stanfordnlp/CoreNLP · warning

QuoteAttribution doCoreference: Null pronounCorefMap

Error message

QuoteAttribution doCoreference: Null pronounCorefMap

What it means

Sieve.doCoreference resolves pronouns to people using pronounCorefMap, which is populated by a prior coreference sieve pass. If the map was never initialized (coreference not run or setup skipped), the method logs a warning and returns null instead of throwing, leaving the quote without a resolved speaker.

Solutions

1. Ensure the coreference sieve stage runs before sieves that call doCoreference so pronounCorefMap is populated.
2. Check the QuoteAttributionAnnotator options — do not disable coreference (e.g. markCoref=false) when sieves need it.
3. Guard call sites: treat a null return from doCoreference as 'no speaker' rather than propagating it.
4. If invoking Sieve directly, initialize pronounCorefMap yourself before calling doCoreference.

Example fix

// before
Person speaker = sieve.doCoreference(key, quote);
// after
if (corefRun) {
  Person speaker = sieve.doCoreference(key, quote);
  if (speaker == null) { /* quote skipped: coref unavailable */ }
}

if (pronounCorefMap == null) { throw new IllegalStateException("Run the coreference sieve before doCoreference"); }

if (pronounCorefMap == null) return null; // caller must null-check

Person speaker = sieve.doCoreference(key, quote);
if (speaker == null) { log.warn("coreference map missing or no antecedent; quote left unattributed"); }

When it happens

Trigger: Calling doCoreference (directly or via quote attribution sieves) before the pass that builds pronounCorefMap, or constructing Sieve subclasses without running the coreference stage.

Common situations: Custom pipelines that call quote-attribution sieves out of order; disabling the coreference sieve in AnnotatorPool options; reusing Sieve objects across documents without re-initialization.

Related errors

• Empty top speakers list for
• attempt to get word when sentence and lattice are null!
• Coreference is not implemented for Arabic
• Coreference is not implemented for Spanish
• after W derivative, index() != x.length()

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/677f4961870675ad. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/quoteattribution/Sieves/Sieve.java:101

      int currQuoteParagraph = QuoteAttributionUtils.getQuoteParagraphIndex(doc, currQuote);
      if (currQuoteParagraph == quoteParagraph) {
        quoteNames.addAll(scanForNames(new Pair<>(currQuote.get(CoreAnnotations.TokenBeginAnnotation.class), currQuote.get(CoreAnnotations.TokenEndAnnotation.class))).first);
      } else {
        break;
      }
    }

    Set<Person> namesInParagraph = new HashSet<>();
    for (String name : quoteNames) {
      namesInParagraph.addAll(characterMap.get(name));
    }
    // log.info("getNamesInParagraph for " + quote.toShorterString() + " got " + namesInParagraph);
    return namesInParagraph;
  }

  public Person doCoreference(int corefMapKey, CoreMap quote) {
    if (pronounCorefMap == null) {
      log.warn("QuoteAttribution doCoreference: Null pronounCorefMap");
      return null;
    }
    Set<Person> quoteNames = new HashSet<>();
    if (quote != null) {
      quoteNames = getNamesInParagraph(quote);
    }
    // log.info("Names in paragraph are: " + quoteNames);
    String referent = pronounCorefMap.get(corefMapKey);
    // log.info("pronounCorefMap lookup referent of charOffsetBegin " + corefMapKey + " is " + referent);
    Person candidate = resolveAmbiguities(referent);
    // log.info("Candidate is: " + candidate);
    if (candidate != null && ! quoteNames.contains(candidate)) {
      return candidate;
    }
    return null;
  }

  private static class TokenNode {

View on GitHub (pinned to 1b7edd19c4)