stanfordnlp/CoreNLP · error · IllegalArgumentException

Invalid hypGraph

Error message

Invalid hypGraph 

What it means

Alignment.makeFromIndexArray throws IllegalArgumentException("Invalid hypGraph " + hypGraph) when the hypothesis SemanticGraph is null or empty. The index array is sized against the hyp graph, so an empty/null hyp graph makes the alignment undefined.

Solutions

1. Verify hypGraph != null && hypGraph.size() > 0 before the call
2. Also ensure indexes.length equals hypGraph.size() (the next precondition checked)
3. Filter out empty hypothesis graphs earlier in the pipeline

Example fix

// before
Alignment a = Alignment.makeFromIndexArray(txtGraph, hypGraph, idx, 1.0, "j");
// after
if (hypGraph != null && hypGraph.size() > 0 && idx.length == hypGraph.size()) {
  Alignment a = Alignment.makeFromIndexArray(txtGraph, hypGraph, idx, 1.0, "j");
}

if (hypGraph != null && hypGraph.size() > 0 && indexes != null && indexes.length == hypGraph.size()) { Alignment a = Alignment.makeFromIndexArray(txtGraph, hypGraph, indexes, score, just); }

boolean validHyp = h != null && !h.isEmpty() && indexes != null && indexes.length == h.size();

try { a = Alignment.makeFromIndexArray(txt, hyp, idx, score, just); } catch (IllegalArgumentException e) { log.warn("Alignment skipped: " + e.getMessage()); }

When it happens

Trigger: Calling makeFromIndexArray with hypGraph == null or a zero-vertex SemanticGraph, typically from an empty hypothesis sentence parse.

Common situations: Entailment/RTE pipelines feeding unpopulated hypothesis graphs after failed parsing; null propagation through match/alignment chains.

Related errors

• Invalid txtGraph
• conditionalLogProbGivenPrevious requires given one less…
• conditionalLogProbsGivenPrevious requires given one less…
• conditionalLogProbGivenFirst requires of one less than…
• unnormalizedConditionalLogProbGivenFirst requires of one…

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/36045ef6e557f908. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/semgraph/semgrex/Alignment.java:162

  /**
   * Constructs and returns a new Alignment from the given hypothesis
   * {@code SemanticGraph} to the given text (passage) SemanticGraph, using
   * the given array of indexes.  The i'th node of the array should contain the
   * index of the node in the text (passage) SemanticGraph to which the i'th
   * node in the hypothesis SemanticGraph is aligned, or -1 if it is aligned to
   * NO_WORD.
   */
  public static Alignment makeFromIndexArray(SemanticGraph txtGraph,
                                             SemanticGraph hypGraph,
                                             int[] indexes,
                                             double score,
                                             String justification) {
    if (txtGraph == null || txtGraph.isEmpty())
      throw new IllegalArgumentException("Invalid txtGraph " + txtGraph);
    if (hypGraph == null || hypGraph.isEmpty())
      throw new IllegalArgumentException("Invalid hypGraph " + hypGraph);
    if (indexes == null)
      throw new IllegalArgumentException("Null index array");
    if (indexes.length != hypGraph.size())
      throw new IllegalArgumentException("Index array length " + indexes.length +
                                         " does not match hypGraph size " + hypGraph.size());
    Map<IndexedWord, IndexedWord> map =
      Generics.newHashMap();
    for (int i = 0; i < indexes.length; i++) {
      IndexedWord hypNode = hypGraph.getNodeByIndex(i);
      IndexedWord txtNode = IndexedWord.NO_WORD;
      if (indexes[i] >= 0)
        txtNode = txtGraph.getNodeByIndex(indexes[i]);
      map.put(hypNode, txtNode);
    }
    return new Alignment(map, score, justification);
  }

  public static Alignment makeFromIndexArray(SemanticGraph txtGraph,

View on GitHub (pinned to 1b7edd19c4)