使用Sentence Detector评估工具时,OpenNLP如何计算漏报?

时间:2015-01-22 06:22:19

标签: java nlp information-retrieval opennlp

我找到了OpenNLP的部分代码,它向我展示了如何计算真正的正值。我想知道在使用OpenNLP句子评估工具时如何计算假阴性值。换句话说,使用精度,召回和F测量在评估工具方面意味着什么?用例子解释会很精彩。 (对不起我的愚蠢问题,我对NLP来说是全新的......)

public static int countTruePositives(final Object[] references, final Object[] predictions) {



List<Object> predListSpans = new ArrayList<Object>(predictions.length);
Collections.addAll(predListSpans, predictions);
int truePositives = 0;
Object matchedItem = null;

for (int referenceIndex = 0; referenceIndex < references.length; referenceIndex++) {
  Object referenceName = references[referenceIndex];

  for (int predIndex = 0; predIndex < predListSpans.size(); predIndex++) {

    if (referenceName.equals(predListSpans.get(predIndex))) {
      matchedItem = predListSpans.get(predIndex);
      truePositives++;
    }
  }
  if (matchedItem != null) {
    predListSpans.remove(matchedItem);
  }
}
return truePositives;

1 个答案:

答案 0 :(得分:0)