Why does Java Wordnet similarity crash on second call?

Question

As stated here I use the Java version of Wordnet::similarity.

This is the server side code. I recieve a List with two elements (i.e. two sentences) and then compute the similarity matrix for those two sentences

String wordNetDirectory = System.getenv("WNHOME");
JWS ws = new JWS(wordNetDirectory,"3.0");
Resnik res = ws.getResnik();
while(true) {
    try {
        Socket clientSocket = serverSocket.accept();
        System.out.println("Socket Established...");

        ObjectOutputStream outToClient = new ObjectOutputStream(clientSocket.getOutputStream());
        ObjectInputStream inFromClient = new ObjectInputStream(clientSocket.getInputStream());

        List<List<String>> sents;
        sents = (List<List<String>>) inFromClient.readObject();
        System.out.println("recieved");

        List<String> s1 = sents.get(0);
        List<String> s2 = sents.get(1);

        Double sum = 0.0;
        for(String w1 : s1) {
            for(String w2 : s2) {
                if(w1.equals(w2)) sum += 1;
                else sum += res.max(w1, w2, "n"); // <- Server.java:54
            }
        }
        sum /= (s1.size() * s2.size());
        outToClient.writeObject(sum);
        System.out.println("wrote " + sum);
    } catch (IOException e) {
    } catch (ClassNotFoundException e) {
    }
}

The client queries the server for sentence similarity in a serial way (i.e. it sends another request after the similarity score of the previous pair is recieved)

Here's the output I get on server side

Socket Established...
recieved
wrote 0.111357828694864
Socket Established...

Exception in thread "main" java.lang.IllegalArgumentException
at edu.mit.jwi.item.IndexWordID.<init>(IndexWordID.java:62)
at edu.mit.jwi.CachingDictionary.getIndexWord(CachingDictionary.java:173)
at edu.sussex.nlp.jws.Resnik.res(Resnik.java:133)
at edu.sussex.nlp.jws.Resnik.max(Resnik.java:313)
at Server.start(Server.java:54)
at Main.main(Main.java:7)

I don't understand why it crashes when connection is established for the second time.

Answer 1

I think it may caused by a illegal word. If w1 or w2 is empty or all whitespace or does not appear in wordnet word list, you will get this exception when you fetch the wordID.

You get the exception in IndexWordID class, so I think this is the most possible problem. You could change the string you put to have a try, or give the code and string argument where you call the method for more information. You can see this doc http://projects.csail.mit.edu/jwi/api/edu/mit/jwi/item/IndexWordID.html