Closed mgeorgescu closed 10 years ago
We treat substances as "common nouns", not NEs, and I think that should apply to drugs as well. Named entities (NEs) are individuals, such as a specific animal ("Lassie"), as opposed to a class of entities ("dog"). Consensus substance examples: gold, water, milk.
Remember, to enter upper-case concepts such as LSD in the AMR Editor, enter !LSD (with a !) to overwrite the AMR Editor's default assumption that upper-case items are strings (and not concepts).
I agree that we currently don't normalize synonyms such as methamphetamine, meth, crystal, desoxyephedrine to a canonical form. We do correct typos and expand true abbreviations such as "Dr." -> doctor and "Mass." -> "Massachusetts", but we keep acronyms such as CEO and LSD and we keep clipped forms such as meth, phone, exam, flu, fax, gym instead of expanding to methamphetamine, telephone, examination, influenza, facsimile, gymnasium.
A pharmaceutical brand name like Tylenol would be product
, correct? I guess it makes sense to distinguish those from general kinds of drugs, like aspirin.
DF data contains different names of drugs:
e.g. LSD, meth, crystal, maryjane
We would opt for treating them as common nouns.
If we need to treat them as NEs: The concept for the NE should be "drug"/"substance" even if it's not part of the NE list. Or should we opt for "thing"?
meth > methamphetamine
Note: On issue #68, the approach was to map certain slang expressions to standard form: 'em > they wanna > want
So point 2 could be contradictory to the approach on #68
Any feedback is appreciated. Thank you.