounter <CollinsRelation>();
f1s = new ClassicCounter <CollinsRelation>();
precisions2 = new ClassicCounter <CollinsRelation>();
recalls2 = new ClassicCounter <CollinsRelation>();
pnums2 = new ClassicCounter <CollinsRelation>();
rnums2 = new ClassicCounter <CollinsRelation>();
}
/// <summary>
/// Construct a new
/// <c>CollocationFinder</c>
/// over the
/// <c>Tree</c>
/// t.
/// </summary>
/// <param name="t">parse tree</param>
/// <param name="w">wordnet connection</param>
/// <param name="hf">
///
/// <see cref="IHeadFinder"/>
/// to use
/// </param>
public CollocationFinder(Tree t, IWordNetConnection w, IHeadFinder hf)
{
CoordinationTransformer transformer = new CoordinationTransformer(hf);
this.wnConnect = w;
this.qTree = transformer.TransformTree(t);
this.collocationCollector = Generics.NewArrayList();
this.hf = hf;
this.GetCollocationsList();
if (Debug)
{
log.Info("Collected collocations: " + collocationCollector);
}
}
/// <summary>Build a custom binarizer for Trees.</summary>
/// <param name="hf">the HeadFinder to use in binarization</param>
/// <param name="tlp">the TreebankLanguagePack to use</param>
/// <param name="insideFactor">whether to do inside markovization</param>
/// <param name="markovFactor">whether to markovize the binary rules</param>
/// <param name="markovOrder">the markov order to use; only relevant with markovFactor=true</param>
/// <param name="useWrappingLabels">whether to use state names (labels) that allow wrapping from right to left</param>
/// <param name="unaryAtTop">
/// Whether to actually materialize the unary that rewrites
/// a passive state to the active rule at the top of an original local
/// tree. This is used only when compaction is happening
/// </param>
/// <param name="selectiveSplitThreshold">if selective split is used, this will be the threshold used to decide which state splits to keep</param>
/// <param name="markFinalStates">whether or not to make the state names (labels) of the final active states distinctive</param>
/// <param name="noRebinarization">if true, a node which already has exactly two children is not altered</param>
public TreeBinarizer(IHeadFinder hf, ITreebankLanguagePack tlp, bool insideFactor, bool markovFactor, int markovOrder, bool useWrappingLabels, bool unaryAtTop, double selectiveSplitThreshold, bool markFinalStates, bool simpleLabels, bool noRebinarization
)
{
this.hf = hf;
this.tlp = tlp;
this.tf = new LabeledScoredTreeFactory(new CategoryWordTagFactory());
this.insideFactor = insideFactor;
this.markovFactor = markovFactor;
this.markovOrder = markovOrder;
this.useWrappingLabels = useWrappingLabels;
this.unaryAtTop = unaryAtTop;
this.selectiveSplitThreshold = selectiveSplitThreshold;
this.markFinalStates = markFinalStates;
this.simpleLabels = simpleLabels;
this.noRebinarization = noRebinarization;
}
private static void DependencyObjectifyHelper <E>(Tree t, Tree root, IHeadFinder hf, ICollection <E> c, IDependencyTyper <E> typer)
{
if (t.IsLeaf() || t.IsPreTerminal())
{
return;
}
Tree headDtr = hf.DetermineHead(t);
foreach (Tree child in t.Children())
{
DependencyObjectifyHelper(child, root, hf, c, typer);
if (child != headDtr)
{
c.Add(typer.MakeDependency(headDtr, child, root));
}
}
}
// do all con/dis-juncts have to be considered to determine a match?
// i.e. true if conj and not negated or disj and negated
public CoordinationMatcher(CoordinationPattern n, Tree root, Tree tree,
IdentityDictionary <Tree, Tree> nodesToParents,
Dictionary <string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder) :
base(root, tree, nodesToParents, namesToNodes, variableStrings, headFinder)
{
myNode = n;
children = new TregexMatcher[myNode.children.Count];
// lazy initialize the children... don't set children[i] yet
//for (int i = 0; i < children.length; i++) {
// TregexPattern node = myNode.children.get(i);
// children[i] = node.matcher(root, tree, nodesToParents,
// namesToNodes, variableStrings);
//}
currChild = 0;
considerAll = myNode.isConj ^ myNode.IsNegated();
}
/// <summary>
/// Uses the specified {@link HeadFinder <code>HeadFinder</code>}
/// to determine the heads for this node and all its descendants,
/// and to store references to the head word node and head tag node
/// in this node's {@link CoreLabel <code>CoreLabel</code>} and the
/// <code>CoreLabel</code>s of all its descendants.
///
/// Note that, in contrast to {@link Tree#percolateHeads
/// <code>Tree.percolateHeads()</code>}, which assumes {@link
/// edu.stanford.nlp.ling.CategoryWordTag
/// <code>CategoryWordTag</code>} labels and therefore stores head
/// words and head tags merely as <code>string</code>s, this
/// method stores references to the actual nodes. This mitigates
/// potential problems in sentences which contain the same word more than once.
/// </summary>
/// <param name="hf">The headfinding algorithm to use</param>
public override void PercolateHeads(IHeadFinder hf)
{
if (IsLeaf())
{
TreeGraphNode hwn = HeadWordNode();
if (hwn == null)
{
SetHeadWordNode(this);
}
}
else
{
foreach (Tree child in Children())
{
child.PercolateHeads(hf);
}
TreeGraphNode head = SafeCast(hf.DetermineHead(this, _parent));
if (head != null)
{
TreeGraphNode hwn = head.HeadWordNode();
if (hwn == null && head.IsLeaf())
{
// below us is a leaf
SetHeadWordNode(head);
}
else
{
SetHeadWordNode(hwn);
}
TreeGraphNode htn = head.HeadTagNode();
if (htn == null && head.IsLeaf())
{
// below us is a leaf
SetHeadTagNode(this);
}
else
{
SetHeadTagNode(htn);
}
}
}
}
/// <summary>Set language-specific options according to flags.</summary>
/// <remarks>
/// Set language-specific options according to flags. This routine should process the option starting in args[i] (which
/// might potentially be several arguments long if it takes arguments). It should return the index after the last index
/// it consumed in processing. In particular, if it cannot process the current option, the return value should be i.
/// <p/>
/// Generic options are processed separately by
/// <see cref="Options.SetOption(string[], int)"/>
/// , and implementations of this
/// method do not have to worry about them. The Options class handles routing options. TreebankParserParams that extend
/// this class should call super when overriding this method.
/// </remarks>
/// <param name="args"/>
/// <param name="i"/>
public override int SetOptionFlag(string[] args, int i)
{
if (Sharpen.Runtime.EqualsIgnoreCase(args[i], "-headFinder") && (i + 1 < args.Length))
{
try
{
IHeadFinder hf = (IHeadFinder)System.Activator.CreateInstance(Sharpen.Runtime.GetType(args[i + 1]));
SetHeadFinder(hf);
optionsString.Append("HeadFinder: " + args[i + 1] + "\n");
}
catch (Exception e)
{
log.Info(e);
log.Info(this.GetType().FullName + ": Could not load head finder " + args[i + 1]);
}
i += 2;
}
return(i);
}
public TreeAnnotatorAndBinarizer(IHeadFinder annotationHF, IHeadFinder binarizationHF, ITreebankLangParserParams tlpParams, bool forceCNF, bool insideFactor, bool doSubcategorization, Options op)
{
this.trainOptions = op.trainOptions;
if (doSubcategorization)
{
annotator = new TreeAnnotator(annotationHF, tlpParams, op);
}
else
{
annotator = new TreeAnnotatorAndBinarizer.TreeNullAnnotator(annotationHF);
}
binarizer = new TreeBinarizer(binarizationHF, tlpParams.TreebankLanguagePack(), insideFactor, trainOptions.markovFactor, trainOptions.markovOrder, trainOptions.CompactGrammar() > 0, trainOptions.CompactGrammar() > 1, trainOptions.HselCut, trainOptions
.markFinalStates, trainOptions.simpleBinarizedLabels, trainOptions.noRebinarization);
if (trainOptions.selectivePostSplit)
{
postSplitter = new PostSplitter(tlpParams, op);
}
else
{
postSplitter = null;
}
this.tf = new LabeledScoredTreeFactory(new CategoryWordTagFactory());
this.tlp = tlpParams.TreebankLanguagePack();
this.forceCNF = forceCNF;
if (trainOptions.printAnnotatedRuleCounts)
{
annotatedRuleCounts = new ClassicCounter <Tree>();
}
else
{
annotatedRuleCounts = null;
}
if (trainOptions.printAnnotatedStateCounts)
{
annotatedStateCounts = new ClassicCounter <string>();
}
else
{
annotatedStateCounts = null;
}
}
public MentionContext(Util.Span span, Util.Span headSpan, int entityId, IParse parse, string extentType, string nameType,
int mentionIndex, int mentionsInSentence, int mentionIndexInDocument, int sentenceIndex, IHeadFinder headFinder)
: base(span, headSpan, entityId, parse, extentType, nameType, headFinder)
{
NounPhraseSentenceIndex = mentionIndex;
MaxNounPhraseSentenceIndex = mentionsInSentence;
NounPhraseDocumentIndex = mentionIndexInDocument;
SentenceNumber = sentenceIndex;
IndexSpan = parse.Span;
PreviousToken = parse.PreviousToken;
NextToken = parse.NextToken;
Head = headFinder.GetLastHead(parse);
List<IParse> headTokens = Head.Tokens;
Tokens = headTokens.ToArray();
NextTokenBasal = Head.NextToken;
//System.err.println("MentionContext.init: "+ent+" "+ent.getEntityId()+" head="+head);
//mNonDescriptorStart = 0;
InitializeHeads(headFinder.GetHeadIndex(Head));
mGender = Similarity.GenderEnum.Unknown;
GenderProbability = 0d;
_number = Similarity.NumberEnum.Unknown;
NumberProbability = 0d;
}
/// <summary>
/// Returns the maximal projection of <code>head</code> in
/// <code>root</code> given a {@link HeadFinder}
/// </summary>
public static Tree MaximalProjection(Tree head, Tree root, IHeadFinder hf)
{
Tree projection = head;
if (projection == root)
{
return root;
}
Tree parent = projection.Parent(root);
while (hf.DetermineHead(parent) == projection)
{
projection = parent;
if (projection == root)
{
return root;
}
parent = projection.Parent(root);
}
return projection;
}
/// <summary>
/// Uses the specified {@link HeadFinder <code>HeadFinder</code>}
/// to determine the heads for this node and all its descendants,
/// and to store references to the head word node and head tag node
/// in this node's {@link CoreLabel <code>CoreLabel</code>} and the
/// <code>CoreLabel</code>s of all its descendants.
///
/// Note that, in contrast to {@link Tree#percolateHeads
/// <code>Tree.percolateHeads()</code>}, which assumes {@link
/// edu.stanford.nlp.ling.CategoryWordTag
/// <code>CategoryWordTag</code>} labels and therefore stores head
/// words and head tags merely as <code>string</code>s, this
/// method stores references to the actual nodes. This mitigates
/// potential problems in sentences which contain the same word more than once.
/// </summary>
/// <param name="hf">The headfinding algorithm to use</param>
public override void PercolateHeads(IHeadFinder hf)
{
if (IsLeaf())
{
TreeGraphNode hwn = HeadWordNode();
if (hwn == null)
{
SetHeadWordNode(this);
}
}
else
{
foreach (Tree child in Children())
{
child.PercolateHeads(hf);
}
TreeGraphNode head = SafeCast(hf.DetermineHead(this, _parent));
if (head != null)
{
TreeGraphNode hwn = head.HeadWordNode();
if (hwn == null && head.IsLeaf())
{
// below us is a leaf
SetHeadWordNode(head);
}
else
{
SetHeadWordNode(hwn);
}
TreeGraphNode htn = head.HeadTagNode();
if (htn == null && head.IsLeaf())
{
// below us is a leaf
SetHeadTagNode(this);
}
else
{
SetHeadTagNode(htn);
}
}
}
}
public abstract TregexMatcher Matcher(Tree root, Tree tree,
IdentityDictionary<Tree, Tree> nodesToParents,
Dictionary<string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder);
/// <summary>
/// Get a {@link TregexMatcher} for this pattern on this tree.
/// Any Relations which use heads of trees should use the provided HeadFinder.
/// </summary>
/// <param name="t">a tree to match on</param>
/// <param name="headFinder">a HeadFinder to use when matching</param>
/// <returns>a TregexMatcher</returns>
public TregexMatcher Matcher(Tree t, IHeadFinder headFinder)
{
return Matcher(t, t, null, new Dictionary<string, Tree>(), new VariableStrings(), headFinder);
}
/// <summary>
/// Return a GrammaticalStructureFactory suitable for this language/treebank.
/// (To be overridden in subclasses.)
/// </summary>
/// <returns>A GrammaticalStructureFactory suitable for this language/treebank</returns>
public IGrammaticalStructureFactory GrammaticalStructureFactory(Predicate<string> puncFilt,
IHeadFinder typedDependencyHf)
{
return GrammaticalStructureFactory();
}
// do all con/dis-juncts have to be considered to determine a match?
// i.e. true if conj and not negated or disj and negated
public CoordinationMatcher(CoordinationPattern n, Tree root, Tree tree,
IdentityDictionary<Tree, Tree> nodesToParents,
Dictionary<string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder) :
base(root, tree, nodesToParents, namesToNodes, variableStrings, headFinder)
{
myNode = n;
children = new TregexMatcher[myNode.children.Count];
// lazy initialize the children... don't set children[i] yet
//for (int i = 0; i < children.length; i++) {
// TregexPattern node = myNode.children.get(i);
// children[i] = node.matcher(root, tree, nodesToParents,
// namesToNodes, variableStrings);
//}
currChild = 0;
considerAll = myNode.isConj ^ myNode.IsNegated();
}
private ShallowParseMentionFinder(IHeadFinder headFinder)
{
HeadFinder = headFinder;
PrenominalNamedEntitiesCollection = true;
CoordinatedNounPhrasesCollection = true;
}
/// <summary>
/// A compiler that uses this HeadFinder and the default basicCategoryFunction
/// </summary>
/// <param name="headFinder">the HeadFinder</param>
public TregexPatternCompiler(IHeadFinder headFinder) :
this(headFinder, DEFAULT_BASIC_CAT_FUNCTION)
{
}
/// <summary>
/// Constructs context information for the specified mention.
/// </summary>
/// <param name="mention">
/// The mention object on which this object is based.
/// </param>
/// <param name="mentionIndexInSentence">
/// The mention's position in the sentence.
/// </param>
/// <param name="mentionsInSentence">
/// The number of mentions in the sentence.
/// </param>
/// <param name="mentionIndexInDocument">
/// The index of this mention with respect to the document.
/// </param>
/// <param name="sentenceIndex">
/// The index of the sentence which contains this mention.
/// </param>
/// <param name="headFinder">
/// An object which provides head information.
/// </param>
public MentionContext(Mention mention, int mentionIndexInSentence, int mentionsInSentence, int mentionIndexInDocument, int sentenceIndex, IHeadFinder headFinder)
: this(mention.Span, mention.HeadSpan, mention.Id, mention.Parse, mention.Type, mention.NameType, mentionIndexInSentence, mentionsInSentence, mentionIndexInDocument, sentenceIndex, headFinder)
{
}
/// <summary>
/// Creates a new mention finder with the specified head finder.
/// </summary>
/// <param name="headFinder">
/// The head finder.
/// </param>
private PennTreebankMentionFinder(IHeadFinder headFinder)
{
PrenominalNamedEntitiesCollection = false;
CoordinatedNounPhrasesCollection = true;
HeadFinder = headFinder;
}
/// <summary>
/// default constructor
/// </summary>
public CoordinationTransformer(IHeadFinder hf)
{
this.headFinder = hf;
}
public DescriptionMatcher(DescriptionPattern n, Tree root, Tree tree,
IdentityDictionary<Tree, Tree> nodesToParents,
Dictionary<string, Tree> namesToNodes,
VariableStrings variableStrings,
IHeadFinder headFinder) :
base(root, tree, nodesToParents, namesToNodes, variableStrings, headFinder)
{
myNode = n;
// no need to reset anything - everything starts out as null or false.
// lazy initialization of children to save time.
// resetChildIter();
}
/// <summary>
/// This gets used by GrammaticalStructureFactory (by reflection). DON'T DELETE.
/// </summary>
/// <param name="t">Parse tree to make grammatical structure from</param>
/// <param name="puncFilter">Filter to remove punctuation dependencies</param>
/// <param name="hf">HeadFinder to use when building it</param>
public EnglishGrammaticalStructure(Tree t, Predicate<string> puncFilter, IHeadFinder hf) :
this(t, puncFilter, hf, true)
{
}
/// <summary>
/// A compiler that uses this HeadFinder and this basicCategoryFunction
/// </summary>
/// <param name="headFinder">the HeadFinder</param>
/// <param name="basicCatFunction">The function mapping strings to strings</param>
public TregexPatternCompiler(IHeadFinder headFinder,
AbstractTreebankLanguagePack.BasicCategoryStringFunction basicCatFunction)
{
this.headFinder = headFinder;
this.basicCatFunction = basicCatFunction;
}