private void getTbSentences_Click(object sender, EventArgs e)
{
ClearList();
string mrgFile = null;
try
{
mrgFile = mrgFileCombo.Text;
mrgFile = mrgFile.Substring(0, mrgFile.IndexOf(' '));
mrgFile = _treeBankEngine.GetFullMrgPath(mrgFile);
}
catch (Exception)
{
MessageBox.Show("Invalid MRG file");
return;
}
try
{
foreach (int sentNum in _treeBankEngine.GetSentenceNumbers(mrgFile))
{
AppendOutput(_treeBankEngine.GetParseTree(mrgFile, sentNum));
}
}
catch (Exception ex)
{
MessageBox.Show(ex.ToString());
}
}
/// <summary>
/// Marks argument nodes from the current node in the corresponding parse from a different TreeBank. This is used when
/// transferring PropBank annotations to parse trees other than those distributed in the TreeBank (e.g., those produced
/// by an automatic syntactic parser).
/// </summary>
/// <param name="treeBankEngine">Initialized TreeBank engine from which to pull the parse tree to mark PropBank arguments within</param>
/// <returns>PropBank node, or null if all arguments couldn't be minimally transferred to the other parse tree. An argument
/// is minimally transferred if the corresponding node in the other parse tree subsumes precisely the same text as the node in the
/// current parse tree. Sometimes this is not possible due to parse errors.</returns>
public PropBankNode MarkArgumentNodesIn(TreeBankEngine treeBankEngine)
{
if (!IsRoot)
{
throw new Exception("Attempted to transform non-root node");
}
// get mrg file in other tree bank
string treeBankMrgFile = treeBankEngine.GetFullMrgPath(MrgFile.Substring(MrgFile.LastIndexOf(Path.DirectorySeparatorChar) + 1));
// need a PropBank root to mark arguments within
PropBankNode pbRoot = new PropBankNode(treeBankEngine.GetParseTree(treeBankMrgFile, SentenceNumber));
// make sure we got the right sentence
if (pbRoot.SurfaceText != SurfaceText)
{
throw new Exception("Failed to convert root to Charniak-parsed version");
}
// Add information to root. Ignore leaf number and argument info for now - we'll set them at the end.
treeBankMrgFile = treeBankMrgFile.Substring(treeBankEngine.MrgPath.Length);
VerbInfo pbInfo = Information;
pbRoot.Information = new VerbInfo(pbInfo.Verb, treeBankMrgFile, pbInfo.SentenceNumber,
-1, pbInfo.Tagger, pbInfo.RoleSetId,
pbInfo.VForm, pbInfo.VTense, pbInfo.VAspect,
pbInfo.VPerson, pbInfo.VVoice, "");
// transfer all argument node lists
foreach (PropBankLabeledNodeCollection nodeCollection in LabeledNodeCollections)
{
// new node collection
PropBankLabeledNodeCollection otherNodeCollection = new PropBankLabeledNodeCollection(new PropBankNodeLabel(nodeCollection.Label.Type, nodeCollection.Label.Feature, nodeCollection.Label.Confidence));
// get single nodes
foreach (PropBankNode singleNode in nodeCollection.SingleNodes)
{
if (!singleNode.IsNullElement)
{
// get argument node from other parse tree
PropBankNode otherArgNode = (PropBankNode)pbRoot.GetMinimallySubsumingNode(singleNode.FirstToken, singleNode.LastToken);
if (otherArgNode == null)
{
return(null);
}
otherNodeCollection.AddSingleNode(otherArgNode);
}
}
// get split arguments
foreach (List <TreeBankNode> splitArg in nodeCollection.SplitNodes)
{
List <TreeBankNode> otherSplitArg = new List <TreeBankNode>();
// get each node in the split argument
foreach (PropBankNode splitArgNode in splitArg)
{
if (!splitArgNode.IsNullElement)
{
// get split node in other tree
PropBankNode otherSplitArgNode = (PropBankNode)pbRoot.GetMinimallySubsumingNode(splitArgNode.FirstToken, splitArgNode.LastToken);
if (otherSplitArgNode == null)
{
return(null);
}
otherSplitArg.Add(otherSplitArgNode);
}
}
// if only one node of the split arg was non-null, at that node as a single
if (otherSplitArg.Count == 1)
{
otherNodeCollection.AddSingleNode(otherSplitArg.First());
}
// otherwise, add the split arg normally
else if (otherSplitArg.Count >= 2)
{
otherNodeCollection.AddSplitNode(otherSplitArg);
}
}
// add coref list if we found non-null nodes
if (otherNodeCollection.SingleNodes.Count > 0 || otherNodeCollection.SplitNodes.Count > 0)
{
pbRoot.LabeledNodeCollections.Add(otherNodeCollection);
}
}
// return null if we didn't find any argument node lists with non-null nodes
if (pbRoot.LabeledNodeCollections.Count == 0)
{
return(null);
}
// set leaf number and argument information
pbRoot.Information.LeafNumber = pbRoot.PredicateNodes.First().Leaves[0].LeafNumber;
pbRoot.Information.LabeledNodeLocations = pbRoot.LabeledNodeLocations;
return(pbRoot);
}
/// <summary>
/// Marks argument nodes from the current node in the corresponding parse from a different TreeBank. This is used when
/// transferring NomBank annotations to parse trees other than those distributed in the TreeBank (e.g., those produced
/// by an automatic syntactic parser).
/// </summary>
/// <param name="treeBankEngine">Initialized TreeBank engine from which to pull the parse tree to mark NomBank arguments within</param>
/// <returns>NomBank node, or null if all arguments couldn't be minimally transferred to the other parse tree. An argument
/// is minimally transferred if the corresponding node in the other parse tree subsumes precisely the same text as the node
/// in the current parse tree. Sometimes this is not possible due to parse errors.</returns>
public NomBankNode MarkArgumentNodesIn(TreeBankEngine treeBankEngine)
{
// make sure we're marking arguments using a root node
if (!IsRoot)
{
throw new Exception("Must pass root node");
}
// get mrg file in other tree bank
string treeBankMrgFile = treeBankEngine.GetFullMrgPath(MrgFile);
// need a NomBank root to mark arguments within
NomBankNode nbRoot = new NomBankNode(treeBankEngine.GetParseTree(treeBankMrgFile, SentenceNumber));
// make sure we got the right sentence
if (nbRoot.SurfaceText != SurfaceText)
{
throw new Exception("Failed to get same parse tree");
}
// Add information to root. Ignore leaf number and argument information - we'll set them at the end.
treeBankMrgFile = treeBankMrgFile.Substring(treeBankEngine.MrgPath.Length).Trim(Path.DirectorySeparatorChar);
NounInfo currInfo = Information;
nbRoot.Information = new NounInfo(currInfo.Noun, treeBankMrgFile, currInfo.SentenceNumber, -1, currInfo.RoleSetId, "");
// transfer all argument node lists
foreach (NomBankLabeledNodeCollection corefList in LabeledNodeCollections)
{
// new node list
NomBankLabeledNodeCollection otherCorefList = new NomBankLabeledNodeCollection(corefList.Label.Copy());
// get single nodes
foreach (NomBankNode singleNode in corefList.SingleNodes)
{
if (!singleNode.IsNullElement)
{
// get argument node from other parse tree
NomBankNode otherArgNode = nbRoot.GetMinimallySubsumingNode(singleNode.FirstToken, singleNode.LastToken) as NomBankNode;
if (otherArgNode == null)
{
return(null);
}
otherCorefList.AddSingleNode(otherArgNode);
}
}
// get split arguments
foreach (List <TreeBankNode> splitNode in corefList.SplitNodes)
{
List <TreeBankNode> otherSplitArg = new List <TreeBankNode>();
// get each node in the split argument
foreach (NomBankNode node in splitNode)
{
if (!node.IsNullElement)
{
// get split node in other tree
NomBankNode otherSplitArgNode = nbRoot.GetMinimallySubsumingNode(node.FirstToken, node.LastToken) as NomBankNode;
if (otherSplitArgNode == null)
{
return(null);
}
otherSplitArg.Add(otherSplitArgNode);
}
}
// if only one node of the split arg was non-null, at that node as a single
if (otherSplitArg.Count == 1)
{
otherCorefList.AddSingleNode(otherSplitArg.First());
}
// otherwise, add the split arg normally
else if (otherSplitArg.Count >= 2)
{
otherCorefList.AddSplitNode(otherSplitArg);
}
}
// make sure all hyphen indexes were applied
if (otherCorefList.Label.HyphenIndexes.Count != otherCorefList.AppliedIndexes.Count)
{
throw new Exception("Not all hyphen indexes were applied");
}
// add coref list if we found non-null nodes
if (otherCorefList.SingleNodes.Count > 0 || otherCorefList.SplitNodes.Count > 0)
{
nbRoot.LabeledNodeCollections.Add(otherCorefList);
}
}
// return null if we didn't find any argument node lists with non-null nodes
if (nbRoot.LabeledNodeCollections.Count == 0)
{
return(null);
}
// set leaf number and argument locations in the information object
nbRoot.Information.LeafNumber = nbRoot.PredicateNode.Leaves[0].LeafNumber;
nbRoot.Information.LabeledNodeLocations = nbRoot.LabeledNodeLocations;
return(nbRoot);
}