stanfordnlp/CoreNLP · error

: Word contains malformed morph annotation

Error message

%s: Word contains malformed morph annotation: %s

What it means

ArabicTreeNormalizer.normalizeWholeTree strips morphological analyses appended to leaf values with the MORPHO_MARK separator. If a leaf splits into other than exactly two parts, the tree value is malformed and the normalizer logs this error and leaves the leaf unchanged rather than crashing.

Solutions

1. Inspect the offending leaf value printed in the message and remove duplicate/stray morph mark characters
2. Ensure each annotated leaf has the form word + morphoMark + morphFeatures with no extra marks
3. Re-export or re-clean the treebank file; re-run annotation from the raw source
4. If intentional multi-part values are needed, pre-split them before normalization

Example fix

// before (leaf value)
"word	NAUF	def	nom"
// after
"word	NAUF def nom"  // exactly one MORPHO_MARK separating word and features

String v = leaf.value();
if (v != null && v.contains(MorphoFeatureSpecification.MORPHO_MARK)) {
  String[] toks = v.split(MorphoFeatureSpecification.MORPHO_MARK);
  if (toks.length != 2) throw new IllegalArgumentException("bad morph annotation: " + v);
}

try {
  tree = normalizer.normalizeWholeTree(tree, treeFactory);
} catch (Exception e) {
  log.warn("skipping malformed tree: " + e.getMessage());
}

When it happens

Trigger: Normalizing an Arabic tree whose leaf value contains the morpho mark character but not exactly one occurrence separating word and morph features (e.g. 'word morphA morphB' or a stray mark with nothing after it).

Common situations: Preprocessed Arabic treebanks where the separator character appears inside the word itself; hand-edited trees; double-morph-annotation from running a pipeline twice.

Understand the failure class

Background: "Invalid ... format", "must be in format X", "does not look like a ..." — invalid argument format errors across CLI tools and libraries — this error's family across 17 libraries.

• Parsing and encoding errors: unexpected token, malformed input — why parsers reject input and how to find the real culprit.

Related errors

• Arabic does not support feature type: " + feat.toString()
• attempt to get word when sentence and lattice are null!
• Bad number put into wordToNumber. Word is: \"" + input +…
• Bad number put into wordToNumber. Word is: \"" + curPart +…
• Can't return head of null or leaf Tree.

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/cc705d4300084d76. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/trees/international/arabic/ArabicTreeNormalizer.java:133

      normalizedString = super.normalizeNonterminal(category);
    }

    return normalizedString.intern();
  }

  @Override
  public Tree normalizeWholeTree(Tree tree, TreeFactory tf) {
    tree = tree.prune(emptyFilter, tf).spliceOut(aOverAFilter, tf);

    for (Tree t : tree) {

      if(t.isLeaf()) {
        //Strip off morphological analyses and place them in the OriginalTextAnnotation, which is
        //specified by HasContext.
        if(t.value().contains(MorphoFeatureSpecification.MORPHO_MARK)) {
          String[] toks = t.value().split(MorphoFeatureSpecification.MORPHO_MARK);
          if (toks.length != 2) {
            log.err(String.format("%s: Word contains malformed morph annotation: %s", this.getClass().getName(), t.value()));
          } else if (t.label() instanceof CoreLabel) {
            CoreLabel cl = (CoreLabel) t.label();
            cl.setValue(toks[0].trim().intern());
            cl.setWord(toks[0].trim().intern());

            Pair<String,String> lemmaMorph = MorphoFeatureSpecification.splitMorphString(toks[0], toks[1]);
            String lemma = lemmaMorph.first();
            String morphAnalysis = lemmaMorph.second();
            if (lemma.equals(toks[0])) {
              cl.setOriginalText(toks[1].trim().intern());
            } else {
              // TODO(spenceg): Does this help?
              String newLemma = lexMapper.map(null, lemma);
              if (newLemma == null || newLemma.trim().isEmpty()) {
                newLemma = lemma;
              }
              String newMorphAnalysis = newLemma + MorphoFeatureSpecification.LEMMA_MARK + morphAnalysis;
              cl.setOriginalText(newMorphAnalysis.intern());

View on GitHub (pinned to 1b7edd19c4)