Drug Disease Relation Extraction from Biomedical Literature Using NLP and Machine Learning

Table 2

Extracted features.

Feature	Example
Frequency features:
Number of named entities	2
Number of drugs	1
Number of diseases	1
Number of verbs between two NE	1
Number of words between NEs	4
Bag-of-Word	Preliminary = 1; evidence = 1; suggests = 1; that = 1; interferons = 1; beta = 1; may = 1; also = 1; induce = 1; regression = 1; of = 1 metastatic = 1; renal = 1; cell = 1; carcinoma = 1;
Lexical features:
Sequence of words of the NE	Interferons beta_ metastatic renal cell carcinoma
Sequence of words between every two NEs	May_also_induce_regression
Sequence of 3 words before each NE	Preliminary_evidence_suggests; Also_induce_regression
Sequence of 3 words after each NE	May_also_induce; null
Morphologic features:
Sequence of lemmas of the words between every two NEs	May_also_induce_regression
Sequence of lemmas of the 3 words before each NE	Preliminary_evidence_suggest; Also_induce_regression
Sequence of lemmas of the 3 words after each NE	May_also_induce; null
Syntactic features:
Sequence of POS of NE	NNS_NN_JJ_JJ_NN_NN
Sequence of POS of words between every two NEs	MD_RB_VB_NN
Sequence of POS of 3 words before each NE	JJ_NN_VBS; RB_VB_NN
Sequence of POS of 3 words after each NE	MD_RB_VB; NULL
Verbs sequence among every two NEs	Induce;
First verb preceding every NE	Suggest; induce
First verb after every NE	Induce; null
Semantic features:
Semantic type sequence	TREAT_DIS