stanfordnlp/CoreNLP · error · IllegalArgumentException

Invalid tag pattern: " + pattern + " for annotation key " +…

Error message

Invalid tag pattern: " + pattern + " for annotation key " + annoKeyString

What it means

After resolving the annotation key, the pattern part of the config must match TAG_ATTR_PATTERN (tag or tag/attribute). When it does not match and attrOnly is set (attribute required), addAnnotationPatterns throws IllegalArgumentException because it cannot build a tag/attribute matcher.

Solutions

Rewrite the pattern to the expected form "annotationKey,tagRegex[/attrRegex]" so it matches TAG_ATTR_PATTERN.
If you only need attribute matching, supply a valid attribute specification with a non-empty tag part.
Escape regex metacharacters so the tag portion parses as a single tag pattern.

Example fix

// before
props.setProperty("clean.xmltokenannotationpatterns", "word,/speaker=*");
// after
props.setProperty("clean.xmltokenannotationpatterns", "word, speaker|.*");

Defensive patterns

Strategy: validation

Validate before calling

String[] parts = patternSpec.split(",", 2);
if (parts.length < 2 || !parts[1].matches("[^/]+(/[^/]*)?")) throw new IllegalArgumentException("Bad CleanXml pattern: " + patternSpec);

Try / catch

try { annotator.setTokenAnnotationPatterns(specs); } catch (IllegalArgumentException e) { if (e.getMessage().startsWith("Invalid tag pattern")) { log.error("Malformed tag pattern: " + e.getMessage()); } else { throw e; } }

Prevention

Follow the documented format annotationKey,tagRegex[/attrRegex] exactly.
Escape regex metacharacters in tag names before embedding them in patterns.
Test each pattern against sample XML in a unit test.

When it happens

Trigger: Calling setDocAnnotationPatterns/setTokenAnnotationPatterns/setSectionAnnotationPatterns with a pattern string whose tag portion fails TAG_ATTR_PATTERN when an attribute is required (attrOnly mode), e.g. a malformed "word,/type=x" style pattern or empty/invalid tag regex.

Common situations: Misplaced separators in the comma-separated config; attribute-only patterns supplied where the tag part is empty; regexes containing characters that break the expected tag[/attr] structure.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/8090f50b6c43ceb0. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/pipeline/CleanXmlAnnotator.java:346

      String[] annoPattern = annoPatternString.split("\\s*=\\s*", 2);
      if (annoPattern.length != 2) {
        throw new IllegalArgumentException("Invalid annotation to tag pattern: " + annoPatternString);
      }
      String annoKeyString = annoPattern[0];
      String pattern = annoPattern[1];
      Class annoKey = EnvLookup.lookupAnnotationKeyWithClassname(null, annoKeyString);
      if (annoKey == null) {
        throw new IllegalArgumentException("Cannot resolve annotation key " + annoKeyString);
      }
      Matcher m = TAG_ATTR_PATTERN.matcher(pattern);
      if (m.matches()) {
        Pattern tagPattern = toCaseInsensitivePattern(m.group(1));
        Pattern attrPattern = toCaseInsensitivePattern(m.group(2));
        annotationPatterns.add(annoKey, Pair.makePair(tagPattern, attrPattern));
      } else {
        if (attrOnly) {
          // attribute is require
          throw new IllegalArgumentException("Invalid tag pattern: " + pattern + " for annotation key " + annoKeyString);
        } else {
          Pattern tagPattern = toCaseInsensitivePattern(pattern);
          annotationPatterns.add(annoKey, Pair.makePair(tagPattern, null));
        }
      }
    }
  }

  /**
   * Helper method to set the TokenBeginAnnotation and TokenEndAnnotation of every token.
   */
  public void setTokenBeginTokenEnd(List<CoreLabel> tokensList) {
    int tokenIndex = 0;
    for (CoreLabel token : tokensList) {
      token.set(CoreAnnotations.TokenBeginAnnotation.class, tokenIndex);
      token.set(CoreAnnotations.TokenEndAnnotation.class, tokenIndex+1);
      tokenIndex++;
    }

View on GitHub (pinned to 1b7edd19c4)