Stanford Parser questions

Question

I am writing a project that works with NLP (natural language parser). I am using the stanford parser.

I create a thread pool that takes sentences and run the parser with them. When I create one thread its all works fine, but when I create more, I get errors. The "test" procedure is finding words that have some connections. If I do an synchronized its supposed to work like one thread but still I get errors. My problem is that I have errors on this code:

public synchronized String test(String s,LexicalizedParser lp )
{

    if (s.isEmpty()) return "";
    if (s.length()>80) return "";
    System.out.println(s);
    String[] sent = s.split(" ");
 Tree parse = (Tree) lp.apply(Arrays.asList(sent));
TreebankLanguagePack tlp = new PennTreebankLanguagePack();
GrammaticalStructureFactory gsf = tlp.grammaticalStructureFactory();
GrammaticalStructure gs = gsf.newGrammaticalStructure(parse);
Collection tdl = gs.typedDependenciesCollapsed();
List list = new ArrayList(tdl);


//for (int i=0;i<list.size();i++)
//System.out.println(list.get(1).toString());

//remove scops and numbers like sbj(screen-4,good-6)->screen good

 Pattern p = Pattern.compile(".*\\((.*?)\\-\\d+,(.*?)\\-\\d+\\).*");

       if (list.size()>2){
    // Split input with the pattern
        Matcher m = p.matcher(list.get(1).toString());
        //check if the result have more than  1 groups
       if (m.find()&& m.groupCount()>1){
           if (m.groupCount()>1)
           {
               System.out.println(list);
 return  m.group(1)+m.group(2);
    }}
}
        return "";

}

the errors that I have are:

at blogsOpinions.ParserText.(ParserText.java:47) at blogsOpinions.ThreadPoolTest$1.run(ThreadPoolTest.java:50) at blogsOpinions.ThreadPool$PooledThread.run(ThreadPoolTest.java:196) Recovering using fall through strategy: will construct an (X ...) tree. Exception in thread "PooledThread-21" java.lang.ClassCastException: java.lang.String cannot be cast to edu.stanford.nlp.ling.HasWord

at edu.stanford.nlp.parser.lexparser.LexicalizedParser.apply(LexicalizedParser.java:289) at blogsOpinions.ParserText.test(ParserText.java:174) at blogsOpinions.ParserText.insertDb(ParserText.java:76) at blogsOpinions.ParserText.(ParserText.java:47) at blogsOpinions.ThreadPoolTest$1.run(ThreadPoolTest.java:50) at blogsOpinions.ThreadPool$PooledThread.run(ThreadPoolTest.java:196)

and how can i get the discription of the subject like the screen is very good, and I want to get screen good from the list the I get and not like list.get(1) .

Answer 1

You can't call LexicalizedParser.parse on a List of String s; it expects a list of HasWord objects. It's much easier to call the apply method on your input string. This will also run a proper tokenizer on your input (instead of your simple split on spaces).

To get relations such as subjectness out of the returned Tree , call its dependencies member.

Answer 2

Hm, I witnessed the same stack trace. Turned out I was loading two instances of the LexicalizedParser in the same JVM. This seemed to be the problem. When I made sure only one instance is created, I was able to call lp.apply(Arrays.asList(sent)) just fine.