Epsilon-Lee/sentiment-benchmark

benchmarking basic models for sentiment classification

sentiment-benchmark

benchmarking basic models for sentiment classification

Models & Results

EmbAvg

Adam lr=0.001 Epoch=100

• Evaluation on devDataset -- Accuracy 0.325 Loss per batch 1.87712673855
• Evaluation on testDataset -- Accuracy 0.287 Loss per batch 1.89027359867

RNNHidAvg

Adam lr=0.001 Epoch=100

• Evaluation on devDataset -- Accuracy 0.395 Loss per batch 1.75760738373
• Evaluation on testDataset -- Accuracy 0.411 Loss per batch 1.74928137207