Target-dependent sentiment classification on Twitter (test)
78.61AccuracyTNet-ATT(+AS)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TNet-ATT(+AS)Attention Supervision (both)=true, Attention instead of CNN=true2019.06 | 78.61 | 77.72 | — | |
| TNet-ATT(+ASa)Attention Supervision (correct)=true, Attention instead of CNN=true2019.06 | 78.47 | 77.63 | — | |
| TNet-ATT(+ASm)Attention Supervision (incorrect)=true, Attention instead of CNN=true2019.06 | 78.08 | 77.42 | — | |
| TNetsource=re-implementation2019.06 | 77.6 | 76.82 | — | |
| TNet(+KT)Keep Training=true2019.06 | 77.54 | 76.78 | — | |
| TNet-ATTAttention instead of CNN=true2019.06 | 77.46 | 76.53 | — | |
| TNet-ATT(+KT)Keep Training=true, Attention instead of CNN=true2019.06 | 77.46 | 76.58 | — | |
| TNet (original)source=Li et al., 20182019.06 | 74.97 | 73.6 | — | |
| HDGCNInitialized embeddings=GloVe2021.06 | 73.41 | — | 71.52 | |
| TNet-LFInitialized embeddings=GloVe2021.06 | 72.98 | — | 71.43 | |
| HDGCNInitialized embeddings=BERT-base2021.06 | 72.69 | — | 71.23 | |
| AOAInitialized embeddings=GloVe2021.06 | 72.3 | — | 70.2 | |
| ASGCN-DGInitialized embeddings=GloVe2021.06 | 72.15 | — | 70.4 | |
| ASGCN-DTInitialized embeddings=GloVe2021.06 | 71.53 | — | 69.68 | |
| TC-LSTM2015.12 | 71.5 | 69.5 | — | |
| Target-dep+Lexicon=With sentiment lexicon features2015.12 | 71.1 | 69.9 | — | |
| ASCNNInitialized embeddings=GloVe2021.06 | 71.05 | — | 69.45 | |
| TD-LSTM2015.12 | 70.8 | 69 | — | |
| Target-dep2015.12 | 69.7 | 68 | — | |
| MN(+AS)Attention Supervision (both)=true2019.06 | 69.64 | 67.88 | — | |
| MN(+ASa)Attention Supervision (correct)=true2019.06 | 69.17 | 67.47 | — | |
| MN(+ASm)Attention Supervision (incorrect)=true2019.06 | 68.9 | 67.2 | — | |
| MN(+KT)Keep Training=true2019.06 | 67.78 | 66.18 | — | |
| MNsource=re-implementation2019.06 | 67.71 | 66.17 | — | |
| LSTM2015.12 | 66.5 | 64.7 | — | |
| AdaRNN-comb2015.12 | 66.3 | 65.9 | — | |
| AdaRNN-w/E2015.12 | 65.8 | 65.5 | — | |
| AdaRNN-w/oE2015.12 | 64.9 | 64.4 | — | |
| SVM-dep2015.12 | 63.4 | 63.3 | — | |
| Recursive NN2015.12 | 63 | 62.8 | — | |
| SVM-indep2015.12 | 62.7 | 60.2 | — |