nlp

AI needs to provide context to be useful.

• NLP is a pipeline.

• many methods can be considered NLP

• national library of medicine did early

• processing language into tokens is difficult

  • language has lots of ambiguity that makes it hard to process.
  • every word counts in unstructured data
    • judging the presence of a particular feature is very difficult.

• word extraction is not adequate

• medical language often uses short abbreviated words that do not follow normal English rules.

• data must be hand annotated to train supervised classifiers.

• models can be evaluated quantitatively or qualitatively to determine performance