Neural sentence embedding models for semantic similarity estimation in the biomedical domain

Benchmark Model Rank Results
sentence-embeddings-for-biomedical-texts-on-biossesSupervised combination of: Jaccard, Q-gram, sent2vec, Paragraph vector DM, skip-thoughts, fastText#1Pearson Correlation: 0.871
sentence-embeddings-for-biomedical-texts-on-biossesUnsupervised combination (mean) of: Jaccard, q-gram, Paragraph vector (PV-DBOW) and sent2vec#2Pearson Correlation: 0.846
sentence-embeddings-for-biomedical-texts-on-biossesParagraph vector (PV-DM)#3Pearson Correlation: 0.819
sentence-embeddings-for-biomedical-texts-on-biossesParagraph vector (PV-DBOW)#5Pearson Correlation: 0.804
sentence-embeddings-for-biomedical-texts-on-biossesSent2vec#6Pearson Correlation: 0.798
sentence-embeddings-for-biomedical-texts-on-biossesfastText (skip-gram, max pooling)#8Pearson Correlation: 0.766
sentence-embeddings-for-biomedical-texts-on-biossesQ-gram (q = 3)#9Pearson Correlation: 0.723
sentence-embeddings-for-biomedical-texts-on-biossesSkip-thoughts#10Pearson Correlation: 0.485
sentence-embeddings-for-biomedical-texts-on-biossesfastText (CBOW, max pooling)#13Pearson Correlation: 0.253