Conclusion

• CBR systems can be quite elaborate and are domain specific.
• Like all software, the more important it is, the more time you can spend developing and maintaining it.
• You can also make it a recommender system, so that it doesn't just give answers but suggestions.
• Take Home Points
  1. CBR systems are based around cases.
  2. The similarity metric is very important.
  3. There is a real art to representing the cases.
  4. There is a real art to developing the similarity metric.
  5. Each CBR system is domain specific.
• Reading: For this lecture skim Janet Kolodner's Case-Based Reasoning and read Case Based Reasoning Wiki.
• The reading for next week is Russell and Norvig's Constraint Satisfaction Problems chapter (pp. 258-286).