Table 8 Comparative performance analysis of Chinese Pre-trained language models in automotive news classification.

From: Qwen TextCNN and BERT models for enhanced multilabel news classification in mobile apps

Category

WWM BERT

BERT+TextCNN hybrid

Prec.

Rec.

F1

Prec.

Rec.

F1

Vehicle technical specifications

New car analysis

0.995

0.980

0.988

–

–

–

Spy photos

0.927

0.981

0.953

–

–

–

Declaration images

0.929

0.963

0.946

–

–

–

New car pricing

0.935

0.903

0.919

–

–

–

Pre-sale

0.906

0.897

0.902

–

–

–

New car launch

0.932

0.866

0.898

–

–

–

Real car exposure

0.857

0.923

0.889

–

–

–

Configuration exposure

0.842

0.889

0.865

–

–

–

Official images

0.917

0.815

0.863

–

–

–

Pre-heating

0.902

0.798

0.847

–

–

–

Release appearance

0.871

0.802

0.835

–

–

–

New car arrival

0.727

0.800

0.762

–

–

–

Consumer guidance

Vehicle brand analysis

–

–

–

0.969

0.963

0.966

Comparison guide

–

–

–

0.955

0.957

0.956

Single vehicle guide

–

–

–

0.911

0.927

0.919

Car purchase manual

–

–

–

0.939

0.900

0.919

Purchase techniques

–

–

–

0.939

0.859

0.897

Evaluation guide

–

–

–

0.904

0.906

0.905

Test drive

–

–

–

0.876

0.919

0.897

Multiple vehicle guide

–

–

–

0.928

0.852

0.888

Marketing guide

–

–

–

0.971

0.745

0.843

Vehicle showcase

–

–

–

0.886

0.795

0.838

Car sharing

–

–

–

0.945

0.729

0.823

Dealership pricing

–

–

–

0.829

0.782

0.805