public virtual string SegmentString(string line)
{
IList <CoreLabel> labeledSequence = SegmentStringToIOB(line);
string segmentedString = IOBUtils.IOBToString(labeledSequence, prefixMarker, suffixMarker);
return(segmentedString);
}
private IList <CoreLabel> SegmentStringToIOB(string line)
{
IList <CoreLabel> tokenList;
if (tf == null)
{
// Whitespace tokenization.
tokenList = IOBUtils.StringToIOB(line);
}
else
{
IList <CoreLabel> tokens = tf.GetTokenizer(new StringReader(line)).Tokenize();
tokenList = IOBUtils.StringToIOB(tokens, null, false, tf, line);
}
IOBUtils.LabelDomain(tokenList, domain);
tokenList = classifier.Classify(tokenList);
return(tokenList);
}
public virtual IList <CoreLabel> SegmentStringToTokenList(string line)
{
IList <CoreLabel> tokenList = CollectionUtils.MakeList();
IList <CoreLabel> labeledSequence = SegmentStringToIOB(line);
foreach (IntPair span in IOBUtils.TokenSpansForIOB(labeledSequence))
{
CoreLabel token = new CoreLabel();
string text = IOBUtils.IOBToString(labeledSequence, prefixMarker, suffixMarker, span.GetSource(), span.GetTarget());
token.SetWord(text);
token.SetValue(text);
token.Set(typeof(CoreAnnotations.TextAnnotation), text);
token.Set(typeof(CoreAnnotations.ArabicSegAnnotation), "1");
int start = labeledSequence[span.GetSource()].BeginPosition();
int end = labeledSequence[span.GetTarget() - 1].EndPosition();
token.SetOriginalText(Sharpen.Runtime.Substring(line, start, end));
token.Set(typeof(CoreAnnotations.CharacterOffsetBeginAnnotation), start);
token.Set(typeof(CoreAnnotations.CharacterOffsetEndAnnotation), end);
tokenList.Add(token);
}
return(tokenList);
}
/// <summary>
/// Evaluate accuracy when the input is gold segmented text *with* segmentation
/// markers and morphological analyses.
/// </summary>
/// <remarks>
/// Evaluate accuracy when the input is gold segmented text *with* segmentation
/// markers and morphological analyses. In other words, the evaluation file has the
/// same format as the training data.
/// </remarks>
/// <param name="pwOut"/>
private void Evaluate(PrintWriter pwOut)
{
log.Info("Starting evaluation...");
bool hasSegmentationMarkers = true;
bool hasTags = true;
IDocumentReaderAndWriter <CoreLabel> docReader = new ArabicDocumentReaderAndWriter(hasSegmentationMarkers, hasTags, hasDomainLabels, domain, tf);
ObjectBank <IList <CoreLabel> > lines = classifier.MakeObjectBankFromFile(flags.testFile, docReader);
PrintWriter tedEvalGoldTree = null;
PrintWriter tedEvalParseTree = null;
PrintWriter tedEvalGoldSeg = null;
PrintWriter tedEvalParseSeg = null;
if (tedEvalPrefix != null)
{
try
{
tedEvalGoldTree = new PrintWriter(tedEvalPrefix + "_gold.ftree");
tedEvalGoldSeg = new PrintWriter(tedEvalPrefix + "_gold.segmentation");
tedEvalParseTree = new PrintWriter(tedEvalPrefix + "_parse.ftree");
tedEvalParseSeg = new PrintWriter(tedEvalPrefix + "_parse.segmentation");
}
catch (FileNotFoundException e)
{
System.Console.Error.Printf("%s: %s%n", typeof(Edu.Stanford.Nlp.International.Arabic.Process.ArabicSegmenter).FullName, e.Message);
}
}
ICounter <string> labelTotal = new ClassicCounter <string>();
ICounter <string> labelCorrect = new ClassicCounter <string>();
int total = 0;
int correct = 0;
foreach (IList <CoreLabel> line in lines)
{
string[] inputTokens = TedEvalSanitize(IOBUtils.IOBToString(line).ReplaceAll(":", "#pm#")).Split(" ");
string[] goldTokens = TedEvalSanitize(IOBUtils.IOBToString(line, ":")).Split(" ");
line = classifier.Classify(line);
string[] parseTokens = TedEvalSanitize(IOBUtils.IOBToString(line, ":")).Split(" ");
foreach (CoreLabel label in line)
{
// Do not evaluate labeling of whitespace
string observation = label.Get(typeof(CoreAnnotations.CharAnnotation));
if (!observation.Equals(IOBUtils.GetBoundaryCharacter()))
{
total++;
string hypothesis = label.Get(typeof(CoreAnnotations.AnswerAnnotation));
string reference = label.Get(typeof(CoreAnnotations.GoldAnswerAnnotation));
labelTotal.IncrementCount(reference);
if (hypothesis.Equals(reference))
{
correct++;
labelCorrect.IncrementCount(reference);
}
}
}
if (tedEvalParseSeg != null)
{
tedEvalGoldTree.Printf("(root");
tedEvalParseTree.Printf("(root");
int safeLength = inputTokens.Length;
if (inputTokens.Length != goldTokens.Length)
{
log.Info("In generating TEDEval files: Input and gold do not have the same number of tokens");
log.Info(" (ignoring any extras)");
log.Info(" input: " + Arrays.ToString(inputTokens));
log.Info(" gold: " + Arrays.ToString(goldTokens));
safeLength = Math.Min(inputTokens.Length, goldTokens.Length);
}
if (inputTokens.Length != parseTokens.Length)
{
log.Info("In generating TEDEval files: Input and parse do not have the same number of tokens");
log.Info(" (ignoring any extras)");
log.Info(" input: " + Arrays.ToString(inputTokens));
log.Info(" parse: " + Arrays.ToString(parseTokens));
safeLength = Math.Min(inputTokens.Length, parseTokens.Length);
}
for (int i = 0; i < safeLength; i++)
{
foreach (string segment in goldTokens[i].Split(":"))
{
tedEvalGoldTree.Printf(" (seg %s)", segment);
}
tedEvalGoldSeg.Printf("%s\t%s%n", inputTokens[i], goldTokens[i]);
foreach (string segment_1 in parseTokens[i].Split(":"))
{
tedEvalParseTree.Printf(" (seg %s)", segment_1);
}
tedEvalParseSeg.Printf("%s\t%s%n", inputTokens[i], parseTokens[i]);
}
tedEvalGoldTree.Printf(")%n");
tedEvalGoldSeg.Println();
tedEvalParseTree.Printf(")%n");
tedEvalParseSeg.Println();
}
}
double accuracy = ((double)correct) / ((double)total);
accuracy *= 100.0;
pwOut.Println("EVALUATION RESULTS");
pwOut.Printf("#datums:\t%d%n", total);
pwOut.Printf("#correct:\t%d%n", correct);
pwOut.Printf("accuracy:\t%.2f%n", accuracy);
pwOut.Println("==================");
// Output the per label accuracies
pwOut.Println("PER LABEL ACCURACIES");
foreach (string refLabel in labelTotal.KeySet())
{
double nTotal = labelTotal.GetCount(refLabel);
double nCorrect = labelCorrect.GetCount(refLabel);
double acc = (nCorrect / nTotal) * 100.0;
pwOut.Printf(" %s\t%.2f%n", refLabel, acc);
}
if (tedEvalParseSeg != null)
{
tedEvalGoldTree.Close();
tedEvalGoldSeg.Close();
tedEvalParseTree.Close();
tedEvalParseSeg.Close();
}
}
public IList <CoreLabel> Apply(string @in)
{
IList <CoreLabel> tokenList;
string lineDomain = string.Empty;
if (this._enclosing.inputHasDomainLabels)
{
string[] domainAndData = @in.Split("\\s+", 2);
if (domainAndData.Length < 2)
{
ArabicDocumentReaderAndWriter.log.Info("Missing domain label or text: ");
ArabicDocumentReaderAndWriter.log.Info(@in);
}
else
{
lineDomain = domainAndData[0];
@in = domainAndData[1];
}
}
else
{
lineDomain = this._enclosing.inputDomain;
}
if (this._enclosing.inputHasTags)
{
string[] toks = @in.Split("\\s+");
IList <CoreLabel> input = new List <CoreLabel>(toks.Length);
string tagDelim = Pattern.Quote(ArabicDocumentReaderAndWriter.tagDelimiter);
string rewDelim = Pattern.Quote(ArabicDocumentReaderAndWriter.rewriteDelimiter);
foreach (string wordTag in toks)
{
string[] wordTagPair = wordTag.Split(tagDelim);
System.Diagnostics.Debug.Assert(wordTagPair.Length == 2);
string[] rewritePair = wordTagPair[0].Split(rewDelim);
System.Diagnostics.Debug.Assert(rewritePair.Length == 1 || rewritePair.Length == 2);
string raw = rewritePair[0];
string rewritten = raw;
if (rewritePair.Length == 2)
{
rewritten = rewritePair[1];
}
CoreLabel cl = new CoreLabel();
if (this._enclosing.tf != null)
{
IList <CoreLabel> lexListRaw = this._enclosing.tf.GetTokenizer(new StringReader(raw)).Tokenize();
IList <CoreLabel> lexListRewritten = this._enclosing.tf.GetTokenizer(new StringReader(rewritten)).Tokenize();
if (lexListRewritten.Count != lexListRaw.Count)
{
System.Console.Error.Printf("%s: Different number of tokens in raw and rewritten: %s>>>%s%n", this.GetType().FullName, raw, rewritten);
lexListRewritten = lexListRaw;
}
if (lexListRaw.IsEmpty())
{
continue;
}
else
{
if (lexListRaw.Count == 1)
{
raw = lexListRaw[0].Value();
rewritten = lexListRewritten[0].Value();
}
else
{
if (lexListRaw.Count > 1)
{
string secondWord = lexListRaw[1].Value();
if (secondWord.Equals(this._enclosing.segMarker.ToString()))
{
// Special case for the null marker in the vocalized section
raw = lexListRaw[0].Value() + this._enclosing.segMarker;
rewritten = lexListRewritten[0].Value() + this._enclosing.segMarker;
}
else
{
System.Console.Error.Printf("%s: Raw token generates multiple segments: %s%n", this.GetType().FullName, raw);
raw = lexListRaw[0].Value();
rewritten = lexListRewritten[0].Value();
}
}
}
}
}
cl.SetValue(raw);
cl.SetWord(raw);
cl.SetTag(wordTagPair[1]);
cl.Set(typeof(CoreAnnotations.DomainAnnotation), lineDomain);
cl.Set(typeof(ArabicDocumentReaderAndWriter.RewrittenArabicAnnotation), rewritten);
input.Add(cl);
}
tokenList = IOBUtils.StringToIOB(input, this._enclosing.segMarker, true, this._enclosing.shouldStripRewrites);
}
else
{
if (this._enclosing.tf == null)
{
tokenList = IOBUtils.StringToIOB(@in, this._enclosing.segMarker);
}
else
{
IList <CoreLabel> line = this._enclosing.tf.GetTokenizer(new StringReader(@in)).Tokenize();
tokenList = IOBUtils.StringToIOB(line, this._enclosing.segMarker, false);
}
}
if (this._enclosing.inputHasDomainLabels && !this._enclosing.inputHasTags)
{
IOBUtils.LabelDomain(tokenList, lineDomain);
}
else
{
if (!this._enclosing.inputHasDomainLabels)
{
IOBUtils.LabelDomain(tokenList, this._enclosing.inputDomain);
}
}
return(tokenList);
}