stanfordnlp/CoreNLP · error · RuntimeIOException

Couldn't load

Error message

Couldn't load {file}

What it means

loadGenderNumber throws RuntimeIOException("Couldn't load <file>") when the serialized gender/number map cannot be read via IOUtils.readObjectFromURLOrClasspathOrFileSystem, or ClassNotFoundException if the serialized data was produced by an incompatible class version. It means the gender-number dictionary resource failed to load or deserialize.

Solutions

1. Match the CoreNLP models version to the library version (re-download models jar)
2. Verify the gender file path exists and is readable
3. Use classpath:/ URL for the bundled resource
4. If ClassNotFoundException, regenerate the serialized gender file with edu.stanford.nlp.dcoref.util.ConvertGenderFile using the current library version

Example fix

// before
<!-- corenlp 3.8 jar with corenlp models 3.4 (serialized class mismatch) -->
// after
<dependency>stanford-corenlp</dependency>
<dependency>stanford-corenlp-models</dependency> <!-- same version as jar -->

String f = props.getProperty("coref.gender.map") /* or default */;
if (f != null && !IOUtils.existsInClasspathOrFileSystem(f)) {
    throw new IllegalStateException("Gender map not found: " + f);
}

try {
    new Dictionaries(props);
} catch (RuntimeIOException e) {
    // also covers ClassNotFoundException cause for serialized map
    log.error("Gender map load failed (check model/lib version match): " + e.getMessage(), e);
    throw e;
}

When it happens

Trigger: Constructing Dictionaries when the gender file is missing/unreadable, or the Java-serialized map's classes don't match the current CoreNLP serialVersionUID (model/code version mismatch).

Common situations: Old CoreNLP model files with a newer jar (deserialization fails), missing models jar, truncated file, or wrong path in coref.gender properties.

Related errors

• Error loading classifier from
• Error reading saved links
• Error creating data exporter
• Error setting up training
• Error initializing FeatureExtractorRunner

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/31bae7757c6ae40c. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/dcoref/Dictionaries.java:396

  /**
   * Load Bergsma and Lin (2006) gender and number list.
   * <br>
   * The list is converted from raw text and numbers to a serialized
   * map, which saves quite a bit of time loading.
   * See edu.stanford.nlp.dcoref.util.ConvertGenderFile
   */
  private void loadGenderNumber(String file, String neutralWordsFile) {
    try {
      getWordsFromFile(neutralWordsFile, neutralWords, false);
    } catch (IOException e) {
      throw new RuntimeIOException("Couldn't load " + neutralWordsFile);
    }
    try {
      Map<List<String>, Gender> temp = IOUtils.readObjectFromURLOrClasspathOrFileSystem(file);
      genderNumber.putAll(temp);
    } catch (IOException | ClassNotFoundException e) {
      throw new RuntimeIOException("Couldn't load " + file);
    }
  }

  private static void loadCorefDict(String[] file,
      ArrayList<Counter<Pair<String, String>>> dict) {

    for(int i = 0; i < 4; i++){
      dict.add(new ClassicCounter<>());

      BufferedReader reader = null;
      try {
        reader = IOUtils.readerFromString(file[i]);
        // Skip the first line (header)
        reader.readLine();

        while(reader.ready()) {
          String[] split = reader.readLine().split("\t");
          dict.get(i).setCount(new Pair<>(split[0], split[1]), Double.parseDouble(split[2]));

View on GitHub (pinned to 1b7edd19c4)