Cross-domain comparison for feature-set `baseline'
optimal variances per pair
optimal variances for training set
train\eval
ps
tg
jh
ws
ps
44.6%
28.4%
30.9%
25.4%
tg
33.2%
43.1%
34.0%
24.1%
jh
38.0%
33.7%
51.3%
27.5%
ws
33.0%
29.3%
32.2%
40.4%
train\eval
ps
tg
jh
ws
ps
44.6%
26.9%
30.9%
25.2%
tg
32.2%
43.1%
32.5%
23.7%
jh
35.0%
32.5%
51.3%
26.5%
ws
33.0%
28.5%
31.2%
40.4%
Cross-domain comparison for feature-set `1to3par'
optimal variances per pair
optimal variances for training set
train\eval
ps
tg
jh
ws
ps
53.0%
29.2%
31.0%
24.6%
tg
33.3%
47.9%
33.2%
23.5%
jh
39.1%
35.2%
57.4%
29.2%
ws
33.6%
28.0%
33.3%
44.5%
train\eval
ps
tg
jh
ws
ps
53.0%
29.2%
28.9%
24.1%
tg
31.6%
47.9%
30.4%
23.0%
jh
38.0%
34.1%
57.4%
28.5%
ws
33.6%
27.3%
33.2%
44.5%
Cross-domain comparison for feature-set `lename-l'
optimal variances per pair
optimal variances for training set
train\eval
ps
tg
jh
ws
ps
52.0%
28.4%
31.3%
25.2%
tg
35.3%
48.9%
34.2%
24.5%
jh
38.6%
35.1%
57.0%
27.8%
ws
34.8%
29.1%
32.1%
44.2%
train\eval
ps
tg
jh
ws
ps
52.0%
28.4%
31.3%
24.8%
tg
33.8%
48.9%
34.0%
24.2%
jh
38.0%
35.1%
57.0%
27.6%
ws
34.2%
28.8%
32.1%
44.2%