public Configuration(ICoreMap sentence)
{
this.stack = new List <int>();
this.buffer = new List <int>();
this.tree = new DependencyTree();
this.sentence = sentence;
}
public Configuration(Edu.Stanford.Nlp.Parser.Nndep.Configuration config)
{
stack = new List <int>(config.stack);
buffer = new List <int>(config.buffer);
tree = new DependencyTree(config.tree);
sentence = new CoreLabel(config.sentence);
}
public virtual bool HasOtherChild(int k, DependencyTree goldTree)
{
for (int i = 1; i <= tree.n; ++i)
{
if (goldTree.GetHead(i) == k && tree.GetHead(i) != k)
{
return(true);
}
}
return(false);
}
internal override bool IsOracle(Configuration c, string t, DependencyTree dTree)
{
if (!CanApply(c, t))
{
return(false);
}
if (t.StartsWith("L") && !dTree.GetLabel(c.GetStack(1)).Equals(Sharpen.Runtime.Substring(t, 2, t.Length - 1)))
{
return(false);
}
if (t.StartsWith("R") && !dTree.GetLabel(c.GetStack(0)).Equals(Sharpen.Runtime.Substring(t, 2, t.Length - 1)))
{
return(false);
}
Configuration ct = new Configuration(c);
Apply(ct, t);
return(CanReach(ct, dTree));
}
// O(n) implementation
public override string GetOracle(Configuration c, DependencyTree dTree)
{
int w1 = c.GetStack(1);
int w2 = c.GetStack(0);
if (w1 > 0 && dTree.GetHead(w1) == w2)
{
return("L(" + dTree.GetLabel(w1) + ')');
}
else
{
if (w1 >= 0 && dTree.GetHead(w2) == w1 && !c.HasOtherChild(w2, dTree))
{
return("R(" + dTree.GetLabel(w2) + ')');
}
else
{
return("S");
}
}
}
public static void WriteConllFile(string outFile, IList <ICoreMap> sentences, IList <DependencyTree> trees)
{
try
{
PrintWriter output = IOUtils.GetPrintWriter(outFile);
for (int i = 0; i < sentences.Count; i++)
{
ICoreMap sentence = sentences[i];
DependencyTree tree = trees[i];
IList <CoreLabel> tokens = sentence.Get(typeof(CoreAnnotations.TokensAnnotation));
for (int j = 1; j <= size; ++j)
{
CoreLabel token = tokens[j - 1];
output.Printf("%d\t%s\t_\t%s\t%s\t_\t%d\t%s\t_\t_%n", j, token.Word(), token.Tag(), token.Tag(), tree.GetHead(j), tree.GetLabel(j));
}
output.Println();
}
output.Close();
}
catch (Exception e)
{
throw new RuntimeIOException(e);
}
}
/// <summary> /// Determine whether applying the given transition in the given /// configuration tree will leave in us a state in which we can reach /// the gold tree. /// </summary> /// <remarks> /// Determine whether applying the given transition in the given /// configuration tree will leave in us a state in which we can reach /// the gold tree. (Useful for building a dynamic oracle.) /// </remarks> internal abstract bool IsOracle(Configuration c, string t, DependencyTree dTree);
/// <summary> /// Provide a static-oracle recommendation for the next parsing step /// to take. /// </summary> /// <param name="c">Current parser configuration</param> /// <param name="dTree">Gold tree which parser needs to reach</param> /// <returns>Transition string</returns> public abstract string GetOracle(Configuration c, DependencyTree dTree);
// TODO replace with GrammaticalStructure#readCoNLLGrammaticalStructureCollection
public static void LoadConllFile(string inFile, IList <ICoreMap> sents, IList <DependencyTree> trees, bool unlabeled, bool cPOS)
{
CoreLabelTokenFactory tf = new CoreLabelTokenFactory(false);
try
{
using (BufferedReader reader = IOUtils.ReaderFromString(inFile))
{
IList <CoreLabel> sentenceTokens = new List <CoreLabel>();
DependencyTree tree = new DependencyTree();
foreach (string line in IOUtils.GetLineIterable(reader, false))
{
string[] splits = line.Split("\t");
if (splits.Length < 10)
{
if (sentenceTokens.Count > 0)
{
trees.Add(tree);
ICoreMap sentence = new CoreLabel();
sentence.Set(typeof(CoreAnnotations.TokensAnnotation), sentenceTokens);
sents.Add(sentence);
tree = new DependencyTree();
sentenceTokens = new List <CoreLabel>();
}
}
else
{
string word = splits[1];
string pos = cPOS ? splits[3] : splits[4];
string depType = splits[7];
int head = -1;
try
{
head = System.Convert.ToInt32(splits[6]);
}
catch (NumberFormatException)
{
continue;
}
CoreLabel token = tf.MakeToken(word, 0, 0);
token.SetTag(pos);
token.Set(typeof(CoreAnnotations.CoNLLDepParentIndexAnnotation), head);
token.Set(typeof(CoreAnnotations.CoNLLDepTypeAnnotation), depType);
sentenceTokens.Add(token);
if (!unlabeled)
{
tree.Add(head, depType);
}
else
{
tree.Add(head, Config.Unknown);
}
}
}
}
}
catch (IOException e)
{
throw new RuntimeIOException(e);
}
}
// NOTE: need to check the correctness again.
private static bool CanReach(Configuration c, DependencyTree dTree)
{
int n = c.GetSentenceSize();
for (int i = 1; i <= n; ++i)
{
if (c.GetHead(i) != Config.Nonexist && c.GetHead(i) != dTree.GetHead(i))
{
return(false);
}
}
bool[] inBuffer = new bool[n + 1];
bool[] depInList = new bool[n + 1];
int[] leftL = new int[n + 2];
int[] rightL = new int[n + 2];
for (int i_1 = 0; i_1 < c.GetBufferSize(); ++i_1)
{
inBuffer[c.buffer[i_1]] = true;
}
int nLeft = c.GetStackSize();
for (int i_2 = 0; i_2 < nLeft; ++i_2)
{
int x = c.stack[i_2];
leftL[nLeft - i_2] = x;
if (x > 0)
{
depInList[dTree.GetHead(x)] = true;
}
}
int nRight = 1;
rightL[nRight] = leftL[1];
for (int i_3 = 0; i_3 < c.GetBufferSize(); ++i_3)
{
// boolean inList = false;
int x = c.buffer[i_3];
if (!inBuffer[dTree.GetHead(x)] || depInList[x])
{
rightL[++nRight] = x;
depInList[dTree.GetHead(x)] = true;
}
}
int[][] g = new int[][] { };
for (int i_4 = 1; i_4 <= nLeft; ++i_4)
{
for (int j = 1; j <= nRight; ++j)
{
g[i_4][j] = -1;
}
}
g[1][1] = leftL[1];
for (int i_5 = 1; i_5 <= nLeft; ++i_5)
{
for (int j_1 = 1; j_1 <= nRight; ++j_1)
{
if (g[i_5][j_1] != -1)
{
int x = g[i_5][j_1];
if (j_1 < nRight && dTree.GetHead(rightL[j_1 + 1]) == x)
{
g[i_5][j_1 + 1] = x;
}
if (j_1 < nRight && dTree.GetHead(x) == rightL[j_1 + 1])
{
g[i_5][j_1 + 1] = rightL[j_1 + 1];
}
if (i_5 < nLeft && dTree.GetHead(leftL[i_5 + 1]) == x)
{
g[i_5 + 1][j_1] = x;
}
if (i_5 < nLeft && dTree.GetHead(x) == leftL[i_5 + 1])
{
g[i_5 + 1][j_1] = leftL[i_5 + 1];
}
}
}
}
return(g[nLeft][nRight] != -1);
}
