Skip to main content

Table 5 The organized groups of SVM- and CRF-based scoring calculated from Twitter text data

From: Classification-based spoken text selection for LVCSR language modeling

Group

SVM-based scoring

CRF-based scoring

C0

CELL-TRN data

CELL-TRN data

C1

Score ≥4.0

Score ≥ 0.9

C2

4.0>score ≥3.0

0.9>score ≥0.8

C3

3.0>score ≥2.0

0.8>score ≥0.7

C4

2.0>score ≥1.0

0.7>score ≥0.6

C5

1.0>score ≥0.0

0.6>score ≥0.5

C6

0.0>score >−1.0

0.5>score ≥0.4

C7

−1.0≥ score >−2.0

0.4>score ≥0.3

C8

−2.0≥ score >−3.0

0.3>score ≥0.2

C9

−3.0≥ score >−4.0

0.2>score ≥0.1

C10

Score ≤−4.0

Score <0.1