Java language detection with langdetect - how to load profiles?

I have the same problem. You can load the profiles from the LangDetect jar using JarUrlConnection and JarEntry. Note in this example I am using Java 7 resource management.

    String dirname = "profiles/";
    Enumeration<URL> en = Detector.class.getClassLoader().getResources(
            dirname);
    List<String> profiles = new ArrayList<>();
    if (en.hasMoreElements()) {
        URL url = en.nextElement();
        JarURLConnection urlcon = (JarURLConnection) url.openConnection();
        try (JarFile jar = urlcon.getJarFile();) {
            Enumeration<JarEntry> entries = jar.entries();
            while (entries.hasMoreElements()) {
                String entry = entries.nextElement().getName();
                if (entry.startsWith(dirname)) {
                    try (InputStream in = Detector.class.getClassLoader()
                            .getResourceAsStream(entry);) {
                        profiles.add(IOUtils.toString(in));
                    }
                }
            }
        }
    }

    DetectorFactory.loadProfile(profiles);
    Detector detector = DetectorFactory.create();
    detector.append(text);
    String langDetected = detector.detect();
    System.out.println(langDetected);

Since no maven-support was available, and the mechanism to load profiles was not perfect (since you you need to define files instead of resources), I created a fork which solves that problem:

https://github.com/galan/language-detector

I mailed the original author, so he can fork/maintain the changes, but no luck - seems the project is abandoned.

Here is an example of how to use it now (own profiles can be written where necessary):

DetectorFactory.loadProfile(new DefaultProfile()); // SmProfile is also available
Detector detector = DetectorFactory.create();
detector.append(input);
String result = detector.detect();
// maybe work with detector.getProbabilities()

I don't like the static approach the DetectorFactory uses, but I won't rewrite the full project, you have to create your own fork/pull request :)

Looks like the library only accepts files. You can either change the code and try submitting the changes upstream. Or write your resource to a temp file and get it to load that.