Table 9.

Effect of each feature on system performance II. The first column shows the values when only the word and preceding class features were used in the SVM learning. The other columns shows the values when the word and preceding class features plus one other feature were used in the learning. The parenthesized values are p-values. The values in bold have a statistically significant difference from the base value. A difference is labeled statistically significant when the p-value is less than 0.05 on the Wilcoxon signed-ranks sum test (two-sided).

	word+pc. (base)	word+pc. +POS	word+pc. +orth.	word+pc. +pre.	word+pc. +suf.	word+pc. +dic.
Precision	0.8000	0.7813 (0.004)	0.7886 (0.014)	0.7867 (0.020)	0.8014 (0.770)	0.7964 (0.084)
Recall	0.5509	0.6423 (0.002)	0.6786 (0.002)	0.6374 (0.002)	0.7035 (0.002)	0.6410 (0.002)
Balanced f-score	0.6524	0.7118 (0.002)	0.7295 (0.002)	0.7041 (0.002)	0.7492 (0.002)	0.7102 (0.002)