Research Article
Research on Application of Intelligent Corpus Annotation of Entity Extraction with Construction of Knowledge Graph
Table 5
The evaluation of different deep learning entity extraction algorithms.
| Experiment content | Model | P | R | F1 |
| Different word vectors | BiLSTM-CRF | 0.810 | 0.806 | 0.807 | Word2vec-BiLSTM-CRF | 0.819 | 0.829 | 0.823 | Bert-BiLSTM-CRF | 0.873 | 0.898 | 0.885 |
| Different downstream models | RoBERTa-WWM-CRF | 0.778 | 0.846 | 0.810 | RoBERTa-WWM-LSTM-CRF | 0.812 | 0.868 | 0.839 | RoBERTa-WWM-BiGRU-CRF | 0.829 | 0.904 | 0.865 |
| This experimental model | RoBERTa-WWM-BiLSTM-CRF | 0.889 | 0.900 | 0.894 |
|
|