// ************************************************************************* //
// OUTPUT FUNCTIONS
// ************************************************************************* //
// This method is equivalent to calling "outputAlignmentResults (null, null)" , but faster;
public Stats calculateOCRaccuracy() {
if (alignment.isEmpty()) {
System.out.println(
"RecursiveAlignmentTool::calculateOCRaccuracy(): There are no words aligned. Aligner must be run prior to calling calculateOCRaccuracy");
return null;
}
// align each segments characters individually
long numError = 0;
long total = 0;
long ocrLength = 0;
for (int kk = 0; kk < alignment.size(); kk++) {
AlignedSequence alignedTerm = alignment.get(kk);
String refout = alignedTerm.getReference();
String candidateout = alignedTerm.getCandidate();
if (candidateout != null) {
ocrLength++;
}
if (refout != null) {
total++;
if (candidateout == null || !refout.equals(candidateout)) {
numError++;
}
}
}
Stats st = new Stats((total - numError), ocrLength, total);
return st;
// double accuracy = 1.0 - (numError / (double) total);
// System.out.println("OCR accuracy: " + wordAccuracy + " (matching words= " + (total -
// numError) + " reference length= " + total + ")");
}
// Alignment results are written to outputfile
// if both arguments are null, then this method simply calculates OCR accuracy and outputs
// nothing.
public Stats outputAlignmentResults(String outputfile, String errorsFilename) {
long numError = 0;
long total = 0;
int inc = 1;
int lineWidth = 0;
StringBuffer candBuffer = new StringBuffer(10000);
StringBuffer refBuffer = new StringBuffer(10000);
boolean colFormat = OUTPUT_FORMAT.toLowerCase().startsWith("c");
// System.out.println(OUTPUT_FORMAT);
String NULL_STRING = "null";
if (WORD_LEVEL_ALIGNMENT) {
lineWidth = 20;
NULL_STRING = "null";
} else {
lineWidth = 100;
NULL_STRING = "@";
}
long counter = 0;
long ocrLen = 0;
try {
Writer writer = null;
Writer errorWriter = null;
if (outputfile != null) {
// writer = new BufferedWriter(new FileWriter(new File(outputfile)));
writer =
new BufferedWriter(
new OutputStreamWriter(new FileOutputStream(new File(outputfile)), "UTF8"));
}
if (errorsFilename != null) {
// errorWriter = new BufferedWriter(new FileWriter(new File(errorsFilename)));
errorWriter =
new BufferedWriter(
new OutputStreamWriter(new FileOutputStream(new File(errorsFilename)), "UTF8"));
}
for (int kk = 0; kk < alignment.size(); kk++) {
AlignedSequence alignedTerm = alignment.get(kk);
String refout = alignedTerm.getReference();
String candidateout = alignedTerm.getCandidate();
if (candidateout != null) {
ocrLen++;
}
String ref = refout == null ? NULL_STRING : refout;
String candidate = candidateout == null ? NULL_STRING : candidateout;
if (writer != null) {
if (colFormat) {
writer.append(candidate + "\t" + ref + "\n");
} else {
if (counter == lineWidth) {
writer.append("OCR:\t" + candBuffer + "\n");
writer.append("GT :\t" + refBuffer + "\n\n");
candBuffer.delete(0, candBuffer.length());
refBuffer.delete(0, refBuffer.length());
counter = 0;
// kk--;
}
if (!WORD_LEVEL_ALIGNMENT) {
candBuffer.append(candidate);
refBuffer.append(ref);
} else {
String tabbedCand = candidate, tabbedRef = ref;
if (candidate.length() > ref.length()) {
for (int t = 0; t < (candidate.length() - ref.length()); t++) {
tabbedRef += " ";
}
} else if (candidate.length() < ref.length()) {
for (int t = 0; t < (ref.length() - candidate.length()); t++) {
tabbedCand += " ";
}
}
candBuffer.append(tabbedCand).append("\t");
refBuffer.append(tabbedRef).append("\t");
}
counter++;
}
}
if (refout != null) {
total++;
if (candidateout == null || !refout.equals(candidateout)) {
numError++;
if (errorWriter != null) {
errorWriter.append(candidateout + "\t" + refout + "\t" + (kk + inc) + "\n");
}
}
}
}
if (!colFormat) {
writer.append("OCR:\t" + candBuffer + "\n");
writer.append("GT :\t" + refBuffer + "\n\n");
}
if (writer != null) {
writer.close();
}
if (errorWriter != null) {
errorWriter.close();
}
} catch (Exception e) {
System.out.println("Error. Can not write the file: " + outputfile);
}
Stats st = new Stats((total - numError), ocrLen, total);
return st;
// System.out.println("OCR accuracy: " + wordAccuracy + " (matching words= " + (total -
// numError) + " reference length= " + total + ")");
}
