Table 1.
Performances of different embedding and classification models
| Embedding models | Diversity | Average F1 | |
|---|---|---|---|
| Proposed | Baseline | ||
| (a) English news | |||
| all-MiniLM-L12-v2 | .7834 | .9064 | .8641 |
| all-distilroberta-v1 | .7767 | .8864 | .8214 |
| all-mpnet-base-v2 | .7653 | .8812 | .8199 |
| (b) Chinese news | |||
| paraphrase-multilingual-MiniLM-L12-v2 | .7962 | .8824 | .7625 |
| distiluse-base-multilingual-cased-v1 | .7321 | .8801 | .7832 |
| distiluse-base-multilingual-cased-v2 | .7415 | .8695 | .7346 |
Values in bold denote the results of the best performing models.