// returns Pair<node,foot>
private Pair <Tree, Tree> CopyHelper(Tree node, IDictionary <string, Tree> newNamesToNodes, ITreeFactory treeFactory, ILabelFactory labelFactory)
{
Tree clone;
Tree newFoot = null;
if (node.IsLeaf())
{
if (node == foot)
{
// found the foot node; pass it up.
clone = treeFactory.NewTreeNode(node.Label(), new List <Tree>(0));
newFoot = clone;
}
else
{
clone = treeFactory.NewLeaf(labelFactory.NewLabel(node.Label()));
}
}
else
{
IList <Tree> newChildren = new List <Tree>(node.Children().Length);
foreach (Tree child in node.Children())
{
Pair <Tree, Tree> newChild = CopyHelper(child, newNamesToNodes, treeFactory, labelFactory);
newChildren.Add(newChild.First());
if (newChild.Second() != null)
{
if (newFoot != null)
{
log.Info("Error -- two feet found when copying auxiliary tree " + tree.ToString() + "; using last foot found.");
}
newFoot = newChild.Second();
}
}
clone = treeFactory.NewTreeNode(labelFactory.NewLabel(node.Label()), newChildren);
}
if (nodesToNames.Contains(node))
{
newNamesToNodes[nodesToNames[node]] = clone;
}
return(new Pair <Tree, Tree>(clone, newFoot));
}
/// <summary>Turns a sentence into a flat phrasal tree.</summary>
/// <remarks>
/// Turns a sentence into a flat phrasal tree.
/// The structure is S -> tag*. And then each tag goes to a word.
/// The tag is either found from the label or made "WD".
/// The tag and phrasal node have a StringLabel.
/// </remarks>
/// <param name="s">The Sentence to make the Tree from</param>
/// <param name="lf">The LabelFactory with which to create the new Tree labels</param>
/// <returns>The one phrasal level Tree</returns>
public static Tree ToFlatTree <_T0>(IList <_T0> s, ILabelFactory lf)
where _T0 : IHasWord
{
IList <Tree> daughters = new List <Tree>(s.Count);
foreach (IHasWord word in s)
{
Tree wordNode = new LabeledScoredTreeNode(lf.NewLabel(word.Word()));
if (word is TaggedWord)
{
TaggedWord taggedWord = (TaggedWord)word;
wordNode = new LabeledScoredTreeNode(new StringLabel(taggedWord.Tag()), Java.Util.Collections.SingletonList(wordNode));
}
else
{
wordNode = new LabeledScoredTreeNode(lf.NewLabel("WD"), Java.Util.Collections.SingletonList(wordNode));
}
daughters.Add(wordNode);
}
return(new LabeledScoredTreeNode(new StringLabel("S"), daughters));
}
/// <summary>Add -TMP when not present within an NP</summary>
/// <param name="tree">The tree to add temporal info to.</param>
private void AddTMP9(Tree tree)
{
// do the head chain under it
Tree ht = headFinder.DetermineHead(tree);
// special fix for possessives! -- make noun before head
if (ht.Value().Equals("POS"))
{
int j = tree.ObjectIndexOf(ht);
if (j > 0)
{
ht = tree.GetChild(j - 1);
}
}
// Note: this next bit changes the tree label, rather
// than creating a new tree node. Beware!
if (ht.IsPreTerminal() || ht.Value().StartsWith("NP") || ht.Value().StartsWith("PP") || ht.Value().StartsWith("ADVP"))
{
if (!TmpPattern.Matcher(ht.Value()).Matches())
{
ILabelFactory lf = ht.LabelFactory();
// log.info("TMP: Changing " + ht.value() + " to " +
// ht.value() + "-TMP");
ht.SetLabel(lf.NewLabel(ht.Value() + "-TMP"));
}
if (ht.Value().StartsWith("NP") || ht.Value().StartsWith("PP") || ht.Value().StartsWith("ADVP"))
{
AddTMP9(ht);
}
}
// do the NPs under it (which may or may not be the head chain
Tree[] kidlets = tree.Children();
foreach (Tree kidlet in kidlets)
{
ht = kidlet;
ILabelFactory lf;
if (tree.IsPrePreTerminal() && !TmpPattern.Matcher(ht.Value()).Matches())
{
// log.info("TMP: Changing " + ht.value() + " to " +
// ht.value() + "-TMP");
lf = ht.LabelFactory();
// Note: this next bit changes the tree label, rather
// than creating a new tree node. Beware!
ht.SetLabel(lf.NewLabel(ht.Value() + "-TMP"));
}
else
{
if (ht.Value().StartsWith("NP"))
{
// don't add -TMP twice!
if (!TmpPattern.Matcher(ht.Value()).Matches())
{
lf = ht.LabelFactory();
// log.info("TMP: Changing " + ht.value() + " to " +
// ht.value() + "-TMP");
// Note: this next bit changes the tree label, rather
// than creating a new tree node. Beware!
ht.SetLabel(lf.NewLabel(ht.Value() + "-TMP"));
}
AddTMP9(ht);
}
}
}
}
/// <summary>
/// If things match, this method destructively changes the children list
/// of the tree t. When this method is called, t is an NP and there must
/// be at least two children to the right of ccIndex.
/// </summary>
/// <param name="t">The tree to transform a conjunction in</param>
/// <param name="ccIndex">The index of the CC child</param>
/// <returns>t</returns>
private static Tree TransformCc(Tree t, int ccIndex)
{
// use the factories of t to create new nodes
ITreeFactory tf = t.TreeFactory();
ILabelFactory lf = t.Label().LabelFactory();
Tree[] ccSiblings = t.Children();
//check if other CC
var ccPositions = new List <int>();
for (int i = ccIndex + 1; i < ccSiblings.Length; i++)
{
if (ccSiblings[i].Value().StartsWith(PartsOfSpeech.CoordinatingConjunction) && i < ccSiblings.Length - 1)
{
// second conjunct to ensure that a CC we add isn't the last child
ccPositions.Add(i);
}
}
// a CC b c ... -> (a CC b) c ... with b not a DT
string beforeSibling = ccSiblings[ccIndex - 1].Value();
if (ccIndex == 1 &&
(beforeSibling == PartsOfSpeech.Determiner ||
beforeSibling == PartsOfSpeech.Adjective ||
beforeSibling == PartsOfSpeech.Adverb ||
!(ccSiblings[ccIndex + 1].Value() == PartsOfSpeech.Determiner)) &&
!(beforeSibling.StartsWith("NP") ||
beforeSibling.Equals("ADJP") ||
beforeSibling == PartsOfSpeech.NounPlural))
{
// && (ccSiblings.Length == ccIndex + 3 || !ccPositions.isEmpty())) { // something like "soya or maize oil"
string leftHead = GetHeadTag(ccSiblings[ccIndex - 1]);
//create a new tree to be inserted as first child of t
Tree left = tf.NewTreeNode(lf.NewLabel(leftHead), null);
for (int i = 0; i < ccIndex + 2; i++)
{
left.AddChild(ccSiblings[i]);
}
// remove all the children of t before ccIndex+2
for (int i = 0; i < ccIndex + 2; i++)
{
t.RemoveChild(0);
}
// if stuff after (like "soya or maize oil and vegetables")
// we need to put the tree in another tree
if (ccPositions.Any())
{
bool comma = false;
int index = ccPositions[0];
if (ccSiblings[index - 1].Value() == PartsOfSpeech.Comma)
{
//to handle the case of a comma ("soya and maize oil, and vegetables")
index = index - 1;
comma = true;
}
string head = GetHeadTag(ccSiblings[index - 1]);
if (ccIndex + 2 < index)
{
Tree tree = tf.NewTreeNode(lf.NewLabel(head), null);
tree.AddChild(0, left);
int k = 1;
for (int j = ccIndex + 2; j < index; j++)
{
t.RemoveChild(0);
tree.AddChild(k, ccSiblings[j]);
k++;
}
t.AddChild(0, tree);
}
else
{
t.AddChild(0, left);
}
Tree rightTree = tf.NewTreeNode(lf.NewLabel(Noun), null);
int start = 2;
if (comma)
{
start++;
}
while (start < t.NumChildren())
{
Tree sib = t.GetChild(start);
t.RemoveChild(start);
rightTree.AddChild(sib);
}
t.AddChild(rightTree);
}
else
{
t.AddChild(0, left);
}
}
// DT a CC b c -> DT (a CC b) c
else if (ccIndex == 2 && ccSiblings[0].Value().StartsWith("DT") &&
ccSiblings[ccIndex - 1].Value() != PartsOfSpeech.NounPlural &&
(ccSiblings.Length == 5 || (ccPositions.Any() && ccPositions[0] == 5)))
{
string head = GetHeadTag(ccSiblings[ccIndex - 1]);
//create a new tree to be inserted as second child of t (after the determiner
Tree child = tf.NewTreeNode(lf.NewLabel(head), null);
for (int i = 1; i < ccIndex + 2; i++)
{
child.AddChild(ccSiblings[i]);
}
// remove all the children of t between the determiner and ccIndex+2
for (int i = 1; i < ccIndex + 2; i++)
{
t.RemoveChild(1);
}
t.AddChild(1, child);
}
// ... a, b CC c ... -> ... (a, b CC c) ...
else if (ccIndex > 2 && ccSiblings[ccIndex - 2].Value() == PartsOfSpeech.Comma &&
ccSiblings[ccIndex - 1].Value() != PartsOfSpeech.NounPlural)
{
string head = GetHeadTag(ccSiblings[ccIndex - 1]);
Tree child = tf.NewTreeNode(lf.NewLabel(head), null);
for (int j = ccIndex - 3; j < ccIndex + 2; j++)
{
child.AddChild(ccSiblings[j]);
}
int i = ccIndex - 4;
while (i > 0 && ccSiblings[i].Value() == PartsOfSpeech.Comma)
{
child.AddChild(0, ccSiblings[i]); // add the comma
child.AddChild(0, ccSiblings[i - 1]); // add the word before the comma
i = i - 2;
}
if (i < 0)
{
i = -1;
}
// remove the old children
for (int j = i + 1; j < ccIndex + 2; j++)
{
t.RemoveChild(i + 1);
}
// put the new tree
t.AddChild(i + 1, child);
}
// something like "the new phone book and tour guide" -> multiple heads
// we want (NP the new phone book) (CC and) (NP tour guide)
else
{
bool commaLeft = false;
bool commaRight = false;
bool preconj = false;
int indexBegin = 0;
Tree conjT = tf.NewTreeNode(lf.NewLabel(PartsOfSpeech.CoordinatingConjunction), null);
// create the left tree
string leftHead = GetHeadTag(ccSiblings[ccIndex - 1]);
Tree left = tf.NewTreeNode(lf.NewLabel(leftHead), null);
// handle the case of a preconjunct (either, both, neither)
Tree first = ccSiblings[0];
string leaf = first.FirstChild().Value().ToLower();
if (leaf.Equals("either") || leaf.Equals("neither") || leaf.Equals("both"))
{
preconj = true;
indexBegin = 1;
conjT.AddChild(first.FirstChild());
}
for (int i = indexBegin; i < ccIndex - 1; i++)
{
left.AddChild(ccSiblings[i]);
}
// handle the case of a comma ("GM soya and maize, and food ingredients")
if (ccSiblings[ccIndex - 1].Value() == PartsOfSpeech.Comma)
{
commaLeft = true;
}
else
{
left.AddChild(ccSiblings[ccIndex - 1]);
}
// create the CC tree
Tree cc = ccSiblings[ccIndex];
// create the right tree
int nextCc;
if (!ccPositions.Any())
{
nextCc = ccSiblings.Length;
}
else
{
nextCc = ccPositions[0];
}
string rightHead = GetHeadTag(ccSiblings[nextCc - 1]);
Tree right = tf.NewTreeNode(lf.NewLabel(rightHead), null);
for (int i = ccIndex + 1; i < nextCc - 1; i++)
{
right.AddChild(ccSiblings[i]);
}
// handle the case of a comma ("GM soya and maize, and food ingredients")
if (ccSiblings[nextCc - 1].Value() == PartsOfSpeech.Comma)
{
commaRight = true;
}
else
{
right.AddChild(ccSiblings[nextCc - 1]);
}
// put trees together in old t, first we remove the old nodes
for (int i = 0; i < nextCc; i++)
{
t.RemoveChild(0);
}
if (ccPositions.Any())
{
// need an extra level
Tree tree = tf.NewTreeNode(lf.NewLabel(Noun), null);
if (preconj)
{
tree.AddChild(conjT);
}
if (left.NumChildren() > 0)
{
tree.AddChild(left);
}
if (commaLeft)
{
tree.AddChild(ccSiblings[ccIndex - 1]);
}
tree.AddChild(cc);
if (right.NumChildren() > 0)
{
tree.AddChild(right);
}
if (commaRight)
{
t.AddChild(0, ccSiblings[nextCc - 1]);
}
t.AddChild(0, tree);
}
else
{
if (preconj)
{
t.AddChild(conjT);
}
if (left.NumChildren() > 0)
{
t.AddChild(left);
}
if (commaLeft)
{
t.AddChild(ccSiblings[ccIndex - 1]);
}
t.AddChild(cc);
if (right.NumChildren() > 0)
{
t.AddChild(right);
}
if (commaRight)
{
t.AddChild(ccSiblings[nextCc - 1]);
}
}
}
return(t);
}
public override Tree NewLeaf(string word)
{
return(new LabeledScoredTreeNode(lf.NewLabel(word)));
}
public Tree NewLeaf(string word)
{
return(NewLeaf(mlf.NewLabel(word)));
}
/// <summary>
/// Create a new <code>TreeGraphNode</code> with the supplied label
/// </summary>
/// <param name="label">the label for this node</param>
public TreeGraphNode(ILabel label)
{
this._label = (CoreLabel)Mlf.NewLabel(label);
}
/// <summary>
/// If things match, this method destructively changes the children list
/// of the tree t.
/// </summary>
/// <remarks>
/// If things match, this method destructively changes the children list
/// of the tree t. When this method is called, t is an NP and there must
/// be at least two children to the right of ccIndex.
/// </remarks>
/// <param name="t">The tree to transform a conjunction in</param>
/// <param name="ccIndex">The index of the CC child</param>
/// <returns>t</returns>
private static Tree TransformCC(Tree t, int ccIndex)
{
if (Verbose)
{
log.Info("transformCC in: " + t);
}
//System.out.println(ccIndex);
// use the factories of t to create new nodes
ITreeFactory tf = t.TreeFactory();
ILabelFactory lf = t.Label().LabelFactory();
Tree[] ccSiblings = t.Children();
//check if other CC
IList <int> ccPositions = new List <int>();
for (int i = ccIndex + 1; i < ccSiblings.Length; i++)
{
if (ccSiblings[i].Value().StartsWith("CC") && i < ccSiblings.Length - 1)
{
// second conjunct to ensure that a CC we add isn't the last child
ccPositions.Add(int.Parse(i));
}
}
// a CC b c ... -> (a CC b) c ... with b not a DT
string beforeSibling = ccSiblings[ccIndex - 1].Value();
if (ccIndex == 1 && (beforeSibling.Equals("DT") || beforeSibling.Equals("JJ") || beforeSibling.Equals("RB") || !(ccSiblings[ccIndex + 1].Value().Equals("DT"))) && !(beforeSibling.StartsWith("NP") || beforeSibling.Equals("ADJP") || beforeSibling
.Equals("NNS")))
{
// && (ccSiblings.length == ccIndex + 3 || !ccPositions.isEmpty())) { // something like "soya or maize oil"
string leftHead = GetHeadTag(ccSiblings[ccIndex - 1]);
//create a new tree to be inserted as first child of t
Tree left = tf.NewTreeNode(lf.NewLabel(leftHead), null);
for (int i_1 = 0; i_1 < ccIndex + 2; i_1++)
{
left.AddChild(ccSiblings[i_1]);
}
if (Verbose)
{
System.Console.Out.WriteLine("print left tree");
left.PennPrint();
System.Console.Out.WriteLine();
}
// remove all the children of t before ccIndex+2
for (int i_2 = 0; i_2 < ccIndex + 2; i_2++)
{
t.RemoveChild(0);
}
if (Verbose)
{
if (t.NumChildren() == 0)
{
System.Console.Out.WriteLine("Youch! No t children");
}
}
// if stuff after (like "soya or maize oil and vegetables")
// we need to put the tree in another tree
if (!ccPositions.IsEmpty())
{
bool comma = false;
int index = ccPositions[0];
if (Verbose)
{
log.Info("more CC index " + index);
}
if (ccSiblings[index - 1].Value().Equals(","))
{
//to handle the case of a comma ("soya and maize oil, and vegetables")
index = index - 1;
comma = true;
}
if (Verbose)
{
log.Info("more CC index " + index);
}
string head = GetHeadTag(ccSiblings[index - 1]);
if (ccIndex + 2 < index)
{
Tree tree = tf.NewTreeNode(lf.NewLabel(head), null);
tree.AddChild(0, left);
int k = 1;
for (int j = ccIndex + 2; j < index; j++)
{
if (Verbose)
{
ccSiblings[j].PennPrint();
}
t.RemoveChild(0);
tree.AddChild(k, ccSiblings[j]);
k++;
}
if (Verbose)
{
System.Console.Out.WriteLine("print t");
t.PennPrint();
System.Console.Out.WriteLine("print tree");
tree.PennPrint();
System.Console.Out.WriteLine();
}
t.AddChild(0, tree);
}
else
{
t.AddChild(0, left);
}
Tree rightTree = tf.NewTreeNode(lf.NewLabel("NP"), null);
int start = 2;
if (comma)
{
start++;
}
while (start < t.NumChildren())
{
Tree sib = t.GetChild(start);
t.RemoveChild(start);
rightTree.AddChild(sib);
}
t.AddChild(rightTree);
}
else
{
t.AddChild(0, left);
}
}
else
{
// DT a CC b c -> DT (a CC b) c
if (ccIndex == 2 && ccSiblings[0].Value().StartsWith("DT") && !ccSiblings[ccIndex - 1].Value().Equals("NNS") && (ccSiblings.Length == 5 || (!ccPositions.IsEmpty() && ccPositions[0] == 5)))
{
string head = GetHeadTag(ccSiblings[ccIndex - 1]);
//create a new tree to be inserted as second child of t (after the determiner
Tree child = tf.NewTreeNode(lf.NewLabel(head), null);
for (int i_1 = 1; i_1 < ccIndex + 2; i_1++)
{
child.AddChild(ccSiblings[i_1]);
}
if (Verbose)
{
if (child.NumChildren() == 0)
{
System.Console.Out.WriteLine("Youch! No child children");
}
}
// remove all the children of t between the determiner and ccIndex+2
//System.out.println("print left tree");
//child.pennPrint();
for (int i_2 = 1; i_2 < ccIndex + 2; i_2++)
{
t.RemoveChild(1);
}
t.AddChild(1, child);
}
else
{
// ... a, b CC c ... -> ... (a, b CC c) ...
if (ccIndex > 2 && ccSiblings[ccIndex - 2].Value().Equals(",") && !ccSiblings[ccIndex - 1].Value().Equals("NNS"))
{
string head = GetHeadTag(ccSiblings[ccIndex - 1]);
Tree child = tf.NewTreeNode(lf.NewLabel(head), null);
for (int i_1 = ccIndex - 3; i_1 < ccIndex + 2; i_1++)
{
child.AddChild(ccSiblings[i_1]);
}
if (Verbose)
{
if (child.NumChildren() == 0)
{
System.Console.Out.WriteLine("Youch! No child children");
}
}
int i_2 = ccIndex - 4;
while (i_2 > 0 && ccSiblings[i_2].Value().Equals(","))
{
child.AddChild(0, ccSiblings[i_2]);
// add the comma
child.AddChild(0, ccSiblings[i_2 - 1]);
// add the word before the comma
i_2 = i_2 - 2;
}
if (i_2 < 0)
{
i_2 = -1;
}
// remove the old children
for (int j = i_2 + 1; j < ccIndex + 2; j++)
{
t.RemoveChild(i_2 + 1);
}
// put the new tree
t.AddChild(i_2 + 1, child);
}
else
{
// something like "the new phone book and tour guide" -> multiple heads
// we want (NP the new phone book) (CC and) (NP tour guide)
bool commaLeft = false;
bool commaRight = false;
bool preconj = false;
int indexBegin = 0;
Tree conjT = tf.NewTreeNode(lf.NewLabel("CC"), null);
// create the left tree
string leftHead = GetHeadTag(ccSiblings[ccIndex - 1]);
Tree left = tf.NewTreeNode(lf.NewLabel(leftHead), null);
// handle the case of a preconjunct (either, both, neither)
Tree first = ccSiblings[0];
string leaf = first.FirstChild().Value().ToLower();
if (leaf.Equals("either") || leaf.Equals("neither") || leaf.Equals("both"))
{
preconj = true;
indexBegin = 1;
conjT.AddChild(first.FirstChild());
}
for (int i_1 = indexBegin; i_1 < ccIndex - 1; i_1++)
{
left.AddChild(ccSiblings[i_1]);
}
// handle the case of a comma ("GM soya and maize, and food ingredients")
if (ccSiblings[ccIndex - 1].Value().Equals(","))
{
commaLeft = true;
}
else
{
left.AddChild(ccSiblings[ccIndex - 1]);
}
// create the CC tree
Tree cc = ccSiblings[ccIndex];
// create the right tree
int nextCC;
if (ccPositions.IsEmpty())
{
nextCC = ccSiblings.Length;
}
else
{
nextCC = ccPositions[0];
}
string rightHead = GetHeadTag(ccSiblings[nextCC - 1]);
Tree right = tf.NewTreeNode(lf.NewLabel(rightHead), null);
for (int i_2 = ccIndex + 1; i_2 < nextCC - 1; i_2++)
{
right.AddChild(ccSiblings[i_2]);
}
// handle the case of a comma ("GM soya and maize, and food ingredients")
if (ccSiblings[nextCC - 1].Value().Equals(","))
{
commaRight = true;
}
else
{
right.AddChild(ccSiblings[nextCC - 1]);
}
if (Verbose)
{
if (left.NumChildren() == 0)
{
System.Console.Out.WriteLine("Youch! No left children");
}
if (right.NumChildren() == 0)
{
System.Console.Out.WriteLine("Youch! No right children");
}
}
// put trees together in old t, first we remove the old nodes
for (int i_3 = 0; i_3 < nextCC; i_3++)
{
t.RemoveChild(0);
}
if (!ccPositions.IsEmpty())
{
// need an extra level
Tree tree = tf.NewTreeNode(lf.NewLabel("NP"), null);
if (preconj)
{
tree.AddChild(conjT);
}
if (left.NumChildren() > 0)
{
tree.AddChild(left);
}
if (commaLeft)
{
tree.AddChild(ccSiblings[ccIndex - 1]);
}
tree.AddChild(cc);
if (right.NumChildren() > 0)
{
tree.AddChild(right);
}
if (commaRight)
{
t.AddChild(0, ccSiblings[nextCC - 1]);
}
t.AddChild(0, tree);
}
else
{
if (preconj)
{
t.AddChild(conjT);
}
if (left.NumChildren() > 0)
{
t.AddChild(left);
}
if (commaLeft)
{
t.AddChild(ccSiblings[ccIndex - 1]);
}
t.AddChild(cc);
if (right.NumChildren() > 0)
{
t.AddChild(right);
}
if (commaRight)
{
t.AddChild(ccSiblings[nextCC - 1]);
}
}
}
}
}
if (Verbose)
{
log.Info("transformCC out: " + t);
}
return(t);
}
/// <summary>
/// Create a new
/// <c>TreeGraphNode</c>
/// with the supplied
/// label.
/// </summary>
/// <param name="label">the label for this node.</param>
public TreeGraphNode(ILabel label)
{
// = null;
this.label = (CoreLabel)mlf.NewLabel(label);
}