Basic models of IR were discussed today. In evaluating these models (metrics) following is required/expected :
1) Allowing partial matches. Returning documents which miss a few keywords . This brings us to next point. Which words to ignore or give importance to.
2) Allotting weights to keywords.
3) Relevance should not depend on size. (repeating a set of words again and again should not make a file more relevant to another file).
