Table 2.

Performance numbers on LR over all feature construction approaches for detecting comparative reviews.

Feature construction	Precision	Recall	F 1
Bow	0.7297	0.6658	0.6907
Bow + Cmp	0.8298	0.7013	0.7523
Bow + Sug	0.7983	0.65	0.7124
Bow + Cmp + Sug	0.7821	0.6961	0.731
Sug + Per	0	0	0
Mix	0.7611	0.7132	0.7302
Bow + Per	0.8175	0.6842	0.7409
Mix + Per	0.7861	0.6789	0.724
Mix + Per + Sug	0.7347	0.6947	0.7092
Mix + Sug + Cmp	0.7377	0.6816	0.7047
Mix + Sug + Cmp + Per	0.7551	0.6618	0.7003