stanfordnlp/CoreNLP · error · IOException

Could not parse input format

Error message

Could not parse input format: ${inputFormat}

What it means

getDocument() switches on the request's inputFormat property; any value other than the handled cases (text, serialized, etc.) hits the default branch and throws IOException with the unrecognized value. The server cannot decode the POSTed request body.

Solutions

1. Set inputFormat to text or serialized (check the supported cases in getDocument for your version)
2. Remove the inputFormat property to use the default (text)
3. If sending structured input, use the Protobuf serialized format produced by CoreNLP's own serializer

Example fix

// before
curl -d '...' 'localhost:9000/?properties={"inputFormat":"json","annotators":"ner"}'
// after
curl -d '...' 'localhost:9000/?properties={"inputFormat":"text","annotators":"ner"}'

String in = props.getProperty("inputFormat", "text");
if (!in.equals("text") && !in.equals("serialized")) {
  throw new IllegalArgumentException("inputFormat must be text or serialized, got: " + in);
}

try {
  postToServer(props, body);
} catch (IOException e) {
  if (e.getMessage().startsWith("Could not parse input format")) {
    props.setProperty("inputFormat", "text");
    postToServer(props, body);
  } else throw e;
}

When it happens

Trigger: POSTing to StanfordCoreNLPServer with the inputFormat property set to an unhandled string (e.g. 'json' as input, 'xml', or a typo like 'txt').

Common situations: Assuming the server accepts JSON input (it accepts JSON only as output); typos in inputFormat; clients echoing back output formats as input formats.

Understand the failure class

Background: Invalid enum value errors: "Unknown type", "Invalid scope", "must be one of" — when a string is not on the library's allowed list — this error's family across 23 libraries.

Related errors

• Unhandled input format for scenegraph
• Array lengths don't match
• Attempt to make ObjectBank with empty file list
• Called headPreTerminal on a leaf:
• Can't have null components!

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/327d9352e69ff64a. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/pipeline/StanfordCoreNLPServer.java:343

        // We use to trim. But now we don't. It seems like doing that is illegitimate. text = text.trim();

        // Read the annotation
        Annotation annotation = new Annotation(text);
        // Set the date (if provided)
        if (date != null) {
          annotation.set(CoreAnnotations.DocDateAnnotation.class, date);
        }
        return annotation;
      case "serialized":
        String inputSerializerName = props.getProperty("inputSerializer", ProtobufAnnotationSerializer.class.getName());
        if (!inputSerializerName.equals(ProtobufAnnotationSerializer.class.getName())) {
          throw new IOException("Specifying an inputSerializer other than ProtobufAnnotationSerializer is now deprecated for security reasons.  See https://github.com/stanfordnlp/CoreNLP/security/advisories/GHSA-wv35-hv9v-526p  If you have need for a different class, please post about your use case on the CoreNLP github.");
        }
        AnnotationSerializer serializer = new ProtobufAnnotationSerializer();
        Pair<Annotation, InputStream> pair = serializer.read(httpExchange.getRequestBody());
        return pair.first;
      default:
        throw new IOException("Could not parse input format: " + inputFormat);
    }
  }

  private String getContentType(Headers headers) {
    String contentType = URL_ENCODED;
    if (headers.containsKey("Content-type")) {
      contentType = headers.getFirst("Content-type").split(";")[0].trim();
    }
    return contentType;
  }

  private String getEncoding(Headers headers) {
    // The default encoding by the HTTP standard is ISO-8859-1, but most
    // real users of CoreNLP would likely assume UTF-8 by default.
    String defaultEncoding = this.strict ? "ISO-8859-1" : "UTF-8";
    if (headers.containsKey("Content-type")) {
      String[] charsetPair = Arrays.stream(headers.getFirst("Content-type").split(";"))
          .map(x -> x.split("="))

View on GitHub (pinned to 1b7edd19c4)