Deploying LanguageTool as a Java Servlet for Grammar Checking

Using LanguageTool for grammar and spelling checks is convenient; the following steps demonstrate how to deploy a LanguageTool servlet.

First, clone the source repository:

git clone [email protected]:languagetool/languagetool.git

Because the project is large, the clone may take a while. After cloning, compile it with Maven:

$ mvn clean test
$ ./build.sh languagetool-standalone -DskipTests

When the build finishes (typically 5–30 minutes), the target directory contains three main JAR files: languagetool.jar, languagetool-server.jar, and languagetool-commandline.jar, plus many supporting files.

You can verify the build by running a command‑line test:

$ echo "I was do book a tickt last night" | java -jar languagetool-commandline.jar -l en-US -m en

The intentionally incorrect sentence triggers a JSON response describing the grammar and spelling issues, e.g.:

{
  "software": {"name":"LanguageTool","version":"4.0-SNAPSHOT"},
  "matches":[{
    "message":"Consider using a past participle here: \"done\".",
    "shortMessage":"Possible agreement error",
    "offset":6,"length":2,
    "rule":{"id":"BEEN_PART_AGREEMENT","description":"Agreement: 'been' or 'was' + past tense"}
  }]
}

Next, create a servlet project and copy all files from the lib directory, together with the three main JARs, into WEB-INF/lib. Inside the servlet, add the following code to invoke LanguageTool:

protected void doPost(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException {
    JLanguageTool lt = new JLanguageTool(Languages.getLanguageForShortCode("en-US"));
    List<RuleMatch> list = lt.check(request.getParameter("text"));
    StringBuilder s = new StringBuilder();
    for (RuleMatch rm : list) {
        s += String.format("message => %s, from => %d, to => %d<br>", rm.getMessage(), rm.getFromPos(), rm.getToPos());
    }
    response.setContentType("text/plain");
    PrintWriter pw = response.getWriter();
    pw.write(s.toString());
}

When running the servlet, an error may appear stating that 'en-US' is not a language code known to LanguageTool. The cause is that the META‑INF resources containing language definitions were not copied.

Inspect the compiled output: language classes reside under org/languagetool/language together with many .properties files. To make them available, package the META‑INF files into a separate JAR: $ zip -r languages.jar META-INF/org/ Copy languages.jar into WEB-INF/lib. After adding this JAR, the servlet works correctly. The servlet code can be simplified by using the concrete language class:

JLanguageTool lt = new JLanguageTool(new AmericanEnglish());

Finally, test the deployed servlet with curl:

$ curl -d "text=I was do book a tickt last night" http://localhost:8080/demo

The response contains messages such as "Consider using a past participle here: done " and "Possible spelling mistake found", confirming that the LanguageTool servlet is operational.