Sentiment Classification on SemEval Twitter Sentiment Analysis 2014 (test)
0.748Avg F1 ScoreEnsemble model
Evaluation Results
| Method | Links | |
|---|---|---|
| Ensemble modelcomponents=10 CNNs and 10 LSTMs2017.04 | 0.748 | |
| Previous best historical scores2017.04 | 0.744 | |
| CNNEmbedding=word2vec, Convolution size=[1,2,3]2017.04 | 0.735 | |
| CNNEmbedding=fasttext, Convolution size=[3,4,5]2017.04 | 0.733 | |
| LSTMEmbedding=word2vec2017.04 | 0.733 | |
| CNNEmbedding=word2vec, Convolution size=[5,6,7]2017.04 | 0.732 | |
| LSTMEmbedding=fasttext2017.04 | 0.73 | |
| LSTMEmbedding=glove2017.04 | 0.73 | |
| LSTMEmbedding=word2vec, Fully connected layer=false2017.04 | 0.725 | |
| CNNEmbedding=word2vec, Convolution size=[3,4,5], Fully connected layer=false2017.04 | 0.724 | |
| CNNEmbedding=word2vec, Convolution size=[3,4,5]2017.04 | 0.723 | |
| LSTMEmbedding=word2vec, Distant training=false2017.04 | 0.719 | |
| CNNEmbedding=word2vec, Convolution size=[3,4,5], Distant training=false2017.04 | 0.716 | |
| CNNEmbedding=glove, Convolution size=[3,4,5]2017.04 | 0.714 | |
| CNNEmbedding=word2vec, Convolution size=[3,4,5], Class weights=false2017.04 | 0.679 | |
| LSTMEmbedding=word2vec, Class weights=false2017.04 | 0.661 | |
| Logistic regressionbaseline=1-3 grams2017.04 | 0.629 |