Skip to Main Content
Table 3: 
Results of Chinese word segmentation.
ModelsTag SetCTB-5CTB-7CTB-9
F1segPsegRsegF1segPsegRsegF1segPsegRseg
LSTM+MLP {B, M, E, S98.47 98.26 98.69 95.45 96.44 96.45 97.11 97.19 97.04 
LSTM+CRF {B, M, E, S98.48 98.33 98.63 96.46 96.45 96.47 97.15 97.18 97.12 
LSTM+MLP {app, seg98.40 98.14 98.66 96.41 96.53 96.29 97.09 97.16 97.02 
 
Joint-SegOnly {app, seg98.50 98.30 98.71 96.50 96.67 96.34 97.09 97.15 97.04 
Joint-Binary {app,dep98.45 98.16 98.74 96.57 96.66 96.49 97.10 97.16 97.04 
Joint-Multi {app,dep1, ⋯ ,depK98.48 98.17 98.80 96.64 96.68 96.60 97.20 97.31 97.19 
Joint-Multi-BERT {app,dep1, ⋯ ,depK98.46 98.12 98.81 97.06 97.05 97.08 97.63 97.68 97.58 
ModelsTag SetCTB-5CTB-7CTB-9
F1segPsegRsegF1segPsegRsegF1segPsegRseg
LSTM+MLP {B, M, E, S98.47 98.26 98.69 95.45 96.44 96.45 97.11 97.19 97.04 
LSTM+CRF {B, M, E, S98.48 98.33 98.63 96.46 96.45 96.47 97.15 97.18 97.12 
LSTM+MLP {app, seg98.40 98.14 98.66 96.41 96.53 96.29 97.09 97.16 97.02 
 
Joint-SegOnly {app, seg98.50 98.30 98.71 96.50 96.67 96.34 97.09 97.15 97.04 
Joint-Binary {app,dep98.45 98.16 98.74 96.57 96.66 96.49 97.10 97.16 97.04 
Joint-Multi {app,dep1, ⋯ ,depK98.48 98.17 98.80 96.64 96.68 96.60 97.20 97.31 97.19 
Joint-Multi-BERT {app,dep1, ⋯ ,depK98.46 98.12 98.81 97.06 97.05 97.08 97.63 97.68 97.58 

The upper part refers the models based on sequence labeling.

The lower part refers our proposed joint models which are detailed in Section 4.4.

The proposed joint models achieve near or better F1seg than models trained only on Chinese word segmentation.

F1seg, Pseg, and Rseg are the F1, precision, and recall of CWS, respectively.

Close Modal

or Create an Account

Close Modal
Close Modal