/**
* Add an array of objects reporting View's Constituents' surface form and character offsets. May
* make deserialization to TextAnnotation problematic, as the relevant methods deduce token
* character offsets directly from list of token strings and raw text.
*
* @param fieldName name to give to this field
* @param view view whose character offsets will be serialized
* @param json Json object to which resulting array will be added
*/
private static void writeTokenOffsets(String fieldName, View view, JsonObject json) {
JsonArray offsetArray = new JsonArray();
for (Constituent c : view.getConstituents()) {
JsonObject cJ = new JsonObject();
writeString(FORM, c.getSurfaceForm(), cJ);
writeInt(STARTCHAROFFSET, c.getStartCharOffset(), cJ);
writeInt(ENDCHAROFFSET, c.getEndCharOffset(), cJ);
offsetArray.add(cJ);
}
json.add(fieldName, offsetArray);
}
protected String getNERString() {
List<Constituent> constituents = new ArrayList<>(view.getConstituents());
Collections.sort(constituents, TextAnnotationUtilities.constituentStartComparator);
StringBuilder sb = new StringBuilder();
String text = textAnnotation.getText();
int where = 0;
for (Constituent c : constituents) {
int start = c.getStartCharOffset();
String startstring = text.substring(where, start);
sb.append(startstring)
.append("[")
.append(c.getLabel())
.append(" ")
.append(c.getTokenizedSurfaceForm())
.append(" ] ");
where = c.getEndCharOffset();
}
return sb.toString();
}
