/// <summary>Extracts info about one relation mention</summary>
private static AceEventMention ParseEventMention(INode node, AceDocument doc)
{
string id = GetAttributeValue(node, "ID");
AceCharSeq extent = ParseCharSeq(GetChildByName(node, "extent"));
AceCharSeq anchor = ParseCharSeq(GetChildByName(node, "anchor"));
// create the mention
AceEventMention mention = new AceEventMention(id, extent, anchor);
// find the mention args
IList <INode> args = GetChildrenByName(node, "event_mention_argument");
foreach (INode arg in args)
{
string role = GetAttributeValue(arg, "ROLE");
string refid = GetAttributeValue(arg, "REFID");
AceEntityMention am = doc.GetEntityMention(refid);
if (am != null)
{
am.AddEventMention(mention);
mention.AddArg(am, role);
}
}
return(mention);
}
public virtual void AddMention(AceEventMention m)
{
mMentions.Add(m);
m.SetParent(this);
}
/// <summary>Parses one ACE specification</summary>
/// <returns>Simply displays the events to stdout</returns>
/// <exception cref="System.IO.IOException"/>
/// <exception cref="Org.Xml.Sax.SAXException"/>
/// <exception cref="Javax.Xml.Parsers.ParserConfigurationException"/>
public static AceDocument ParseDocument(File f)
{
// parse the Dom document
IDocument document = ReadDocument(f);
//
// create the ACE document object
//
INode docElement = document.GetElementsByTagName("document").Item(0);
AceDocument aceDoc = new AceDocument(GetAttributeValue(docElement, "DOCID"));
//
// read all entities
//
INodeList entities = document.GetElementsByTagName("entity");
int entityCount = 0;
for (int i = 0; i < entities.GetLength(); i++)
{
INode node = entities.Item(i);
//
// the entity type and subtype
//
string id = GetAttributeValue(node, "ID");
string type = GetAttributeValue(node, "TYPE");
string subtype = GetAttributeValue(node, "SUBTYPE");
string cls = GetAttributeValue(node, "CLASS");
// create the entity
AceEntity entity = new AceEntity(id, type, subtype, cls);
aceDoc.AddEntity(entity);
// fetch all mentions of this event
IList <INode> mentions = GetChildrenByName(node, "entity_mention");
// parse all its mentions
foreach (INode mention1 in mentions)
{
AceEntityMention mention = ParseEntityMention(mention1);
entity.AddMention(mention);
aceDoc.AddEntityMention(mention);
}
entityCount++;
}
//log.info("Parsed " + entityCount + " XML entities.");
//
// read all relations
//
INodeList relations = document.GetElementsByTagName("relation");
for (int i_1 = 0; i_1 < relations.GetLength(); i_1++)
{
INode node = relations.Item(i_1);
//
// the relation type, subtype, tense, and modality
//
string id = GetAttributeValue(node, "ID");
string type = GetAttributeValue(node, "TYPE");
string subtype = GetAttributeValue(node, "SUBTYPE");
string modality = GetAttributeValue(node, "MODALITY");
string tense = GetAttributeValue(node, "TENSE");
// create the relation
AceRelation relation = new AceRelation(id, type, subtype, modality, tense);
aceDoc.AddRelation(relation);
// XXX: fetch relation_arguments here!
// fetch all mentions of this relation
IList <INode> mentions = GetChildrenByName(node, "relation_mention");
// traverse all mentions
foreach (INode mention1 in mentions)
{
AceRelationMention mention = ParseRelationMention(mention1, aceDoc);
relation.AddMention(mention);
aceDoc.AddRelationMention(mention);
}
}
//
// read all events
//
INodeList events = document.GetElementsByTagName("event");
for (int i_2 = 0; i_2 < events.GetLength(); i_2++)
{
INode node = events.Item(i_2);
//
// the event type, subtype, tense, and modality
//
string id = GetAttributeValue(node, "ID");
string type = GetAttributeValue(node, "TYPE");
string subtype = GetAttributeValue(node, "SUBTYPE");
string modality = GetAttributeValue(node, "MODALITY");
string polarity = GetAttributeValue(node, "POLARITY");
string genericity = GetAttributeValue(node, "GENERICITY");
string tense = GetAttributeValue(node, "TENSE");
// create the event
AceEvent @event = new AceEvent(id, type, subtype, modality, polarity, genericity, tense);
aceDoc.AddEvent(@event);
// fetch all mentions of this relation
IList <INode> mentions = GetChildrenByName(node, "event_mention");
// traverse all mentions
foreach (INode mention1 in mentions)
{
AceEventMention mention = ParseEventMention(mention1, aceDoc);
@event.AddMention(mention);
aceDoc.AddEventMention(mention);
}
}
return(aceDoc);
}
public virtual void AddEventMention(AceEventMention rm)
{
mEventMentions.Add(rm);
}
/// <summary>Parses an ACE document.</summary>
/// <remarks>
/// Parses an ACE document. Works in the following steps: (a) reads both the
/// XML annotations; (b) reads the tokens; (c) matches the tokens against the
/// annotations (d) constructs mSentenceEntityMentions and
/// mRelationEntityMentions
/// </remarks>
/// <exception cref="System.IO.IOException"/>
/// <exception cref="Org.Xml.Sax.SAXException"/>
/// <exception cref="Javax.Xml.Parsers.ParserConfigurationException"/>
public static Edu.Stanford.Nlp.IE.Machinereading.Domains.Ace.Reader.AceDocument ParseDocument(string prefix, bool usePredictedBoundaries)
{
mLog.Fine("Reading document " + prefix);
Edu.Stanford.Nlp.IE.Machinereading.Domains.Ace.Reader.AceDocument doc = null;
//
// read the ACE XML annotations
//
if (usePredictedBoundaries == false)
{
doc = AceDomReader.ParseDocument(new File(prefix + XmlExt));
}
else
{
// log.info("Parsed " + doc.getEntityMentions().size() +
// " entities in document " + prefix);
//
// will use the predicted entity boundaries (see below)
//
int lastSlash = prefix.LastIndexOf(File.separator);
System.Diagnostics.Debug.Assert((lastSlash > 0 && lastSlash < prefix.Length - 1));
string id = Sharpen.Runtime.Substring(prefix, lastSlash + 1);
// log.info(id + ": " + prefix);
doc = new Edu.Stanford.Nlp.IE.Machinereading.Domains.Ace.Reader.AceDocument(id);
}
doc.SetPrefix(prefix);
//
// read the raw byte stream
//
string trueCasedFileName = prefix + OrigExt + ".truecase";
if ((new File(trueCasedFileName).Exists()))
{
mLog.Severe("Using truecased file: " + trueCasedFileName);
doc.ReadRawBytes(trueCasedFileName);
}
else
{
doc.ReadRawBytes(prefix + OrigExt);
}
//
// read the AceTokens
//
int offsetToSubtract = 0;
IList <IList <AceToken> > sentences = AceSentenceSegmenter.TokenizeAndSegmentSentences(prefix);
doc.SetSentences(sentences);
foreach (IList <AceToken> sentence in sentences)
{
foreach (AceToken token in sentence)
{
offsetToSubtract = token.AdjustPhrasePositions(offsetToSubtract, token.GetLiteral());
doc.AddToken(token);
}
}
//
// match char sequences to phrases
//
doc.MatchCharSeqs(prefix);
//
// construct the mEntityMentions matrix
//
ICollection <string> entityKeys = doc.mEntityMentions.Keys;
int sentence_1;
foreach (string key in entityKeys)
{
AceEntityMention em = doc.mEntityMentions[key];
sentence_1 = doc.mTokens[em.GetHead().GetTokenStart()].GetSentence();
// adjust the number of rows if necessary
while (sentence_1 >= doc.mSentenceEntityMentions.Count)
{
doc.mSentenceEntityMentions.Add(new List <AceEntityMention>());
doc.mSentenceRelationMentions.Add(new List <AceRelationMention>());
doc.mSentenceEventMentions.Add(new List <AceEventMention>());
}
// store the entity mentions in increasing order:
// (a) of the start position of their head
// (b) if start is the same, in increasing order of the head end
List <AceEntityMention> sentEnts = doc.mSentenceEntityMentions[sentence_1];
bool added = false;
for (int i = 0; i < sentEnts.Count; i++)
{
AceEntityMention crt = sentEnts[i];
if ((crt.GetHead().GetTokenStart() > em.GetHead().GetTokenStart()) || (crt.GetHead().GetTokenStart() == em.GetHead().GetTokenStart() && crt.GetHead().GetTokenEnd() > em.GetHead().GetTokenEnd()))
{
sentEnts.Add(i, em);
added = true;
break;
}
}
if (!added)
{
sentEnts.Add(em);
}
}
//
// construct the mRelationMentions matrix
//
ICollection <string> relKeys = doc.mRelationMentions.Keys;
foreach (string key_1 in relKeys)
{
AceRelationMention rm = doc.mRelationMentions[key_1];
sentence_1 = doc.mTokens[rm.GetArg(0).GetHead().GetTokenStart()].GetSentence();
//
// no need to adjust the number of rows: was done above
//
// store the relation mentions in increasing order
// (a) of the start position of their head, or
// (b) if start is the same, in increasing order of ends
List <AceRelationMention> sentRels = doc.mSentenceRelationMentions[sentence_1];
bool added = false;
for (int i = 0; i < sentRels.Count; i++)
{
AceRelationMention crt = sentRels[i];
if ((crt.GetMinTokenStart() > rm.GetMinTokenStart()) || (crt.GetMinTokenStart() == rm.GetMinTokenStart() && crt.GetMaxTokenEnd() > rm.GetMaxTokenEnd()))
{
sentRels.Add(i, rm);
added = true;
break;
}
}
if (!added)
{
sentRels.Add(rm);
}
}
//
// construct the mEventMentions matrix
//
ICollection <string> eventKeys = doc.mEventMentions.Keys;
foreach (string key_2 in eventKeys)
{
AceEventMention em = doc.mEventMentions[key_2];
sentence_1 = doc.mTokens[em.GetMinTokenStart()].GetSentence();
/*
* adjust the number of rows if necessary -- if you're wondering why we do
* this here again, (after we've done it for entities) it's because we can
* have an event with no entities near the end of the document and thus
* won't have created rows in mSentence*Mentions
*/
while (sentence_1 >= doc.mSentenceEntityMentions.Count)
{
doc.mSentenceEntityMentions.Add(new List <AceEntityMention>());
doc.mSentenceRelationMentions.Add(new List <AceRelationMention>());
doc.mSentenceEventMentions.Add(new List <AceEventMention>());
}
// store the event mentions in increasing order
// (a) first, event mentions with no arguments
// (b) then by the start position of their head, or
// (c) if start is the same, in increasing order of ends
List <AceEventMention> sentEvents = doc.mSentenceEventMentions[sentence_1];
bool added = false;
for (int i = 0; i < sentEvents.Count; i++)
{
AceEventMention crt = sentEvents[i];
if ((crt.GetMinTokenStart() > em.GetMinTokenStart()) || (crt.GetMinTokenStart() == em.GetMinTokenStart() && crt.GetMaxTokenEnd() > em.GetMaxTokenEnd()))
{
sentEvents.Add(i, em);
added = true;
break;
}
}
if (!added)
{
sentEvents.Add(em);
}
}
return(doc);
}
/// <summary>Matches all relevant mentions, i.e.</summary>
/// <remarks>
/// Matches all relevant mentions, i.e. entities and anchors, to tokens Note:
/// entity mentions may match with multiple tokens!
/// </remarks>
public virtual void MatchCharSeqs(string filePrefix)
{
//
// match the head and extent of entity mentions
//
ICollection <string> keys = mEntityMentions.Keys;
foreach (string key in keys)
{
AceEntityMention m = mEntityMentions[key];
//
// match the head charseq to 1+ phrase(s)
//
try
{
m.GetHead().Match(mTokens);
}
catch (MatchException)
{
mLog.Severe("READER ERROR: Failed to match entity mention head: " + "[" + m.GetHead().GetText() + ", " + m.GetHead().GetByteStart() + ", " + m.GetHead().GetByteEnd() + "]");
mLog.Severe("Document tokens: " + TokensWithByteSpan(m.GetHead().GetByteStart(), m.GetHead().GetByteEnd()));
mLog.Severe("Document prefix: " + filePrefix);
System.Environment.Exit(1);
}
//
// match the extent charseq to 1+ phrase(s)
//
try
{
m.GetExtent().Match(mTokens);
}
catch (MatchException)
{
mLog.Severe("READER ERROR: Failed to match entity mention extent: " + "[" + m.GetExtent().GetText() + ", " + m.GetExtent().GetByteStart() + ", " + m.GetExtent().GetByteEnd() + "]");
mLog.Severe("Document tokens: " + TokensWithByteSpan(m.GetExtent().GetByteStart(), m.GetExtent().GetByteEnd()));
System.Environment.Exit(1);
}
//
// set the head word of the mention
//
m.DetectHeadToken(this);
}
// we need to do this for events as well since they may not have any AceEntityMentions associated with them (if they have no arguments)
ICollection <string> eventKeys = mEventMentions.Keys;
foreach (string key_1 in eventKeys)
{
AceEventMention m = mEventMentions[key_1];
//
// match the extent charseq to 1+ phrase(s)
//
try
{
m.GetExtent().Match(mTokens);
}
catch (MatchException)
{
mLog.Severe("READER ERROR: Failed to match event mention extent: " + "[" + m.GetExtent().GetText() + ", " + m.GetExtent().GetByteStart() + ", " + m.GetExtent().GetByteEnd() + "]");
mLog.Severe("Document tokens: " + TokensWithByteSpan(m.GetExtent().GetByteStart(), m.GetExtent().GetByteEnd()));
System.Environment.Exit(1);
}
}
}
public virtual void AddEventMention(AceEventMention e)
{
mEventMentions[e.GetId()] = e;
}