stanfordnlp/CoreNLP · error · IllegalArgumentException

You mixed CoreLabels with " + hw.getClass() + "? Why would…

Error message

You mixed CoreLabels with " + hw.getClass() + "?  Why would you do that?

What it means

In the XML output path with lemmas enabled, every token must be a CoreLabel so the lemma annotation can be read; if some tokens are CoreLabels and others are not (or none are), the tagger throws IllegalArgumentException 'You mixed CoreLabels with <class>? Why would you do that?'. It enforces homogeneous CoreLabel tokens when lemmas are requested.

Solutions

Make every token in the sentence a CoreLabel when lemmas are enabled in output.
Set lemmas on each CoreLabel (label.setLemma(...)) before output if precomputed.
Disable lemma output in the output format configuration if you only have plain words.
Validate homogeneity: all tokens instanceof CoreLabel before the call.

Example fix

// before
List<HasWord> sent = new ArrayList<>();
sent.add(new CoreLabel());           // word 1 is CoreLabel
sent.add(new Word("dog"));           // word 2 is not -> throws
// after
List<HasWord> sent = new ArrayList<>();
CoreLabel c1 = new CoreLabel(); c1.setWord("The");
CoreLabel c2 = new CoreLabel(); c2.setWord("dog");
sent.add(c1); sent.add(c2);

Defensive patterns

Strategy: type-guard

Validate before calling

boolean homogeneousCoreLabels(java.util.List<? extends edu.stanford.nlp.ling.HasWord> sent) {
  return sent.stream().allMatch(w -> w instanceof edu.stanford.nlp.ling.CoreLabel);
}

Type guard

if (outputLemmas && sent.stream().anyMatch(w -> !(w instanceof CoreLabel)))
  throw new IllegalArgumentException("Lemma output requires all-CoreLabel tokens");

Try / catch

try {
  tagger.outputTaggedSentence(sent, ...);
} catch (IllegalArgumentException e) {
  log.error("Mixed token types with lemma output: {}", e.getMessage());
}

Prevention

When lemmas are enabled, enforce CoreLabel-only token lists
Set lemma annotations before output if precomputed
Disable lemma output when using plain word tokens

When it happens

Trigger: XML output with outputLemmas=true (hasCoreLabels true) and a sentence where any element is not a CoreLabel; also the verbose TSV path at line 1407 where the first token is a CoreLabel but later tokens are not.

Common situations: Mixing token factories in one sentence (first token CoreLabel, rest Word); enabling lemmatization options in output config while supplying custom token types; hand-assembled sentences after pipeline tokens were converted to plain Words.

Understand the failure class

Background: Type mismatch errors: IllegalArgumentException, TypeError and type guards across 150 open-source libraries — this error's family across 150 libraries.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/302ac2f094e1515a. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/tagger/maxent/MaxentTagger.java:1383

                                    int sentNum, boolean outputLemmas) {
    if (sentence == null) {
      return "";
    }
    boolean hasCoreLabels = sentence.size() > 0 && sentence.get(0) instanceof CoreLabel;
    StringBuilder sb = new StringBuilder();
    sb.append("<sentence id=\"").append(sentNum).append("\">\n");
    int wordIndex = 0;
    for (HasWord hw : sentence) {
      String word = hw.word();
      if ( ! (hw instanceof HasTag)) {
        throw new IllegalArgumentException("Expected HasTags, got " +
                                           hw.getClass());
      }
      String tag = ((HasTag) hw).tag();
      sb.append("  <word wid=\"").append(wordIndex).append("\" pos=\"").append(XMLUtils.escapeAttributeXML(tag)).append("\"");
      if (outputLemmas && hasCoreLabels) {
        if ( ! (hw instanceof CoreLabel)) {
          throw new IllegalArgumentException("You mixed CoreLabels with " +
                                             hw.getClass() + "?  " +
                                             "Why would you do that?");
        }
        CoreLabel label = (CoreLabel) hw;
        String lemma = label.lemma();
        if (lemma != null) {
          sb.append(" lemma=\"").append(XMLUtils.escapeAttributeXML(lemma)).append('\"');
        }
      }
      sb.append(">").append(XMLUtils.escapeElementXML(word)).append("</word>\n");
      ++wordIndex;
    }
    sb.append("</sentence>\n");
    return sb.toString();
  }

  private static String getTsvWords(boolean verbose, boolean outputLemmas,
                                    List<? extends HasWord> sentence) {

View on GitHub (pinned to 1b7edd19c4)