Table 5.

Abbreviation detection

Top 10	[1]	w 2	[1-2]	w 2	[1-3]	w 2
1	Contains period	0.30	Contains period	0.35	S2	5885.83
2	All upper case	0.02	C(Lnorm, •)	0.18	S3	4855.66
3	Contains digit	0.01	logλ	0.13	S4	1999.51
4	-	-	C(¬Lnorm, ¬•)	0.12	S2, S3	1798.60
5	-	-	C(¬Lnorm, •)	0.09	logλ	1180.39
6	-	-	C(Lnorm, ¥)	0.09	S5	894.98
7	-	-	All upper case	0.02	S4, S5	715.70
8	-	-	Contains digit	8.16E-5	S2, S5	617.98
9	-	-	-	-	S2, S4, S5	474.86
10	-	-	-	-	S3, S4, S5	256.81
Top 10	[1-4]	w 2	[1-5]	w 2	[1-6]	w 2
1	S2	1063.78	S5	1027.15	LT	952.62
2	S3	962.33	S4, S5	914.02	Mean-LT	952.62
3	S2, S3	507.82	S2, S5	610.69	All upper case	549.64
4	S4	391.68	S2, S4, S5	527.28	S3, S4, S5	529.85
5	S3, S4, S5	379.70	S2	463.94	S3, S5	521.60
6	S3, S5	325.68	S3, S4, S5	274.81	erforderl.	403.54
7	S5	265.62	S3, S5	253.30	pathol.	392.23
8	S4, S5	222.55	Mean-LT	145.91	verschiebl.	375.40
9	logλ	143.67	LT	145.91	d-lat.	358.11
10	S2, S5	129.90	S2, S4	90.13	entzündl.	345.21

Top 10 feature rankings per feature set (1 Rule-based features; 2 Statistical features; 3 Scaling features; 4 Language-dependent features; 5 Length features; 6 Word type features). Length (LT); w²: Weight based feature relevance criterion.