Table 6 The results of the BERT model's shallow labeling (without text enhancement) by category.

From: Research on automatic labeling of imbalanced texts of customer complaints based on text enhancement and layer-by-layer semantic matching

Serial number

Accuracy

Recall

F1 score

1

1.0000

0.2500

0.4000

2

0.8750

0.9545

0.9130

3

1.0000

0.9961

0.9980

4

0.5333

0.8889

0.6667

5

0.0000

0.0000

0.0000

6

0.7647

0.8667

0.8125

7

0.8824

0.9375

0.9091

8

1.0000

0.6842

0.8125

9

0.8889

0.5000

0.6400

10

0.5000

0.5000

0.5000

11

0.6667

0.8065

0.7299

12

0.5000

0.8571

0.6316

13

1.0000

0.5000

0.6667

14

1.0000

0.8776

0.9348

15

0.9800

0.9899

0.9849

16

0.8571

0.9231

0.8889

Back to article page