Table 4:

The tokens with the most negative and most positive weights in a logistic regression model trained to distinguish technical abstracts from PLS.

Token	Weight	Token	Weight
0	−7.262	people	4.681
.	−6.126	review	4.551
%	−5.379	We	4.461
CI	−4.986	This	3.413
;	−4.821	that	2.943
95	−4.593	The	2.836
significant	−4.273	side	2.722
R	−3.726	who	2.671
1	−3.685	blood	2.515
There	−3.477	found	2.514
bias	−3.303	searched	2.407
criteria	−3.263	The	2.114
outcome	−3.247	results	2.098
(	−3.195	their	2.022
inclusion	−3.148	current	1.984