What causes "Treebank: Bad tree in treebank!: " in stanfordnlp/CoreNLP?

Locate the offending tree (it is printed in the message) and fix or remove it from the corpus.

stanfordnlp/CoreNLP · error · IllegalStateException

Treebank: Bad tree in treebank!:

Error message

Treebank: Bad tree in treebank!:

What it means

During textualSummary statistics collection, Treebank validates tree structure: a subtree must be either a preterminal (single child) or a normal internal node; anything else (e.g. a non-preterminal leaf or a malformed structure) is reported as a bad tree via IllegalStateException.

Solutions

Locate the offending tree (it is printed in the message) and fix or remove it from the corpus.
Pre-validate the treebank by reading and traversing all trees before summary/training.
Re-export the data from the original source in correct PTB bracketed format.

Example fix

// before
treebank.textualSummary(); // throws on bad tree
// after
for (Tree t : treebank) { validateTree(t); } // fix data first
treebank.textualSummary();

Defensive patterns

Strategy: validation

Validate before calling

for (Tree t : treebank) { for (Tree sub : t) { if (sub.isLeaf() && !sub.parent(t).isPreTerminal()) { log.warn("bad tree: " + sub); } } }

Try / catch

try { summary = treebank.textualSummary(); } catch (IllegalStateException e) { /* quarantine the reported tree and retry */ }

Prevention

Validate corpus files conform to PTB bracketed format before loading.
Regenerate trees from trusted parsers/tools, not hand-edited files.
Check one offending tree at a time using the message's subtree dump.

When it happens

Trigger: Calling treebank.textualSummary() (as done after training/eval setup) when the loaded treebank contains malformed trees — typically trees where a nonterminal node has zero children or leaves appear at non-preterminal positions.

Common situations: Corrupted or non-penn-treebank-conformant corpus files; trees produced by buggy external tools; empty/whitespace-damaged files read into the Treebank.

Understand the failure class

Background: Schema validation failed / invalid input schema: payload rejected because its shape doesn't match the expected schema — this error's family across 28 libraries.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/1b324c40ee1bc91a. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/trees/Treebank.java:410

          if (tlp != null && tlp.isPunctuationTag(subtree.value())) {
            puncts.incrementCount(subtree.firstChild().value());
          }
        } else if (subtree.isPhrasal()) {
          boolean hasLeafChild = false;
          for (Tree kt : subtree.children()) {
            if (kt.isLeaf()) {
              hasLeafChild = true;
            }
          }
          if (hasLeafChild) {
            numPreTerminalWithMultipleChildren++;
            if (preTerminalMultipleChildrenEg == null) {
              preTerminalMultipleChildrenEg = subtree;
            }
          }
          cats.incrementCount(subtree.value());
        } else {
          throw new IllegalStateException("Treebank: Bad tree in treebank!: " + subtree);
        }
      }
    }
    StringWriter sw = new StringWriter(2000);
    PrintWriter pw = new PrintWriter(sw);
    NumberFormat nf = NumberFormat.getNumberInstance();
    nf.setMaximumFractionDigits(0);
    pw.println("Treebank has " + numTrees + " trees (" + numTreesLE40 + " of length <= 40) and " + numWords + " words (tokens)");
    if (numTrees > 0) {
      if (numTags != numWords) {
        pw.println("  Warning! numTags differs and is " + numTags);
      }
      if (roots.size() == 1) {
        String root = (String) roots.keySet().toArray()[0];
        pw.println("  The root category is: " + root);
      } else {
        pw.println("  Warning! " + roots.size() + " different roots in treebank: " + Counters.toString(roots, nf));
        pw.println("  Examples:");

View on GitHub (pinned to 1b7edd19c4)