/// <summary>
/// Reads in a single ACE*.apf.xml file and convert it to RelationSentence
/// objects.
/// </summary>
/// <remarks>
/// Reads in a single ACE*.apf.xml file and convert it to RelationSentence
/// objects. However, you probably should call parse() instead.
/// </remarks>
/// <param name="prefix">
/// prefix of ACE filename to read (e.g.
/// "/u/mcclosky/scr/data/ACE2005/english_test/bc/CNN_CF_20030827.1630.01"
/// ) (no ".apf.xml" extension)
/// </param>
/// <returns>list of RelationSentence objects</returns>
/// <exception cref="System.IO.IOException"/>
/// <exception cref="Org.Xml.Sax.SAXException"/>
/// <exception cref="Javax.Xml.Parsers.ParserConfigurationException"/>
private IList <ICoreMap> ReadDocument(string prefix, Annotation corpus)
{
logger.Info("Reading document: " + prefix);
IList <ICoreMap> results = new List <ICoreMap>();
AceDocument aceDocument;
if (aceVersion.Equals("ACE2004"))
{
aceDocument = AceDocument.ParseDocument(prefix, false, aceVersion);
}
else
{
aceDocument = AceDocument.ParseDocument(prefix, false);
}
string docId = aceDocument.GetId();
// map entity mention ID strings to their EntityMention counterparts
IDictionary <string, EntityMention> entityMentionMap = Generics.NewHashMap();
/*
* for (int sentenceIndex = 0; sentenceIndex < aceDocument.getSentenceCount(); sentenceIndex++) {
* List<AceToken> tokens = aceDocument.getSentence(sentenceIndex);
* StringBuffer b = new StringBuffer();
* for(AceToken t: tokens) b.append(t.getLiteral() + " " );
* logger.info("SENTENCE: " + b.toString());
* }
*/
int tokenOffset = 0;
for (int sentenceIndex = 0; sentenceIndex < aceDocument.GetSentenceCount(); sentenceIndex++)
{
IList <AceToken> tokens = aceDocument.GetSentence(sentenceIndex);
IList <CoreLabel> words = new List <CoreLabel>();
StringBuilder textContent = new StringBuilder();
for (int i = 0; i < tokens.Count; i++)
{
CoreLabel l = new CoreLabel();
l.SetWord(tokens[i].GetLiteral());
l.Set(typeof(CoreAnnotations.ValueAnnotation), l.Word());
l.Set(typeof(CoreAnnotations.CharacterOffsetBeginAnnotation), tokens[i].GetByteStart());
l.Set(typeof(CoreAnnotations.CharacterOffsetEndAnnotation), tokens[i].GetByteEnd());
words.Add(l);
if (i > 0)
{
textContent.Append(" ");
}
textContent.Append(tokens[i].GetLiteral());
}
// skip "sentences" that are really just SGML tags (which come from using the RobustTokenizer)
if (words.Count == 1)
{
string word = words[0].Word();
if (word.StartsWith("<") && word.EndsWith(">"))
{
tokenOffset += tokens.Count;
continue;
}
}
ICoreMap sentence = new Annotation(textContent.ToString());
sentence.Set(typeof(CoreAnnotations.DocIDAnnotation), docId);
sentence.Set(typeof(CoreAnnotations.TokensAnnotation), words);
logger.Info("Reading sentence: \"" + textContent + "\"");
IList <AceEntityMention> entityMentions = aceDocument.GetEntityMentions(sentenceIndex);
IList <AceRelationMention> relationMentions = aceDocument.GetRelationMentions(sentenceIndex);
IList <AceEventMention> eventMentions = aceDocument.GetEventMentions(sentenceIndex);
// convert entity mentions
foreach (AceEntityMention aceEntityMention in entityMentions)
{
string corefID = string.Empty;
foreach (string entityID in aceDocument.GetKeySetEntities())
{
AceEntity e = aceDocument.GetEntity(entityID);
if (e.GetMentions().Contains(aceEntityMention))
{
corefID = entityID;
break;
}
}
EntityMention convertedMention = ConvertAceEntityMention(aceEntityMention, docId, sentence, tokenOffset, corefID);
// EntityMention convertedMention = convertAceEntityMention(aceEntityMention, docId, sentence, tokenOffset);
entityCounts.IncrementCount(convertedMention.GetType());
logger.Info("CONVERTED MENTION HEAD SPAN: " + convertedMention.GetHead());
logger.Info("CONVERTED ENTITY MENTION: " + convertedMention);
AnnotationUtils.AddEntityMention(sentence, convertedMention);
entityMentionMap[aceEntityMention.GetId()] = convertedMention;
}
// TODO: make Entity objects as needed
// convert relation mentions
foreach (AceRelationMention aceRelationMention in relationMentions)
{
RelationMention convertedMention = ConvertAceRelationMention(aceRelationMention, docId, sentence, entityMentionMap);
if (convertedMention != null)
{
relationCounts.IncrementCount(convertedMention.GetType());
logger.Info("CONVERTED RELATION MENTION: " + convertedMention);
AnnotationUtils.AddRelationMention(sentence, convertedMention);
}
}
// TODO: make Relation objects
// convert EventMentions
foreach (AceEventMention aceEventMention in eventMentions)
{
EventMention convertedMention = ConvertAceEventMention(aceEventMention, docId, sentence, entityMentionMap, tokenOffset);
if (convertedMention != null)
{
eventCounts.IncrementCount(convertedMention.GetType());
logger.Info("CONVERTED EVENT MENTION: " + convertedMention);
AnnotationUtils.AddEventMention(sentence, convertedMention);
}
}
// TODO: make Event objects
results.Add(sentence);
tokenOffset += tokens.Count;
}
return(results);
}