stanfordnlp/CoreNLP · error · IllegalStateException

Could not parse CoNLL file

Error message

Could not parse CoNLL file

What it means

While reading CoNLL-format data lines, readDataset dispatches on the field at index 3 (SUBJECT/OBJECT/'-'). If fields[1] and fields[3] don't match any known role pattern, the line doesn't conform to the expected KBP CoNLL schema and it throws an IllegalStateException.

Solutions

1. Validate each line has the expected tab-separated columns before feeding the file
2. Fix rows whose role column is not SUBJECT/OBJECT/'-' (fields[3])
3. Ensure the file uses tabs, not spaces, as delimiters
4. Cross-check your file against the official KBP example input format

Example fix

// before (space-delimited, fields shift)
John_SUBJ ... OBJECT ... // line read with split("\t") yields wrong roles
// after
// ensure tab separation: "tok\tO\tO\tSUBJECT\tPERSON"
readDataset("fixed.conll");

int lineNo = 0;
for (String line : Files.readAllLines(Paths.get(file))) {
  lineNo++;
  if (line.isEmpty() || line.startsWith("#")) continue;
  String[] f = line.split("\t", -1);
  if (f.length < 5) throw new IllegalStateException("line " + lineNo + " has " + f.length + " columns");
  boolean ok = "SUBJECT".equals(f[3]) || "OBJECT".equals(f[3])
      || ("-".equals(f[1]) && "-".equals(f[3]));
  if (!ok) throw new IllegalStateException("line " + lineNo + " bad role: " + f[3]);
}

try {
  readDataset(path);
} catch (IllegalStateException e) {
  log.error("CoNLL parse failed — check tabs/columns: " + e.getMessage());
  throw e;
}

When it happens

Trigger: A data line whose 4th column (role) is neither SUBJECT, OBJECT, nor the '-' pair ('-' in fields[1] and fields[3]) — e.g. malformed rows, wrong delimiter, or a truncated file with shifted columns.

Common situations: Hand-edited or merged CoNLL files; tab vs space delimiters altering column positions; rows from a different CoNLL task schema.

Understand the failure class

Background: "Invalid ... format", "must be in format X", "does not look like a ..." — invalid argument format errors across CLI tools and libraries — this error's family across 17 libraries.

Related errors

• cannot be cast into a…
• First line of input file should be header definition
• Gabor sucks at logic and he should feel bad about it
• cannot be cast into a
• Shouldn't happen:

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/6f1f90469a9a1d6d. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/ie/KBPRelationExtractor.java:383

    while ( (line = reader.readLine()) != null ) {
      String[] fields = line.split("\t");
      if (relation == null) {
        // Case: read the relation
        assert fields.length == 1;
        relation = fields[0];
      } else if (fields.length == 9) {
        // Case: read a token
        tokens.add(fields[0]);
        if ("SUBJECT".equals(fields[1])) {
          subject = new Span(Math.min(subject.start(), i), Math.max(subject.end(), i + 1));
          subjectNER = valueOf(fields[2].toUpperCase(Locale.ROOT));
        } else if ("OBJECT".equals(fields[3])) {
          object = new Span(Math.min(object.start(), i), Math.max(object.end(), i + 1));
          objectNER = valueOf(fields[4].toUpperCase(Locale.ROOT));
        } else if ("-".equals(fields[1]) && "-".equals(fields[3])) {
          // do nothing
        } else {
          throw new IllegalStateException("Could not parse CoNLL file");
        }
        i += 1;
      } else if (StringUtils.isNullOrEmpty(line.trim())) {
        // Case: commit a sentence
        examples.add(Pair.makePair(new KBPInput(
            subject,
            object,
            subjectNER,
            objectNER,
            new Sentence(tokens)
        ), relation));

        // (clear the variables)
        i = 0;
        relation = null;
        tokens = new ArrayList<>();
        subject = new Span(Integer.MAX_VALUE, Integer.MIN_VALUE);
        object = new Span(Integer.MAX_VALUE, Integer.MIN_VALUE);

View on GitHub (pinned to 1b7edd19c4)