/**
* Use a {@link SharedDefaultGazetteer} to duplicate this gazetteer by sharing the internal FSM
* rather than re-loading the lists.
*/
@Override
public Resource duplicate(Factory.DuplicationContext ctx) throws ResourceInstantiationException {
return Factory.createResource(
SharedDefaultGazetteer.class.getName(),
Utils.featureMap(SharedDefaultGazetteer.SDEF_GAZ_BOOTSTRAP_GAZETTEER_PROPERTY_NAME, this),
Factory.duplicate(this.getFeatures(), ctx),
this.getName());
}
protected void doExecute(Document theDocument) throws ExecutionException {
interrupted = false;
if (theDocument == null) {
throw new ExecutionException("No document to process!");
}
AnnotationSet outputAS = theDocument.getAnnotations(getOutputAnnotationSet());
if (containingType == null || containingType.isEmpty()) {
annotateText(document, outputAS, 0, document.getContent().size());
} else {
AnnotationSet inputAS = null;
if (inputASName == null || inputASName.isEmpty()) {
inputAS = theDocument.getAnnotations();
} else {
inputAS = theDocument.getAnnotations(inputASName);
}
AnnotationSet containingAnns = inputAS.get(containingType);
for (Annotation containingAnn : containingAnns) {
annotateText(
document, outputAS, gate.Utils.start(containingAnn), gate.Utils.end(containingAnn));
}
}
}
private static Document readDocument(String gateDocumentString)
throws ResourceInstantiationException {
Document gateDocument =
(Document)
Factory.createResource(
"gate.corpora.DocumentImpl",
Utils.featureMap(
"stringContent",
gateDocumentString,
"mimeType",
"text/xml",
"encoding",
"UTF-8"));
return gateDocument;
}
@Override
public void execute() throws ExecutionException {
Document doc = getDocument();
AnnotationSet as = doc.getAnnotations(getAnnotationSetName());
AnnotationSet tocs = as.get(getTokenAnnotationTypeName());
try {
for (Annotation t : tocs) {
String content = Utils.stringFor(doc, t);
String val = getOrthographyValue(content);
if (val != null) t.getFeatures().put("orth", val);
}
} catch (Exception e) {
throw new ExecutionException(e);
}
}
/**
* Method is executed after the init() method has finished its execution. <br>
*
* @throws ExecutionException
*/
public void execute() throws ExecutionException {
// lets start the progress and initialize the progress counter
fireProgressChanged(0);
// If no document provided to process throw an exception
if (document == null) {
fireProcessFinished();
throw new GateRuntimeException("No document to process!");
}
// langugage ID feature Name
if (languageIdFeatureName == null || languageIdFeatureName.trim().length() == 0)
languageIdFeatureName = "lang";
/* Default behaviour: classify the text of the whole document and
* store the result as a document feature. */
if ((annotationType == null) || (annotationType.length() == 0)) {
String docText = document.getContent().toString();
Classification classification = classifier.classify(docText);
document.getFeatures().put(languageIdFeatureName, classification.bestCategory());
}
/* Optional behaviour: classify the text underlying each annotation
* and store each results as an annotation feature. */
else {
AnnotationSet annotations = document.getAnnotations(annotationSetName).get(annotationType);
for (Annotation annotation : annotations) {
String text = Utils.stringFor(document, annotation);
Classification classification = classifier.classify(text);
annotation.getFeatures().put(languageIdFeatureName, classification.bestCategory());
}
}
// process finished, acknowledge user about this.
fireProcessFinished();
}
// carry out the actual annotations on the given span of text in the
// document.
protected void annotateText(Document doc, AnnotationSet outputAS, long from, long to) {
String text = "";
try {
text = doc.getContent().getContent(from, to).toString();
} catch (InvalidOffsetException ex) {
throw new GateRuntimeException("Unexpected offset exception, offsets are " + from + "/" + to);
}
// send the text to the service and get back the response
// System.out.println("Annotating text: "+text);
// System.out.println("Starting offset is "+from);
// NOTE: there is a bug in the TagMe service which causes offset errors
// if we use the tweet mode and there are certain patterns in the tweet.
// The approach recommended by Francesco Piccinno is to replace those
// patterns by spaces.
if (getIsTweet()) {
logger.debug("Text before cleaning: >>" + text + "<<");
// replace
text = text.replaceAll(patternStringRT3, " ");
text = text.replaceAll(patternStringRT2, " ");
text = text.replaceAll(patternHashTag, " $1");
// now replace the remaining patterns by spaces
StringBuilder sb = new StringBuilder(text);
Matcher m = patternUrl.matcher(text);
while (m.find()) {
int start = m.start();
int end = m.end();
sb.replace(start, end, nSpaces(end - start));
}
m = patternUser.matcher(text);
while (m.find()) {
int start = m.start();
int end = m.end();
sb.replace(start, end, nSpaces(end - start));
}
text = sb.toString();
logger.debug("Text after cleaning: >>" + text + "<<");
}
TagMeAnnotation[] tagmeAnnotations = getTagMeAnnotations(text);
for (TagMeAnnotation tagmeAnn : tagmeAnnotations) {
if (tagmeAnn.rho >= minrho) {
FeatureMap fm = Factory.newFeatureMap();
fm.put("tagMeId", tagmeAnn.id);
fm.put("title", tagmeAnn.title);
fm.put("rho", tagmeAnn.rho);
fm.put("spot", tagmeAnn.spot);
fm.put("link_probability", tagmeAnn.link_probability);
if (tagmeAnn.title == null) {
throw new GateRuntimeException("Odd: got a null title from the TagMe service" + tagmeAnn);
} else {
fm.put("inst", "http://dbpedia.org/resource/" + recodeForDbp38(tagmeAnn.title));
}
try {
gate.Utils.addAnn(
outputAS, from + tagmeAnn.start, from + tagmeAnn.end, getOutputAnnotationType(), fm);
} catch (Exception ex) {
System.err.println(
"Got an exception in document " + doc.getName() + ": " + ex.getLocalizedMessage());
ex.printStackTrace(System.err);
System.err.println("from=" + from + ", to=" + to + " TagMeAnn=" + tagmeAnn);
}
}
}
}
