stanfordnlp/CoreNLP · error · RuntimeException

Error: Attempted to run Tregex on sentence without a…

Error message

Error: Attempted to run Tregex on sentence without a constituency parse.  To use this method you must annotate the document with a constituency parse using the 'parse' annotator.

What it means

CoreSentence.tregexResultTrees(TregexPattern) needs the sentence's constituency parse tree; tregex operates on Tree objects. If constituencyParse() returns null (the 'parse' annotator was not run) a RuntimeException is thrown explaining that the 'parse' annotator is required.

Solutions

1. Add the 'parse' annotator to the pipeline's annotators list.
2. If constituency parsing is too slow, configure a fast constituency model or accept depparse and rewrite queries against dependency output.
3. Guard calls with sentence.constituencyParse() != null before running Tregex.
4. Use CoreMapExpressionExtractor / dependency-based alternatives if only dependency structure is needed.

Example fix

// before
props.setProperty("annotators", "tokenize,ssplit,pos,lemma,ner");
Tree t = sentence.tregexResultTrees("NP < NN").get(0);
// after
props.setProperty("annotators", "tokenize,ssplit,pos,lemma,ner,parse");
Tree t = sentence.tregexResultTrees("NP < NN").get(0);

if (sentence.constituencyParse() == null) { throw new IllegalStateException("Run pipeline with the 'parse' annotator before Tregex queries"); }

Optional<Tree> parse = Optional.ofNullable(sentence.constituencyParse());
parse.ifPresent(t -> runTregex(t));

try { trees = sentence.tregexResultTrees(pattern); } catch (RuntimeException e) { if (e.getMessage().startsWith("Error: Attempted to run Tregex")) { trees = Collections.emptyList(); log.warn("No constituency parse; skipping Tregex"); } else { throw e; } }

When it happens

Trigger: Calling sentence.tregexResultTrees(pattern) (or the String-pattern overload, and helpers like sentenceParseMatches) on a pipeline that lacks the 'parse' (or a constituency-providing) annotator in its annotator list.

Common situations: Building a pipeline with only tokenize/ssplit/pos/lemma/ner and then running Tregex queries; switching from a dependency-only (depparse) pipeline to constituency-based code without adding 'parse'; memory-constrained setups where 'parse' was removed.

Related errors

• Error: cannot process tregex operations with no…
• format error in embeddings
• format error unexpected featureFactory line:
• Need to supply a parser model with -model
• Need to supply an output filename with -output

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/72015921b0d4487d. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/pipeline/CoreSentence.java:105

  /** list of ner tags **/
  public List<String> nerTags() { return tokens().stream().map(token -> token.ner()).collect(Collectors.toList()); }

  /** constituency parse **/
  public Tree constituencyParse() {
    return sentenceCoreMap.get(TreeCoreAnnotations.TreeAnnotation.class);
  }

  /** Tregex - find subtrees of interest with a general Tregex pattern **/
  public List<Tree> tregexResultTrees(String s) {
    // the patterns are cached by computeIfAbsent, so we don't wastefully recompile a TregexPattern every sentence
    return tregexResultTrees(patternCache.computeIfAbsent(s, compilePattern));
  }

  public List<Tree> tregexResultTrees(TregexPattern p) {
    // throw a RuntimeException if no constituency parse available to signal to user to use "parse" annotator
    if (constituencyParse() == null)
      throw new RuntimeException("Error: Attempted to run Tregex on sentence without a constituency parse.  " +
              "To use this method you must annotate the document with a constituency parse using the 'parse' " +
              "annotator.");
    List<Tree> results = new ArrayList<>();
    TregexMatcher matcher = p.matcher(constituencyParse());
    while (matcher.find()) {
      results.add(matcher.getMatch());
    }
    return results;
  }

  public List<String> tregexResults(TregexPattern p) {
    return tregexResultTrees(p).stream().map(treeToSpanString).collect(Collectors.toList());
  }

  public List<String> tregexResults(String s) {
    return tregexResultTrees(s).stream().map(treeToSpanString).collect(Collectors.toList());
  }

View on GitHub (pinned to 1b7edd19c4)