Using fuzzy logic to identify related questions

artkohn · **Joined:** Fri Aug 21, 2009 7:48 pm **Posts:** 1

I am creating an application where users type in a question, and then I want the database to use a sort of fuzzy logic to examine the words in the question and indentify the most closely matched previously asked questions. Is there any set of open source code that will help me get started with this? Or is there a better way to do this?

Cameron McKenzie · **Posted:** Sat Aug 22, 2009 3:33 pm

You can do pretty powerful queries and fuzzy logic using the Criteria API and using the 'like' abilities for matching at the beginning, end, or anywhere in a given word or field.

Here's a nice little tutorial on using the Hibernate Criteria API:

http://www.hiberbook.com/HiberBookWeb/learn.jsp?tutorial=09howtousethecriteriaapi

It's not plain english, but this query below using the criteria API to find a User match that has an id greater than (gt) 2, less than (ld) 8, has a notNull email address, and a String match at the end of a field on '.com' It's all highly readable by a Java professional, and can allow you or the end user to really control the types of matches you get.

Code:

Session session = HibernateUtil.beginTransaction();
  Criterion c1 = Restrictions.gt("id", (long)2);
  Criterion c2 = Restrictions.lt("id", (long)8);
  Criterion c3 = Restrictions.isNotNull("emailAddress");
  User user = new User();
  user.setEmailAddress(".com");
  Example c4 = Example.create(user);
  c4.enableLike(MatchMode.END);
  c4.ignoreCase();
  Criteria criteria = session.createCriteria(User.class);
  criteria.add(c1);
  criteria.add(c2);
  criteria.add(c3);
  criteria.add(c4);
  List results = criteria.list();
  HibernateUtil.commitTransaction();
  for (int i = 0; i<results.size(); i++) {
    System.out.println(results.get(i).toString());
  }