What causes "Unparsable sentence: " in stanfordnlp/CoreNLP?

Wrap rpq.parse in a try/catch or check the boolean and skip/log unparsable sentences instead of letting main throw

stanfordnlp/CoreNLP · error · RuntimeException

Unparsable sentence:

Error message

Unparsable sentence:

What it means

When the reranking parser fails to find a parse for a sentence, ParseAndPrintMatrices.main() throws RuntimeException("Unparsable sentence: ...") including the token list. With no parse produced there are no deep trees or matrices to print, so the tool aborts that run.

Solutions

Wrap rpq.parse in a try/catch or check the boolean and skip/log unparsable sentences instead of letting main throw
Pre-filter sentences: length limits, strip noise, ensure non-empty token lists
Increase parser limits (e.g. maximum sentence length / beam size options) for long sentences

Example fix

// before
if (!rpq.parse(sentence)) {
  throw new RuntimeException("Unparsable sentence: " + sentence);
}
// after
if (!rpq.parse(sentence)) {
  System.err.println("Skipping unparsable sentence: " + sentence);
  continue;
}

Defensive patterns

Strategy: try-catch

Validate before calling

List<HasWord> tokens = sentence;
if (tokens.isEmpty() || tokens.size() > maxSentenceLength) {
  continue; // skip sentences the parser cannot handle
}

Try / catch

if (!rpq.parse(sentence)) {
  System.err.println("Skipping unparsable sentence: " + sentence);
  continue;
}
// or wrap the whole loop:
try {
  processAllSentences();
} catch (RuntimeException e) {
  if (e.getMessage().startsWith("Unparsable sentence")) { logAndSkip(e); } else { throw e; }
}

Prevention

Pre-filter input: remove empty/overly long lines and non-text noise
Log skipped sentences instead of aborting the batch
Tune parser limits (length, beam) for your data domain
Run a small sample of the input through the parser first

When it happens

Trigger: rpq.parse(sentence) returns false inside the DocumentPreprocessor loop — typically sentences that exceed parser length/beam limits, contain only punctuation or unknown tokens, or are malformed input text in the input file.

Common situations: Feeding long sentences or noisy text (web data, OCR output) to the parser; empty lines producing degenerate token lists; unusually low nThreads/beam settings causing parse failures.

Understand the failure class

Background: 'Something went wrong' / 'Request failed (500)' / 'HTTP error! status: 404' — what failed HTTP requests actually mean and how to find the real cause — this error's family across 28 libraries.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/9a1b558760af0539. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/parser/dvparser/ParseAndPrintMatrices.java:116

    DVModel model = DVParser.getModelFromLexicalizedParser(parser);

    File outputFile = new File(outputPath);
    FileSystem.checkNotExistsOrFail(outputFile);
    FileSystem.mkdirOrFail(outputFile);

    int count = 0;
    if (inputPath != null) {
      Reader input = new BufferedReader(new FileReader(inputPath));
      DocumentPreprocessor processor = new DocumentPreprocessor(input);
      for (List<HasWord> sentence : processor) {
        count++; // index from 1
        ParserQuery pq = parser.parserQuery();
        if (!(pq instanceof RerankingParserQuery)) {
          throw new IllegalArgumentException("Expected a RerankingParserQuery");
        }
        RerankingParserQuery rpq = (RerankingParserQuery) pq;
        if (!rpq.parse(sentence)) {
          throw new RuntimeException("Unparsable sentence: " + sentence);
        }
        RerankerQuery reranker = rpq.rerankerQuery();
        if (!(reranker instanceof DVModelReranker.Query)) {
          throw new IllegalArgumentException("Expected a DVModelReranker");
        }
        DeepTree deepTree = ((DVModelReranker.Query) reranker).getDeepTrees().get(0);
        IdentityHashMap<Tree, SimpleMatrix> vectors = deepTree.getVectors();

        for (Map.Entry<Tree, SimpleMatrix> entry : vectors.entrySet()) {
          log.info(entry.getKey() + "   " +  entry.getValue());
        }

        FileWriter fout = new FileWriter(outputPath + File.separator + "sentence" + count + ".txt");
        BufferedWriter bout = new BufferedWriter(fout);

        bout.write(SentenceUtils.listToString(sentence));
        bout.newLine();
        bout.write(deepTree.getTree().toString());

View on GitHub (pinned to 1b7edd19c4)