Distant Supervision with Transductive Learning for Adverse Drug Reaction Identification from Electronic Medical Records

Table 2

Types of feature extraction for a given sentence. Here, the first character is either (syntactically lemmatized lexicon) or (surface lexicon), and the second character is either (phrase) or (word). BOW stands for bag-of-words. CUI C0033487 is a UMLS concept of propofol. C0031469 is a UMLS concept of phenylephrine. CUI C0020649 is a UMLS concept of hypotension.

Sentences	Types	Example of feature representation	Example of drug-event candidates (d, p, e, y)
On arrival here, propofol was held due to hypotension.	L–P	C0033487 be-hold-due-to C0020649	(C0033487, be-hold-due-to, C0020649, ADR)
	L–W	C0033487 be hold due to C0020649	NA
	S–P	C0033487 was-held-due-to C0020649	(C0033487, was-held-due-to, C0020649, ADR)
	S–W	C0033487 was held due to C0020649	NA
	BOW	On arrival here, propofol was held due to hypotension.	NA
Phenylephrine drip was started for hypotension.	L–P	C0031469 be-start-for C0020649	(C0031469, be-start-for, C0020649, IND)
	L–W	C0031469 be start for C0020649	NA
	S–P	C0031469 was-started-for C0020649	(C0031469, was-started-for, C0020649, IND)
	S–W	C0031469 was started for C0020649	NA
	BOW	Phenylephrine drip was started for hypotension.	NA