AMIA Now NLP Tutorial Exercises
These pages and files were part of a live demo and tutorial on NLP given as part of AMIA Now 2010.
- http://gskinner.com/RegExr/ - interface to try RegEx's
Find everyone with Atrial fibrillation
- First, cut and paste the ECG impressions above into the RegEx browser
- Then, create regular expressions to find all instances of Atrial fibrillation
- Find everyone with evidence of a heart attack (myocardial infarction) or active ischemia, using the same dataset
Cut and paste the clinical note into the RegEx browser
- Find all occurences of "peripheral vascular disease" (and its synonyms). How many did you find?
- Has this person had a heart attack yet?
Task 1: Compare the differences between concept and word searching
We have generated a test version of a evidenced-based medicine / medication info search that we use from our EMR.
Notice there is a radio button to select concept vs. word searches. This will change the results in the "KM Documents" hits you see. This is searching the KnowledgeMap database of documents (about 15,000 documents from medical and resident curricula at Vanderbilt). Try any search, but here are a few examples to try:
- mad cow disease
- hepatolenticular degeneration
Task 2: Peruse KM concept index for a sample document
Look at the sample h&p document attached and the concept indices output from KMCI.
Eight output formats are available: full XML, limited XML, pipe-delimited, and several HTML outputs. The most "human readable" is attached below.