Table 6 Comparison results of overall accuracy, precision, recall, and F1-score for OLI
From: Social media text analytics of Malayalam–English code-mixed using deep learning
Model | Word embedding method | Precision | Recall | F1-score | Accuracy |
|---|---|---|---|---|---|
CNN | Word2Vec FastText | 0.9901 0.9834 | 0.9895 0.9810 | 0.9896 0.9816 | 0.9895 0.9810 |
LSTM | Word2Vec FastText | 0.9930 0.9905 | 0.9930 0.9900 | 0.9929 0.9901 | 0.9930 0.9900 |
GRU | Word2Vec FastText | 0.9965 0.9898 | 0.9965 0.9890 | 0.9964 0.9891 | 0.9965 0.9890 |
BiLSTM | Word2Vec FastText | 0.9937 0.9965 | 0.9935 0.9965 | 0.9935 0.9964 | 0.9935 0.9965 |
BiGRU | Word2Vec FastText | 0.9965 0.9970 | 0.9965 0.9970 | 0.9964 0.9969 | 0.9965 0.9970 |
BiLSTM + CNN | Word2Vec FastText | 0.9960 0.9969 | 0.9960 0.9970 | 0.9959 0.9969 | 0.9960 0.9970 |
BiGRU + CNN | Word2Vec FastText | 0.9960 0.9852 | 0.9960 0.9835 | 0.9959 0.9839 | 0.9960 0.9835 |
LSTM + CNN | Word2Vec FastText | 0.9950 0.9964 | 0.9950 0.9965 | 0.9949 0.9964 | 0.9950 0.9965 |
GRU + CNN | Word2Vec FastText | 0.99701 0.9955 | 0.9970 0.9955 | 0.9969 0.9954 | 0.9970 0.9955 |
Transformer based classification model | XLM-R | 0.9904 | 0.9900 | 0.9901 | 0.9900 |