Homework: Information Retrieval

            (Introduction to Medical Informatics)
         (http://www.cpmc.columbia.edu/edu/textbook)

Assignment:

Log onto Medline and try to find articles only from 1993
about aspirin causing kidney failure. Your search must do
better than "aspirin and kidney failure."

1. First try to use the keywords alone. To do this, suffix
     every phrase with .DE. (this tells it to look only in
     the keyword fields).
     
     (kidney failure).DE.

2. Then try to use the free text search method alone. To do
     this, suffix every word with .TI,AB. (this tells it to
     look only in the title and abstract fields, avoiding
     the keywords).
     
     (kidney failure).TI,AB.

Hints:
     
     kidney failure means one word after the other
     kidney with failure means same sentence
     kidney same failure means same paragraph
     kidney and failure means same document
     not failure means documents without failure
     kidney or failure looks for either word in the document
     
     Each "search" (query) is numbered. You can combine
     searches by using the search numbers.
          
          1: kidney
          2: 1 with failure
     
     You can select articles from 1993 using
          ... and 1993.YR.
     
     approximate synonyms (there will be many others)
          renal = kidney
          failure = insufficiency = compromise = impairment
     
     Show the results of a search with the d option.

Answer format:

Give both queries (keyword version and free text version)
and give a list of the articles returned for each query (you
can give main author and the first 5 words of the title).

Do your best to estimate the recall and precision for each
query, showing what numbers you chose and why you chose
them.


Issues for class discussion on Homework: Information Retrieval

1. cannot assume that coders will use the keywords as you
     expect them; they do not know your context

2. keyword search does not require listing multiple
     synonyms; you only need to find the correct one

3. recall vs precision: if you get too few documents, then
     broaden the search by using more general terms (eg
     kidney disease for kidney failure; NSAI for aspirin) or
     leaving out some criteria (eg kidney and aspirin)