stanfordnlp/CoreNLP · error · IllegalStateException

Cannot deserialize KBP triples with this method!

Error message

Cannot deserialize KBP triples with this method!

What it means

ProtobufAnnotationSerializer.fromProto(CoreMap) refuses to deserialize sentences that carry KBP relation triples (RelationTriplesAnnotation from the kbp annotator). The lossy sentence-level conversion cannot reconstruct these triples, so their presence triggers an explicit IllegalStateException instead of silently dropping data.

Solutions

  1. Use the document-level deserialization path (fromProto(CoreNLPProtos.Document)) so KBP triples are reconstructed.
  2. Remove the kbp annotator from the pipeline, or clear KBP triple fields, before sentence-level serialization.
  3. Catch IllegalStateException and re-issue deserialization at the document level.

Example fix

// before
CoreMap sentence = serializer.fromProto(sentenceProto); // throws when KBP triples present
// after
Annotation doc = serializer.fromProto(docProto); // full-document deserialization preserves KBP triples
Defensive patterns

Strategy: validation

Validate before calling

if (sentenceProto.getKbpTripleCount() > 0) {
  throw new IllegalArgumentException("Sentence proto carries KBP triples; use document-level deserialization");
}

Try / catch

try {
  cm = serializer.fromProto(sentenceProto);
} catch (IllegalStateException e) {
  cm = serializer.fromProto(documentProto).get(CoreAnnotations.SentencesAnnotation.class).get(idx);
}

Prevention

When it happens

Trigger: Calling ProtobufAnnotationSerializer.fromProto(CoreNLPProtos.Sentence) on a Sentence proto whose getKbpTripleCount() > 0, i.e. the pipeline was run with the kbp annotator and only the sentence protos were serialized.

Common situations: Streaming sentence protos from a server annotated with kbp, then deserializing with the lossy sentence-only API; mixed pipelines that enable kbp/openie but still use the older sentence-level serialization helpers.

Understand the failure class

Background: UnsupportedOperationException and "is not supported" errors: when a library deliberately refuses a call — this error's family across 30 libraries.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/406ec35e4b89004e. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/pipeline/ProtobufAnnotationSerializer.java:1584

    List<CoreLabel> tokens = proto.getTokenList().stream().map(this::fromProto).collect(Collectors.toList());
    lossySentence.set(TokensAnnotation.class, tokens);
    setSentenceTokenAnnotations(lossySentence, proto, tokens, null);
    // Add entailed sentences
    if (proto.getEntailedSentenceCount() > 0) {
      List<SentenceFragment> entailedSentences = proto.getEntailedSentenceList().stream().map(frag -> fromProto(frag, lossySentence.get(CollapsedDependenciesAnnotation.class))).collect(Collectors.toList());
      lossySentence.set(NaturalLogicAnnotations.EntailedSentencesAnnotation.class, entailedSentences);
    }
    // Add entailed clauses
    if (proto.getEntailedClauseCount() > 0) {
      List<SentenceFragment> entailedClauses = proto.getEntailedClauseList().stream().map(frag -> fromProto(frag, lossySentence.get(CollapsedDependenciesAnnotation.class))).collect(Collectors.toList());
      lossySentence.set(NaturalLogicAnnotations.EntailedClausesAnnotation.class, entailedClauses);
    }
    // Add relation triples
    if (proto.getOpenieTripleCount() > 0) {
      throw new IllegalStateException("Cannot deserialize OpenIE triples with this method!");
    }
    if (proto.getKbpTripleCount() > 0) {
      throw new IllegalStateException("Cannot deserialize KBP triples with this method!");
    }
    // Add chinese characters
    if (proto.getCharacterCount() > 0) {
      List<CoreLabel> sentenceCharacters = proto.getCharacterList().stream().map(this::fromProto).collect(Collectors.toList());
      lossySentence.set(SegmenterCoreAnnotations.CharactersAnnotation.class, sentenceCharacters);
    }
    // Add text -- missing by default as it's populated from the Document
    lossySentence.set(TextAnnotation.class, recoverOriginalText(tokens, proto));

    // add section info
    if (proto.hasSectionName())
      lossySentence.set(SectionAnnotation.class, proto.getSectionName());
    if (proto.hasSectionDate())
      lossySentence.set(SectionDateAnnotation.class, proto.getSectionDate());
    if (proto.hasSectionAuthor())
      lossySentence.set(AuthorAnnotation.class, proto.getSectionAuthor());
    if (proto.hasSectionIndex())
      lossySentence.set(SectionIndexAnnotation.class, proto.getSectionIndex());

View on GitHub (pinned to 1b7edd19c4)