Sentence Relatedness on SICK (test + train)
0.61Spearman CorrelationFastSent
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FastSentObjective=Predict adjacent sentences2016.02 | 0.61 | 0.72 | |
| DictRep BOW+embs.Aggregation=Bag-of-Words, Embeddings=Pre-trained2016.02 | 0.61 | 0.7 | |
| Skipgram2016.02 | 0.6 | 0.69 | |
| CBOW2016.02 | 0.6 | 0.69 | |
| FastSent+AEObjective=Adjacent sentence prediction + Denoising autoencoder2016.02 | 0.6 | 0.65 | |
| CPHRASE2016.02 | 0.6 | 0.72 | |
| SkipThought2016.02 | 0.57 | 0.6 | |
| DictRep BOWAggregation=Bag-of-Words, Embeddings=None2016.02 | 0.57 | 0.66 | |
| CaptionRep BOWAggregation=Bag-of-Words2016.02 | 0.56 | 0.65 | |
| CaptionRep RNNAggregation=RNN2016.02 | 0.53 | 0.62 | |
| Unigram TFIDF2016.02 | 0.52 | 0.58 | |
| DictRep RNNAggregation=RNN, Embeddings=None2016.02 | 0.49 | 0.56 | |
| DictRep RNN+embs.Aggregation=RNN, Embeddings=Pre-trained2016.02 | 0.49 | 0.59 | |
| SAE+embs.Embeddings=Pre-trained2016.02 | 0.47 | 0.49 | |
| NMT En to FrTarget Language=French2016.02 | 0.47 | 0.49 | |
| SDAEEmbeddings=None2016.02 | 0.46 | 0.46 | |
| SDAE+embs.Embeddings=Pre-trained2016.02 | 0.46 | 0.46 | |
| NMT En to DeTarget Language=German2016.02 | 0.46 | 0.46 | |
| Paragraph Vec DMArchitecture=DM2016.02 | 0.44 | 0.46 | |
| Paragraph Vec DBOWArchitecture=DBOW2016.02 | 0.42 | 0.46 | |
| SAEEmbeddings=None2016.02 | 0.32 | 0.31 |