stanfordnlp/CoreNLP · error · RuntimeException

Unexpected number of field

Error message

Unexpected number of field {fields.length}, expected >= {FIELDS_MIN} for line ({filename},{lineCnt}): {line}

What it means

DocumentIterator.next() splits each non-blank line of the CoNLL file on the delimiter and requires at least FIELDS_MIN tab-separated fields to extract document id, part number, word, etc. A line with fewer fields than FIELDS_MIN is malformed, so it throws a RuntimeException including the file name, line count, and raw line to pinpoint the bad row.

Solutions

1. Open the file at the reported line (from the message: filename,lineCnt) and fix or remove the malformed row, restoring all required tab-separated columns.
2. Ensure no text editor or transfer step (e.g. git autocrlf, upload normalization) stripped trailing tabs from the corpus.
3. Confirm the file is the correct CoNLL-2011 format matching the FIELDS_MIN expectation; re-obtain the original corpus if not.
4. Run a pre-check script that counts fields per line and reports lines below the required count before running the reader.
5. If generating files yourself, terminate each row with the full column set including trailing empty fields.

Example fix

// before (line stripped to 2 fields)
// XYZ	1

// after (full CoNLL row with all FIELDS_MIN columns)
// XYZ	1	1	word	POS	PARSE	PRED	COREF...

// Pre-check every line has enough tab-separated fields
try (BufferedReader r = Files.newBufferedReader(Paths.get(corpusPath))) {
  String line; int n = 0;
  while ((line = r.readLine()) != null) {
    if (!line.trim().isEmpty() && line.split("\t").length < 10 /* FIELDS_MIN */)
      throw new IllegalStateException("Line " + (n+1) + " has too few fields: " + line);
    n++;
  }
}

When it happens

Trigger: Calling DocumentIterator (via CoNLL2011DocumentReader.nextDoc / dcoref pipeline) over a corpus file that contains a line with too few columns — e.g. a blank-ish line with only a word, a header line, or a row where trailing tabs were stripped.

Common situations: Editors or upload pipelines trimming trailing tabs from TSV/CoNLL files; mixing CoNLL formats (CoNLL-2010 vs 2011 vs 2012 have different column counts); accidentally feeding a non-CoNLL text file as the corpus path.

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

• Error extracting labelled spans for column
• adjustFinalToken: Unexpected final char: |
• Bad data format:
• Both parse.model and parse.executable properties must be…
• Cannot cast " + classname + " into " + type.getName()

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/b1284646b1558a81. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/dcoref/CoNLL2011DocumentReader.java:658

          line = line.trim();
          if (line.length() != 0) {
            if (line.startsWith(docStart)) {
              // Start of new document
              if (document != null) {
                logger.warning("Unexpected begin document at line (\" + filename + \",\" + lineCnt + \")");
              }
              document = new Document();
              document.documentIdPart = line.substring(docStartLength);
            } else if (line.startsWith("#end document")) {
              annotateDocument(document);
              docCnt++;
              return document;
              // End of document
            } else {
              assert document != null;
              String[] fields = delimiterPattern.split(line);
              if (fields.length < FIELDS_MIN) {
                throw new RuntimeException("Unexpected number of field " + fields.length +
                        ", expected >= " + FIELDS_MIN + " for line (" + filename + "," + lineCnt + "): " + line);
              }
              String curDocId = fields[FIELD_DOC_ID];
              String partNo = fields[FIELD_PART_NO];
              if (document.getDocumentID() == null) {
                document.setDocumentID(curDocId);
                document.setPartNo(partNo);
              } else {
                // Check documentID didn't suddenly change on us
                assert(document.getDocumentID().equals(curDocId));
                assert(document.getPartNo().equals(partNo));
              }
              curSentWords.add(fields);
            }
          } else {
            // Current sentence has ended, new sentence is about to be started
            if (curSentWords.size() > 0) {
              assert document != null;

View on GitHub (pinned to 1b7edd19c4)